AN ABSTRACT OF THE DISSERTATION OF
Amin Mirkouei for the degree of Doctor of Philosophy in Industrial Engineering presented on June 6, 2016 Title: Techno-Economic Optimization and Environmental Impact Analysis for a Mixed-Mode Upstream and Midstream Forest Biomass to Bio-Products Supply Chain
Abstract approved:
Karl R. Haapala
Growing awareness and concern within society over the use of and reliance on fossil fuels has stimulated research efforts in identifying, developing, and selecting alternative energy sources and energy technologies. Bioenergy represents a promising replacement for conventional energy, due to reduced environmental impacts and broad applicability.
Sustainable
energy
challenges,
however,
require
innovative
manufacturing technologies and practices to mitigate energy and material consumption. This research aims to facilitate sustainable production of bioenergy from forest biomass and to promote deployment of novel processing equipment such as transportable biorefineries. The study integrates knowledge from the renewable energy production and supply chain management disciplines to evaluate economic and environmental targets of bioenergy production with the use of the multi-criteria decision making approach. The presented approach herein includes qualitative and quantitative methods to address the existing challenges and gaps in the bioenergy manufacturing system. The qualitative method employs decision tree analysis to classify the potential biomass
harvesting sites, considering biomass quality and availability. The quantitative method proposes mathematical models to optimize the upstream and midstream biomass-tobioenergy supply chain cost, using mixed bio-refinery modes (transportable and fixed) and transportation pathways (traditional and new). The supply chain environmental impacts are assessed by considering the carbon footprint of the harvesting, collection, size reduction, transportation activities, and bio-refinery processing. While transportable bio-refineries are shown to reduce biomass-to-bioenergy supply chain costs, production and deployment of transportable bio-refineries are limited due to operational challenges associated with undeveloped mixed-mode bioenergy supply chains, as well as supply uncertainty. A case study for northwest Oregon, USA is undertaken using actual data to verify the proposed approach.
©Copyright by Amin Mirkouei June 6, 2016 All Rights Reserved
Techno-Economic Optimization and Environmental Impact Analysis for a Mixed-Mode Upstream and Midstream Forest Biomass to Bio-Products Supply Chain
by Amin Mirkouei
A DISSERTATION submitted to
Oregon State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
Presented June 06, 2016 Commencement June 2017
Doctor of Philosophy dissertation of Amin Mirkouei presented on June 6, 2016 APPROVED:
Major Professor, representing Industrial Engineering
Head of the School of Mechanical, Industrial and Manufacturing Engineering
Dean of the Graduate School
I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my dissertation to any reader upon request.
Amin Mirkouei, Author
ACKNOWLEDGEMENTS
First, I would like to say thank you to my parents and my sister, and my family for all of their guidance and support in all aspects of my life. They are the light of my life and my inspiration to be a better person every day. Their love gave me the necessary strength to overcome all challenges I faced on my program.
I would like to express my sincere gratitude to Dr. Karl. R. Haapala (Associate Professor at School of Mechanical, Industrial, and Manufacturing Engineering) for his invaluable advice, support, and guidance throughout the PhD program. His supervision and inspiration greatly improved my research skills.
Special thanks should be given to my graduate advisory committee Dr. John Sessions (Chair and Professor at Department of Forest Engineering, Resources and Management), Dr. Ganti Murthy (Associate Professor at Department of Biological and Ecological Engineering), Dr. David Porter (Associate Professor at School of Mechanical, Industrial, and Manufacturing Engineering), and Dr. Zhaohui Wu (Professor at College of Business) at Oregon State University, for their valuable insights and advice during this research.
In the end, I wish to express gratitude to the School of Mechanical, Industrial, and Manufacturing Engineering at Oregon State University for support of this research.
TABLE OF CONTENTS Page Chapter 1:
Introduction ............................................................................................. 1
Research Motivation ...................................................................................... 1 Background .................................................................................................... 2 Research Objectives ....................................................................................... 6 Research Tasks ............................................................................................... 7 Dissertation Outline........................................................................................ 7 References ............................................................................................................... 11 Chapter 2:
A Review and Future Directions in Techno-Economic Modeling and
Optimization of Upstream Forest Biomass to Bio-Oil Supply Chains ....................... 16 Abstract ........................................................................................................ 16 Introduction .................................................................................................. 16 Review Methodology ................................................................................... 25 Narrative Literature Review Method ........................................................ 26 Systematic Literature Review Method ..................................................... 27 Narrative Literature Review ......................................................................... 29 Harvesting and Collection ........................................................................ 30 Logistics.................................................................................................... 34 Pretreatment .............................................................................................. 40 Storage ...................................................................................................... 47 Techno-Economic Modeling .................................................................... 51 Systematic Literature Review ...................................................................... 57 Analysis of Publication Data .................................................................... 57 Analysis of Citation Data ......................................................................... 62 Analysis of Keywords .............................................................................. 63 Analysis of Research Methodologies ....................................................... 66 Discussion .................................................................................................... 66 Conclusions and Future Direction ................................................................ 70 Acknowledgment ..................................................................................................... 73 References ............................................................................................................... 73
TABLE OF CONTENTS (Continued) Page Chapter 3:
Multi-criteria Decision Making for Sustainable Bio-Oil Production
using a Mixed Supply Chain ....................................................................................... 94 Abstract ........................................................................................................ 94 Introduction .................................................................................................. 95 Background .................................................................................................. 99 Methodology .............................................................................................. 104 Phase 1. Qualitative Analysis Method .................................................... 105 Phase 2. Quantitative Analysis Method .................................................. 107 Case Study .................................................................................................. 110 Phase 1. Classification of Harvesting Areas ........................................... 110 Phase 2. Mathematical Optimization Model .......................................... 112 Computational Results............................................................................ 115 Sensitivity Analysis .................................................................................... 116 Effect of Transportable Bio-refinery Cost .............................................. 117 Effect of Available Amount of Forest Biomass ..................................... 118 Conclusions ................................................................................................ 119 Acknowledgment ................................................................................................... 122 Annex A: Nomenclature ........................................................................................ 122 References ............................................................................................................. 124 Computational Codes ................................................................................................ 129 Chapter 4:
Evolutionary Decision Making for Biomass-based Energy Supply
Chains
............................................................................................................. 135 Abstract ...................................................................................................... 135 Introduction ................................................................................................ 136 Background ................................................................................................ 140 Methodology .............................................................................................. 148 Economic Analysis (Phase 1) ................................................................. 149 Environmental Impact Analysis (Phase 2) ............................................. 153 Application of the Method ......................................................................... 158
TABLE OF CONTENTS (Continued) Page Results and Discussion ............................................................................... 161 Sensitivity Analysis .................................................................................... 165 Effect of Mobile Bio-refinery Cost ........................................................ 165 Effect of Available Amount of Forest Biomass ..................................... 167 Conclusions ................................................................................................ 169 Acknowledgment ................................................................................................... 171 Annex A: Nomenclature ........................................................................................... 172 References
............................................................................................................. 174
Computational Codes ................................................................................................ 177 Chapter 5:
Summary and Conclusions ................................................................. 182
Summary .................................................................................................... 182 Conclusions ................................................................................................ 183 Contributions .............................................................................................. 184 Opportunities for Future Research ............................................................. 186 Appendices ............................................................................................................. 189 Appendix A: Reducing the Cost and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains .................................................................. 191 Appendix B: A Network Model to Optimize Upstream and Midstream Biomass-toBioenergy Supply Chain Costs ................................................................................. 205 Appendix C: Evolutionary Optimization of Bioenergy Supply Chain Cost with Uncertain Forest Biomass Quality and Availability ................................................. 217 Appendix D: Reducing Greenhouse Gas Emissions for Sustainable Bio-Oil Production Using a Mixed Supply Chain ................................................................. 224 Appendix E: Integration of Machine-Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process ............................ 234
LIST OF FIGURES Figure .......................................................................................................................Page Figure 2.1. Upstream segment of the general biomass-to-bioenergy supply chain: a) conventional structure and b) proposed structure (paper sections are also indicated) ..................................................................................................................................... 27 Figure 2.2. Classification framework for quantitative method [78] ........................... 28 Figure 2.3. Classification framework for qualitative method [78] ............................. 29 Figure 2.4. Harvesting and collection of forest biomass: a) harvested biomass and b) collection operation system .................................................................................... 30 Figure 2.5. Forest biomass supply logistics for energy production: a) grinding and loading and b) transportation (chip van) ..................................................................... 39 Figure 2.6. General biomass pyrolysis conversion and post-conversion processing [38] .............................................................................................................................. 42 Figure 2.7. A mobile fast pyrolysis bio-refinery (Courtesy of Phillip C. Badger, Renewable Oil International LLC) ............................................................................. 47 Figure 2.8. Renewable energy storage solutions using a) biomass (chips) and b) biooil (pyrolysis oil) ......................................................................................................... 50 Figure 2.9. Increase in publications of biomass-to-bio-oil SC research (Jan. 2000 to June 2015, *estimated for July-Dec. 2015) ................................................................ 58 Figure 2.10. Increase in citations each year (Jan. 2000 to June 2015, *estimated for July-Dec. 2015) ........................................................................................................... 62 Figure 2.11. Bibliometric map of keywords (density visualization from VOSviewer software) ..................................................................................................................... 65 Figure 3.1. Energy consumption in the United States [4] ........................................... 96 Figure 3.2. Upstream and midstream biomass-to-bioenergy supply chain entities .... 97 Figure 3.3. Integrated multi-criteria decision-making approach for mixed bioenergy supply chains ............................................................................................................. 105 Figure 3.4. A schematic of a mixed-mode and mixed-pathway biomass-to-bio-oil supply chain (question marks indicate location and asset decision points) .............. 110 Figure 3.5. Supervised classification analysis for providing biomass from potential harvesting sites (R results) ........................................................................................ 112
LIST OF FIGURES (Continued) Figure .......................................................................................................................Page Figure 3.6. Effect of transportable bio-refinery cost on annual supply chain costs: Case 1, transportable bio-refinery costs are reduced by 50%, and Case 2, transportable bio-refinery costs are increased by 50% ............................................. 118 Figure 3.7. Effect of available amount of forest biomass on annual supply chain costs: Case 3, the amount of biomass is decreased by 50%, and Case 4, the amount of biomass increased by 50% ........................................................................................ 120 Figure 4.1. Conventional biomass-based energy supply chain entities .................... 137 Figure 4.2. A mixed biomass-to-bio-oil supply chain (mobile bio-refinery image courtesy Phillip C. Badger; fixed bio-refinery image courtesy UPM Lappeenranta Bio-refinery, Finland) ............................................................................................... 138 Figure 4.3. Pyrolysis conversion process principles ................................................. 145 Figure 4.4. Decision support system for sustainable bioenergy production ............. 148 Figure 4.5. Cradle to grave global warming potential for bio-oil production and consumption .............................................................................................................. 164 Figure 4.6. Effect of mobile bio-refinery cost on environmental and economic measures .................................................................................................................... 166 Figure 4.7. Effect of the available amount of forest biomass on environmental and economic measures ................................................................................................... 168
LIST OF TABLES Table ........................................................................................................................Page Table 2.1. Previously reported bio-oil production costs ............................................ 45 Table 2.2. Attributes of several types of biomass and bio-oil [56] ............................. 46 Table 2.3. Current and future generation bio-oil production technologies [175,176] 46 Table 2.4. Techno-economic studies for bioenergy production via fast pyrolysis (1991-2016)................................................................................................................. 55 Table 2.5. Portable and fixed bio-refinery characteristics [57,148] ........................... 57 Table 2.6. Top ten journals identified based on number of records (Jan. 2000 to June 2015) ........................................................................................................................... 59 Table 2.7. Top ten countries based on number of records (between Jan. 2000 and June 2015) ........................................................................................................................... 60 Table 2.8. Most productive scholars based on number of records (Jan. 2000 to June 2015) ........................................................................................................................... 60 Table 2.9. Ten most productive organizations based on number of records (Jan. 2000 to June 2015) ............................................................................................................... 61 Table 2.10. Ten most common research areas based on number of records (Jan. 2000 to June 2015) ............................................................................................................... 62 Table 2.11. Ten most-cited journals (between Jan. 2000 and June 2015) .................. 63 Table 2.12. Ten most-cited studies and scholars (between Jan. 2000 and June 2015) 64 Table 2.13 Classification of the ten most-cited studies .............................................. 66 Table 3.1. Selected bio-refinery attributes [37]…………………………………….114 Table 3.2. Annual truck costs based on capacity [25]……………………………...114 Table 3.3. Annual costs of upstream entities [25,37]………………………………115 Table 3.4. Effect of transportable bio-refinery cost on the overall annual cost…….117 Table 3.5. Effect of available amount of forest biomass on the overall annual cost.119
LIST OF TABLES (Continued) Table ........................................................................................................................Page Table 4.1. Example of training set for quality and accessibility of harvesting sites . 150 Table 4.2. Identified weights from the support vector machine method (R results) 150 Table 4.3. Example of testing set of harvesting site ................................................. 151 Table 4.4. Cradle-to-grave global warming potential for bio-oil production and consumption over 20-year time period [2], [3] ......................................................... 157 Table 4.5. Base case transportation details ............................................................... 160 Table 4.6. Collection site attributes .......................................................................... 161 Table 4.7. Effect of mobile bio-refinery cost on the overall annual cost.................. 165 Table 4.8. Effect of available amount of forest biomass on the overall annual cost 167
1 Chapter 1:
Introduction
Research Motivation According to the United States (U.S.) Energy Information Administration, an average of 7.4 million barrels of crude oil per day was imported in 2015, representing a value of about $370 million per day [1]. Renewable energy sources have been suggested as part of a comprehensive strategy to cut the use of oil in half by 2030, in addition to phasing out the use of coal [2]. This aggressive goal has been driven by the following national priorities [3]:
“Dramatically reduce dependence on foreign oil”
“Promote the use of diverse domestic, and sustainable energy resources”
“Establish an advanced bio-industry and create jobs”
“Reduce carbon emissions from energy production and consumption”
Biomass represents a promising renewable resource due to its abundance and low price, but only 45% is used due to logistical challenges and immature production technologies [4]. Thus, attention must be placed on the development of biomass-based energy resources, such as the conversion of forest harvest residues in the Pacific Northwest, which can reduce wildfire risks, improve the soils and land management, and promote domestic bioenergy production. Biomass includes forest resources, agricultural residues, perennial grasses, woody crops, and wastes (e.g., food wastes, urban wood waste, and municipal solid wastes), as well as algae and other resources [4]. Biomass, unlike other renewable energy resources, can be converted to chemicals and
2 transportation fuels (e.g., biodiesel and bioethanol) and also can be converted to power [3].
According U.S. Department of Energy, two-thirds of the U.S. oil consumption are in the transportation sector, which generates one-third of the U.S. greenhouse gas (GHG) emissions [3]. Special attention should be placed on biomass-based energy because biomass is a key renewable resource that can address the mentioned national priorities, including supplementing fossil-based energy. On the other hand, replacing fossil energy imports with bioenergy offers a significant opportunity for domestic job creation [5]. Recent studies reported that the transition to renewable energy sources would create 1 million jobs by 2030 and 2 million jobs by 2050, even after offsetting the loss of fossil fuel jobs [5].
Background Forest biomass is the most abundant and low price type of biomass, and is composed of lignin, hemicellulose, and cellulose [6]. In addition to its low price and abundance, forest biomass (e.g., western juniper and tan-oak in Pacific Northwest) is preferable for several other reasons: the possibility of using non-edible biomass, the ability to obtain a higher net energy yield for the production of fuels and chemicals, and the ability to avoid unnecessary land use change [7]. Bio-oil from forest biomass is a potential source of a number of valuable chemicals. More than 350 products of forest biomass pyrolysis have been identified, including acetic acids, tars, esters, aromatics, sugars, alcohols, turpentine, and methanol [6]. Therefore, investigation into the development of low-cost
3 separation techniques is needed to enable commercialization of these chemicals [6]. The commercial competitiveness of bio-oil depends on logistics factors (e.g., availability and industrial practices), the scale and location of bio-refineries, and other factors [8].
Various aspects of the upstream segment of biomass to bioenergy SC have previously been discussed by De Meyer et al. [9], and it is essential to develop efficient technologies to assist stakeholders to overcome the identified challenges [10]. Secondgeneration conversion technologies, which use nonfood crops such as forest biomass (stumps, leaves, bark, roots, small stems, and branches of live and dead trees) are currently under development, but exhibit unclear technical and economic performance [11]. Researchers continue to evaluate the advantages of conversion technologies in terms of energy efficiency [12]. Additionally, several researchers argued and underlined that conversion and pretreatment technology (e.g., pyrolysis) development can mitigate logistics issues related to biomass to bio-oil SC networks [13].
Bio-oil is a low-grade liquid fuel that can be used in industrial heating (e.g., furnaces, fueling heaters, and boilers), industrial turbines, stationary diesel engines, and upgrading to transport fuels [14]. Additionally, it can serve as a source of several chemicals [15]. Bio-oil has limitations compared to biofuel, however, such as low energy density and corrosive properties that are harmful to existing engines [16]. The high post-conversion processing cost is an essential issue to be considered in producing higher quality hydrocarbon fuel from bio-oil. The physical properties of bio-oil, such
4 as lower heating value (half of conventional fuel oil), poor volatility, solids content, high viscosity, corrosiveness, and incompatibility with conventional fuels limit the range of bio-oil. Fast pyrolysis process represents a conversion technology for local deployment and is economically attractive for small-scale bio-oil production from forest biomass [17].
Bals et al. [18] reported that the fast pyrolysis process is nearly energy neutral, and that the major costs are capital and drying costs of forest biomass. They reported that the total energy demand during fast pyrolysis is 1.83 MJ/kg; and the estimated selling price of bio-oil and bio-char are about $176/Mg and $61/Mg, respectively. Upgrading biooil to transport fuel requires a post-conversion, full deoxygenation process through hydro-treating and catalytic vapor cracking. Currently, the high production costs are not competitive with fossil fuel production costs. Another alternative, the production of hydrogen from biomass through pyrolysis, has been extensively investigated [19], since hydrogen and CO2 can be efficiently produced from the water-soluble fraction of bio-oil [20].
Bio-char is another product of the pyrolysis process, which is a valuable fuel for industrial application since it has a high energy content (about 30 GJ/ton) [21]. Biochar is also used as a soil amendment to improve soil health [22]. Qian et al. [23] reported that the effective utilization of bio-char through thermochemical techniques can improve the economic viability and environmental benefits of biomass. The main applications of biochar include catalysts (for syngas cleaning and synthesis process)
5 [24], soil amendment (for improvement of productivity and soil health) [22], fuel for fuel cells [25], sorbent of contaminant for mitigation of environmental issues [26], and activated carbon for reduction of hydrophobic contaminants [28].
In addition to developing cost-effective production technology, an optimal and robust bioenergy system is essential to support a competitive biomass-based energy market [29]. The large number of studies in the field of biomass-to-bioenergy SC presenting quantitative assessments indicate the importance of such methods (e.g., cost calculation, Geographic Information Systems (GIS), simulation, and mathematical optimization) to overcome the barriers that inhibit the development of bioenergy sector [9].
Mathematical optimization is an appropriate method to incorporate various decision levels (especially strategic and tactical) into models to optimize criteria (e.g., profit, cost, GHG emissions, and energy consumption) and facilitate the search for a desired solution [30]. Mathematical optimization modeling in the upstream segment of biomass-to-bioenergy SCs refers to identifying high yield resources, transportation configurations, coordination of entities (e.g., collection, grinding, delivery, and storage), and efficient technologies [9]. Mathematical optimization is the most common quantitative method applied over the last decade to address economic objectives. Optimization models are represented by objective functions (e.g., linear or nonlinear), decision variables (e.g., binary, integer, and continuous), constraints (e.g., capacity, conservation flow, and resource availability), and other parameters [31]. For instance,
6 optimization models can find the optimal locations and transport pathways with the assistance of binary variables in biomass energy SCs. The main optimization models are deterministic, stochastic, and multi-objective; each has several benefits and drawbacks. Deterministic models are the most common method in optimization programming and represent the vast majority of objectives (e.g., minimize the cost or maximize the profit) in biomass logistics. Stochastic models are uncommon and hard to solve due to incorporating uncertainties. A few studies in this field have applied multi-objective models due to the complexity of simultaneously handling several objectives (e.g., economic, environmental, and social). Further discussion about stochastic and multi-objective modeling are provided by Awudu and Zhang [32].
Research Objectives The goal of this research is to support the substitution of fossil-based energy and products with biomass-based energy and bio-products to address the cross-cutting sustainability factors (i.e., economic, environmental, and social) that influence the biomass-to-bioenergy supply chain (B2BSC). This work mainly focuses on addressing logistical challenges, e.g., high handling, transportation, and storage costs in traditional supply chain, by utilizing pretreatment technology (transportable pyrolysis refinery) in the upstream segment of B2BSCs. A transportable pyrolysis refinery is a truckmounted unit that can travel to farms and forests, where underutilized biomass residue is available. The transportable bio-refinery enables conversion of low-energy density biomass to high-energy density intermediate bio-products for ease transport and handling.
7
Research Tasks To meet this objective, multi-criteria decision making methods are developed, including a qualitative method (i.e., decision making analysis to find the potential harvesting sites) and quantitative methods (i.e., machine learning, artificial intelligence, operations research, and life cycle assessment). To demonstrate the application of these methods, a case study, is investigated for the Pacific Northwest, using real data. The presented case considers a small portion (20-50 thousand dry tons) of available biomass in Northwest Oregon, which is a fraction of that available in the U.S. each year (around 370 million dry tons) [33].
Dissertation Outline This dissertation is reported in manuscript format and includes five chapters and five appendices, and a total of eight manuscripts. Chapter 1 provides the motivation, background, objectives, and tasks of this research.
Chapter 2 is an article submitted to the Journal of Renewable and Sustainable Energy Review and titled “A Review and Future Directions in Techno-Economic Modeling and Optimization of Upstream Forest Biomass to Bio-oil Supply Chains.” This article reviews a techno-economic modeling and optimization efforts targeted at the upstream segment of the forest biomass to bio-oil supply chain.
8 Chapter 3 is an article to be submitted to the Journal of Cleaner Production and titled “Multi-criteria Decision Making for Sustainable Bio-Oil Production using a Mixed Supply Chain.” This article presents a multi-criteria decision making system, including both qualitative (decision tree analysis) and quantitative (deterministic optimization model) that was developed for sustainable bio-oil production using a mixed supply chain.
Chapter 4 is an article to be submitted to Environmental Science and Technology and titled “Evolutionary Decision Making for Biomass-based Energy Supply Chains.” This article provides a multi-criteria decision making methodology, comprised of various methods (i.e., machine learning and artificial intelligence techniques, stochastic optimization, and life cycle assessment) to incorporate uncertainty parameters and address real-world barriers in commercialization and global warming potential.
Chapter 5 presents the summary, conclusions, and contributions of this research, and proposes opportunities for future work.
In addition, this dissertation has five appendices, which are articles published in archival journal and peer-reviewed proceedings. Appendix A is a journal article published in the Journal of Cleaner Production (December 2015) and titled “Reducing the Cost and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains.” In this article, a mathematical model was proposed capable of helping decision makers in determining the optimal combination and location of fixed refineries
9 and mobile refineries for a known quantity of woody biomass and a given set of harvesting locations by considering capital and operational cost of refineries, and transportation costs. The environmental impacts of the integrated supply chains were assessed by considering the carbon footprint of the refinery infrastructure and transportation activities.
Appendix B is a conference article published in the Proceedings of ASME 2015 Manufacturing Science and Engineering Conference and titled “A Network Model to Optimize Upstream and Midstream Biomass-to-Bioenergy Supply Chain Costs.” This article provides a decision support system to facilitate sustainable production of bioenergy from forest biomass and to promote deployment of novel processing equipment (mobile bio-refinery). The method employs two phases: 1) classification of potential biomass harvesting sites and 2) optimization of the supply network through a mixed integer linear programming model that minimizes the costs of upstream and midstream supply chain segments.
Appendix C is a conference article published in the Proceedings of IIE 2016 Industrial and Systems Engineering Research Conference and titled “Evolutionary Optimization of Bioenergy Supply Chain Cost with Uncertain Forest Biomass Quality and Availability.” This article proposes a multi-criteria decision making method capable of helping investigators to incorporate uncertainty in B2BSC decision support systems. The method employs two quantitative methods: 1) support vector machine method was used to predict the pattern of uncertainty parameters and 2) a stochastic programming
10 method was used to assess the role of a transportable bio-refinery and the impact of real-world uncertainties in B2BSCs.
Appendix D is a conference article published in the Proceedings of ASME 2016 Design for Manufacturing and the Life Cycle Conference and titled “Reducing Greenhouse Gas Emissions for Sustainable Bio-Oil Production Using a Mixed Supply Chain.” This article explores global warming potential of the proposed cost-optimal forest biomassto-bio-oil mixed supply chain in Chapter 3. Therefore, life cycle assessment method was used for cradle-to-grave system to evaluate the reduction of GHG emissions compared to a more traditional supply chain.
Appendix E is a conference article published in the Proceedings of Flexible Automation and Intelligent Manufacturing Conference 2014 and titled “Integration of MachineLearning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process.” This article investigates the selection of the best supplier for a biomass supply chain (BSC) network through decision-tree analysis for defining the type of biomass feedstock materials for biofuel production and integration of machine learning techniques with a mathematical programming model to select the most appropriate feedstock suppliers.
11 References [1] U.S. EIA, “Total Crude Oil and Products Imports from All Countries,” 2016. [Online]. Available: https://www.eia.gov/. [Accessed: 25-Jan-2016]. [2] Union of Concerned Scientists, “The Promise of Biomass Clean Power and Fuel If
Handled
Right,”
2012.
[Online].
Available:
https://www.google.com/#q=The+Promise+o+f+Biomass+Clean+Power+and+Fuel% E2%80%94If+Handled+Right. [Accessed: 10-Jan-2016]. [3]
US DOE, “Bioenergy Walkthrough - Bioenergy Technology Office,” 2014.
[4]
USDOE, U.S. billion-ton update: biomass supply for a bioenergy and
bioproducts industry. Oak Ridge National Laboratory, 2011. [5]
N. Sadasivam, “Economy Would Gain Two Million New Jobs in Low-Carbon
Transition,
Study
Says
|
InsideClimate
News.”
[Online].
Available:
http://insideclimatenews.org/news/18112015/low-carbon-economy-may-create-2million-jobs-study-finds-clean-energy. [Accessed: 18-Feb-2016]. [6]
D. M. Alonso, J. Q. Bond, and J. A. Dumesic, “Catalytic conversion of biomass
to biofuels,” Green Chem., vol. 12, no. 9, p. 1493, 2010. [7]
J. R. Bartle and A. Abadi, “Toward Sustainable Production of Second
Generation Bioenergy Feedstocks†,” Energy Fuels, vol. 24, no. 1, pp. 2–9, 2009. [8]
A. V. Bridgwater, A. J. Toft, and J. G. Brammer, “A techno-economic
comparison of power production by biomass fast pyrolysis with gasification and combustion,” Renew. Sustain. Energy Rev., vol. 6, no. 3, pp. 181–246, 2002. [9]
A. De Meyer, D. Cattrysse, J. Rasinmäki, and J. Van Orshoven, “Methods to
optimise the design and management of biomass-for-bioenergy supply chains: A review,” Renew. Sustain. Energy Rev., vol. 31, pp. 657–670, 2014. [10]
S. Akhtari, T. Sowlati, and K. Day, “Economic feasibility of utilizing forest
biomass in district energy systems – A review,” Renew. Sustain. Energy Rev., vol. 33, pp. 117–127, May 2014. [11]
O. Akgul, A. Zamboni, F. Bezzo, N. Shah, and L. G. Papageorgiou,
“Optimization-based approaches for bioethanol supply chains,” Ind. Eng. Chem. Res., vol. 50, no. 9, pp. 4927–4938, 2010.
12 [12]
A. Uslu, A. P. C. Faaij, and P. C. A. Bergman, “Pre-treatment technologies, and
their effect on international bioenergy supply chain logistics. Techno-economic evaluation of torrefaction, fast pyrolysis and pelletisation,” Energy, vol. 33, no. 8, pp. 1206–1223, Aug. 2008. [13]
D. Yue, F. You, and S. W. Snyder, “Biomass-to-bioenergy and biofuel supply
chain optimization: Overview, key issues and challenges,” Comput. Chem. Eng., vol. 66, pp. 36–56, 2014. [14]
Z. Yang, A. Kumar, and R. L. Huhnke, “Review of recent developments to
improve storage and transportation stability of bio-oil,” Renew. Sustain. Energy Rev., vol. 50, pp. 859–870, Oct. 2015. [15]
P. C. Badger and P. Fransham, “Use of mobile fast pyrolysis plants to densify
biomass and reduce biomass handling costs—A preliminary assessment,” Biomass Bioenergy, vol. 30, no. 4, pp. 321–325, Apr. 2006. [16]
T. P. Vispute, G. W. Huber, and others, “Breaking the chemical and engineering
barriers to lignocellulosic biofuels,” Int. Sugar J., vol. 110, no. 1311, 2008. [17]
S. Gold and S. Seuring, “Supply Chain and Logistics Issues of Bio-Energy
Production,” J. Clean. Prod., vol. 19, no. 1, pp. 32–42, 2011. [18]
B. D. Bals and B. E. Dale, “Developing a model for assessing biomass
processing technologies within a local biomass processing depot,” Bioresour. Technol., vol. 106, pp. 161–169, 2012. [19]
S. Ayalur Chattanathan, S. Adhikari, and N. Abdoulmoumine, “A review on
current status of hydrogen production from bio-oil,” Renew. Sustain. Energy Rev., vol. 16, no. 5, pp. 2366–2372, Jun. 2012. [20]
D. Wang, S. Czernik, D. Montané, M. Mann, and E. Chornet, “Biomass to
Hydrogen via Fast Pyrolysis and Catalytic Steam Reforming of the Pyrolysis Oil or Its Fractions,” Ind. Eng. Chem. Res., vol. 36, no. 5, pp. 1507–1518, May 1997. [21]
M. Van de Velden, J. Baeyens, and I. Boukis, “Modeling CFB biomass
pyrolysis reactors,” Biomass Bioenergy, vol. 32, no. 2, pp. 128–139, 2008. [22]
D. Granatstein, C. E. Kruger, H. Collins, S. Galinato, M. Garcia-Perez, and J.
Yoder, “Use of biochar from the pyrolysis of waste organic material as a soil
13 amendment. Final project report,” Cent. Sustain. Agric. Nat. Resour. Wash. State Univ. Wenatchee WA, 2009. [23]
K. Qian, A. Kumar, H. Zhang, D. Bellmer, and R. Huhnke, “Recent advances
in utilization of biochar,” Renew. Sustain. Energy Rev., vol. 42, pp. 1055–1064, Feb. 2015. [24]
S. Zhang, M. Asadullah, L. Dong, H.-L. Tay, and C.-Z. Li, “An advanced
biomass gasification technology with integrated catalytic hot gas cleaning. Part II: Tar reforming using char as a catalyst or as a catalyst support,” Fuel, vol. 112, pp. 646– 653, Oct. 2013. [25]
S. Y. Ahn, S. Y. Eom, Y. H. Rhie, Y. M. Sung, C. E. Moon, G. M. Choi, and
D. J. Kim, “Utilization of wood biomass char in a direct carbon fuel cell (DCFC) system,” Appl. Energy, vol. 105, pp. 207–216, May 2013. [26]
M. Ahmad, A. U. Rajapaksha, J. E. Lim, M. Zhang, N. Bolan, D. Mohan, M.
Vithanage, S. S. Lee, and Y. S. Ok, “Biochar as a sorbent for contaminant management in soil and water: A review,” Chemosphere, vol. 99, pp. 19–33, Mar. 2014. [27]
M. V. Gil, M. Martínez, S. García, F. Rubiera, J. J. Pis, and C. Pevida,
“Response surface methodology as an efficient tool for optimizing carbon adsorbents for CO2 capture,” Fuel Process. Technol., vol. 106, pp. 55–61, Feb. 2013. [28]
P. Oleszczuk, S. E. Hale, J. Lehmann, and G. Cornelissen, “Activated carbon
and biochar amendments decrease pore-water concentrations of polycyclic aromatic hydrocarbons (PAHs) in sewage sludge,” Bioresour. Technol., vol. 111, pp. 84–91, May 2012. [29]
M. de Lourdes Bravo, M. M. Naim, and A. Potter, “Key issues of the upstream
segment of biofuels supply chain: a qualitative analysis,” Logist. Res., vol. 5, no. 1–2, pp. 21–31, 2012. [30]
D. Yue, S. Pandya, and F. You, “Integrating Hybrid Life Cycle Assessment
with Multi-objective Optimization: A Modeling Framework,” Environ. Sci. Technol., 2016. [31]
C. Cambero and T. Sowlati, “Assessment and optimization of forest biomass
supply chains from economic, social and environmental perspectives – A review of literature,” Renew. Sustain. Energy Rev., vol. 36, pp. 62–73, Aug. 2014.
14 [32]
I. Awudu and J. Zhang, “Uncertainties and sustainability concepts in biofuel
supply chain management: A review,” Renew. Sustain. Energy Rev., vol. 16, no. 2, pp. 1359–1368, Feb. 2012. [33]
R. D. Perlack, L. M. Eaton, A. F. Turhollow Jr, M. H. Langholtz, C. C. Brandt,
M. E. Downing, R. L. Graham, L. L. Wright, J. M. Kavkewitz, A. M. Shamey, and others, “US billion-ton update: biomass supply for a bioenergy and bioproducts industry,” 2011.
15
Chapter 2: A Review and Future Directions in TechnoEconomic Modeling and Optimization of Upstream Forest Biomass to Bio-Oil Supply Chains
Submitted to Renewable and Sustainable Energy Reviews (September 2015)
16 Chapter 2:
A Review and Future Directions in Techno-Economic Modeling
and Optimization of Upstream Forest Biomass to Bio-Oil Supply Chains
Abstract Recent interest in biomass supply chain management has stimulated research efforts in the industry and academic communities. Techno-economic modeling and optimization efforts targeted on the upstream segment of the forest biomass to bio-oil supply chain are reviewed. Key components of upstream supply chain decision making are then presented through an overview and classification of the existing methods and contributions. There is a need to classify and analyze the relevant methodologies and approaches identified in prior studies, and to subsequently assess their usefulness through empirical research and case-based analysis. Both narrative and systematic literature reviews are performed using qualitative analysis and classic bibliometric techniques to demonstrate the scope of current papers and the call for future needs. It is found, due to growing demands for bioenergy, future biomass-to-bioenergy supply chains should draw upon existing research toward the development of efficient and effective forest biomass supply chain networks. It is further concluded that a new generation of pretreatment technologies is needed for techno-economic optimization of upstream forest biomass value chains.
Introduction Motivation: Since the early 1970s, when the world had its first energy shortage crisis and recognized a looming environmental crisis, interest in developing new sources of
17 energy has risen significantly [1]. A spike in oil prices (2007-2008) created a more recent wave of interest in renewable energy sources [2]. In the United States (U.S.), federal laws, such as the Energy Independence and Security Act of 2007, as well as state level renewable energy standards support greater use of biomass to produce energy [3]. Both industry and academic communities have sought substitutions for fossil-based energy sources. In the 1990s, renewable energy resources, including biomass, hydroelectric, geothermal, solar, and wind, were introduced as novel replacements for fossil-based energy resources (e.g., coal and crude oil) [4]. The growing number of investigations reported in peer-reviewed proceedings and archival journals, as well as online discussions dedicated to the topic, confirms global interest in and support of renewable energy sources (e.g., biogas and biofuels) [5–7]. According to the U.S. Department of Energy [8], 10% of energy in the U.S. is derived from renewable energy sources. Biomass, a renewable and bio-degradable energy resource, provides 50% of renewable energy needs in the U.S.; consequently, biomass plays a key role in the energy industry.
Biofuels represent a promising substitute for conventional fuels for mobility and heating applications; however, availability, quality, and variability of biomass greatly affect energy self-sufficiency [9]. Additionally, recent awareness among the internal and external stakeholders over the need to alleviate the level of greenhouse gas (GHG) emissions has motivated research efforts in logistics and management of biomass supply chains (SCs) [10–12]. As witnessed over the past several years with the development of new fossil energy sources (oil sands and shale), high volatility of
18 energy prices is another concern, and is a manifestation of phenomena resulting from constantly increasing demand for a limited supply of fossil energy resources. In spite of its environmental benefits, forest biomass is too costly to compete with fossil fuels in many countries [13,14]. Lower fossil fuel prices would further impair the competitiveness of bioenergy. With the recent price volatility and energy slump, it is prudent to systematically design and optimize the upstream segment of biomass-tobioenergy SCs. Such efforts directly affect the sustainability and robustness of other activities in the entire value chain [15–17].
Challenges: Increasing bioenergy availability is crucial due to its significant advantages (e.g., reduction of dependence on imported energy and GHG emissions). Supply challenges, (e.g., low energy density and bulky), are major barriers that prevent using forest biomass to produce bioenergy [18]. Borard et al. [2] suggested that wood product market prices and harvest costs are key factors in determining the feasibility of woody biomass utilization. Market prices for competing products, such as wood pulp, can affect biomass availability. Yemshanov et al. [19] observed that biomass supply costs are the biggest constraint to widespread use of forest harvest residues for bioenergy commercialization. Most investigations into biomass processing and bioenergy production focus on the economic aspect [20].
The economic aspect is influenced by several factors. First, biomass has the low-energy density, which results in higher costs for transport and delivery to refinery facilities [20,21]. Second, biomass feedstocks are dispersed and, thus, collection of significant
19 amounts requires traversing large areas [22]. Third, biomass requires pretreatment (e.g., drying and pyrolysis) to reduce moisture content and to enhance its energy density [23,24], which increases production time and cost. Fourth, the seasonal nature and annual variability of biomass require careful planning and scheduling to ensure adequate quality and quantities of the biomass feedstocks [25]. Therefore, design and optimization of the SC network are key challenges of economically feasible conversion of biomass to bioenergy [26].
Several optimization studies have focused on the development of decision support systems for forest biomass SCs [27–29]. Most of these studies have considered economic performance [30–32] through modeling, planning, and management of the value chains. Techno-economic analysis and optimization can help decision makers to deal with a wide range of decision levels (i.e., operational, tactical, strategic). Parker et al. [33] developed a mixed integer linear programming (MILP) optimization model to assess potential biofuel supply across the western U.S.; they explored spatial information, including different feedstocks, potential refinery locations, and transportation networks to maximize the profit in biofuel SCs. Gold and Seuring [34] discussed the principal issues of SCs and logistics for bioenergy production identified by the existing literature. Various review studies report criteria, modeling approach, assumptions, restrictions, uncertainties, and future work to design a proper biomass-tobioenergy SC structure [17,35–37].
20 Background: Lignocellulosic biomass is the most abundant type of biomass, and is composed of lignin (15-20%), hemicellulose (25-35%), and cellulose (40-50%) [38]. In addition to its abundance, lignocellulosic biomass (e.g., western juniper in central Oregon, USA) is preferable for several other reasons: the possibility of using nonedible biomass, the ability to obtain a higher net energy yield for the production of fuels and chemicals, and the ability to avoid undue land use change [39,40]. Bio-oil from lignocellulosic biomass is a potential source of a number of valuable chemicals. More than 350 products of lignocellulosic pyrolysis have been identified, including tars, acetic acids, alcohols, esters, aromatics, sugars, turpentine, and methanol [38]. Therefore, investigation into the development of low-cost separation techniques is needed to enable commercialization of these chemicals [38]. Radlein [41] reported a detailed review on the production of chemicals from bio-oils. The commercial competitiveness of bio-oil depends on logistics factors (e.g., availability and industrial practices), the scale and location of bio-refineries, and other factors [42,43].
Various aspects of the upstream segment of biomass to bioenergy SC have previously been discussed [6], and it is essential to develop efficient technologies to assist stakeholders to overcome the identified challenges [44,13,35]. Second-generation conversion technologies, which use nonfood crops such as forest biomass (roots, stumps, bark, leaves, small stems, and branches of live and dead trees) are currently under development, but exhibit unclear technical and economic performance [20,34]. Researchers continue to evaluate the advantages of conversion technologies in terms of energy efficiency [45]. Additionally, several researchers argued and underlined that
21 conversion technology and pretreatment technique (e.g., size reduction and drying) developments can mitigate SC management and logistics issues of biomass to bio-oil SC networks [17,46,47]. Fast pyrolysis process represents a conversion technology for local deployment and is economically attractive for small-scale bio-oil production from lignocellulosic biomass [48,49,34].
Bals et al. [50] reported that the fast pyrolysis process is nearly energy neutral, and that the major costs are capital costs and costs of drying the forest biomass. They reported that the total energy demand during fast pyrolysis is 1.83 MJ/kg; the estimated selling price of bio-oil and bio-char are about $176/Mg and $61/Mg, respectively. Upgrading the bio-oil to renewable transport fuel requires a post-conversion, full deoxygenation process through hydrotreating and catalytic vapor cracking. Currently, the high production costs are not competitive with fossil fuel production costs. Another alternative, the production of hydrogen from biomass through pyrolysis, has been extensively investigated [51], since hydrogen and CO2 can be efficiently produced from the water-soluble fraction of bio-oil [52]. Czernik at al. [53] reported that 1.3 Gt/year of biomass can produce 100 Mt/year of hydrogen.
Bio-oil is a low-grade liquid fuel that can be used in industrial heating (e.g., fueling heaters, furnaces, and boilers), industrial turbines, stationary dual-fuel diesel engines, and upgrading to transport fuels [54,55]. Additionally, it can serve as a source of several chemicals and bio-refining feedstock [56,57]. Bio-oil has limitations, however, such as low energy density and corrosive properties that are harmful to existing engines [58].
22 The high post-conversion processing cost is an essential issue to be considered in producing higher-quality hydrocarbon fuel from bio-oil. The physical properties of biooil such as low heating value (half of conventional fuel oil), poor volatility, solids content, high viscosity, coking, corrosiveness, and incompatibility with conventional fuels limit the range of bio-oil applications [54]. Standard bio-oil properties are necessary for commercial application. In order to realize industrial advancements and commercialization of bio-oil-based fuels and chemicals, considerable work is required to overcome the techno-economic challenges of SC logistics, bio-oil production, handling, and upgrading.
Bio-char is another product of the pyrolysis process, which is a valuable fuel for industrial application since it has a high energy content (about 30 GJ/ton) [59]. Biochar is also used as a soil amendment to improve soil health [60]. Qian et al. [61] reported that the effective utilization of bio-char through thermochemical techniques can improve the economic viability and environmental benefits of biomass. The main applications of biochar include catalysts (for syngas cleaning and synthesis process) [62], soil amendment (for improvement of productivity and soil health) [60], fuel for fuel cells [63], sorbent of contaminant for mitigation of environmental issues [64], gas storage for CO2 and H2 [65], activated carbon for reduction of hydrophobic contaminants [66]. Qian et al. [61] provided an overview of bio-char structure and properties, production methods, and other applications.
23 In addition to developing cost-effective production technology, an optimal and robust bioenergy system is essential to support a competitive biomass-based energy market [67]. The large number of studies in the field of biomass-to-bioenergy SC that present quantitative assessments indicate the importance of such methods (e.g., cost calculation, geographic information systems (GIS), simulation, and optimization) to overcome the barriers that inhibit the development of the bioenergy sector [6]. More detailed information about the applied quantitative methods and models to evaluate biomass supply chain performance has been previously reported by Ba et al. [68]. The cost calculation method is the simplest approach, and involves summing partial costs of different entities of the SC using spreadsheets for automating calculation. GIS plays an important role as an effective tool and a spatial database that can provide inputs related to biomass availability, biomass logistics, and biofuel facility location selection for spatial and temporal optimization models, however, GIS is not capable of directly optimizing an objective function [68,69]. Simulation methods are more appropriate than the cost calculation and GIS-based methods due to high modeling flexibility and ease of understanding/modifying the model. In addition, the broad choice of simulation software can ease decision making. In spite of the benefits, supply chain simulation modeling is not appropriate for calculating optimal cost, nor for strategic and tactical decisions. Thus, mathematical modeling and operations research approaches (e.g., deterministic, stochastic, and multi-objective optimization) are commonly used to incorporate various decision levels (especially strategic and tactical) into models (e.g., to optimize profit, cost, GHG emissions, and energy consumption) and to facilitate the search for a desired solution [70].
24
Mathematical optimization has been widely applied over the last decade to address economic objectives in the upstream segment of biomass-to-bioenergy SCs [6,17]. These studies have enabled identification of high yield resources, transportation configurations, coordination of entities (e.g., collection, grinding, delivery, and storage), and efficient technologies. Mathematical optimization models are represented by objective functions (e.g., linear or nonlinear), decision variables (e.g., binary, integer, and continuous), constraints (e.g., capacity, conservation flow, and resource availability), and other parameters [35]. Such models can be used to find the optimal locations and transport pathways with the assistance of binary variables in biomass-tobioenergy SCs. Deterministic models represent the vast majority of mathematical optimization objectives (e.g., minimizing cost or maximizing profit) in biomass logistics. Stochastic models are less common due to challenges in incorporating uncertainties (e.g., difficulty in developing computational algorithms). Few studies have applied multi-objective models for biomass supply chain optimization due to the complexity of simultaneously handling several objectives (e.g., economic, environmental, and social). Further discussion of stochastic and multi-objective modeling are provided by Awudu and Zhang [36] and Cambero and Sowlati [35]. In particular, mathematical optimization has been demonstrated as a viable approach to aid techno-economic analysis, as discussed in greater detail in the narrative review (Section 3.2).
25 Objective: Biomass SC optimization approaches that incorporate various decision levels (i.e., strategic, tactical, and operational) to evaluate sustainability performance have not been found in the literature that integrate conversion processes with the forest biomass SC. Thus, the specific objective of this study is to investigate the nature of the current state-of-the-science and to identify potential future directions for cost effective development of conversion technology for biomass-to-bioenergy SCs. The presented study analyzes and compares distributed and centralized processing network systems that process forest biomass into bio-oil. This study helped to define the topics explored in the literature review. The literature review was conducted using the narrative and systematic techniques and analytical and empirical methods to explore existing research methodologies and approaches. Additionally, this study identifies challenges and benefits of forest biomass supply, as well as simultaneous internal and external stakeholder needs.
Review Methodology The methodology applied in this literature review addresses the existing methods with respect to the state-of-science and consists of two parts. First, a narrative literature review was conducted by analyzing the contents of publications in the upstream segment of biomass to bioenergy SCs, along with techno-economic modeling appearing over the past four decades. Second, a systematic literature review was conducted using quantitative and qualitative methods for publications appearing from January 1, 2000 to June 30, 2015.
26 Narrative Literature Review Method The narrative literature review method identifies the purpose of the research, explores key concepts, and defines chronological advancements in reported research. Narrative review addresses the challenges from the existing articles and proposes solutions to overcome some of the challenges. Herein, it identifies the evolution of research in the upstream segment of the biomass to bioenergy SC, with a focus on technical and economic aspects. The current challenges in forest biomass and bioenergy SC management are identified through the basic concepts, major work, and key contributions. Several studies reported that the upstream (e.g., harvesting, extraction, and transportation) forest fuel SC is influential on the economic viability of biomass to bioenergy SC [71–73].
The conventional structure of biomass to bioenergy upstream SCs includes harvesting and collection, logistics, and storage. In the earlier studies, undertaken using the conventional upstream segment structure, researchers have not considered pretreatment [72,71,18]. Figure 2.1a represents the conventional structure of the upstream segment of biomass-to-bioenergy SCs. Several studies [74–76] illustrated the current technology and economic development related to the upstream biomass feedstock SC (e.g., pulpwood and wood residues); these compared the “State of Technology” in 2014 (current state) and 2017 (target design case). Figure 2.1b shows the proposed structure, which is composed of four main processes: harvesting and collection, logistics, pretreatment, and storage. Pretreatment can convert biomass to intermediate bioproducts such as bio-oil and bio-char. The pretreatment process is still insufficiently
27 robust to accommodate technology modeling, optimization, and sustainability aspects of biomass to bio-oil SC implementation [17,47].
a)
Harvesting & Collection
Harvesting & Collection
Logistics
Logistics
Storage
Pretreatment
Biomass Flow Bio-oil Flow
Storage
b)
Figure 2.1. Upstream segment of the general biomass-to-bioenergy supply chain: a) conventional structure and b) proposed structure (paper sections are also indicated)
Systematic Literature Review Method Since scholars have limited time to keep up-to-date in broad technological research and development, review studies play a key role in bridging science and engineering research. The systematic review (SR) method presents a series of evaluations, strategies, and analyses that assist in identifying key characteristics of previous studies and designing future studies. One of the main features of SR is reduction of author bias in comparison with narrative reviews, which often support the authors’ point of view. Thus, complementary systematic and narrative reviews were performed herein to analyze the publications, citations, and adopted research methods in the upstream biomass-to-bioenergy SC domain. These techniques have been previously applied for
28 performing literature reviews in sustainable SC management [77,78]. No prior studies in the bioenergy SC domain have been identified that perform a SR. Thus, quantitative and qualitative methods were developed in the SR applied herein for analyzing recent publication related to biomass-to-bioenergy upstream SC management. The quantitative method presents an analysis of publication data, citation data, and keywords to provide a novel classification for available studies. Figure 2.2 summarizes the dimensions of publication and citation data analysis used herein as a quantitative method.
Keywords Numbers Date Date Journals Publication Data
Citation Data
Most Cited Journals
Authors Most Cited Authors Geography Citation Trend Publication Trend
Figure 2.2. Classification framework for quantitative method [78]
The qualitative method provides a comprehensive characterization of current literature for classifying the research methodologies. Wacker [79] categorized qualitative literature review methods in two main categories (analytical and empirical) and six subcategories. Several studies have used and modified Wacker’s classification by
29 identifying additional subcategories [80–82]. Figure 2.3 presents the identified categories and subcategories that were used in the qualitative method herein.
Conceptual Analytical
Mathematical Statistical Experimental
Action Researches Empirical
Statistical Sampling Case studies
Figure 2.3. Classification framework for qualitative method [78]
Narrative Literature Review The narrative review conducted herein identifies the qualitative evidence of the need for research to improve upstream forest biomass-to-bioenergy system performance and related mathematical modeling approaches. Biomass can be classified as lignocellulose (e.g., wood), triglycerides (e.g., vegetable oil), amorphous sugars (e.g., glucose), and starch (e.g., corn). Lignocellulosic biomass is the most abundant, cheapest, and fastgrowing type [83,84]. The forest industry is one of the main suppliers of lignocellulosic feedstocks [85].
Forest biomass could support the mitigation of anthropogenic carbon emissions as a renewable energy source and could help forest communities generate jobs and develop local energy sources [86,35]. Forest biomass can be converted to the different fuel types
30 (e.g., solid, liquid, and gaseous) that has higher bulk density [13,87], but energy generation through forest biomass is confronted by several challenges. Some of the main barriers of using forest biomass as an energy resource are high moisture content, low energy density, seasonal variation, conversion technology limitations, and storage space requirements [88,16]. The upstream forest SC is responsible for biomass harvesting and collection, logistics, pretreatment, and storage, as described in the following sections.
Harvesting and Collection Harvesting and collection involves several operations, such as felling, bundling, skidding or forwarding, unloading at landing, grinding or chipping, and loading for truck transport. Harvesting and collection (Figure 2.4a and 2.4b) of forest biomass is a key step in bioenergy production because it will determine the modes of transportation and storage. Harvesting and collection systems for forest harvest residue (FHR) increased rapidly in the mid-1980s [89] to reduce fire risks, prevent smoke pollution, and create renewable energy sources [26].
a)
b)
Figure 2.4. Harvesting and collection of forest biomass: a) harvested biomass and b) collection operation system
31
Harvesting and collection are highly influential in defining SC cost effectiveness [90]. Harvesting and collection operations can either be coupled or de-coupled. On steeper terrain, harvesting of merchantable logs and biomass is done simultaneously because it is cost prohibitive to return for branches and tops. On flatter terrain, tops and branches left in the field dry more quickly than piled forest residues.
Forest residue availability may be limited by the season or extreme weather, safety, environmental aspects, or site accessibility [2]. Further, the harvest season and operation time can affect biomass yield and energy density [37]. Additionally, increasing the demand for forest biomass coupled with inappropriate harvesting methods have had negative impacts on biodiversity, soil, and water conservation [91]. Spinelli et al. [92] noted that possible alternative methods to employ in forest fuel logistics are transporting loose residue, chipping, or bundling (at collection sites). Hauling loose residue is preferable only for short distances. The major advantages of mechanical particle-size reduction (e.g., grinding or chipping) are reducing the heat requirements for conversion (especially thermochemical process such as fast pyrolysis) and increasing biomass bulk density and yields for ease of handling and transport [93– 95]. Grinding forest biomass is an energy-intensive, expensive process with a $26.12 per oven-dry tonne impact range [96]. Bit type, screen size, and knife-edge bits are major parameters during grinding that affect the bulk density and fuel consumption [97]. Sokhansanj et al. [98] reported that the cost of grinding biomass is nearly $11/MJ,
32 using three grinders for 450 Gg of biomass (500 collection sites with 800 Mg biomass at each site), but costs can vary depending on equipment used and feedstock conditions.
Early studies explored harvesting operation systems, collection of large quantities of biomass for energy generation, and improvement of profitability through optimal product allocation [99,100]. Later studies examined the cost/benefit tradeoffs of integrated harvesting systems and developed industrial case studies using the integrated methods and approaches, e.g., economic targets and environmental needs [101,102,35]. Zamora-Cristales and Sessions developed a forest residue collection model to estimate the cost of biomass collection, using forwarders and excavator-base loaders [90]. Their results indicate that excavator-based loader alone, is the most cost-effective approach for less than 50 meters distance, and two forwarders and one excavator-based loader is usually the most cost-effective approach for beyond 50 meters distance, depending upon mobilization costs. They estimated the optimal collection cost between 15 and 350 meters are from $7.2 to $27.5 per dry ton. Thus, the rational biomass collection is essential to design an optimal biomass supply chain network. Muth et al. [103] developed biomass SCs that use advanced preprocessing strategies for forest thinning to achieve high quality and low cost biomass in bio-refinery. Puttock [104] developed a model to estimate the production costs for seven integrated harvesting systems from six countries (i.e., Canada, United Kingdom, United States, Denmark, Finland, and Sweden). The author concluded that production costs of fuelwood are highly dependent on the type of harvesting and collection system used.
33 The harvesting approach fundamentally affects the upstream decision making in terms of producing the desired shape, size, quantity, and quality of biomass at the correct time [71]. Grisso et al. [105] noted that the number of annual harvest days has a significant impact on economics of harvesting, collecting, and storing of biomass. The number of harvest machines required and storage capacity required are dependent upon the number of harvest days. Consequently, a shorter harvest period is a critical variable in scheduling of harvest operations that results in reduced capital and operational costs of harvest machines and storage due to a reduced harvest volume. Pan et al. [26] evaluated harvesting cost and productivity of a whole-tree harvesting system to examine the feasibility of using forest biomass for bioenergy production. They argued that tree size is not a significant economic factor in most harvesting and collection operations (e.g., felling, skidding, grinding, and loading). Laitila and Väätäinen [106] and Röser et al. [107] looked at the various forest biomass supply systems for fuel production, and reported that harvesting and chipping at the roadside was the most cost-efficient supply system. Additionally, Zamora-Cristales et al. developed a stochastic simulation model to analyze the economics of mobile chipper and trucks under uncertainty for bioenergy production from forest biomass [108]. They noted road characteristics (e.g., road gradient, surface, width, and horizontal alignment) and processing locations are important factors that affect truck and chipper standing time, and consequently biomass operation costs.
One of the challenges in the utilization of forest biomass for energy is the environmental sustainability of biomass energy [34]. The Scottish Agricultural
34 College, Finland, and Sweden have developed guidelines for biomass harvesting [109]. These guidelines emphasize retention, disposal, redistribution, burning and mulching, as well as management practices for biomass extraction. The long-term effects of harvest operations and forest biomass removal are on soil productivity, water quality, and habitat [110–112]. Achat et al. [113] quantified the overall effects of removing forest biomass, especially harvest residues (branches) on nutrient outputs, biological soil fertility, and tree growth. Although found to be the most cost-efficient in studies by Laitila, Väätäinen, Röser, and others, their study indicated that whole-tree harvesting has negative effects on soil properties and forest ecosystems. Since present forest biomass harvest operation and forest management guidelines are insufficient, it is essential to develop a novel biomass guideline by integrating existing guidelines with new findings to ensure a sustainable SC [114]. Different approaches are required for each type of harvesting system, which results in different types of forest biomass SCs in terms of handling (e.g., sticks or chips), transportation (e.g., truck or rail), and storage (e.g., covered or uncovered). Since the harvesting approach affects other activities in the biomass-to-bioenergy SC, the main objective is promoting sustainability through mitigating economic, environmental, and social concerns.
Logistics Transportation is a main activity linking the individual entities within biomass-tobioenergy SCs. Transportation costs are assumed to be a major cost driver associated with obtaining biomass [115]. Transportation attributes, including transport mode (e.g., truck, rail, and ship), vehicle capacity, and distance traveled, have a substantial effect
35 on transportation costs [13]. The delivery cost of forest biomass depends on several other factors, such as road type and road conditions (travel speed), type of biomass (bulk density), harvesting methods, and processing operations [116]. For instance, reliable estimation of biomass pile volume allows decision makers to deploy the processing operations appropriately, and develop an efficient SC [117]. Eriksson and Björheden [118] pointed out that optimizing biofuel production for forest biomass essentially means minimizing transportation cost. Thus, forest biomass logistics costs are a key component of the overall cost of forest activities and bioenergy production [119–121].
Forest biomass can be delivered to bioenergy production facilities by truck, rail, and ship. Trucks transport about 90% of forestry products to mills in the U.S. [122]. Searcy et al. [121] reviewed the associated costs of transferring biomass feedstocks by truck, rail, and ship for small- and large-project sizes in detail. Tractor-trailers and fixed trucks are two types of trucks for transporting forest biomass. Tractor-trailers use standard highway road tractors, which are usually about 12,000 to 20,000 lbs. (5.5 to 9 metric tons). These types of trucks are designed for greater capacity and higher versatility, compared with fixed trucks [120]. Fixed-trucks are usually shorter than a tractor-trailer and designed for tighter areas and high maneuverability, however, they have the lower payload capacity. Additionally, different trailer options are available for hauling forest biomass, including log trailers, container trailers, and chip vans. Log trailers are designed to haul logs or bundled trees with higher payload capacities. Container trailers are designed to hold bulk material using sturdy walls; they have less
36 capacity than log trailers and chip vans. Chip vans (Figure 2.5b) are generally enclosed box trailers approximately 8.5 feet (2.62 m) in width, 8.5 feet (2.62 m) in height, and of various lengths (32-53 ft., 9.8-16.1 m) for hauling chipped or ground products. A drop center can add 10% additional volume [123]. Chip vans are considered to be the most cost-efficient mode of forest biomass transport because of higher payload capacity. Spinelli et al. [69] and Johansson et al. [71] reported the costs of different transportation methods and identified the conditions that can make one preferable over the others. In general, bundling is the least-efficient method and transporting loose residue is the cheapest method, when the travel distance is within 24 miles (40 km) [73].
According to the U.S. Department of Transportation [124], legal load limits on highway roads in the U.S. for tractor-trailer are 80,000 lbs. (36.3 metric tons) gross vehicle weight. To operate above the legal limit, special permits must be obtained [125]. The tare weight of a tractor-trailer is between 26,000 and 30,000 lbs. (11.8 and 13.7 metric tons), including six to ten tandem axles. Badger and Fransham [56] reported that chip vans typically haul between 50,000 and 54,000 lbs. (22.7 and 24.5 metric tons) per load in the U.S. Zamora-Cristales and Sessions explored the economic feasibility of using single trailer and double trailers in three U.S. States (i.e., Washington, Oregon, and California) from the grinder at the landing to the bio-refinery [126]. Their results indicate that double trailers can be a promising alternative under limited conditions only in Washington and Oregon due to State transportation regulations. Additionally, the energy value of a trailer vanload for green wood chips is generally between 210 and
37 270 GJ, but depends heavily on moisture content [123,127]. Thus moisture management is an important part of the biomass SC, both for transportation optimization and, when used for boiler fuel, to improve biomass combustion value [127].
One of the most challenging problems in transportation planning is a road network optimization to minimize total transportation cost. As forest biomass has low value and is bulky, road network optimization plays a key role in efficient and cost effective forest fuel supply [128]. Alam et al. [116] developed a road network optimization model to assist woody biomass supply decision-making for sustainable bioenergy production in northwestern Ontario. They sought to provide efficient and effective woody biomass supply logistics for energy production by minimizing transportation time and cost. Berwick et al. [129] developed a software model which estimates the transportation costs by considering different trucks, equipment, and input prices for different trip characteristics and configurations. One application of the proposed model is forest biomass transport. Zamora-Cristales et al. [130] reported truck/machine interaction to be a key variable limiting biomass conversion efficiency. Comminution (a size reduction technique for lignocellulosic biomass feedstocks) and transportation can be decoupled, but the cost advantages depend upon the situation, e.g., available processing options, location (in-forest yard or bioenergy facility), and transportation options [131]. Since forest biomass has a low bulk density [93], processing biomass through chipping, grinding, and shredding can increase the bulk density [120]. However, comminution of forest biomass has a negative effect on durability and longevity during storage.
38 Additionally, to ensure a continuous supply and meet growing demands during winter and spring seasons, it is essential to use covered storage.
Guzmán [132] evaluated the technical and economic feasibility of producing wood chips under different conditions. Guzmán also explored different variables such as working hours and transportation cost to investigate woodchip production in Chile. Processing of forest biomass can increase the bulk density of wood chips and grindings, which is essential to improve handling efficiency. Processing can occur at any step in the upstream forest biomass SC, although it is lower cost when integrated with harvesting and collection [48,71]. Gold and Seuring [34] reported that the most common processing options after harvesting and collection are drying, baling, chipping, and grinding (Figure 2.5a). Zamora-Cristales et al. [133] demonstrated that bulk density of dry grindings could be increased more than 20% through vertical blowing into trailers. Kanzian et al. [134] developed a method for forest biomass logistics supply through a combination of geographic information systems (GIS) and linear programming. The results indicated that direct transport of forest biomass and chipping at the destination is the lowest-cost fuelwood SC system. Additionally, biomass can be chipped at or near the harvesting site and transported via chip trucks at a slightly higher cost. The authors argued that harvesting residues can only be recommended for large-scale refineries because of poor quality and low biomass yield. A prior study found that the roll-off approach would improve economic efficiency and forest residue accessibility [135]. Han et al. [136] and Harrill et al. [137] proposed a new approach using a roll-off truck paired with a small skid-steer loader to collect and
39 transport slash to the centralized processing areas for bioenergy production. They applied the proposed approach to quantify the operational costs of removing biomass. Since transportation costs significantly increase with a slight increase in truck travel time and distance [138], the roll-off truck approach is appropriate for short distances. Further studies are essential to explore the application of this approach into forest biomass collection and transportation [136]. Bisson et al. [139] evaluated the use of a modified off-highway dump truck to carry shuttle residues from roadside piles to a centralized landing where a high capacity grinder loaded trailers pulled by 6 x 6 truck tractors to a transfer point. The trailers were then dropped at main roads for pickup by on-highway 6 x 4 truck tractors.
a)
b)
Figure 2.5. Forest biomass supply logistics for energy production: a) grinding and loading and b) transportation (chip van)
Production rates, costs, and fuel consumption rates for various grinding and transport configurations were calculated by Johnson et al. [140] for grinding harvest residues in the inland West United States. For the base conditions of the study, grinding at the landing and transporting ground material in chip vans to the refinery was the lowest
40 cost option, followed by grinding and shuttling to an intermediate access point for loading into a large truck. Shuttling loose material to an intermediate point before grinding was the most costly option. Zamora-Cristales et al. [131] developed an economic optimization model for processing and transporting forest biomass by considering options for comminution at each residue pile, processing in central yards, and processing at the bioenergy refinery using different types of trucks, trailers, and grinders. Forest residues could be ground or bundled. Their software model used a combination of simulation and mixed integer programming to identify the most costefficient system.
Pretreatment Pretreatment process is one of the essential steps for further efficient conversion of forest biomass [141]. Since lignocellulose is composed of cellulose, lignin, and hemicellulose, a pretreatment process is required to make these substances accessible for further conversion to be used for the production of energy (e.g., heat, power, and transport fuel), chemicals, or other purposes [142]. There are many pretreatment technology types for forest biomass have been introduced in the literature, including mechanical, chemical, and biological, as well as the various combinations of these technologies [142]. Pretreatment conversion technologies (e.g., thermolysis or pyrolysis processes) for forest biomass impacts the economic, environmental, and social aspects of the bioenergy production system [34]. Thermochemical technologies (e.g., combustion, hydrothermal liquefaction, gasification, and pyrolysis) are one of the most suitable pretreatment conversion processes in terms of simplicity and cost-
41 efficiency to produce heat, electricity, and fuels [38]. The cost of pretreatment varies with the capacity of facilities, the type of conversion technology, and type of biomass. Conversion efficiency and scale of conversion are two key factors that affect economic feasibility of bioenergy production. Wright and Brown [143] provided detailed information about the optimal size of bio-refineries, using different technologies such as gasification or pyrolysis, for different types of biomass. Additionally, to increase production yield and reduce the required heat, biomass required mechanical particle size reduction and drying through thermochemical processes to feeding feedstock into a pyrolysis reactor.
The pyrolysis process, by definition, involves chemical change brought on by heating biomass feedstock in the absence of oxygen [56]. It has been classified into three types, i.e., slow, intermediate, and fast pyrolysis, based on the length of reaction time [17]. In this review, the terms fast pyrolysis and pyrolysis are used interchangeably. Pyrolysis oil, or bio-oil, is produced from biomass feedstock mainly through fast pyrolysis processes that condense a mixture of oxygenated hydrocarbons and water [144]. The characteristics of the fast pyrolysis process include very high heating rates, reaction temperatures of around 500°C (932°F) in the vapor phase, short vapor residence times (typically less than 2s), and rapid cooling of the vapors to produce bio-oil [56]. The fast pyrolysis process decomposes the biomass to char, light gases, and a vapor phase of oxygenated hydrocarbons and water. Figure 2.6 illustrates the general biomass pyrolysis conversion and upgrading process as a post-conversion process. Bio-oil can be upgraded to biofuels through hydrotreating and hydrocracking processes to break
42 down long-chain hydrocarbons and reduce oxygen content [17,84,145]. Several parameters affect the pyrolysis process, including biomass species, temperature, particle size, feed rate, and residence time of volatiles. Additionally, internal bio-oil consumption, instead of diesel to provide electricity for pyrolysis plant and selling biochar, can reduce the cost up to18% [55]. Using bio-char instead of electricity consumption has a significant impact on bio-oil commercialization.
Solid Separation
Processes
Pyrolysis Reaction
Hydrotreating and Refining
Size Reduction
Drying
Products
Gas/Liquid Separation
Biomass
Cooling
Gas
Liquid
Bio-char
Gas
Bio-oil
Chemicals
Biofuels
Figure 2.6. General biomass pyrolysis conversion and post-conversion processing [38]
Luo et al. [146] reported that the biomass species and temperature play a significant role in pyrolysis efficiency and effectiveness, and noted the optimal temperature to produce the highest quality bio-oil is about 773 K (500 °C or 932 °F) within roughly one second. The yields of the three main products reported for fast pyrolysis process are 50-75% bio-oil, 15-25% biochar, and 10-20% syngas (non-condensable gas) [147,148]. More detailed information about the properties of pyrolysis oil from wood have been previously reported [149–151].
43 Bio-oil is a viscous mixture of oxygenated hydrocarbons and water. Bio-oil has 1200 kg/m3 bulk density, 20-30% water content, a heating value between 16 and 19 MJ/kg, and a product value of about $4.70/GJ [57]. Despite the addressed shortcomings, biooils can be upgraded into several chemicals, transportation fuels, and also hydrogen [52,152]. Chattanathan et al. [51] reported the major hydrogen production techniques, including the bio-oil steam reforming technique, and the important factors that are known in hydrogen production (e.g., temperature, catalyst type, and carbon ratio). The cost of production of hydrogen from bio-oil is reported to be $27.42/GJ ($3.25/kg or $0.86/liter) [153]. Czernik and French [154] estimated the cost of hydrogen production using fast pyrolysis bio-oil at $4.26/kg, of which bio-oil contributed to 56.3% of the production cost.
Fast pyrolysis production technology has developed rapidly since the late 1970s and reached near commercial status in the 1990s [155]. Czernik and Bridgwater [156] argued that the fast pyrolysis process has met commercialization targets for chemical products; and is being developed for producing liquid fuels. Several companies, such as Red Arrow Products Co. (U.S.), DynaMotive (Canada), BTG (The Netherlands), and Fortum (Finland), constructed different types of fast pyrolysis facilities, e.g., circulating fluidized bed and bubbling fluidized bed technologies, with different capacities ranging from 10-100 tons/day. Fast pyrolysis is currently the only simple, profitable, and efficient route for lignocellulosic biomass [54,84], and has an average annual return on investment of 3.5% [50]. The high uncertainty in the selling price of bio-char and bio-oil, however, reduces the viability of pyrolysis on a regional scale
44 [50]. The capital and operating costs of fast pyrolysis refineries were estimated to be about $48.3 million and $9.6 million, respectively, to convert 550 dry tons/day of forest chips to 426 tons/day of bio-oil [145]. Venderbosch et al. [157] discussed the principles, the main conversion technologies, and the economic viability of fast pyrolysis processes. They also reviewed bio-oil applications and technology development for production of biofuels and chemicals from bio-oil. Pu et al. [158] examined the major chemical constituents of lignocellulosic biomass and reviewed the recent advances in the conversion of biomass to biofuels. Ringer et al. [145] reported bio-oil production cost, selling price, and any applicable attributes of bio-oil from various reports. Rogers and Brammer [159] compared the selling price of bio-oil with gas oil and heavy fuel oil produced using different refinery sizes. Table 2.1 presents estimated bio-oil production costs as reported in the literature for different refinery sizes.
Fast pyrolysis concentrates forest biomass energy content into a smaller volume, which facilitates storage and transport [157,170]. Converting bio-oil to biofuels further increases the energy density of biomass, and reduces the high oxygen content (deoxygenation) and the formation C-C bonds. Deoxygenation of biomass, which can be accomplished via elimination of H2O and/or CO2, is crucial since oxygen reduces the heat content and inhibits blending of fuels [171]. Advantages of pyrolysis as a pretreatment technology include easier, less costly handling, higher bulk density, and removal of char [172,173]. Bio-oil has an energy density of six times that of green forest biomass at 45% moisture content [56]. Bradley [174] reported the advantages of
45 bio-oil compared to crude oil, e.g., its tendency to separate and sink when spilled in water. Bio-oil properties and process yield depend on biomass feedstock type, product collection methods, and conversion process type (e.g., rapid heating and cooling, reaction temperature, heat transfer rate, and residence time [58]), among other factors. Table 2.2 presents the attributes for several types of biomass and bio-oil, based on a study by Badger and Fransham [56].
Table 2.1. Previously reported bio-oil production costs Bio-oil cost Refinery size Year (metric ton/day) $/gal $/GJ 0.59 7.30 1000 1992 Solantausta et al. [160] 0.41 5.00 1000 1994 Cottam and Bridgwater [161] 0.50 6.10 1000 1994 Gregoire and Bain [162] 1.73 21.20 2.4 2000 Islam and Ani [163] 0.82 10.10 24 2000 Islam and Ani [163] 1.21 14.50 100 2002 Mullaney et al. [147] 0.89 10.60 400 2002 Mullaney et al. [147] 0.77 9.50 48-Wellman 2004 Peacocke et al. [164] 0.65 8.00 48-BTG 2004 Peacocke et al. [164] 0.43 5.10 2000 2005 Marker [165] 0.55 6.77 500 2005 Marker [165] 0.62 7.62 550 2006 Ringer et al. [145] 0.67 6.00 132 2008 Uslu et al. [45] €0.70- €0.73 €8.70- €9.30 - 2008 Van de Velden et al. [59] 0.74 4.04 200 2009 Dynamotive [166] 0.94 11.54 100 2010 Badger et al. [148] 0.48 6.00 - 2010 Czernik et al. [53] 0.83 10.19 2000 2010 Wright et al. [95] £0.733 £9.00 400 2012 Rogers and Brammer [159] 0.59 7.24 2000 2012 Jones and Male [167] 1.76 21.73 5-15 ovt1 2013 Brown et al. [168] 0.78 ($76/ton) 9.57 - 2014 Czernik and French [154] 1.15 14.11 13.6 2015 Mirkouei et al. [29] 1.10 13.5 50 2016 Mirkouei et al. [27] 0.31 2.09 N/A 1992 U.S. EIA [169] (Fuel Oil No. 6) 0.99 6.26 N/A 2015 U.S. EIA [169] (Fuel Oil No. 6) Assumes 4.55 kg/gal (1.2 kg/liter) density of bio-oil and 17.9 MJ/kg high heat value of bio-oil; 1ovt=oven dry ton; 2157.9 MJ/gal high heat value of fuel oil No. 6 Study
Current state-of-the-art and next generation conversion technologies, and their attributes for bio-oil production are shown in Table 2.3. Fast pyrolysis has been the
46 focus for the past 15 years, and is the focus of this literature review, because other technologies have either been in early development or are too costly. For instance, hydrothermal liquefaction is an alternative to fast pyrolysis for bio-oil production, but is more expensive since it requires higher pressures and longer residence times and uses high moisture feedstocks such as animal manure. However, the process produces a lower oxygen content bio-oil, has a higher process yield, and requires less additional processing than pyrolysis [175].
Table 2.2. Attributes of several types of biomass and bio-oil [56] Density (kg/m3)
MC (% wb)
350 Green whole tree chips 400 Solid wood, low density (Douglasfir) 865 Solid wood, high density (Oak) 1200 Bio-oil MC %wb: Moisture content on a percentage wet basis; EDR: Energy density ratio of bio-oil to other forms of biomass
Energy Density
EDR
45 12
MJ/kg1 10.7 17.1
GJ/m3 3.7 6.8
1/6 1/3
12 -
17.1 18.0
14.7 21.6
2/3 1
Table 2.3. Current and future generation bio-oil production technologies [175,176] Current
Under development
Technology Fast pyrolysis Bio-oil stabilization Hydroprocessing Catalytic fast pyrolysis Hydrothermal liquefaction Hydropyrolysis
Temperature (°C) ~500°C 150-250 300-350 ~500 ~375 ~375
Pressure (MPa) 0.1 10 20 0.1 ~20 1-5
Several organizations, such as Renewable Oil International (ROI) LLC [177], have developed technologies to convert biomass-based renewable resources to liquid or gas products through different intermediate technologies. Figure 2.7 shows a firstgeneration mobile pyrolysis refinery (trailer-mounted units) built by ROI in 2003.
47 Larger units would incorporate two or more smaller units hooked together. Badger and Fransham [56] discussed some of the broad applications of mobile bio-refineries, e.g., locating in close proximity to biomass resources and transportation of high energy density product (bio-oil) instead of low energy density product (forest biomass) to centralized facilities, which would, consequently, reduce handling, transportation, and storage costs to roughly half for bio-oil, compared to biomass feedstock in the form of chips.
Figure 2.7. A mobile fast pyrolysis bio-refinery (Courtesy of Phillip C. Badger, Renewable Oil International LLC)
Storage Capital and operational costs of storage facilities lead to higher logistics costs [134], thus a key challenge in forest biomass logistics is storage, especially in winter and spring seasons. This problem has received little attention in biomass fuel SC research, however, and researchers usually ignore the effects of biomass storage on the overall SC cost [178]. Various types of storage for biomass resources and associated material losses have been reported [17,178,179]. Rentizelas et al. [178] evaluated three methods
48 frequently used for biomass storage and applied them in a case study to compare the overall cost and material loss rate. In the first method, the storage facility is attached to bio-refinery and uses hot air generated by the refinery to reduce moisture content and avoid quality degradation of the biomass. In the second method, a covered, metalroofed storage facility is used without biomass drying. Finally, an ambient storage method was considered, which involved covering the biomass with a plastic film. The positive effect of using the first method depends on the biomass moisture level; forest biomass typically has a moisture content of at least 50-60% (wet basis), immediately post-harvest, and varies depending on the season. Additionally, since the storage space required for biomass feedstock is determined by the amount of biomass inventory, minimizing holding inventory is essential to reducing storage costs and develop an effective SC network. Rigdon et al. [180] evaluated the effect of various storage conditions on biomass constituents (e.g., cellulose, hemicellulose, and lignin). Their results indicated a dramatic decrease in ethanol production from 0.2 to 0.02 g/L when using uncovered storage over six-month period, while ethanol yields remained relatively stable when using covered storage. Therefore, covering biomass during storage is essential for maintaining high biofuel yields, especially for ethanol production. In addition to degradation (1% material loss/month), the common lowest cost option, ambient (uncovered) storage, also raises potential health risks, mainly due to high water content [178]. Covered storage has a higher cost, but lower biomass degradation (0.5% material loss/month). A high quality, closed warehouse with hot air drying can be used for storage, and exhibits negligible material loss [179].
49 Generally, an SC storage problem can be developed as a warehouse location problem, where the capital and operational costs of storage are optimized using different approaches. Biomass storage facilities (Figure 2.8a) are often located near harvesting sites or bio-refinery sites. Some studies have considered siting intermediate storage facilities between harvesting sites and bio-refineries [181,182]. An intermediate storage facility can increase overall cost due to the additional handling and transportation. Studies have explored siting the storage facility adjacent to the biorefinery [183]. Reducing biomass moisture content prevents potential safety and health issues, thus one of the major advantages of co-locating storage and the bio-refinery is the ability to dry biomass using waste process heat. Pettersson and Nordfjell [184] examined the fuel quality of logging residues by considering moisture content, dry matter losses, ash content, and calorific value, before and after storage and handling. Murphy et al. [185] developed a model for predicting the moisture and energy content of forest biomass over a 16-month period in Ireland. They pointed out that the key factors affecting biomass moisture content change are biomass type, storage type, and evapotranspiration. Larson et al. [186] analyzed the potential impacts of biomass feedstock storage losses on inventory management and plant-gate cost in east Tennessee. They found that last-in, first-out biomass feedstock inventory management can minimize the plant-gate cost. They also argued that harvesting time and location are two key factors influencing inventory and delivery management.
Drums, barrels, tanks, and similar containers can be used to store bio-oil. The two main approaches to store pyrolysis oils are underground storage tanks and external storage
50 tanks (Figure 2.8b). Decision makers need to consider various aspects in selecting an appropriate storage facility, including the mass/volume of substances. Existing complexity and properties of pyrolysis oil (e.g., corrosivity, instability, and high acidity) lead to limitations in storing this product. Biomass pyrolysis oils contain reactive organic compounds that can change the physical properties and increase the molecular weight of the oil during storage [187]. Since pyrolysis oils are corrosive to common storage tank materials (e.g., steel), further investigation is required to mitigate reactions between the oils and the tank materials [188]. Yang et al. [54] reported the recent developments in storing bio-oil, using different methods (e.g., physical and chemical) that can improve bio-oil properties.
a)
b)
Figure 2.8. Renewable energy storage solutions using a) biomass (chips) and b) bio-oil (pyrolysis oil)
Czernik [188] proposed a method that can be used to predict the effects of various storage conditions on pyrolysis oil properties. He also evaluated the impacts of different storage conditions on physical and chemical properties of pyrolysis oil from woody biomass using three storage temperatures: 37, 60, and 90 °C (98.6, 140, and 194 °F, respectively) over different periods of time. He used oils generated using the vortex reactor system at the National Renewable Energy Laboratory (NREL). The results
51 indicated that pH remained constant during storage, while molecular weight, water content, and viscosity increased with storage temperature and time. Czernik also evaluated polyester resin and high-density polyethylene materials for chemical resistance to pyrolysis oils at 20 and 60 °C (68 and 140 °F). The results indicated that both would be suitable materials for construction of pyrolysis oil storage tanks; each only exhibited slight swelling (less than standard). Prior to Czernik, Aubin and Roy [159] and Soltes et al. [160] reported that wood pyrolysis oils can destroy and corrode aluminum and carbon steel at even moderate temperatures. For best performance, pyrolysis oils should be stored in air-free stainless steel and/or polymer tanks at room temperature.
Techno-Economic Modeling Techno-economic assessment has been used by investigators to explore and couple the technical and economics standpoints of the processes involved in bioenergy production from lignocellulosic biomass that represent the largest portion (~60%) of the total SC costs [141]. Techno-economic assessment can define the potential methods as it depends on several factors, such as, biomass type and desired products, as well as the process objectives (e.g., economic assessment or environmental impacts) to determine the economic feasibility [191]. The main purpose of techno-economic modeling is to evaluate investment factors and better define the bottlenecks of process configurations, and consequently to promote the bioenergy yield and reduce capital expenditures [95]. The basic techno-economic modeling includes cash flow and rate of return analysis by employing Aspen Plus software for technical modeling (i.e., mass and energy
52 calculations) and Aspen Icarus software for size and equipment costs, along with a spreadsheet investment analysis [95,145]. A risk analysis can be developed to evaluate the maturity of the technology and accuracy of the economic analyses, which the accuracy is usually ±30% of the actual cost [191]. The total plant investment cost includes following major costs: project investment, fixed operating cost, and variable operating cost, as well as overhead and contingency factors to total installed equipment cost. Swanson et al. [191] reported the general techno-economic model assumptions for biofuels production, including assumption of financial, capital costs, operating costs, feedstock and products, process, and miscellaneous in detail.
Based on the existing techno-economic analysis, pyrolysis process has been reported as being potentially cost-effective pretreatment method for production of transportation fuel from lignocellulosic feedstocks [95,192]. Additionally, fast pyrolysis technology has been commercialized, and there are several companies producing bio-oil, such as Eysen in Ontario and Wisconsin, Dynamotive in Vancouver, Advanced Biorefinery Inc. in Ontario, and Renewable Oil International, LLC in Alabama, as well as research laboratories in National Renewable Energy Laboratory in Golden, Colarado, Iowa State University, and University of Oklahoma [193]. There is a wide range of pyrolysis reactors introduced, which comprehensively reviewed by [155,156,194]. The following reactors have been implemented commercially: bubbling fluidized bed, circulating fluidized bed, mechanically mixed beds, rotating cone reactor, vacuum pyrolysis, and ablative pyrolysis [195]. Prior studies reported that post-conversion methods are immature and there are no commercial post-conversion enterprises for producing
53 transport fuel from forest biomass. Wright et al. [95] examined a techno-economic study on fast pyrolysis of biomass to bio-oil, and upgrading to diesel range fuels in two different scenarios: 1) using self-generated hydrogen from bio-oil and 2) purchasing hydrogen for fuel upgrading. The capital costs of establishing these plants are $287 and $200 million, respectively [95]. Their process design includes the following eight components: chopping and grinding, drying, pyrolysis, solid removal, bio-oil recovery, storage, combustion, and hydroprocessing. Their results indicate transport fuel production cost in Scenario 1 (when hydrogen is produced from the process itself) is higher. Other bioenergy production pathway, such as biochemical or gasification, can be promising, but it requires further techno-economic analysis to design an efficient platform. The results of prior studies indicate a significant difference in technology choices and development levels due to various assumptions used by each study [191]. Table 2.4 represents a number of the techno-economic studies use for bio-oil production via pyrolysis of biomass.
Polagye et al. [57] examined the economic feasibility of biofuel and bioenergy production through techno-economic analysis on four types of bio-refineries (i.e., mobile, transportable, stationary, and relocatable) in order to quantify the transportation-production trade-offs. The main characteristics of these bio-refineries are given in Table 2.5, and other detailed characteristics are provided in their study. Additionally, they provided operational and economic parameters for loading, unloading, chipping, and debarking. Polagye et al. proposed a fast pyrolysis cost chain (annual cost) for different types of facilities. The relocatable bio-refinery can be
54 disassembled and rebuilt to tradeoff between the large production refinery and feedstock availability. Polagye et al. also argued that biofuels production using a mobile or transportable bio-refinery is more costly than using a stationary or relocatable bio-refinery. Biofuel production using a centralized bio-refinery (i.e., stationary and relocatable) is preferred by industry, due to larger capacities that can achieve lower costs in large-scale systems.
55 Table 2.4. Techno-economic studies for bioenergy production via fast pyrolysis (1991-2016) Study
Year
Elliott et al. [196]
Research Overview A review study on thermochemical conversion of biomass to liquid fuel (pyrolysis and liquefaction) was provided for 1983 to 1990. Techno-economic assessment results indicated the pyrolytic process was moving quickly forward in development. A techno-economic study was conducted to estimate cost of using pyrolysis and liquefaction to produce transportation fuel from woody biomass. The estimated Solantausta et al. [160] lowest cost for transportation fuel production was $12/GJ and fuel efficiency is around 50%. Cottam et al. [161] A techno-economic assessment was conducted to investigate and compare economic and technical opportunities for upgrading pyrolysis oil to higher quality fuels. A process model was developed for bio-oil production from wood chips to investigate the characteristics and parameters (equipment cost, plant size, and production Gregoire and Bain [162] capacity) of the process to gain insights about techno-economic evaluation and viability of process developments. A decision support system was developed to facilitate techno-economic assessment of biomass to bioenergy systems. Economic and technical parameters were Mitchell et al. [197] assessed for various feedstocks and conversion technologies. The reported results indicate the relation between the delivered feedstock cost and bioenergy cost. A techno-economic assessment and comparison of pelletizing and pyrolysis were reported for wood fuels production. The results illustrate the need for further Östman et al. [198] development is necessary to promote the quality of pyrolysis oil. Techno-economic assessments of thermal processes (combustion, pyrolysis, and gasification) to generate power (dual fuel diesel engines) from wood chips were Bridgwater et al. [42] reported, considering activities from chip transportation to power supply to the grid. It was found a large pyrolysis plant would have lower cost than gasification. Techno-economic assessments were conducted for different types of pyrolysis (fluidized bed and circulating fluid beds). It was concluded that the bio-oil industry Mullaney et al. [147] was immature and had many challenges for commercialization, such as bio-oil quality, plant size, and bio-oil standards. A study synthesized the relevant issues (technical requirements, bio-oil stability, applications, environmental, and safety) for advancing the pyrolysis technology to Ringer et al. [145] commercialization. A techno-economic analysis was conducted on a bio-oil plant, which included feed handling and drying, pyrolysis, combustion, and recovery. A techno-economic assessment was reported for bio-fuel production from various materials (chips, pellets, bio-oil, and methanol), using different facilities (mobile). Polagye et al. [57] The reported results indicate the cost competitive thinning option for transferring beyond 300km distance are pelletization and fast pyrolysis. A techno-economic comparison of different pretreatment technologies (torrefaction, pelletization, and fast pyrolysis) was reported; torrefaction was found most Uslu et al. [45] promising in terms of cost. Pretreatment technologies were shown to have a significant impact by easing transportation and handling. A techno-economic assessment was presented to evaluate three different preconversion processes (rotating cone reactor pyrolysis, fluidized bed reactor pyrolysis, Magalhães et al. [199] and torrefaction) for bio-oil production from biomass. The results indicate that torrefaction is the most cost-effective option. A techno-economic analysis was conducted to compare the bio-oil production from wood and agriculture using fast pyrolysis. It was reported that agricultural Osamaa et al. [200] residues were more challenging for bioenergy production due to the high amount of alkali metals and nitrogen in the oil. A techno-economic assessment was conducted for transportation fuels production from pyrolysis oil, using two different scenarios: self-produced hydrogen or Wright et al. [95] purchasing the hydrogen. The reported results indicate that fuel cost is almost 30% less when hydrogen is purchased. A techno-economic assessment was used to evaluate three different conversion technologies (pyrolysis, gasification, and biochemical) for liquid fuel production Anex et al. [201] from biomass. The reported results indicate that the fuel from pyrolysis had the lowest cost and biochemical had the highest cost. A techno-economic analysis was presented to compare eight different pyrolysis configurations to identify promising technology that indicate major cost-drivers, i.e., Trippe et al. [202] biomass feedstock and investment costs. The reported results indicate the biosyncrude (bio-oil and biochar) was possible produce in Germany at costs of €35/MWh. A techno-economic analysis was performed to evaluate the bio-oil production, using fast pyrolysis transportable plant with 100 dry ton per day capacity. It was Badger et al. [148] reported the energy cost bio-oil and biochar was valued at $6.35/MMBTU. Also feedstock cost can drastically affect the final bio-oil cost. A techno-economic comparison between different conversion configurations was performed. The reported results indicate that fast pyrolysis has higher internal Ghezzaz et al. [203] return rate, hardwood is the most suitable feedstock for biochemical process, and profitability is dependent on biomass cost and quality. A review study was presented on techno-economic analysis for bio-oil production, as well as a model to evaluate the cost and performance based on the prior Rogers and Brammer [159] scenarios in the literature. The results indicate it is possible to produce bio-oil from biomass with similar costs as distillate fuel oil production in the U.K. A techno-economic analysis was performed for the fast pyrolysis and hydroprocessing pathway to investigate the feasibility of cellulosic biofuels from biomass. It Brown et al. [204] was concluded the minimum fuel selling price of diesel fuel and gasoline produced via fast pyrolysis along with hydroprocessing to be $2.57/gal. A techno-economic analysis was reported for biofuel production using mobile conversion facilities (fast pyrolysis and torrefaction). Results show converting forest Brown et al. [168] residue to more energy dense products (bio-oil and torrefied wood) has lower delivery cost than using conventional wood chip delivery. A techno-economic analysis was provided for bio-oil production from empty fruit bunches via fluidized-bed fast pyrolysis. Several factors (capital cost, payback Do et al. [205] period, and product value) were estimated to evaluate economic feasibility. It was reported that plant size and bio-oil yield are key parameters on the product value. A techno-economic analysis was performed for biofuel production from microalgae using two different conversion pathways (thermal drying prior to catalytic Thilakaratne et al. [206] pyrolysis and mechanical dewatering prior to catalytic pyrolysis). The results show mechanical dewatering has higher energy efficiency than thermal drying. A techno-economic analysis under uncertainty was conducted for several conversion pathways (gasification, pyrolysis, liquefaction, and fermentation) to investigate Zhao et al. [207] economic feasibility of biofuel production. Results show none of the pathways would be profitable, but fast pyrolysis and hydroprocessing has the lowest fuel price. De Jong et al. [192] A comparison on six conversion pathways was provided for bio-jet fuel production, using techno-economic and pioneer plant analysis. Results show none of the
1991 1992 1993 1994 1995 2000 2002 2002 2006 2007 2008 2009 2009 2010 2010 2010 2010 2011 2012 2013 2013 2014 2014 2015 2015
56
Patel et al. [55]
pathways were able to match the petroleum jet fuel price. Their analysis reported that hydrothermal liquefaction and pyrolysis were promising alternatives. A review study reported techno-economic and life cycle assessments for different conversion technologies (gasification, combustion, pyrolysis, liquefaction, carbonization, and co-firing) for bioenergy production. Three major indicators (production costs, functional units, and environmental impacts) were compared. It was concluded that techno-economic assessment on product co-generation and different formation pathways for a product would be useful.
2016
57
Table 2.5. Portable and fixed bio-refinery characteristics [57,148] Mobile Capacity (tons/day) Location Lifetime (years) Salvage Value (%) Mobility Capital ($ thousand) Capital Cost ($ thousand) Operating Cost ($ thousand/year) Setup/Breakdown Time Mileage Charge ($/km) Transport Speed (km/h) Fast Pyrolysis Rate ($/t) Advanced Fast Pyrolysis Rate ($/t)
15 Logging deck 15 0 60 1,472 183 4 hrs. 1 40 162 159
Transportable 100 In-forest 15 0 200 6,031 3,316
Stationary Variable Outsideforest Variable 20 N/A 14,300 3,052
Relocatable 500 Outsideforest 20 0 -
4 days 3 40 77 73
N/A N/A N/A 52 48
2 mos. 100 80 61 58
Systematic Literature Review A systematic review (SR) was used to identify leading journals and scholars and to assess the trends in biomass-to-bioenergy SC research. This SR evaluated publication data, citations, and keywords in a quantitative manner for recent literature. In addition, the key research methodologies employed by the most-cited articles were classified.
Analysis of Publication Data In this study, the authors used the Web of ScienceTM (Thomson-Reuters) to gather the titles, abstracts, and keywords for relevant international conference and journal articles between January 1, 2000 and June 30, 2015. These records were collected into two databases using the following sets of keywords in the Web of ScienceTM queries:
Keyword Set 1: (Lignocellulosic OR Forest OR Wood) AND (Biomass OR Biofuel OR Bioenergy) AND Supply Chain
58 Keyword Set 2: (Forest OR Wood OR Lignocellulosic) AND (BioOil OR Bio-oil OR Pyrolysis Oil) AND (Economic OR Cost OR Supply Chain).
A period of January 1, 2000 to June 30, 2015 is chosen for SR due to the low number of studies in forest biomass to bioenergy SCs before 2000. From 1980 to 1999, only four and eleven records were found from keyword Set 1 and Set 2, respectively. The query for Keyword Set 1 from the Web of ScienceTM Core Collection (1965-present) generated a database of 277 records related to forest biomass SCs. Using Keyword Set 2, 126 records were identified that were related to forest pyrolysis oil SCs. A comparison of the two databases found nine papers in common. VOSviewer software [208] was applied to analyze the database records. From the SR results for the keyword queries, it can be seen that biomass-to-bio-oil SC research is accelerating, with the most rapid growth occurring over the last 5-7 years (Figure 2.9).
*
Number of Records
100
80
60
40
20
0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Figure 2.9. Increase in publications of biomass-to-bio-oil SC research (Jan. 2000 to June 2015, *estimated for July-Dec. 2015)
59 Table 2.6 reports the ten journals with the most records in the databases for both sets of keywords. The top journals in both databases are the International Journal of Biomass and Bioenergy, Renewable Sustainable Energy Reviews, the Journal of Biofuels, Bioproducts & Biorefining (Biofpr), and the Journal of Applied Energy.
Table 2.6. Top ten journals identified based on number of records (Jan. 2000 to June 2015) Keyword Set 1 Source Titles Reco rds 38 Biomass & Bioenergy 17 Renewable Sustainable Energy Reviews 14 Applied Energy Biofuels Bioproducts Biorefining Journal of Cleaner Production Scandinavian Journal of Forest Research Renewable Energy Bioresource Technology Silva Fennica Energies
% of 277 13.7 6.1
Keyword Set 2 Source Titles
Reco rds 11 11
% of 126 8.7 8.7
9
7.1
9
7.1
13
4.7
Biomass & Bioenergy Biofuels Bioproducts Biorefining Renewable Sustainable Energy Reviews Energy Fuels
12
4.3
Bioresource Technology
7
5.6
9
3.2
Green Chemistry
5
3.9
8 8
2.9 2.9
5 4
3.9 3.2
7 7
2.5 2.5
Applied Energy Industrial Engineering Chemistry Research Fuel Journal of Analytical and Applied Pyrolysis
4 3
3.2 2.4
5.1
Table 2.7 reports the 10 countries with the highest level of authorship in this field of research. Most publications in both databases are authored by researchers from the United States, Canada, Finland, and England.
60 Table 2.7. Top ten countries based on number of records (between Jan. 2000 and June 2015) Keyword Set 1 Countries/Territories Records 80 USA 32 Canada 29 Finland 26 Italy 24 Sweden 23 England 19 Austria 14 Netherlands 12 Norway 12 Germany
% of 277 28.9 11.5 10.5 9.4 8.7 8.3 6.9 5.1 4.3 4.3
Keyword Set 2 Countries/Territories Records 42 USA 17 Canada 12 England 8 Finland 7 Germany 6 China 6 Netherlands 6 Italy 5 Spain 4 India
% of 126 33.3 13.5 9.5 6.3 5.5 4.8 4.8 4.8 3.9 3.2
Table 2.8 reports the ten most productive scholars in each research field (database) examined. Taraneh Sowlati (eleven papers) and Amit Kumar (five papers) have the most publications in the forest biomass SC (Keyword Set 1) and bio-oil SC (Keyword Set 2), respectively. Sowlati has an academic background in biomass SC management, mathematical modeling and optimization, life cycle assessment, multi-criteria decision making, and simulation. Kumar has an academic background in energy and environmental modeling, life cycle assessment, and techno-economic assessment of energy systems.
Table 2.8. Most productive scholars based on number of records (Jan. 2000 to June 2015) Keyword Set 1 Authors Records 11 Sowlati, T. 9 Shah, N. 7 Sikanen, L. 6 Spinelli, R. 6 Roser, D. Junginger, M. 6 6 Gonzalez, R. Asikainen, A. 6 5 Saloni, D. 5 Ranta, T.
% of 277 3.9 3.2 2.5 2.2 2.2 2.2 2.2 2.2 1.8 1.8
Keyword Set 2 Authors Records 5 Kumar, A. 4 Solantausta, Y. 4 Czernik, S. 4 Brown, R. 3 Wang, Z. 3 Wang, H. 3 Sadhukhan, J. 3 Oasmaa, A. 3 Ng, K. 3 Bridgwater, A.
% of 126 3.9 3.2 3.2 3.2 2.4 2.4 2.4 2.4 2.4 2.4
61
Table 2.9 presents the ten most productive organizations in forest biomass SC (Keyword Set 1) and bio-oil production (Keyword Set 2) research. The three most productive organizations in each database are the University London Imperial College of Science, Technology and Medicine (sixteen papers), the University of British Columbia (fourteen papers), and the University of Eastern Finland (nine papers); and Iowa State University (seven papers), the University of Alberta (five papers), and the Pacific Northwest National Laboratory (five papers), respectively.
Table 2.9. Ten most productive organizations based on number of records (Jan. 2000 to June 2015) Keyword Set 1 Organizations Reco rds 16 Univ London Imperial Coll Sci Technol Med 14 Univ British Columbia 9 Univ Eastern Finland 8 Univ Padua 8 Texas A&M Univ 8 Finnish Forest Res Inst 7 Univ Utrecht 7 Swedish Univ Agr Sci 6 N Carolina State Univ 6 Lappeenranta Univ Technol
% of 277 5.8 5.1 3.2 2.9 2.9 2.9 2.5 2.5 2.2 2.2
Keyword Set 2 Organizations Reco rds 7 Iowa State Univ Univ Alberta Pacific NW Natl Lab Mississippi State Univ Natl Renewable Energy Lab Aston Univ Washington State Univ Univ Manchester Univ British Columbia VTT, Technical Research Center of Finland
5 5 5 4 4 3 3 3 2
% of 126 5.6 3.9 3.9 3.9 3.2 3.2 2.4 2.4 2.4 1.6
Table 2.10 presents the top ten research areas in each database, based on number of records. Energy fuels, biotechnology applied microbiology, and engineering are the three most common research areas in biomass-to-bio-oil SC research.
62 Table 2.10. Ten most common research areas based on number of records (Jan. 2000 to June 2015) Keyword Set 1 Research Areas Recor ds 149 Energy Fuels 73 Biotechnology Applied Microbiology 68 Engineering
% of 277 53.8 26.4 24.5
Agriculture Environmental Sciences Ecology Forestry
56 50
20.217 18.1
45
16.2
Materials Science Chemistry Thermodynamics Meteorology Atmospheric Sciences
11 9 6 4
3.9 3.2 2.2 1.4
Keyword Set 2 Research Areas Recor ds 70 Energy Fuels 44 Engineering
% of 126 55.6 34.9
Biotechnology Applied Microbiology Chemistry Agriculture
33
26.2
26 23
20.6 18.2
Environmental Sciences Ecology Materials Science Spectroscopy Polymer Science Thermodynamics
13
10.3
6 3 3 2
4.8 2.4 2.4 1.6
Analysis of Citation Data The 394 papers contained in the combined database have been cited by 6957 other publications from January 2000 through June 2015. Figure 2.10 indicates the annual increase in the number of citations for the papers contained in the combined databases. Figure 2.9 and Figure 2.10 illustrate that the numbers of publications and citations have been growing exponentially over the past 15 years.
*
Number of Citations
2000
1500
1000
500
0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Figure 2.10. Increase in citations each year (Jan. 2000 to June 2015, *estimated for July-Dec. 2015)
63
The most frequently cited journals in forest biomass-to-bio-oil SC research between January 2000 and June 2015 are reported in Table 2.11. Energy and Fuels, the International Journal of Biomass and Bioenergy, and the Journal of Biofuels, Bioproducts & Biorefining are the three most-cited journals in the combined database.
Table 2.11. Ten most-cited journals (between Jan. 2000 and June 2015) Source Title Energy & Fuels Biomass & Bioenergy Biofuels, Bioproducts & Biorefining Green Chemistry Renewable & Sustainable Energy Reviews Bioresource Technology ChemSusChem Energy & Environmental Science Energy Conversion & Management Applied Energy
Cumulative Citations 1,025 1,000 692 524 294 278 237 183 175 158
The most frequently cited publications in biomass-to-bio-oil SC research between January 2000 and June 2015 are reported in Table 2.12. The three most-cited publications are by Czernik and Bridgwater, [156] (871 citations); Alonso, Bond, and Dumesic, [38] (454 citations); and Lange, [171] (189 citations).
Analysis of Keywords Figure 2.11 is a bibliometric map highlighting the frequency of keywords used in the most-cited studies from the combined database (394 papers). The map uses clustering and colors to indicate the frequency of occurrence, with cool colors (blue and green) for less frequent keywords and warm colors (yellow, orange, and red) for more frequently used keywords. Supply chain, emission, environmental impact, ghg, yield,
64 bio-oil, forest biomass, renewable energy, crop, sustainability, transportation, storage, logistics, pretreatment, market, pyrolysis, bio-refinery, and location are highlighted by the obtained visual network as the most-frequently used terms in the identified articles. The less-frequently used terms indicate emerging areas of research that may need further attention from investigators who have the relevant expertise.
Table 2.12. Ten most-cited studies and scholars (between Jan. 2000 and June 2015) Authors Czernik, Bridgwater [156] Alonso, Bond, Dumesic [38] Lange [171]
Demirbas, Balat [149]
Luo, Wang, Liao, Zhou, Gu, Cen [146] Koh, Ghazoul [209]
Lin, Huber [83]
Effendi, Gerhauser, Bridgwater [210] Kim, Kim, Dale [211] Gnansounou, Dauriat [212]
Article Title Overview of applications of biomass fast pyrolysis oil Catalytic conversion of biomass to biofuels Lignocellulose conversion: an introduction to chemistry, process and economics Recent advances in the production and utilization trends of bio-fuels: a global perspective Research on biomass fast pyrolysis for liquid fuel
Source Title Energy & Fuels
Total Citations 871
Year 2004
Green Chemistry
454
2010
Biofuels, Bioproducts & Biorefining Energy Conversion and Management
189
2007
166
2006
Biomass & Bioenergy
144
2004
Biofuels, biodiversity, and people: understanding the conflicts and finding opportunities The critical role of heterogeneous catalysis in lignocellulosic biomass conversion Production of renewable phenolic resins by thermochemical conversion of biomass: a review Biofuels, land use change, and greenhouse gas emissions: some unexplored variables Techno-economic analysis of lignocellulosic ethanol: a review
Biological Conservation
134
2008
Energy & Environmental Science
133
2009
Renewable & Sustainable Energy Reviews
126
2008
Environmental Science & Technology Bioresource Technology
121
2009
116
2010
65
Figure 2.11. Bibliometric map of keywords (density visualization from VOSviewer software)
66 Analysis of Research Methodologies The qualitative method is used to classify the ten most-cited studies as shown in Table 2.13 based on the addressed categories and subcategories in Figure 2.3. The results indicate conceptual research methods based on literature review are the most prevalent in the context of analytical studies, while, in the context of empirical studies, experimental design was the most applied research method. Since there is a need to improve bio-oil quality and to optimize processes, many studies have focused on empirical exploration to develop commercial processes, e.g., Lin and Huber [83] and Kim et al. [211].
Table 2.13 Classification of the ten most-cited studies Aut hors [156] [38] [171]
[149]
Analy tical Conce ptual Conce ptual Conce ptual Conce ptual
Experi mental
[146] [209] [83]
[210]
Conce ptual Experi mental Conce ptual Experi mental
[211] [212]
Empir ical
Conce ptual
Classifi cation Lit. Review Lit. Review Lit. Review Lit. Review Exp. Design Lit. Review Exp. Design Lit. Review Exp. Design Lit. Review
Highlight Reviewed scientific developments in application of bio-oil and concluded with suggestions for future developments Reviewed catalytic strategies and addressed the importance of hydrogen in producing biofuels Reviewed the economics of biomass conversion Reviewed recent global advances in production of biofuels Developed new pyrolysis system at feed rate up to 20 kg/h Highlighted positive and negative impacts of biofuel use Developed new catalytic process for production of cost-efficient lignocellulosic biofuels Reviewed production of renewable phenolic resins by thermochemical conversion of biomass Examined several variables (cropping management) that affect GHG emissions of biofuel production Performed techno-economic evaluations of lignocellulosic ethanol production
Discussion The earliest work summarizing forest biomass SC studies can be traced to Hakkila [93], who argued that the consumption pattern of forest products significantly changed after
67 the first energy crisis. Specially, since 1973, fuelwood has been consumed more than any other biomass product (e.g., paper and wood products). Research has mainly focused on understanding the potential challenges of forest biomass SC, logistics management, and costs related to ground preparation, planting, cultivation, harvesting, collection, processing, in-forest and main-road transportation, pretreatment, and storage. Later studies started focusing on assessing forest-based biomass SCs through quantitative models to calculate the delivered costs of biomass and to identify the relative advantages and challenges of biomass-to-bioenergy supply, such as reduced environmental impacts of bioenergy consumption [71,213]. Researchers mainly focused on the appropriate planning, management, and operational approaches through network modeling and optimization to recognize the failure or success of a fledgling industry [214]. Mathematical programming and simulation-based modeling are the two main approaches applied to develop SC models [215]. As reviewed above, key contributions have emerged from a range of disciplines, e.g., operations research, SC management, decision making, and sustainability assessment.
Growing use of renewable energy resources is impacting each of the three dimensions of sustainability across the world, which includes access to low-cost, secure energy (economic dimension), reduced net carbon emissions (environmental dimension), and maintaining and developing rural communities (social dimension). It has been reported that logistics is a key activity that can improve the efficiency and profitability of the forest biomass industry [34,116,121]. This review of the state-of-the-art in forest biomass-to-bio-oil SC research found that economic optimization models have been
68 widely applied to assist forest biomass SC logistics decision making. Allen et al. [71] reported that biomass-based renewable energy logistics costs (e.g., processing and transportation) constitute a significant portion of the total costs of biomass-to-bio-oil SC network. Several studies [42,216,217] reported the costs of transporting bio-oil from distributed pyrolysis bio-refineries to a centralized bio-refinery. They used different methods of transportation (i.e., tanker and pipeline), different maximum loads (24-44 tons tanker capacity and 560 m3/day pipeline capacity), and also reported on the effects of fixed and variable costs, distance, and approaches on transportation costs of bio-oil.
Over the last 30 years, numerous studies have investigated the economic aspect of biomass-to-bioenergy SCs. In general, bioenergy production costs fall under two major categories: supply costs, which include purchasing and transferring feedstocks, and production costs, which include capital and operational costs of energy conversion. Some past studies have investigated various parameters (e.g., type of feedstock, scale of process, time, and temperature) of conversion technologies to meet economic targets. Past studies have found that some biomass feedstocks, such as lignocellulosic feedstocks, are relatively inexpensive, but are difficult to convert to useful bioenergy. On the other hand, sugars and starches are expensive biomass feedstocks but they are more easily converted to useable energy. There is a tradeoff between feedstock supply cost and production cost. Similar to feedstock delivery, lignocellulosic conversion requires large investments ($50-$100/bbl.), however, collection and delivery of lignocellulosic feedstocks are relatively inexpensive. Therefore, cost reduction in
69 lignocellulosic biomass conversion plays a key role in biomass-to-bioenergy SC implementation [218]. From the review undertaken above, it was found that the key parameters and activities that affect cost modeling and optimization of forest biomassto-bioenergy SCs include 1) biomass feedstock attributes, e.g., availability, moisture content, bulk density, purchase price, and yield of biomass [37]; 2) harvesting and collection [109]; 3) logistics [178]; 4) pretreatment [56]; and 5) storage [188]. In particular, it was found that the upstream segment of the forest biomass-to-bio-oil SC is influential on the economic viability of the production system [13,30,219].
One of the main roadblocks for bioenergy production is the lack of economical conversion technologies [83]. In addition to logistics, conversion and post-conversion technologies can affect the economic feasibility of using forest biomass. While capital costs of refinery facilities do not typically scale linearly with capacity [148], the operational cost is more directly impacted by the scale of the system. Thus, there is a significant need for studies focusing on upscaling biomass conversion and postconversion processes to accelerate the commercialization of biomass-to-bioenergy production technologies. This study gives decision-makers a base for substantiating further development of logistics systems and pretreatment technologies.
Biomass quality attributes (e.g., energy content, moisture content, particle size, ash, and contaminant content) affect the processing operations (e.g., chipping, grinding, and sorting), transportation attributes (e.g., type, capacity, and distance), and conversion and post-conversion technologies properties (e.g., resource type, product type, and
70 capacity). Biomass yield and bulk density are two key factors impacting transportation cost – higher yields lead to reduced truck travel distance and higher bulk density reduces the number of truck trips [13]. A standard tanker truck can carry bio-oil with an energy content of 558 GJ (13-18 MJ/kg), which is twice the energy content of wood chips carried by a chip van [120].
Several studies investigated the upstream segment of biomass to bio-oil SC networks. This study extends prior work by incorporating recent research into mixed-mode networks of mobile and fixed refineries to trade off between processing and transportation costs [29,220]. Consequently, successful mobile refinery development will enable locating facilities in the field near biomass sources, thus enabling transport of higher energy density product to a central facility. It was also found that biomassto-bioenergy SC literature has not paid close attention to biomass and bioenergy storage strategies. The lowest cost solution is adopted in most cases, without considering the effects of other biomass storage solutions that can reduce overall cost [178]. Simultaneous consideration of all entities in the upstream segment could lead to strategies for the economic success of a forest fuel development project. Future studies, for instance, should evaluate the biomass storage methods addressed above, as well as novel storage strategies to minimize system costs.
Conclusions and Future Direction The presented study used both a narrative review and a systematic review of the literature to assess the linkages among current studies and to identify the potential
71 technologies and practices that would address existing gaps and future perspectives within the upstream segment of biomass-to-bioenergy SC. From the narrative review, it is clear that most studies that examine the upstream segment of biomass-to-bio-oil SCs focus on isolated problems within three main activities: harvesting, logistics, and storage. From the systematic review, it is apparent that forest biomass-to-bio-oil SC research has been a rapidly growing area of interest over the last 15 years. From both reviews, it is evident that there is a need for more focused research on forest biomassto-bio-oil SC issues, specifically on pretreatment process development (e.g., pyrolysis) and implementation at the industry level. Also, there is a need for more investigation into modeling and optimization of pretreatment as a part of the upstream segment of biomass-to-bioenergy SCs. The review highlights the gap between work in literature and industry practice.
While many studies predominantly explored the complete SC (i.e., upstream, midstream, and downstream) to tackle sustainability issues (i.e., economic, environmental, and social), there is a dearth of literature for detailed analysis of each individual segment of the SC. Exploring and addressing the gaps in each segment can raise the awareness of decision makers and, subsequently, aid in identifying alternative ways of making business more robust and sustainable. In summary, this study reveals some of the gaps in research related to the upstream segment of forest biomass-tobioenergy SCs. Specifically, the following potential paths for a future research are defined:
72
Development of holistic models for integrating pretreatment processes into the upstream segment of forest biomass-to-bioenergy SCs
Exploration of pretreatment processes to identify economic, environmental, technological, and political challenges and barriers to implementation
Development of a holistic techno-economic assessment method to evaluate the investment factors by comparing different system designs and assumptions
Development and implementation of novel pretreatment processes that could be adopted by industry into practice (e.g., transportable bio-refinery facilities)
Exploration of the state-of-the-art within other disciplines to integrate adopted methods and approaches (e.g., mathematical analysis and experimental design)
Exploration of issues related to the existing metrics and measures for optimization based on triple bottom line sustainability.
Further, there is a need to develop a detailed research plan for each of the paths proposed. Investigations into biomass-to-bioenergy SCs are increasing due to internal and external stakeholders’ needs, which include growing demand for bioenergy and for reduction of economic and environmental impacts. This study analyzed the nature of existing publications, citations, keywords, and research methodologies. Limitations of this study include the use of specific combinations of keywords for generating a database (394 papers) of prior research from the Web of ScienceTM. Thus, this approach may have omitted some related papers from consideration. It is hoped this study will lead to production of more sustainable bioenergy and chemical products from forest biomass.
73
Acknowledgment The authors wish to acknowledge Phillip C. Badger (Renewable Oil International) for their valuable input in the development of this literature review.
References [1]
Light AR. Federalism and the Energy Crisis: A View from the States. Publius
1976;6:81–96. doi:10.2307/3329606. [2]
Botard S, Aguilar F, Stelzer H, Gallagher T, Dwyer J. Operational costs and
sensitivity analyses of integrated harvest of solid hardwood products and woody biomass: Case study in central Missouri. 2015. [3]
Aguilar FX, Saunders A. Policy instruments promoting wood for energy uses:
evidence from the continental United States. 2010. [4]
Meckler M. Innovative energy design for the ’90s 1993.
[5]
Carpenter D, Westover TL, Czernik S, Jablonski W. Biomass feedstocks for
renewable fuel production: a review of the impacts of feedstock and pretreatment on the yield and product distribution of fast pyrolysis bio-oils and vapors. Green Chem 2014;16:384–406. [6]
De Meyer A, Cattrysse D, Rasinmäki J, Van Orshoven J. Methods to optimise
the design and management of biomass-for-bioenergy supply chains: A review. Renew Sustain Energy Rev 2014;31:657–670. [7]
Mafakheri F, Nasiri F. Modeling of biomass-to-energy supply chain operations:
Applications, challenges and research directions. Energy Policy 2014;67:116–126. [8]
US DOE/EIA. Monthly Energy Review - December 2014 - mer.pdf 2014.
http://www.eia.gov/totalenergy/data/monthly/pdf/mer.pdf
(accessed
January
24,
2015). [9]
Gowen MM. Biofuel v fossil fuel economics in developing countries: How
green
is
the
4215(89)90068-2.
pasture?
Energy
Policy
1989;17:455–70.
doi:10.1016/0301-
74 [10]
Gold S. Bio-energy supply chains and stakeholders. Mitig Adapt Strateg Glob
Change 2010;16:439–62. doi:10.1007/s11027-010-9272-8. [11]
Jäppinen E, Korpinen O-J, Ranta T. The effects of local biomass availability
and possibilities for truck and train transportation on the greenhouse gas emissions of a small-diameter energy wood supply chain. BioEnergy Res 2013;6:166–177. [12]
Mirkouei A, Haapala KR, Sessions J, Murthy GS. Reducing Greenhouse Gas
Emissions For Sustainable Bio-Oil Production Using A Mixed Supply Chain. Proc 2016 ASME IDETC-CIE 21st Des Manuf Life Cycle Conf Pap DETC2015-59262 August 21-25Charlotte N C USA 2016. [13]
Akhtari S, Sowlati T, Day K. Economic feasibility of utilizing forest biomass
in district energy systems – A review. Renew Sustain Energy Rev 2014;33:117–27. doi:10.1016/j.rser.2014.01.058. [14]
McKendry P. Energy production from biomass (part 2): conversion
technologies. Bioresour Technol 2002;83:47–54. [15]
Elia JA, Floudas CA. Energy Supply Chain Optimization of Hybrid Feedstock
Processes:
A
Review.
Annu
Rev
Chem
Biomol
Eng
2014;5:147–79.
doi:10.1146/annurev-chembioeng-060713-040425. [16]
Sharma B, Ingalls RG, Jones CL, Khanchi A. Biomass supply chain design and
analysis: Basis, overview, modeling, challenges, and future. Renew Sustain Energy Rev 2013;24:608–27. doi:10.1016/j.rser.2013.03.049. [17]
Yue D, You F, Snyder SW. Biomass-to-bioenergy and biofuel supply chain
optimization: Overview, key issues and challenges. Comput Chem Eng 2014;66:36– 56. doi:10.1016/j.compchemeng.2013.11.016. [18]
Castillo-Villar KK. Metaheuristic Algorithms Applied to Bioenergy Supply
Chain Problems: Theory, Review, Challenges, and Future. Energies 2014;7:7640–72. doi:10.3390/en7117640. [19]
Yemshanov D, McKenney DW, Fraleigh S, McConkey B, Huffman T, Smith
S. Cost estimates of post harvest forest biomass supply for Canada. Biomass Bioenergy 2014;69:80–94. [20]
Akgul O, Zamboni A, Bezzo F, Shah N, Papageorgiou LG. Optimization-based
approaches for bioethanol supply chains. Ind Eng Chem Res 2010;50:4927–4938.
75 [21]
Freppaz D, Minciardi R, Robba M, Rovatti M, Sacile R, Taramasso A.
Optimizing forest biomass exploitation for energy supply at a regional level. Biomass Bioenergy 2004;26:15–25. doi:10.1016/S0961-9534(03)00079-5. [22]
USDOE. U.S. billion-ton update: biomass supply for a bioenergy and
bioproducts industry. Oak Ridge National Laboratory; 2011. [23]
Beier C, Caputo J, Groffman PM. Measuring ecosystem capacity to provide
regulating services: forest removal and recovery at Hubbard Brook (USA). Ecol Appl 2015. doi:10.1890/14-1376.1. [24]
You F, Tao L, Graziano DJ, Snyder SW. Optimal design of sustainable
cellulosic biofuel supply chains: Multiobjective optimization coupled with life cycle assessment and input–output analysis. AIChE J 2012;58:1157–1180. [25]
Sokhansanj S, Mani S, Turhollow A, Kumar A, Bransby D, Lynd L, et al.
Large-scale production, harvest and logistics of switchgrass (Panicum virgatum L.) – current technology and envisioning a mature technology. Biofuels Bioprod Biorefining 2009;3:124–41. doi:10.1002/bbb.129. [26]
Pan F, Han H-S, Johnson LR, Elliot WJ, others. Production and cost of
harvesting, processing, and transporting small-diameter (≤ 5 inches) trees for energy. Forest 2007. [27]
Mirkouei A, Haapala KR, Sessions J, Murthy GS. Multi-criteria Decision
Making for Sustainable Bio-Oil Production using a Mixed Supply Chain. ASME J Manuf Sci Eng Rev 2015. [28]
Shabani N, Akhtari S, Sowlati T. Value Chain Optimization of Forest Biomass
for Bioenergy Production: A Review. Renew Sustain Energy Rev 2013;23:299–311. doi:10.1016/j.rser.2013.03.005. [29]
Mirkouei A, Mirzaie P, Haapala KR, Sessions J, Murthy GS. Reducing the Cost
and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains. J Clean Prod 2015. doi:doi:10.1016/j.jclepro.2015.11.023. [30]
Ćosić B, Stanić Z, Duić N. Geographic distribution of economic potential of
agricultural and forest biomass residual for energy use: Case study Croatia. Energy 2011;36:2017–28. doi:10.1016/j.energy.2010.10.009.
76 [31]
Mirkouei A, Haapala KR. Integration of Machine Learning and Mathematical
Programming Methods into the Biomass Feedstock Supplier Selection Process. 24th Int. Conf. Flex. Autom. Intell. Manuf. FAIM May 20-23 2014 San Antonio Tex., Flexible Automation and Intelligent Manufacturing; 2014. [32]
Mirkouei A, Haapala KR. A Network Model to Optimize Upstream and
Midstream Biomass-to-Bioenergy Supply Chain Costs. ASME 2015 Int. Manuf. Sci. Eng. Conf. MSEC MSEC2015-9355 June 8-12 2015, Charlotte, NC: 2015. [33]
Parker N, Tittmann P, Hart Q, Nelson R, Skog K, Schmidt A, et al.
Development of a biorefinery optimized biofuel supply curve for the Western United States. Biomass Bioenergy 2010;34:1597–607. doi:10.1016/j.biombioe.2010.06.007. [34]
Gold S, Seuring S. Supply Chain and Logistics Issues of Bio-Energy
Production. J Clean Prod 2011;19:32–42. doi:10.1016/j.jclepro.2010.08.009. [35]
Cambero C, Sowlati T. Assessment and optimization of forest biomass supply
chains from economic, social and environmental perspectives – A review of literature. Renew Sustain Energy Rev 2014;36:62–73. doi:10.1016/j.rser.2014.04.041. [36]
Awudu I, Zhang J. Uncertainties and sustainability concepts in biofuel supply
chain management: A review. Renew Sustain Energy Rev 2012;16:1359–68. doi:10.1016/j.rser.2011.10.016. [37]
An H, Wilhelm WE, Searcy SW. Biofuel and petroleum-based fuel supply
chain research: a literature review. Biomass Bioenergy 2011;35:3763–3774. [38]
Alonso DM, Bond JQ, Dumesic JA. Catalytic conversion of biomass to
biofuels. Green Chem 2010;12:1493. doi:10.1039/c004654j. [39]
Bartle JR, Abadi A. Toward Sustainable Production of Second Generation
Bioenergy Feedstocks†. Energy Fuels 2009;24:2–9. [40]
Lauer C, McCaulou JC, Sessions J, Capalbo SM. Biomass supply curves for
western juniper in Central Oregon, USA, under alternative business models and policy assumptions. For Policy Econ 2015;59:75–82. [41]
Radlein D. The production of chemicals from fast pyrolysis bio-oils. vol. 1.
CPL Press: Newbury, UK; 1999.
77 [42]
Bridgwater AV, Toft AJ, Brammer JG. A techno-economic comparison of
power production by biomass fast pyrolysis with gasification and combustion. Renew Sustain Energy Rev 2002;6:181–246. [43]
Brammer JG, Lauer M, Bridgwater AV. Opportunities for biomass-derived
“bio-oil” in European heat and power markets. Energy Policy 2006;34:2871–2880. [44]
Giarola S, Shah N, Bezzo F. A comprehensive approach to the design of ethanol
supply chains including carbon trading effects. Bioresour Technol 2012;107:175–185. [45]
Uslu A, Faaij APC, Bergman PCA. Pre-treatment technologies, and their effect
on international bioenergy supply chain logistics. Techno-economic evaluation of torrefaction,
fast
pyrolysis
and
pelletisation.
Energy
2008;33:1206–23.
doi:10.1016/j.energy.2008.03.007. [46]
Kumar D, Murthy GS. Impact of pretreatment and downstream processing
technologies on economics and energy in cellulosic ethanol production. Biotechnol Biofuels 2011;4:27. [47]
You F, Wang B. Life Cycle Optimization of Biomass-to-Liquid Supply Chains
with
Distributed–Centralized
Processing
Networks.
Ind
Eng
Chem
Res
2011;50:10102–10127. doi:10.1021/ie200850t. [48] and
Hamelinck CN, Suurs RAA, Faaij APC. International bioenergy transport costs energy
balance.
Biomass
Bioenergy
2005;29:114–34.
doi:10.1016/j.biombioe.2005.04.002. [49]
Boateng AA, Daugaard DE, Goldberg NM, Hicks KB. Bench-scale fluidized-
bed pyrolysis of switchgrass for bio-oil production. Ind Eng Chem Res 2007;46:1891– 1897. [50]
Bals BD, Dale BE. Developing a model for assessing biomass processing
technologies within a local biomass processing depot. Bioresour Technol 2012;106:161–169. [51]
Ayalur Chattanathan S, Adhikari S, Abdoulmoumine N. A review on current
status of hydrogen production from bio-oil. Renew Sustain Energy Rev 2012;16:2366– 72. doi:10.1016/j.rser.2012.01.051.
78 [52]
Wang D, Czernik S, Montane D, Mann M, Chornet E. Biomass to hydrogen via
fast pyrolysis and catalytic steam reforming of the pyrolysis oil or its fractions. Ind Eng Chem Res 1997;36:1507–1518. [53]
Czernik S, French R, Penev MM. Distributed Bio-Oil Reforming 2010.
[54]
Yang Z, Kumar A, Huhnke RL. Review of recent developments to improve
storage and transportation stability of bio-oil. Renew Sustain Energy Rev 2015;50:859–70. doi:10.1016/j.rser.2015.05.025. [55]
Patel M, Zhang X, Kumar A. Techno-economic and life cycle assessment on
lignocellulosic biomass thermochemical conversion technologies: A review. Renew Sustain Energy Rev 2016;53:1486–1499. [56]
Badger PC, Fransham P. Use of mobile fast pyrolysis plants to densify biomass
and reduce biomass handling costs—A preliminary assessment. Biomass Bioenergy 2006;30:321–5. doi:10.1016/j.biombioe.2005.07.011. [57]
Polagye BL, Hodgson KT, Malte PC. An economic analysis of bio-energy
options using thinnings from overstocked forests. Biomass Bioenergy 2007;31:105– 25. doi:10.1016/j.biombioe.2006.02.005. [58]
Vispute TP, Huber GW, others. Breaking the chemical and engineering barriers
to lignocellulosic biofuels. Int Sugar J 2008;110. [59]
Van de Velden M, Baeyens J, Boukis I. Modeling CFB biomass pyrolysis
reactors. Biomass Bioenergy 2008;32:128–139. [60]
Granatstein D, Kruger CE, Collins H, Galinato S, Garcia-Perez M, Yoder J. Use
of biochar from the pyrolysis of waste organic material as a soil amendment. Final project report. Cent Sustain Agric Nat Resour Wash State Univ Wenatchee WA 2009. [61]
Qian K, Kumar A, Zhang H, Bellmer D, Huhnke R. Recent advances in
utilization
of
biochar.
Renew
Sustain
Energy
Rev
2015;42:1055–64.
doi:10.1016/j.rser.2014.10.074. [62]
Zhang S, Asadullah M, Dong L, Tay H-L, Li C-Z. An advanced biomass
gasification technology with integrated catalytic hot gas cleaning. Part II: Tar reforming using char as a catalyst or as a catalyst support. Fuel 2013;112:646–53. doi:10.1016/j.fuel.2013.03.015.
79 [63]
Ahn SY, Eom SY, Rhie YH, Sung YM, Moon CE, Choi GM, et al. Utilization
of wood biomass char in a direct carbon fuel cell (DCFC) system. Appl Energy 2013;105:207–16. doi:10.1016/j.apenergy.2013.01.023. [64]
Ahmad M, Rajapaksha AU, Lim JE, Zhang M, Bolan N, Mohan D, et al.
Biochar as a sorbent for contaminant management in soil and water: A review. Chemosphere 2014;99:19–33. doi:10.1016/j.chemosphere.2013.10.071. [65]
Gil MV, Martínez M, García S, Rubiera F, Pis JJ, Pevida C. Response surface
methodology as an efficient tool for optimizing carbon adsorbents for CO2 capture. Fuel Process Technol 2013;106:55–61. doi:10.1016/j.fuproc.2012.06.018. [66]
Oleszczuk P, Hale SE, Lehmann J, Cornelissen G. Activated carbon and biochar
amendments decrease pore-water concentrations of polycyclic aromatic hydrocarbons (PAHs)
in
sewage
sludge.
Bioresour
Technol
2012;111:84–91.
doi:10.1016/j.biortech.2012.02.030. [67]
de Lourdes Bravo M, Naim MM, Potter A. Key issues of the upstream segment
of biofuels supply chain: a qualitative analysis. Logist Res 2012;5:21–31. [68]
Ba BH, Prins C, Prodhon C. Models for optimization and performance
evaluation of biomass supply chains: An Operations Research perspective. Renew Energy 2016;87:977–989. [69]
Zhang F, Johnson D, Johnson M, Watkins D, Froese R, Wang J. Decision
support system integrating GIS with simulation and optimisation for a biofuel supply chain. Renew Energy 2016;85:740–8. doi:10.1016/j.renene.2015.07.041. [70]
Yue D, Pandya S, You F. Integrating Hybrid Life Cycle Assessment with Multi-
objective Optimization: A Modeling Framework. Environ Sci Technol 2016. [71]
Allen J, Browne M, Hunter A, Boyd J, Palmer H. Logistics management and
costs of biomass fuel supply. Int J Phys Distrib Logist Manag 1998;28:463–477. [72]
Sedjo RA. The economics of forest-based biomass supply. Energy Policy
1997;25:559–66. doi:10.1016/S0301-4215(97)00045-1. [73]
Wu J, Wang J, McNeel J. Economic modeling of woody biomass utilization for
bioenergy and its application in central Appalachia, USA. Can J For Res 2010;41:165– 179.
80 [74]
Idaho National Laboratory. Feedstock Supply System Design and Economics
for Conversion of Lignocellulosic Biomass to Hydrocarbon Fuels 2014. [75]
DOE. Feedstock Supply System Design and Analysis 2014.
[76]
Dunn JB, Wang M, Wang Z, Cafferty K, Jacobson J, Tan E, et al. Supply Chain
Sustainability Analysis of Fast Pyrolysis and Hydrotreating Bio-Oil to Produce Hydrocarbon Fuels. Argonne National Laboratory (ANL); 2015. [77]
Hassini E, Surti C, Searcy C. A literature review and a case study of sustainable
supply chains with a focus on metrics. Int J Prod Econ 2012;140:69–82. doi:10.1016/j.ijpe.2012.01.042. [78]
Taticchi P, Garengo P, Nudurupati SS, Tonelli F, Pasqualino R. A review of
decision-support tools and performance measurement and sustainable supply chain management. Int J Prod Res 2014:1–22. [79]
Wacker JG. A definition of theory: research guidelines for different theory-
building research methods in operations management. J Oper Manag 1998;16:361–85. doi:10.1016/S0272-6963(98)00019-9. [80]
Tranfield D, Denyer D, Smart P. Towards a Methodology for Developing
Evidence-Informed Management Knowledge by Means of Systematic Review. Br J Manag 2003;14:207–22. doi:10.1111/1467-8551.00375. [81]
Kevin Burgess, Prakash J. Singh, Rana Koroglu. Supply chain management: a
structured literature review and implications for future research. Int J Oper Prod Manag 2006;26:703–29. doi:10.1108/01443570610672202. [82]
Seuring S. A review of modeling approaches for sustainable supply chain
management. Decis Support Syst 2013;54:1513–20. doi:10.1016/j.dss.2012.05.053. [83]
Lin Y-C, Huber GW. The critical role of heterogeneous catalysis in
lignocellulosic
biomass
conversion.
Energy
Env
Sci
2009;2:68–80.
doi:10.1039/B814955K. [84]
Wang H, Male J, Wang Y. Recent advances in hydrotreating of pyrolysis bio-
oil and its oxygen-containing model compounds. Acs Catal 2013;3:1047–1070. [85]
Rafael S, Tarelho L, Monteiro A, Sá E, Miranda AI, Borrego C, et al. Impact
of forest biomass residues to the energy supply chain on regional air quality. Sci Total Environ 2015;505:640–8. doi:10.1016/j.scitotenv.2014.10.049.
81 [86]
Vanhala P, Repo A, Liski J. Forest bioenergy at the cost of carbon
sequestration? Curr Opin Environ Sustain 2013;5:41–46. [87]
McKendry P. Energy production from biomass (part 1): overview of biomass.
Bioresour Technol 2002;83:37–46. doi:10.1016/S0960-8524(01)00118-3. [88]
Zhu X, Li X, Yao Q, Chen Y. Challenges and models in supporting logistics
system design for dedicated-biomass-based bioenergy industry. Bioresour Technol 2011;102:1344–51. doi:10.1016/j.biortech.2010.08.122. [89]
Gingras J-F. Harvesting small trees and forest residues. Biomass Bioenergy
1995;9:153–60. doi:10.1016/0961-9534(95)00087-9. [90]
Mattsson J-E, Mitchell C. IEA bioenergy agreement task IX harvesting and
supply of woody biomass for energy 1992 – 1994. Biomass Bioenergy 1995;9:117–25. doi:10.1016/0961-9534(95)00084-4. [91]
Caputo J. Sustainable forest biomass: promoting renewable energy and forest
stewardship. Environmental and Energy Study Institute; 2009. [92]
Spinelli R, Nati C, Magagnotti N. Recovering logging residue: experiences
from the Italian Eastern Alps. Croat J For Eng 2007;28:1–9. [93]
Hakkila DP. Utilization of Residual Forest Biomass. Util. Residual For.
Biomass, Springer Berlin Heidelberg; 1989, p. 352–477. [94]
Johansson J, Liss J-E, Gullberg T, Björheden R. Transport and handling of
forest energy bundles—advantages and problems. Biomass Bioenergy 2006;30:334– 341. [95]
Wright MM, Daugaard DE, Satrio JA, Brown RC. Techno-economic analysis
of biomass fast pyrolysis to transportation fuels. Fuel 2010;89, Supplement 1:S2–10. doi:10.1016/j.fuel.2010.07.029. [96]
Zamora-Cristales R, Sessions J, Marrs G. Forest Biomass Feedstock Cost
Sensitivity to Grinding Parameters forBio-jet Fuel Production. Renew Energy 2016. [97]
Zamora-Cristales R, Sessions J, Smith D, Marrs G. Effect of grinder
configuration on forest biomass bulk density, particle size distribution and fuel consumption. Biomass Bioenergy 2015;81:44–54.
82 [98]
Sokhansanj S, Kumar A, Turhollow AF. Development and implementation of
integrated biomass supply analysis and logistics model (IBSAL). Biomass Bioenergy 2006;30:838–847. [99]
Beardsell MG, Stuart WB, Mitchell CP. Integrated Harvesting Systems to
Incorporate the Recovery of Logging Residues with the Harvesting of Conventional Forest Products: Existing Equipment for Residue Recovery. School of Forestry and Wildlife Resources, VPI & SU; 1983. [100] Goulding CJ, Twaddle AA. Harvesting whole trees with processing and log allocation (in the forest) to conventional and energy products. Biomass 1990;22:145– 58. doi:10.1016/0144-4565(90)90013-A. [101] Hudson J. Integrated harvesting systems. Biomass Bioenergy 1995;9:141–51. doi:10.1016/0961-9534(95)00086-0. [102] Yue D, You F. Sustainable scheduling of batch processes under economic and environmental criteria with MINLP models and algorithms. Comput Chem Eng 2013;54:44–59. doi:10.1016/j.compchemeng.2013.03.013. [103] Muth D, Jacobson J, Cafferty K, Jeffers R. Feedstock Pathways for Bio-oil and Syngas Conversion Pathways. INL Technical Memorandum, INL/EXT-13-29487; 2013. [104] Puttock GD. Estimating cost for integrated harvesting and related forest management
activities.
Biomass
Bioenergy
1995;8:73–9.
doi:10.1016/0961-
9534(95)00001-N. [105] Grisso RD, McCullough D, Cundiff JS, Judd JD. Harvest schedule to fill storage for year-round delivery of grasses to biorefinery. Biomass Bioenergy 2013;55:331–8. doi:10.1016/j.biombioe.2013.02.027. [106] Laitila J, Väätäinen K. Truck transportation and chipping productivity of whole trees and delimbed energy wood in Finland. Croat J For Eng 2012;33:199–210. [107] Röser D, Sikanen L, Asikainen A, Parikka H, Väätäinen K. Productivity and cost of mechanized energy wood harvesting in Northern Scotland. Biomass Bioenergy 2011;35:4570–4580.
83 [108] Zamora-Cristales R, Boston K, Sessions J, Murphy G. Stochastic simulation and optimization of mobile chipping economics in processing and transport of forest biomass from residues. Silva Fenn 2014;47. doi:10.14214/sf.937. [109] Abbas D, Current D, Phillips M, Rossman R, Hoganson H, Brooks KN. Guidelines for harvesting forest biomass for energy: A synthesis of environmental considerations.
Biomass
Bioenergy
2011;35:4538–46.
doi:10.1016/j.biombioe.2011.06.029. [110] Cook J, Beyea J. Bioenergy in the United States: progress and possibilities1. Biomass Bioenergy 2000;18:441–55. doi:10.1016/S0961-9534(00)00011-8. [111] Grigal DF. An update of forest soils: A technical paper for a generic environmental impact statement on timber harvesting and forest management in Minnesota. Laurent Energy Auth Va MN 2004. [112] Shepard JP. Water quality protection in bioenergy production: the US system of forestry Best Management Practices. Biomass Bioenergy 2006;30:378–84. doi:10.1016/j.biombioe.2005.07.018. [113] Achat DL, Deleuze C, Landmann G, Pousse N, Ranger J, Augusto L. Quantifying consequences of removing harvesting residues on forest soils and tree growth–A meta-analysis. For Ecol Manag 2015;348:124–141. [114] Andersson G, Asikainen A, Björheden R, Hall PW, Hudson JB, Jirjis R, et al. Production of Forest Energy. In: Richardson J, Björheden R, Hakkila P, Lowe AT, Smith CT, editors. Bioenergy Sustain. For., Springer Netherlands; 2002, p. 49–123. [115] Richardson JJ, Spies KA, Rigdon S, York S, Lieu V, Nackley L, et al. Uncertainty in biomass supply estimates: Lessons from a Yakama Nation case study. Biomass Bioenergy 2011;35:3698–707. doi:10.1016/j.biombioe.2011.05.030. [116] Alam MB, Pulkki R, Shahi C. Road network optimization model for supplying woody biomass feedstock for energy production in northwestern Ontario. Open For Sci J 2012;5:1–14. [117] Long JJ, Boston K. An Evaluation of Alternative Measurement Techniques for Estimating the Volume of Logging Residues. For Sci 2014;Volume 60, Number 1, 1 p. 200–204(5). doi:10.5849/forsci.13-501.
84 [118] Eriksson LO, Björheden R. Optimal storing, transport and processing for a forest-fuel supplier. Eur J Oper Res 1989;43:26–33. [119] Caputo AC, Palumbo M, Pelagagge PM, Scacchia F. Economics of biomass energy utilization in combustion and gasification plants: effects of logistic variables. Biomass Bioenergy 2005;28:35–51. doi:10.1016/j.biombioe.2004.04.009. [120] Schroeder R, Jackson B, Ashton S, Hubbard W, Biles L, Mayfield C, et al. Biomass Transportation and Delivery 2007:145–148. [121] Searcy E, Flynn P, Ghafoori E, Kumar A. The Relative Cost of Biomass Energy Transport. Appl. Biochem. Biotechnol., vol. 137, 2007, p. 639–652. [122] Forest Resources Association. Annual pulpwood statistics summary report, 2001-2005. Forest Resources Association; 2006. [123] Sessions J, Tuers K, Boston K, Zamora R, Anderson R. Pricing Forest Biomass for Power Generation. West J Appl For 2013;28:51–56. [124] U.S.
Department
of
Transportation.
Federal
Motor
Carrier
Safety
Administration Rules and regulations 2014. http://www.fmcsa.dot.gov/regulations (accessed May 9, 2015). [125] Sessions J, Balcom J. Determining maximum allowable weights for highway vehicles. For Prod J USA 1989. [126] Zamora-Cristales R, Sessions J. Are double trailers cost effective for transporting forest biomass on steep terrain? Calif Agric 2015;69:177–183. [127] Acuna M, Anttila P, Sikanen L, Prinz R, Asikainen A. Predicting and controlling moisture content to optimise forest biomass logistics. Croat J For Eng 2012;33:225–238. [128] Routa J, Asikainen A, Björheden R, Laitila J, Röser D. Forest energy procurement: state of the art in Finland and Sweden. Wiley Interdiscip Rev Energy Environ 2013;2:602–613. [129] Berwick MD, Farooq M, others. Truck costing model for transportation managers. Mountain-Plains Consortium; 2003. [130] Zamora-Cristales R, Sessions J, Murphy G, Boston K. Economic Impact of Truck–Machine Interference in Forest Biomass Recovery Operations on Steep Terrain. For Prod J 2013;63:162–73. doi:10.13073/FPJ-D-13-00031.
85 [131] Zamora-Cristales R, Sessions J, Boston K, Murphy G. Economic Optimization of Forest Biomass Processing and Transport in the Pacific Northwest USA. For Sci 2015;61:220–234. [132] Antonio Guzmán J. Study of wood chip production from forest residues in Chile. Biomass 1984;5:167–79. doi:10.1016/0144-4565(84)90021-0. [133] Zamora-Cristales R, Sessions J, Smith D, Marrs G. Effect of High Speed Blowing on the Bulk Density of Ground Residues. For Prod J 2014;64:290–299. [134] Kanzian C, Holzleitner F, Stampfer K, Ashton S, others. Regional energy wood logistics–optimizing local fuel supply. Silva Fenn 2009;43:113–128. [135] Rawlings C. A study of how to decrease the costs of collecting, processing and transporting slash. Montana Community Development Corporation; 2004. [136] Han H-S, Halbrook J, Pan F, Salazar L. Economic evaluation of a roll-off trucking system removing forest biomass resulting from shaded fuelbreak treatments. Biomass Bioenergy 2010;34:1006–1016. [137] Harrill H, Han H-S, Pan F. Application of hook-lift trucks in centralized slash grinding
operations.
2009
Counc.
For.
Eng.
COFE
Conf.
Proceedings“Environmentally Sound For. Oper. Lake Tahoe, 2009. [138] Acuna M, Sessions J. A simulated annealing algorithm to solve the log-truck scheduling problem. In: Simulated Annealing: Strategies. Potential Uses and Advantages. NOVA Science; 2014. [139] Bisson J, Han S-K, Han H-S. Evaluating the system logistics of a biomass recovery operation in northern California. Proc Counc. For. Eng. COFE Meet. Missoula MT 11p, 2013. [140] Johnson L, Lippke B, Oneil E. Modeling Biomass Collection and Woods Processing Life-Cycle Analysis*. For Prod J 2012;62:258–272. [141] Gregg DJ, Saddler JN. A review of techno-economic modeling methodology for a wood-to-ethanol process. Biotechnol. Fuels Chem., Springer; 1997, p. 609–623. [142] Harmsen P, Huijgen W, Bermudez L, Bakker R. Literature review of physical and chemical pretreatment processes for lignocellulosic. Biomass 2010:1–49. [143] Wright M, Brown RC. Establishing the optimal sizes of different kinds of biorefineries. Biofuels Bioprod Biorefining 2007;1:191–200.
86 [144] Steele P, Yu F, Gajjela S. Past, Present, and Future Production of Bio-oil. Mississippi State University; 2009. [145] Ringer M, Putsche V, Scahil J. Large-Scale Pyrolysis Oil Production: A Technology Assessment and Economic Analysis. Golden (CO): National Renewable Energy Laboratory; 2006 Nov. Report No. NREL/TP-510-37779. Contract No.: DEAC36-99-GO10337; 2006. [146] Luo Z, Wang S, Liao Y, Zhou J, Gu Y, Cen K. Research on biomass fast pyrolysis
for
liquid
fuel.
Biomass
Bioenergy
2004;26:455–62.
doi:10.1016/j.biombioe.2003.04.001. [147] Mullaney H, Farag I, LaClaire C, Barrett C. Technical, Environmental and Economic Feasibility of Bio-Oil in New Hampshires North Country. UNH Proj 2002. [148] Badger P, Badger S, Puettmann M, Steele P, Cooper J. Techno-economic analysis: Preliminary assessment of pyrolysis oil production costs and material energy balance associated with a transportable fast pyrolysis system. BioResources 2010;6:34–47. [149] Demirbas MF, Balat M. Recent advances on the production and utilization trends of bio-fuels: A global perspective. Energy Convers Manag 2006;47:2371–81. doi:10.1016/j.enconman.2005.11.014. [150] Mohan D, Pittman Charles U, Steele PH. Pyrolysis of Wood/Biomass for Biooil: A Critical Review. Energy Fuels 2006;20:848–89. doi:10.1021/ef0502397. [151] Isahak WNRW, Hisham MW, Yarmo MA, Hin TY. A review on bio-oil production from biomass by using pyrolysis method. Renew Sustain Energy Rev 2012;16:5910–5923. [152] Bridgwater AV, Bridge SA. A review of biomass pyrolysis and pyrolysis technologies. Biomass Pyrolysis Liq. Upgrad. Util., Springer; 1991, p. 11–92. [153] Brown D, Rowe A, Wild P. Techno-economic comparisons of hydrogen and synthetic fuel production using forest residue feedstock. Int J Hydrog Energy 2014;39:12551–12562. [154] Czernik S, French R. Distributed production of hydrogen by auto-thermal reforming of fast pyrolysis bio-oil. Int J Hydrog Energy 2014;39:744–750.
87 [155] Bridgwater AV, Peacocke GVC. Fast pyrolysis processes for biomass. Renew Sustain Energy Rev 2000;4:1–73. [156] Czernik S, Bridgwater AV. Overview of Applications of Biomass Fast Pyrolysis Oil. Energy Fuels 2004;18:590–8. doi:10.1021/ef034067u. [157] Venderbosch RH, Prins W, others. Fast pyrolysis technology development. Biofuels Bioprod Biorefining 2010;4:178. [158] Pu Y, Zhang D, Singh PM, Ragauskas AJ. The new forestry biofuels sector. Biofuels Bioprod Biorefining 2008;2:58–73. [159] Rogers JG, Brammer JG. Estimation of the production cost of fast pyrolysis bio-oil. Biomass Bioenergy 2012;36:208–17. doi:10.1016/j.biombioe.2011.10.028. [160] Solantausta Y, Beckman D, Bridgwater AV, Diebold JP, Elliott DC. Assessment of liquefaction and pyrolysis systems. Biomass Bioenergy 1992;2:279– 297. [161] Cottam M-L, Bridgwater AV. Techno-economic modelling of biomass flash pyrolysis and upgrading systems. Biomass Bioenergy 1994;7:267–273. [162] Gregoire CE, Bain RL. Technoeconomic analysis of the production of biocrude from wood. Biomass Bioenergy 1994;7:275–283. [163] Islam MN, Ani FN. Techno-economics of rice husk pyrolysis, conversion with catalytic treatment to produce liquid fuel. Bioresour Technol 2000;73:67–75. [164] Peacocke GVC, Bridgwater AV, Brammer JG. Techno-economic assessment of power production from the Wellman and BTG fast Pyrolysis processes. Sci Therm Chem Biomass Convers Vic Can Wiss Ber FZKA 2004;7170. [165] Marker TL. Opportunities for biorenewables in oil refineries. UOP LLC; 2005. [166] Dynamotive. Dynamotive Energy Systems Corporation, Canadian BioOil Plant: Summary (USD) 2009. [167] Jones SB, Male JL. Production of Gasoline and Diesel from Biomass via Fast Pyrolysis, Hydrotreating and Hydrocracking: 2011 State of Technology and Projections to 2017. Pacific Northwest National Laboratory; 2012. [168] Brown D, Rowe A, Wild P. A techno-economic analysis of using mobile distributed pyrolysis facilities to deliver a forest residue resource. Bioresour Technol 2013;150:367–376.
88 [169] U.S. EIA. U.S. Residual Fuel Oil Wholesale/Resale Price by Refiners 2016. https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=EMA_EPPR_PWG_ NUS_DPG&f=M (accessed March 6, 2016). [170] Kim J, Realff MJ, Lee JH, Whittaker C, Furtner L. Design of Biomass Processing Network for Biofuel Production Using an Milp Model. Biomass Bioenergy 2011;35:853–871. doi:10.1016/j.biombioe.2010.11.008. [171] Lange J-P. Lignocellulose conversion: an introduction to chemistry, process and economics. Biofuels Bioprod Biorefining 2007;1:39–48. doi:10.1002/bbb.7. [172] Bridgwater AV. Technical and economic assessment of thermal processes for biofuels. Life Cycle Techno-Econ Assess Northeast Biomass Liq Proj NNFCC Proj 2009;8:18. [173] Ng KS, Sadhukhan J. Process integration and economic analysis of bio-oil platform for the production of methanol and combined heat and power. Biomass Bioenergy 2011;35:1153–1169. [174] Bradley D. European Market Study for BioOil (Pyrolysis Oil). Clim Change Solut 2006. [175] Elliott DC. Historical developments in hydroprocessing bio-oils. Energy Fuels 2007;21:1792–1815. [176] Dayton DC. Conversion Technologies for Advanced Biofuels – Bio-Oil Production
2012.
https://www.google.com/search?q=Conversion+Technologies+for+Advanced+Biofue ls+%E2%80%93+Bio-Oil+Production&ie=utf-8&oe=utf-8 (accessed February 19, 2016). [177] ROI.
Renewable
Oil
International
LLC.
Company
2003.
http://www.renewableoil.com/Pages/default.aspx (accessed May 5, 2015). [178] Rentizelas AA, Tolis AJ, Tatsiopoulos IP. Logistics issues of biomass: The storage problem and the multi-biomass supply chain. Renew Sustain Energy Rev 2009;13:887–94. doi:10.1016/j.rser.2008.01.003. [179] Cundiff JS, Dias N, Sherali HD. A linear programming approach for designing a herbaceous biomass delivery system. Bioresour Technol 1997;59:47–55. doi:10.1016/S0960-8524(96)00129-0.
89 [180] Rigdon AR, Jumpponen A, Vadlani PV, Maier DE. Impact of various storage conditions on enzymatic activity, biomass components and conversion to ethanol yields from sorghum biomass used as a bioenergy crop. Bioresour Technol 2013;132:269– 275. [181] Nilsson D, Hansson P-A. Influence of various machinery combinations, fuel proportions and storage capacities on costs for co-handling of straw and reed canary grass to district heating plants. Biomass Bioenergy 2001;20:247–260. [182] Tatsiopoulos IP, Tolis AJ. Economic aspects of the cotton-stalk biomass logistics and comparison of supply chain methods. Biomass Bioenergy 2003;24:199– 214. doi:10.1016/S0961-9534(02)00115-0. [183] Papadopoulos DP, Katsigiannis PA. Biomass energy surveying and technoeconomic assessment of suitable CHP system installations. Biomass Bioenergy 2002;22:105–24. doi:10.1016/S0961-9534(01)00064-2. [184] Pettersson M, Nordfjell T. Fuel quality changes during seasonal storage of compacted logging residues and young trees. Biomass Bioenergy 2007;31:782–92. doi:10.1016/j.biombioe.2007.01.009. [185] Murphy G, Kent T, Kofman PD. Modeling air drying of Sitka spruce (Picea sitchensis) biomass in off-forest storage yards in Ireland. For Prod J 2012;62:443–449. [186] Larson JA, Yu TE, English BC, Jensen KL, Gao Y, Wang C. Effect of outdoor storage losses on feedstock inventory management and plant-gate cost for a switchgrass conversion facility in East Tennessee. Renew Energy 2015;74:803–814. [187] Diebold JP, Czernik S. Additives To Lower and Stabilize the Viscosity of Pyrolysis
Oils
during
Storage.
Energy
Fuels
1997;11:1081–91.
doi:10.1021/ef9700339. [188] Czernik S. Storage of biomass pyrolysis oils. Proc. Spec. Workshop Biomass Pyrolysis Oil Prop. Combust., 1994, p. 26–28. [189] Aubin H, Roy C. STUDY ON THE CORROSIVENESS OP WOOD PYROLYSIS OILS. Pet Sci Technol 1990;8:77–86. [190] Soltes EJ, Lin S-C. Hydroprocessing of biomass tars for liquid engine fuels. Prog Biomass Convers 1984;5:1–68.
90 [191] Swanson RM, Platon A, Satrio JA, Brown RC. Techno-economic analysis of biomass-to-liquids production based on gasification. Fuel 2010;89, Supplement 1:S11– 9. doi:10.1016/j.fuel.2010.07.027. [192] de Jong S, Hoefnagels R, Faaij A, Slade R, Mawhood R, Junginger M. The feasibility of short-term production strategies for renewable jet fuels–a comprehensive techno-economic comparison. Biofuels Bioprod Biorefining 2015;9:778–800. [193] Sorenson CB. A Comparative Financial Analysis of Fast Pyrolysis Plants in Southwest Oregon. The University of Montana Missoula, MT, 2010. [194] Bridgewater AV. Biomass fast pyrolysis. Therm Sci 2004;8:21–50. [195] Rogers JG. A techno-economic assessment of the use of fast pyrolysis bio-oil from UK energy crops in the production of electricity and combined heat and power 2009. https://research.aston.ac.uk/portal/en/theses/a-technoeconomic-assessment-ofthe-use-of-fast-pyrolysis-biooil-from-uk-energy-crops-in-the-production-ofelectricity-and-combined-heat-and-power(5e372078-73fe-4f6d-947d6035410771f8).html (accessed August 23, 2015). [196] Elliott DC, Beckman D, Bridgwater AV, Diebold JP, Gevert SB, Solantausta Y. Developments in direct thermochemical liquefaction of biomass: 1983-1990. Energy Fuels 1991;5:399–410. [197] Mitchell CP, Bridgwater AV, Stevens DJ, Toft AJ, Watters MP. Technoeconomic assessment of biomass to energy. Biomass Bioenergy 1995;9:205– 226. [198] Östman A, Solantausta Y, Beckman D, others. Comparisons of alternative routes for biomass fuel. Biomass (pellets), pyrolysis oil and tall oil pitch. VTT Tied 2000. [199] Magalhães AI, Rodriguez AL, Putra ZA, Thielemans G, others. Technoeconomic assessment of biomass pre-conversion processes as a part of biomass-toliquids line-up. Biofuels Bioprod Biorefining 2009;3:584–600. [200] Oasmaa A, Solantausta Y, Arpiainen V, Kuoppala E, Sipilä K. Fast pyrolysis bio-oils from wood and agricultural residues. Energy Fuels 2009;24:1380–1388.
91 [201] Anex RP, Aden A, Kazi FK, Fortman J, Swanson RM, Wright MM, et al. Techno-economic comparison of biomass-to-transportation fuels via pyrolysis, gasification, and biochemical pathways. Fuel 2010;89:S29–S35. [202] Trippe F, Fröhling M, Schultmann F, Stahl R, Henrich E. Techno-economic analysis of fast pyrolysis as a process step within biomass-to-liquid fuel production. Waste Biomass Valorization 2010;1:415–430. [203] Ghezzaz H, Stuart P, others. Biomass availability and process selection for an integrated forest biorefinery. Pulp Pap Can 2011;112:19–26. [204] Brown TR, Thilakaratne R, Brown RC, Hu G. Techno-economic analysis of biomass to transportation fuels and electricity via fast pyrolysis and hydroprocessing. Fuel 2013;106:463–469. [205] Do TX, Lim Y, Yeo H. Techno-economic analysis of biooil production process from palm empty fruit bunches. Energy Convers Manag 2014;80:525–534. [206] Thilakaratne R, Wright MM, Brown RC. A techno-economic analysis of microalgae remnant catalytic pyrolysis and upgrading to fuels. Fuel 2014;128:104– 112. [207] Zhao X, Brown TR, Tyner WE. Stochastic techno-economic evaluation of cellulosic biofuel pathways. Bioresour Technol 2015;198:755–763. [208] Van Eck NJ, Waltman L. Software survey: VOSviewer, a computer program for bibliometric mapping 2010. doi:10.1007/s11192-009-0146-3. [209] Koh LP, Ghazoul J. Biofuels, biodiversity, and people: Understanding the conflicts
and
finding
opportunities.
Biol
Conserv
2008;141:2450–60.
doi:10.1016/j.biocon.2008.08.005. [210] Effendi A, Gerhauser H, Bridgwater AV. Production of renewable phenolic resins by thermochemical conversion of biomass: A review. Renew Sustain Energy Rev 2008;12:2092–116. doi:10.1016/j.rser.2007.04.008. [211] Kim H, Kim S, Dale BE. Biofuels, Land Use Change, and Greenhouse Gas Emissions: Some Unexplored Variables. Environ Sci Technol 2009;43:961–7. doi:10.1021/es802681k.
92 [212] Gnansounou E, Dauriat A. Techno-economic analysis of lignocellulosic ethanol:
A
review.
Bioresour
Technol
2010;101:4980–91.
doi:10.1016/j.biortech.2010.02.009. [213] Gonzalez RW, Phillips R, Jameel H, Abt R, Pirraglia A, Saloni D, et al. Biomass to energy in the southern United States: Supply chain and delivered cost. BioResources 2011;6:2954–2976. [214] Ballou RH. Business Logistics Managelnent 1992. [215] Akgul O, Zamboni A, Bezzo F, Shah N, Papageorgiou LG. Optimization-based approaches for bioethanol supply chains. Ind Eng Chem Res 2010;50:4927–4938. [216] Rogers JG, Brammer JG. Analysis of transport costs for energy crops for use in biomass pyrolysis plant networks. Biomass Bioenergy 2009;33:1367–1375. [217] Pootakham T, Kumar A. Bio-oil transport by pipeline: A techno-economic assessment. Bioresour Technol 2010;101:7137–7143. [218] Junginger M, Faaij A, Björheden R, Turkenburg WC. Technological learning and cost reductions in wood fuel supply chains in Sweden. Biomass Bioenergy 2005;29:399–418. doi:10.1016/j.biombioe.2005.06.006. [219] Ahtikoski A, Heikkilä J, Alenius V, Siren M. Economic viability of utilizing biomass energy from young stands—The case of Finland. Biomass Bioenergy 2008;32:988–96. doi:10.1016/j.biombioe.2008.01.022. [220] Mirkouei A, Haapala KR, Sessions J, Murthy GS. Evolutionary Optimization of Bioenergy Supply Chain Cost with Uncertain Forest Biomass Quality and Availability. Proc. IIEISERC May 21-24 Anaheim Calif. USA Rev., 2016.
93
CHAPTER 3: MULTI-CRITERIA DECISION MAKING FOR SUSTAINABLE BIO-OIL PRODUCTION USING A MIXED SUPPLY CHAIN
To be Submitted in the Journal of Cleaner Production
94 Chapter 3:
Multi-criteria Decision Making for Sustainable Bio-Oil Production using a Mixed Supply Chain
Abstract Growing awareness and concern within society over the use of and reliance on fossil fuels has stimulated research efforts in identifying, developing, and selecting alternative energy sources and energy technologies. Bioenergy represents a promising replacement for conventional energy, due to its reduced environmental impacts and broad applicability. Sustainable energy challenges, however, require innovative manufacturing technologies and practices to mitigate energy and material consumption. This research aims to facilitate sustainable production of bioenergy from forest biomass and to promote deployment of novel processing equipment, such as transportable biorefineries. The study integrates knowledge from the renewable energy production and supply chain management disciplines to evaluate economic targets of bioenergy production with the use of the multi-criteria decision making approach. The presented approach includes qualitative and quantitative methods to address the existing challenges and gaps in the bioenergy manufacturing system. The qualitative method employs decision tree analysis to classify the potential biomass harvesting sites by considering biomass quality and availability. The quantitative method proposes a mathematical model to optimize upstream and midstream biomass-to-bioenergy supply chain cost using mixed bio-refinery modes (transportable and fixed) and transportation pathways (truck-truck and truck-tanker). While transportable bio-refineries are shown to reduce biomass-to-bioenergy supply chain costs, manufacturing and deployment of
95 the transportable bio-refineries is limited due to interoperability challenges of undeveloped mixed-mode and mixed-pathway bioenergy supply chains and quality uncertainty.
Introduction Research into modeling and techno-economic analysis of renewable energy production has increased over the last few decades due to positive economic, environmental, and social impacts, e.g., access to clean energy and growth of new labor markets [1,2]. Investigators have paid much attention to approaches that meet cost targets and reduce greenhouse gas (GHG) emissions, e.g., emissions of carbon dioxide, methane, and nitrous oxides [3]. Biomass can be harvested and replenished in a relatively short timeframe, and represents one of the primary feedstocks for bioenergy manufacturing. According to the United States (U.S.) Energy Information Administration [4], over 80% of U.S. energy is provided through conventional fuels such as gas, coal, and petroleum (Figure 3.1). Renewable energy sources provide 10% of total U.S. energy consumption, and half of this renewable energy demand is provided through biomass. Biomass plays a key role in the renewable energy industry, thus attention should be placed on optimizing the production, conversion, and use of biomass in bioenergy manufacturing [5,6]. The manufacturing research community has initiated several efforts investigating improved biomass processing methods [7,8] and biomass-tobioenergy supply chain (B2BSC) optimization [9], as introduced briefly below.
96 1% Renewable
9%
10%
9%
Biomass
5%
Petroleum
Hydroelectric
21% Natural Gas
50%
35% 35%
Coal
Wind Geothermal
25% Nuclear Electric Power
Solar
Figure 3.1. Energy consumption in the United States [4]
A successful bioenergy manufacturing industry would benefit society by reducing energy costs, carbon footprint emissions, and the imports of fossil fuels, as well as improving energy security and health benefits [10]. The development and deployment of bioenergy sources are limited, however, due to supply uncertainties and lack of flexible, robust, and cost-effective manufacturing technologies and systems [11,12]. The development of manufacturing technologies, manufacturing systems, and supply chains must occur in concert to realize the expected economic, environmental, and social benefits of an established bioenergy manufacturing industry.
Several challenges must be overcome before achieving an efficient, optimized system that will ensure the viability and sustainability of the bioenergy network at all decision levels. The two key challenges are: 1) minimization of bioenergy production costs and 2) reduction of carbon footprint emissions. High costs of biomass feedstock supply inhibit the development of a strong bioenergy market. Thus, a hybrid method is required that will simultaneously minimize associated costs and carbon footprints, as well as overcome barriers associated with competitive markets for the upstream and
97 midstream segments of B2BSC. However, social aspects of sustainability are not addressed directly in the investigation here.
Optimal and sustainable bioenergy manufacturing systems require the following: (1) highly productive biomass sources, (2) advanced coordination of biomass supply (e.g., collection and transportation) at different decision levels (e.g., strategic, tactical, and operational), and (3) efficient conversion facilities that can mitigate bioenergy production costs [13,14]. In a broad sense, supply chain management considers all logistics network activities, which include challenges of integrating energy processing and manufacturing in B2BSCs. Figure 3.2 indicates the main upstream and midstream entity in the B2BSC network, i.e., harvesting, collection, pre-processing (e.g., drying, size reduction), transportation, and conversion.
Biomass Harvesting Collection
PreBioenergy Transportation Conversion processing Products
Upstream
Midstream
Figure 3.2. Upstream and midstream biomass-to-bioenergy supply chain entities
A motivation of this study is to examine the substitution of conventional energy sources with renewable energy sources through informed technology development, which will lead to improvements of social, environmental, and economic aspects of energy manufacturing, supply, and use. Strategic decision making, in this regard, examines internal and external perspectives from a multidisciplinary perspective. Thus, a detailed strategic decision model of B2BSC costs and environmental impacts is developed here
98 to satisfy conflicting objectives of diverse decision makers. The model can be utilized to consider several decision making objectives as follows:
Biomass source location and collection methods,
Transportation modes and routing, and
Manufacturing technology methods and locations.
In pursuing these objectives, research must identify methods that can consider optimization aspects and overcome challenges of traditional approaches. The presented approach utilizes multi-criteria decision making to explore quantitative and qualitative aspects of B2BSC to assist industry decision makers in determining optimal solutions for minimizing production cost and carbon footprint. This approach reduces the existing challenges in the limited interoperability between quantitative aspects of mathematical models and qualitative aspects of bioenergy manufacturing systems. The general multi-criteria decision support methodology is extended to consider feedstock supply quality, availability, and multiple process outputs. A higher capacity, transportable bio-refinery is evaluated, rather than a lower capacity mobile biorefinery, to investigate the potential advantages of economies of scale in bio-oil production. The presented approach integrates both qualitative and quantitative methods to assess criteria relevant to B2BSCs. This approach considers the effect of harvesting, collection, pre-processing, transportation, bio-refining, and storage activities on the overall annual cost of mixed-mode (transportable and fixed biorefining) and mixed-pathway (truck-truck and truck-tanker transport) B2BSCs.
99 The mixed-mode bio-refinery system combines transportable and fixed bio-refineries and explores the economic aspects of each mode to determine the most economical mode under different circumstances. Mixed-pathway transportation includes the combination of traditional and new pathways, using different equipment for each path. The traditional (truck-truck) pathway employs a single-trailer truck (14 metric ton or 15.4 US ton capacity) to haul biomass from harvesting sites to collection sites (or staging sites) and a double-trailer truck (25 metric ton or 27.5 US ton capacity) to haul biomass from staging sites to the fixed bio-refinery. The new (truck-tanker) pathway employs a single-trailer truck (14 metric ton capacity) to haul biomass from harvesting sites to a transportable bio-refinery, and a tanker truck (23 metric ton or 25.3 US ton capacity) to transfer bio-oil to storage sites. The next section includes an overview of existing studies in the context of bioenergy manufacturing system through the lens of sustainability. Background The academic and industrial communities have been intensively investigating novel uses of bioenergy sources for several decades. Investigators have placed recent scrutiny on the production logistics for relevant bioenergy products, including forest biomass [14,15]. Research has mainly focused on developing quantitative models to represent existing systems and addressing related logistics issues, such as transportation and production planning [3,16]. The results of this work have led to the establishment of various concepts in the domain of traditional bioenergy sources and the integration of key contributions across a broader range of areas including sustainable manufacturing, bioenergy production planning, and supply chain decision making.
100
The need for further investigation is not only through empirical studies, but also through the development of conceptual studies that can integrate methods and tools for improving B2BSC performance [14]. B2BSC development is supported by academia, government and industry [17–19]. Several studies highlight the need to integrate supply chain processes with either conventional energy sources (e.g., gas and diesel) or bioenergy sources (e.g., biogas and biodiesel). Several methods and metrics have been classified by researchers to enhance the performance of bioenergy processing [10].
In understanding the effect of biomass qualities on bio-energy production, conversion, and use, it is important to evaluate the type(s) of biomass to be considered. The major energy-use sources are solid biomass (e.g., crops and straw), wet biomass (e.g., manure and organic waste), oil seeds (e.g., sunflower and rapeseed), and plant materials (e.g., sugar and starch). Products of converted biomass include heat and power, oils and fuels, and chemicals, each of which has specific properties that impact effectiveness in the desired applications [18]. For example, bio-power can be used for heat and power generation and has environmental benefits over conventional power sources, while biofuels are a potential option for transportation systems that can have lower environmental impacts than conventional fuels.
The advantages of using biomass to produce bioenergy have been identified as follows [16]: (1) availability across the world, (2) potential of using agricultural and forestry residues and byproducts, (3) potential of reducing fire hazards, and (4) the carbon-
101 neutral nature of biomass combustion. Identified drawbacks include [19]: (1) high transportation costs and environmental impacts due to high moisture content (low energy density), (2) large storage area requirements due to biomass properties (low material density), and (3) high production costs due to immature manufacturing technologies. Biomass supply is generally coordinated through purchases of feedstocks from suppliers, or through harvesting, collection, and transportation by the company from the forest or field.
Prior research efforts in bioenergy manufacturing systems have applied mathematical programming models that represent existing systems through the objective function and use real world constraints to generate optimal outcomes [16,18,20]. In general, mathematical programming is comprised of four distinct techniques: (1) linear programming, (2) integer programming, (3) mixed integer linear programming, and (4) non-linear programming [15]. Among these techniques, mixed integer linear programming (MILP) is the most common optimization method because it can be applied in different types of decision problems. For bioenergy supply chain optimization, the MILP approach enables consideration of integer variables representing discrete units of equipment.
While the economic objective is most frequently addressed by using mathematical programming approaches [15], the models have differences in specific characteristics, including the objective function, decision variables, and constraints. Different types of objective functions have been considered at the strategic decision level for economic
102 objective problems, e.g., to minimize total cost [6], maximize overall profits [21], maximize net present value [22], maximize revenue [23], minimize risk on investment [24], and minimize transportation and processing costs [6]. Transportation optimization problems consider decision variables such as modes, capacity, and routing. Common alternative biomass transport modes include log trailers, container trailers, and chip vans. Each of these modes is dependent on the circumstance and collection approach in the forest or field [25]. Due to equipment limitations, the collection mode can have a significant effect on biomass availability and collection costs. Similarly, different technologies can be applied in the production of bioenergy from biomass, which affect costs and environmental impacts.
In general, energy production pathways are classified as thermochemical and biochemical [26,27]. Since these processes have different characteristics, investigators need to consider issues such as type of biomass and preferred bio-products when performing cost optimization and sustainability analyses. Biochemical conversion can, for example, convert corn stover to ethanol using biochemical processing. Thermochemical processing involves gasification and pyrolysis based conversion from biomass to bio-oil [28,18]. Gasification converts biomass into a syngas, tar, and a solid char at high temperature under oxygen limited reaction conditions. Fast pyrolysis conversion technology is explored in this research, and converts biomass to bio-oil and bio-char at temperatures between 250 and 550°C in the absence of oxygen [29]. To address several of the disadvantages mentioned above (e.g., high transportation cost and low energy density), a transportable, fast pyrolysis bio-refinery has been developed
103 [28, 29]. The transportable bio-refinery is a truck-mounted unit that can travel to farms and forests, where underutilized residue is available. The transportable bio-refinery enables conversion of low-energy density biomass to high-energy density intermediate bio-products [28]. The main advantages of transportable bio-refineries are reported as follows [32,19]:
Reduced fixed costs (e.g., depreciation, as well as interest, insurance, and taxes), variable costs (e.g., repair and maintenance, fuels and lubricants, and overhead), and labor costs (e.g., wages and benefits) for establishing a bioenergy manufacturing system,
Reduced land area requirements for biomass storage, and
Reduced transportation costs due to transfer of denser biochemicals (e.g., bio-oil) instead of unprocessed biomass.
In addition, transportable bio-refineries enable mixed-pathway transportation, which overcomes several logistical challenges (e.g., tight curves, limited areas to turn around, and narrowness) associated with transferring of underutilized forest harvest residues. Employing both single- and double-trailer trucks in the traditional pathway reduces transportation costs by assembling loads at the staging area to take advantage of the maximum allowable legal weight based on the number of axles and axle spacing [33]. Using tanker trucks in a new mixed pathway configuration decreases transportation costs since they transfer high energy density bio-oil from the transportable bio-refinery, rather than hauling low energy density biomass from harvesting sites to a fixed bio-refinery.
104 Since their initial development, however, these transportable bio-refineries have not been adopted by manufacturing industry, and their benefits from a cost and environmental impact perspective are unclear. In particular, the advantages of using transportable bio-refining technology in place of centralized, or fixed bio-refineries are likely situationally dependent on biomass type, transportation distances, time of the year, policies and regulations, and numerous other factors. Mirkouei and Haapala [34] previously proposed an integrated decision making method to evaluate the potential suppliers and select the best supplier to purchase biomass for bioenergy production through an optimization goal programming model, which addressed three main objectives: cost, quality, and delivery time. Additionally, Mirkouei et al. [6] developed an economic model and applied mixed-mode (mobile and fixed bio-refinery) biomass processing and bio-oil supply in northwest Oregon, which demonstrated the potential cost benefits of mobile bio-refinery technology. Methodology As discussed above, various approaches have been proposed in the literature to design optimal biomass pre-processing and bioenergy manufacturing networks. There are limited frameworks, however, that have combined both qualitative and quantitative analysis techniques. The methodology reported here presents a decision method that integrates qualitative (Decision Tree Analysis) and quantitative (Mathematical Programming) analysis into a decision support system to assess the role of mixed (mode and pathway) bioenergy supply chains (Figure 3.3). The methodology proceeds in two phases to accomplish B2BSC network optimization. In Phase 1, the harvesting areas are classified according to the availability and quality of biomass feedstocks. In Phase
105 2, an MILP model is developed to optimize the total cost of production in a mixed supply chain. Since transportation is one of the main cost drivers in B2BSCs, transporting biomass to a central processing facility may not be cost effective. Thus, using a transportable bio-refinery is a potential alternative in terms of overall cost to fixed bio-refineries, when: 1) the travel time from the forest to the bio-refinery is high, 2) the availability of biomass is low, and 3) the quality of biomass (energy density) is low. The two phases of the methodology are described in more detail below. Qualitative Decision Tree Analysis
Quantitative Mathematical Programming
Decision Support System Figure 3.3. Integrated multi-criteria decision-making approach for mixed bioenergy supply chains
Phase 1. Qualitative Analysis Method Qualitative analysis methods have several advantages compared with quantitative analysis methods, which include the graphical representation of all possible decisions, issues, and consequences, along with the ability to explicitly show all possible alternative results. These advantages assist decision makers to more rapidly compare the various alternatives. The key benefits of the qualitative method in this research study are the consideration of various parameters (e.g., distance, cost, and biomass types), uncertainties (e.g., seasonality, quality, availability, moisture content, and lead time), and processes (e.g., chipping and grinding). The use of quantitative methods (e.g., mathematical programming) requires development of complex models (e.g.,
106 dynamic and stochastic), which are difficult to quantify due to the nature of above mentioned parameters, uncertainties, and processes.
Decision tree (DT) analysis is a method in data mining that is often used to reduce the ambiguity in decision-making processes by representing all solutions in a single view [35]. DT analysis provides a comprehensive resolution of the possible decisions, using a simple and easy to understand graphic design. DT analysis categorizes the parameter values into defined categories based on decision makers’ preferences. It requires two datasets (i.e., training and testing datasets) to implement the analysis, classify the data based on the desired features, and test the result to measure the misclassification error rate. The misclassification error rate is the proportion of misclassified observations, and is computed from the number of errors in training datasets. A training dataset is used to implement the DT analysis and a testing dataset is used to ensure the effectiveness of the classification method. An optimal decision tree has a low misclassification error rate in supervised classification learning [35]. Therefore, supervised classification learning improves the capability of decision-making by allowing investigators to make educated decisions, using historical data.
A decision tree analysis is developed as a qualitative method to evaluate and select the potential biomass harvesting sites based on availability and quality of biomass feedstock. The training and testing datasets include simulated levels for biomass quality and availability (i.e., low, medium, and high) at each harvesting site. Each harvesting site is categorized into different priority levels based on the historical parameter values
107 (i.e., biomass quality and available amount of biomass). Those sites that have high quality and high availability of biomass are classified as the first priority, and the remaining sites are classified in three lower priority levels. One of the key benefits of using this qualitative method is the ability to consider various effective criteria (e.g., distance or moisture content) in the B2BSC network, which would require complex mathematical modeling and computation if decision makers were to explore them quantitatively. Quantitative methods offer a powerful approach for problems amenable to mathematical modeling and optimization, as described in the next section (Phase 2).
Phase 2. Quantitative Analysis Method A mathematical model is developed as a quantitative analysis approach to assess the commercial feasibility of transportable bio-refinery for bioenergy production within the mixed-mode and mixed-pathway B2BSC network. An MILP model is formulated with the aim of minimizing the sum of B2BSC fixed, variable, and labor costs. The mathematical optimization model takes into account the overall annual costs for the combination of transportable and fixed bio-refineries. The final products of transportable bio-refinery processing are bio-oil, which will be transferred to a fixed bio-refinery, and bio-char, which will be transferred to a distribution center. While the focus of this study is on the upstream and midstream segments of the biomass to biooil supply chains, the downstream segment (which includes distribution and demand activities) and bio-char (another bioproduct of pyrolysis) will be considered in future work.
108 Objective Function. The objective function is developed to minimize the total cost (TC), and includes the fixed, variable, and labor costs of harvesting, collection, preprocessing, transportation, bio-refining, and storage (Eq. 1). The model nomenclature is provided in Annex A. Since transportation has significant impacts on total costs, the model considers three truck types to address maneuverability on various forest roads and main roads: 1) Single-trailer trucks transport biomass from harvesting sites to collection sites or transportable bio-refineries, 2) Double-trailer trucks transport biomass from the collection sites to fixed biorefineries, and 3) Tanker trucks transport bio-oil from transportable bio-refineries to fixed biorefineries. 𝑀𝑖𝑛 𝑇𝐶 = ∑ ∑ ∑ [𝐹ℎ + 𝐿ℎ + 𝑉ℎ ] ∗ 𝑖
𝑗
𝑡
+ ∑ ∑ ∑ [𝐹𝑐 + 𝐿𝑐 + 𝑉𝑐 ] ∗ 𝑖
𝑗
𝑡
𝑋𝑖𝑗𝑡 𝑃𝑅𝑐
+ ∑ ∑ ∑ [𝐹𝑝 + 𝐿𝑝 + 𝑉𝑝 ] ∗ 𝑖
𝑗
𝑡
𝑋𝑖𝑗𝑡 𝑃𝑅𝑝
+ ∑ ∑ ∑ [𝐹𝑚 + 𝐿𝑚 + 𝑉𝑚 ] ∗ 𝑗
𝑘
𝑡
+ ∑ ∑ ∑ [𝐹𝑙 + 𝐿𝑙 + 𝑉𝑙 ] ∗ 𝑗
𝑙
𝑡
𝑋𝑖𝑗𝑡 𝑃𝑅ℎ
𝑋𝑗𝑘𝑡 𝑃𝑅𝑚𝑜𝑏
𝑋𝑗𝑙𝑡 𝑃𝑅𝑓𝑖𝑥
+ ∑ ∑ ∑𝐹𝑠_𝑚𝑎𝑠𝑠 ∗ 𝐵𝑙𝑡 + ∑ ∑ ∑𝐹𝑠_𝑜𝑖𝑙 ∗ 𝐵𝑠𝑡 𝑗
𝑙
𝑡
𝑘
𝑠
𝑡
+ ∑ ∑ ∑ [𝐹𝑠_𝑡𝑟𝑛𝑠 + 𝐿𝑠_𝑡𝑟𝑛𝑠 + 𝑉𝑠_𝑡𝑟𝑛𝑠 ] ∗ 𝑖
𝑗
𝑡
𝑋𝑖𝑗𝑡 𝑃𝑅𝑠
𝑋𝑖𝑘𝑡 𝑃𝑅𝑠 𝑖 𝑘 𝑡 𝑋𝑗𝑙𝑡 + ∑ ∑ ∑ [𝐹𝑏_𝑡𝑟𝑛𝑠 + 𝐿𝑏_𝑡𝑟𝑛𝑠 + 𝑉𝑏_𝑡𝑟𝑛𝑠 ] ∗ 𝑃𝑅𝑏 + ∑ ∑ ∑ [𝐹𝑠_𝑡𝑟𝑛𝑠 + 𝐿𝑠_𝑡𝑟𝑛𝑠 + 𝑉𝑠_𝑡𝑟𝑛𝑠 ] ∗
𝑗
𝑙
𝑡
+ ∑ ∑ ∑ [𝐹𝑘_𝑡𝑟𝑛𝑠 + 𝐿𝑘_𝑡𝑟𝑛𝑠 + 𝑉𝑘_𝑡𝑟𝑛𝑠 ] ∗ 𝑘
𝑠
𝑡
𝑌𝑘𝑠𝑡 𝑃𝑅𝑡𝑛𝑘
(3.1)
109
Model Constraints. The following constraints are applied in solving the mathematical optimization model: 1) To conserve flows in and out of node j (Eq. 2) 2) To account for efficiency of conversion (yield) of biomass to bio-oil (Eqs. 3, 4) 3) Transportation arc triggers to allow transport of material between sites (Eqs. 5, 6) 4) To ensure that the amount of biomass for conversion does not exceed the capacity of transportable and fixed bio-refineries (Eqs.7, 8) 5) To ensure that the amount of bio-oil produced does not exceed the storage capacity (Eq. 9) 6) To ensure the amount of transferred biomass is equal to the sum of the available amount of biomass in all harvesting sites (Eq. 10) 7) Non-negativity, binary, and integer constraints to guarantee that the solution is feasible (Eqs. 11-13). ∑ ∑ 𝑋𝑖𝑗𝑡 − ∑ ∑ 𝑋𝑗𝑙𝑡 = 0 𝑖∈𝐼 𝑡∈𝑇
𝑙∈𝐿 𝑡∈𝑇
𝑃 ∗ ∑ ∑ 𝑋𝑖𝑘𝑡 − ∑ ∑ 𝑌𝑘𝑠𝑡 = 0 𝑖∈𝐼 𝑡∈𝑇
𝑠∈𝑆 𝑡∈𝑇
𝑃 ∗ ∑ ∑ 𝑋𝑗𝑙𝑡 − ∑ ∑ 𝑌𝑙𝑠𝑡 = 0 𝑗∈𝐽 𝑡∈𝑇
𝑠∈𝑆 𝑡∈𝑇
∑ ∑ 𝑋𝑗𝑙𝑡 ≤ 𝑀 ∗ 𝐵𝑙𝑡 𝑗∈𝐽 𝑡∈𝑇
∑ ∑ 𝑌𝑘𝑠𝑡 + ∑ ∑ 𝑌𝑙𝑠𝑡 ≤ 𝑀 ∗ 𝐵𝑠𝑡 𝑘∈𝐾 𝑡∈𝑇
𝑙∈𝐿 𝑡∈𝑇
∑ ∑ 𝑋𝑗𝑘𝑡 ≤ 𝐶𝐴𝑃𝑚𝑜𝑏 𝑗∈𝐽 𝑡∈𝑇
∑ ∑ 𝑋𝑗𝑙𝑡 ≤ 𝐶𝐴𝑃𝑙 𝑗∈𝐽 𝑡∈𝑇
∑ ∑ 𝑌𝑙𝑠𝑡 + ∑ ∑ 𝑌𝑘𝑠𝑡 ≤ 𝐶𝐴𝑃𝑠_𝑜𝑖𝑙 𝑙∈𝐿 𝑡∈𝑇
𝑘∈𝐾 𝑡∈𝑇
∑ ∑ ∑ 𝑋𝑖𝑗𝑡 ≥ 𝛼𝑡 𝑖∈𝐼 𝑗∈𝐽 𝑡∈𝑇
∀𝑖 ∈ 𝐼, ∀𝑙 ∈ 𝐿, ∀𝑡 ∈ 𝑇
(3.2)
∀𝑖 ∈ 𝐼, ∀𝑠 ∈ 𝑆, ∀𝑡 ∈ 𝑇
(3.3)
∀𝑗 ∈ 𝐽, ∀𝑠 ∈ 𝑆, ∀𝑡 ∈ 𝑇
(3.4)
∀𝑗 ∈ 𝐽 , ∀𝑡 ∈ 𝑇
(3.5)
∀𝑘 ∈ 𝐾 , ∀𝑙 ∈ 𝐿, ∀𝑡 ∈ 𝑇
(3.6)
∀𝑗 ∈ 𝐽 , ∀𝑡 ∈ 𝑇
(3.7)
∀𝑗 ∈ 𝐽 , ∀𝑡 ∈ 𝑇
(3.8)
∀𝑙 ∈ 𝐿, ∀𝑘 ∈ 𝐾, ∀𝑡 ∈ 𝑇
(3.9)
∀𝑖 ∈ 𝐼, ∀𝑗 ∈ 𝐽, ∀𝑡 ∈ 𝑇
(3.10)
110 𝑋𝑖𝑗𝑡 , 𝑋𝑖𝑘𝑡 , 𝑋𝑗𝑙𝑡 ≥ 0
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖, 𝑗, 𝑘, 𝑙, 𝑎𝑛𝑑 𝑡
(3.11)
𝐵𝑙𝑡 , 𝐵𝑠𝑡 = {0, 1}
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑙, 𝑠, 𝑎𝑛𝑑 𝑡
(3.12)
𝑌𝑙𝑠𝑡 , 𝑌𝑘𝑠𝑡 are integers
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑘, 𝑙, 𝑠, 𝑎𝑛𝑑 𝑡
(3.13)
Case Study A simulated case study is conducted to demonstrate the application of the method described above to B2BSC management (Figure 3.4). In this simulated case, decision makers have a desire to reduce bioenergy production costs. Based on the model, the potential locations for a transportable bio-refinery could be selected from among the harvest areas close to collection sites. Different road types are considered between harvesting sites and fixed bio-refineries, which along with the processing modes, impacts the types of truck used in the supply chain. Distribution Center
Transportable Bio-refinery
Harvesting Sites
Collection or Staging Sites
Tanker
Fixed Bio-refinery
Truck
Figure 3.4. A schematic of a mixed-mode and mixed-pathway biomass-to-bio-oil supply chain (question marks indicate location and asset decision points)
Phase 1. Classification of Harvesting Areas As discussed above, decision tree analysis can be applied as a qualitative method to classify harvesting sites based on the quality and availability of underutilized forest harvest residues. A simulated dataset is used to provide the parameter values for the
111 available amount and quality of biomass for 180 harvesting sites (as a training dataset), containing different types of forest biomass (e.g., slash, chips, and logs). DT analysis classifies the harvesting sites into four priority levels, with sites in the first level considered as potential harvesting sites, since they have high quality and high availability of biomass. The DT analysis first evaluates biomass quality levels and then evaluates the available amount of biomass, using historical data for each harvesting site. In this study, DT analysis is conducted using R, a language for statistical computing [36].
Figure 3.5 illustrates the classification results for the 180 harvesting sites. Of these, 100 sites are found to be in the first four priority levels. In this study, training and testing dataset for developing supervised classification learning include parameter values (biomass quality and availability) of 150 and 30 harvesting sites, respectively. Biomass quality is evaluated based on energy content (low heating value), which is the amount of energy stored in the given biomass. The energy content percentage of biomass in this study is assumed on a dry basis with no energy loss to remove (evaporate) any moisture. The high and medium quality biomass have at least 70% and 50% energy content respectively. The DT analysis only considers those harvesting sites that have high and medium quantities of biomass. Each site with high quantity has at least 150 metric tons (165 US tons) of available biomass, and each site with medium quantity has at least 50 metric tons (55 US tons) of available biomass. With this information, the results of DT analysis indicate that 27, 22, 20, and 31 sites are placed in the first, second, third, and fourth priority level, respectively. The first
112 priority has high quality and quantity of biomass, second priority has medium quality and high quantity of biomass, third priority has high quality and medium quantity of biomass, and fourth priority has medium quality and quantity of biomass. The priority levels are defined by decision makers, which can be justified based on their systems. Besides considering biomass availability and quality, this phase can be expanded to define the appropriate priority level for harvesting sites using other criteria, such as grit contamination of biomass, distance from a bio-refinery, and seasonality. From DT analysis and decision maker evaluation, it is determined that the proposed mixed B2BSC network contains twenty high potential harvesting sites (out of 27), as well as defining five collection sites, two transportable bio-refineries, and a fixed bio-refinery with storage capacity. Quality Medium (> 50%)
High (> 70%) Quantity
Quantity Medium(> 50 tons)
4th Priority
High (> 150 tons)
2nd Priority
Medium(> 50 tons)
3rd Priority
High (> 150 tons)
1st Priority
Figure 3.5. Supervised classification analysis for providing biomass from potential harvesting sites (R results)
Phase 2. Mathematical Optimization Model To create the mathematical model for the simulated supply chain, information and data are collected from relevant prior research [37,25,38,6]. Cost calculations for harvesting, collection, pre-processing, and bioenergy processing are conducted using
113 the machine rate method for the fixed, variable, labor costs [39]. The first step is to determine the purchase price and salvage value, as well as interest, insurance, and tax rates for estimating the fixed cost. The second step is to determine the repair and maintenance, fuel and lubricant consumption, supporting equipment, and other relevant costs to estimate the variable cost. The last step is to calculate the labor cost by considering wages and benefits. All cost values in this study are adjusted for inflation to 2015 using the U.S. Producer Price Index [40]. The assumptions of model formulation are: 1) the available amount of biomass is known, 2) the type of truck for each type of route is known, and 3) the time horizon is one year.
Table 3.1 depicts production rate of transportable and fixed bio-refineries, as well as related fixed, variable, and labor costs. The effective lifetime of a bio-refinery is assumed to be ten years for transportable and fixed bio-refinery [37]. The production yield for biomass-to-bio-oil conversion is dependent on the type of feedstock, its moisture
and
ash
content,
and
the
bio-refinery
process
(e.g.,
fast,
intermediate/conventional, and slow pyrolysis). Conventional pyrolysis requires particle sizes of 5-50 millimeters (0.2-2 inches), as opposed to fast pyrolysis, which requires less than one millimeter particles. Although fast pyrolysis is expected to have higher bio-oil yields (up to 70%), conventional pyrolysis is considered here to achieve lower associated grinding energy costs [41]. The variable costs of transportable and fixed bio-refinery encompasses feedstock loading cost, repair and maintenance cost, purchased energy cost, mobilization cost, and move-in and setup cost; though the last two costs are not associated with fixed bio-refineries.
114 Table 3.1. Selected bio-refinery attributes [37] Bio-refinery Size (dtpda) Annual Fixed Cost ($) Annual Variable Cost ($) Annual Labor Cost ($) Bio-oil Product Yield (%) a dry tons per day
Transportable 50 600,762 89,082 375,357 50
Fixed 200 4,314,095 1,578,218 1,155,334 50
Table 3.2 presents fixed, variable, and labor cost of trucks types that are needed for different types of roads (e.g., in-forest roads and highways). The single-trailer truck has more flexibility and transportability on in-forest roads, while the double-trailer truck has a larger capacity [33]. In the new pathway, the tanker truck has been applied to transfer bio-oil from the transportable bio-refinery to storage. The effective life of a truck is assumed as eight years [25]. Fixed cost includes trailer truck purchase price and salvage value to calculate depreciation cost, as well as interest, insurance, and tax expenses on the capital investment. Variable cost consists of tire, repair and maintenance, fuel and lubricant, and overhead costs. In the traditional pathway, the variable cost also includes the costs of hooking and unhooking the additional trailer, which is added at the collection sites (or staging sites). Hooking and unhooking cost is assumed to be $120 per round trip in this study. Labor cost includes wages and benefits. Table 3.2. Annual truck costs based on capacity [25] Fixed Cost ($) Variable Cost ($) Labor Cost ($) Capacity (metric tons)
Single-trailer Truck 24,633 12,9520 47,360 14
Double-trailer Truck 47,536 233,420 47,360 25
Tanker 44,074 145,332 47,360 23
Fixed, variable, and labor costs of harvesting, collection, and pre-processing activities in the upstream segment of the B2BSC are presented in Table 3.3, and include the associated costs of using a harvester, forwarder, and grinder, respectively [25,37]. The
115 variable cost of grinding includes the costs of the grinder bits, grates, and anvil, repair and maintenance, fuel and lubricant, loader and supporting equipment, and overhead costs. Additionally, annual fixed cost of storage, which are required to store biomass feedstock and bio-oil are considered $81,000 and $170,100, respectively [6].
Table 3.3. Annual costs of upstream entities [25,37] Fixed Cost ($) Variable Cost ($) Labor Cost ($)
Harvesting 111,255 118,032 76,464
Collection 74,178 101,513 65,520
Size Reduction 146,482 508,500 51,705
Computational Results The mathematical optimization model was solved using an optimization solver (Gurobi 5.6.3) with Python 2.7 [42,43]. The model has 121 decision variables (including two binary variables, two integer variables, and 117 continuous variables) and 115 constraints. An optimal solution was found after 15 iterations in almost four seconds using a Windows 8 64-bit Operation System, Intel Core i5 processor (CPU 1.80GHz), and 6GB RAM. The overall annual cost was predicted as $2,418,432, or $0.29/liter ($1.10/gallon). The model results for the Base Case included 20 harvesting sites, five collection sites (staging sites), two transportable bio-refineries, and one storage facility in one year time horizon. Since the throughput of the transportable bio-refinery was assumed to be 50 metric tons (55 US tons) per day, annual production (329 days) is 16,425 metric tons (18,105 US tons). In the Base Case, the amount of forest biomass to be processed was 20,000 dry metric tons (22,046 US tons), based on the actual data obtained from Oregon Department of Forestry for three forest districts, i.e., Forest Grove, Tillamook, and Astoria [44]. The optimal solution indicates that available
116 biomass would be processed in two transportable bio-refineries producing 10,000 metric tons (11,023 US tons or 2,201,433 gallons) of bio-oil. The predicted total annual cost for the Base Case without considering the transportable bio-refinery (Case 0) was considerably higher at $3,069,009, or $0.37/liter ($1.4/gallon). Thus, the transportable bio-refinery reduces the overall cost by over $650,578 (~27%), as shown in Table 3.4.
This analysis illustrates that the role of new transportable bioenergy manufacturing technology provides a benefit to the B2BSC network. In particular, the results demonstrate that overall production costs can be minimized when implementing a transportable bio-refinery. By exploring the model further, it can be shown that the use of transportable bio-refinery is more suitable when the unit transportation costs increase, as discussed below.
Sensitivity Analysis Based on the structure of the mathematical optimization model, several criteria are seen to have significant effects on the overall cost of the system, which can be examined using sensitivity analysis. The main purpose of sensitivity analysis is to assess the effect of basic variables (e.g., continuous and integer variables), non-basic variables (e.g., bio-oil product yield), and right hand side parameters (e.g., bio-refinery capacity and biomass/bio-oil storage capacity) on the optimization results. The presented sensitivity analysis evaluates the effect of two major criteria: transportable bio-refinery cost and available amount biomass. Apart from the Base Case, two different cases are presented below.
117
Effect of Transportable Bio-refinery Cost The effect of transportable bio-refinery cost on the overall cost of the simulated network is investigated using two different cases. The transportable bio-refinery cost is the major cost driver in bio-oil production in the Base Case. Thus, exploring the effect of this attribute is essential to assess the commercial feasibility for bioenergy production. In the first alternative case (Case 1), the bio-refinery costs (i.e., fixed, variable, and labor costs) of the transportable bio-refinery are reduced by 50%. In the second case (Case 2), the bio-refinery costs of the transportable bio-refinery are increased by 50%. Table 3.4 reports the optimum results for each case, as well as the results without considering the transportable bio-refinery in the Base Case (Case 0). Table 3.4. Effect of transportable bio-refinery cost on the overall annual cost Overall Cost ($) Cases Base Case Case 0 (Base Case w/o TRa) Case 1 (-50% cost) Case 2 (+50% cost) a Base Case without transportable bio-refinery
2,418,431 3,069,009 1,769,905 3,066,958
Bio-refinery Cost ($) Fixed Cost Variable Labor Cost Cost 600,762 89,082 375,357 1,181,944 432,388 316,530 300,381 44,541 187,678 901,143 133,623 563,035
The results show that changes in the transportable bio-refinery cost directly impacts the total cost. The overall annual cost decreased in Case 1 by approximately $648,526 (27%) and in Case 2, the annual cost increased by $616,258 (25%). The Case 1 result indicates that the optimal supply chain would consist of two transportable biorefineries. In Case 2, the transportable bio-refinery would be responsible for processing all forest biomass due to higher transportation and storage cost of using the fixed biorefinery compared with the transportable bio-refinery (Figure 3.6). Although the biorefinery costs are higher than Case 2, the increased processing rate of the fixed bio-
118 refinery in Case 0 reduces total annual cost. While transportation cost of the traditional pathway is higher than new pathway, the combination of single- and double-trailer trucks in the traditional pathway reduces the transportation costs compared to using only single trailer trucks. 3.5 Transportation Cost
Total Annual Cost (Millions $)
3 Storage Cost
2.5 Fixed Bio-refinery Cost
2 Transportable Bio-refinery Cost
1.5 Size Reduction Cost
1 Collection Cost
0.5 Harvesting Cost
0 Base Case
Case 0
Case 1
Case 2
Figure 3.6. Effect of transportable bio-refinery cost on annual supply chain costs: Case 1, transportable bio-refinery costs are reduced by 50%, and Case 2, transportable bio-refinery costs are increased by 50%
Effect of Available Amount of Forest Biomass The effect of the available amount of forest biomass is also investigated due to the importance of this attribute in the proposed multi-criteria decision making approach. In Phase 1, the available amount of biomass is the main parameter, along with biomass quality. In Phase 2, changing the amount of available biomass affects the total annual cost of simulated network, due to the direct impacts of this parameters on each entity in the B2BSC. As explained above, the available amount of biomass is assumed to be 20,000 metric tons in the Base Case. In Case 3, the available amount is decreased by
119 50%. In Case 4, the available amount is increased by 50%. Changing the amount of biomass has a direct impact on harvesting, collection, pre-processing, transportation, and bio-refinery costs. Table 3.5 reports the predicted annual cost for each case.
Table 3.5. Effect of available amount of forest biomass on the overall annual cost Cases Base Case Case 3 (-50% amount of biomass) Case 4 (+50% amount of biomass)
Overall Cost ($) 2,418,431 1,672,201 3,516,242
Amount of Available Biomass (metric ton) 20,000 10,000 30,000
The total supply chain cost is found to change directly, but nonlinearly with the amount of available biomass processed. In Case 3, the annual cost of the network decreased by $746,230 (31%), while increasing the amount of biomass in Case 4 increased costs by $1,097,811 (45%). The optimal supply chains utilize one transportable bio-refinery in Case 3, and two transportable bio-refineries in Case 4. The effect of biomass availability on the various components of supply chain costs are shown in Figure 3.7.
Conclusions Growing energy demand and related concerns about energy security and environmental impacts of material extraction and energy generation indicate that new alternative energy sources are needed. Studies investigating biomass-to-bioenergy production report that biomass is expected to occupy a significant fraction of total electricity, heat, and transportation energy sources in the future. Subsequently, development of bioenergy technologies will play a crucial role in energy availability; these technologies represent a key future market for the manufacturing industry. However, current market
120 readiness has led to uncertainties and undeveloped supply chains that inhibit the development of a sustainable and robust bioenergy manufacturing industry.
4 Transportation Cost
Total Annual Cost (Millions $)
3.5 Storage Cost
3
Fixed Bio-refinery Cost
2.5
Transportable Bio-refinery Cost
2 1.5
Size Reduction Cost
1
Collection Cost
0.5
Harvesting Cost
0 Base Case
Case 3
Case 4
Figure 3.7. Effect of available amount of forest biomass on annual supply chain costs: Case 3, the amount of biomass is decreased by 50%, and Case 4, the amount of biomass increased by 50%
The method developed herein presents a decision support system for B2BSC optimization, which includes two phases: classification tree analysis as a qualitative technique and mathematical programming model as a quantitative technique. The qualitative analysis phase employs decision tree (DT) analysis to classify the potential harvesting sites based on the defined parameters (biomass quality and quantity) that have high impacts on bioenergy production. The quantitative analysis phase formulates a mixed-integer linear programming (MILP) model to optimize the annual cost of upstream and midstream segments of B2BSC network. The MILP model considers mixed-modes (transportable and fixed bio-refineries) and mixed-pathways (truck-truck
121 and truck-tanker transportation) for a simulated case, representing a realistic B2BSC network.
It is demonstrated that this decision support approach can be used by decision makers to assist in the design of optimal bioenergy manufacturing systems. The MILP model is used to evaluate the applicability of a transportable bio-refinery in biomass-to-biooil production. In addition, it was found that integration of qualitative and quantitative methods offers a promising approach to supplementing existing methods for B2BSC operation and management through the use of double-trailer trucks in the traditional pathway and transportable bio-refinery technology in the new pathway. The model results seem to indicate the commercial feasibility of using mixed-mode and mixedpathway B2BSC to facilitate bioenergy production of underutilized forest harvest residues and to overcome logistical challenges in biomass transportation.
The results of the presented method have shown that the optimal integration of transportable and fixed bio-refineries can potentially improve the robustness and reduce the overall cost of bioenergy manufacturing systems. This mixed supply chain would be particularly beneficial for meeting sustainability goals, including economic and environmental targets. Prior research has found transportation to be a key source of environmental impacts in the conversion of underutilized forest biomass to bioenergy products. Future work should explore the effect of transportable biorefineries on supply chain sustainability and consider the downstream segment (end
122 use) using actual bioenergy manufacturing system data to assess robustness and accuracy of modeling.
Acknowledgment The authors would like to thank Michael Wilson and Rob Nall of the Oregon Department of Forestry for providing data to support this research.
Annex A: Nomenclature Indices 𝑖 𝑗 𝑘 𝑙 𝑡 ℎ 𝑐 𝑝 𝑚𝑜𝑏 𝑙 𝑠 𝑏 𝑘 𝑀
Set of harvesting sites Set of collection sites Set of transportable bio-refinery sites Set of fixed (non-transportable) bio-refinery sites Set of time periods Harvesting Collection Pre-processing Transportable bio-refinery Fixed bio-refinery Single-trailer truck Double-trailer truck Tanker large positive constant (Big M)
Parameters Annual fixed cost for a harvesting site ($) 𝐹ℎ Annual variable cost for a harvesting site ($) 𝑉ℎ Annual labor cost for a harvesting site ($) 𝐿ℎ Annual fixed cost for a collection site ($) 𝐹𝑐 Annual variable cost for a collection site ($) 𝑉𝑐 Annual labor cost for a collection site ($) 𝐿𝑐 Annual fixed cost for a pre-processing site ($) 𝐹𝑝 Annual variable cost for a pre-processing site ($) 𝑉𝑝 Annual labor cost for a pre-processing site ($) 𝐿𝑝 Annual fixed cost for a transportable bio-refinery ($) 𝐹𝑚
123 𝑉𝑚 𝐿𝑚 𝐹𝑙 𝑉𝑙 𝐿𝑙 𝐹𝑠_𝑚𝑎𝑠𝑠 𝐹𝑠_𝑜𝑖𝑙 𝐹𝑠_𝑡𝑟𝑛𝑠 𝑉𝑠_𝑡𝑟𝑛𝑠 𝐿𝑠_𝑡𝑟𝑛𝑠 𝐹𝑏_𝑡𝑟𝑛𝑠 𝑉𝑏_𝑡𝑟𝑛𝑠 𝐿𝑏_𝑡𝑟𝑛𝑠 𝐹𝑘_𝑡𝑟𝑛𝑠 𝑉𝑘_𝑡𝑟𝑛𝑠 𝐿𝑘_𝑡𝑟𝑛𝑠 𝑃𝑅ℎ 𝑃𝑅𝑐 𝑃𝑅𝑝 𝑃𝑅𝑚𝑜𝑏 𝑃𝑅𝑓𝑖𝑥 𝑃𝑅𝑠 𝑃𝑅𝑏 𝑃𝑅𝑡𝑛𝑘 𝐶𝐴𝑃𝑚𝑜𝑏 𝐶𝐴𝑃𝑙 𝐶𝐴𝑃𝑠−𝑜𝑖𝑙 𝛼 P
Annual variable cost for a transportable bio-refinery ($) Annual labor cost for a transportable bio-refinery ($) Annual fixed cost for a fixed bio-refinery ($) Annual variable cost for a fixed bio-refinery ($) Annual labor cost for a fixed bio-refinery ($) Annual fixed cost of biomass storage ($) Annual fixed cost of bio-oil storage ($) Annual fixed cost of a single-trailer truck ($) Annual variable cost of a single-trailer truck ($) Annual labor cost of a single-trailer truck ($) Annual fixed cost of a double-trailer truck ($) Annual variable cost of a double-trailer truck ($) Annual labor cost of a double-trailer truck ($) Annual fixed cost of tanker truck ($) Annual variable cost of tanker truck ($) Annual labor cost of tanker truck ($) Annual utilization rate of a harvester (metric tons) Annual utilization rate of a forwarder (metric tons) Annual utilization rate of a grinder (metric tons) Annual processing rate of a transportable bio-refinery (metric tons) Annual processing rate of a fixed bio-refinery (metric tons) Annual utilization rate of a single-trailer truck (metric tons) Annual utilization rate of a double-trailer truck (metric tons) Annual utilization rate of a tanker truck (metric tons) Annual capacity of a transportable bio-refinery (metric tons) Annual capacity of a fixed bio-refinery (metric tons) Annual capacity of bio-oil storage (metric tons) Annual available amount of biomass (metric tons) Percentage yield of converting biomass to bio-oil
Continuous Variables Amount of biomass transported from site i to site j at time t 𝑋𝑖𝑗𝑡 Amount of biomass transported from site i to site k at time t 𝑋𝑖𝑘𝑡 Amount of biomass transported from site j to site l at time t 𝑋𝑗𝑙𝑡 Integer Variables Amount of bio-oil transported from site k to site s at time t 𝑌𝑘𝑠𝑡 Amount of bio-oil transported from site l to site s at time t 𝑌𝑙𝑠𝑡 Binary Variables
124 𝐵𝑙𝑡 𝐵𝑠𝑡
Binary variable for transportation from site j to site l at time t Binary variable for transportation from site k and l to site s at time t
References [1]
Idaho National Laboratory, 2014, “Feedstock Supply System Design and
Economics for Conversion of Lignocellulosic Biomass to Hydrocarbon Fuels.” [2]
Jones, S., Meyer, P., Snowden-Swan, L., Padmaperuma, A., Tan, E., Dutta, A.,
Jacobson, J., and Cafferty, K., 2013, “Process Design and Economics for the Conversion of Lignocellulosic Biomass to Hydrocarbon Fuels.” [3]
You, F., and Wang, B., 2011, “Life Cycle Optimization of Biomass-to-Liquid
Supply Chains with Distributed–Centralized Processing Networks,” Ind. Eng. Chem. Res., 50(17), pp. 10102–10127. [4]
USEIA, 2014, Annual energy outlook 2014 with projections to 2040, Energy
Information Administration, United States Department of Energy, Washington D.C., USA. [5]
USEIA, 2011, Annual energy outlook 2011 with projections to 2035, Energy
Information Administration, United States Department of Energy, Washington D.C., USA. [6]
Mirkouei, A., Mirzaie, P., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015,
“Reducing the Cost and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains,” J. Clean. Prod. [7]
Zhang, P. F., Pei, Z. J., Wang, D. H., Wu, X. R., Cong, W. L., Zhang, M., and
Deines, T., 2011, “Ultrasonic vibration-assisted pelleting of cellulosic biomass for biofuel manufacturing,” J. Manuf. Sci. Eng., 133(1), p. 011012. [8]
Zhang, M., Song, X., Zhang, P., Pei, Z. J., Deines, T. W., and Wang, D., 2013,
“Size reduction of cellulosic biomass in biofuel manufacturing: separating the confounding effects of particle size and biomass crystallinity,” J. Manuf. Sci. Eng., 135(2), p. 021006. [9]
Mirkouei, A., and Haapala, K. R., 2015, “A Network Model to Optimize
Upstream and Midstream Biomass-to-Bioenergy Supply Chain Costs,” ASME 2015
125 International Manufacturing Science and Engineering Conference (MSEC), MSEC2015-9355, Charlotte, NC. [10]
Mirkouei, A., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015, “A Review
and Future Directions in Techno-Economic Modeling and Optimization of Upstream Forest Biomass to Bio-oil Supply Chains,” Renew. Sustain. Energy Rev. Rev. [11]
Badger, P., Badger, S., Puettmann, M., Steele, P., and Cooper, J., 2010,
“Techno-economic analysis: Preliminary assessment of pyrolysis oil production costs and material energy balance associated with a transportable fast pyrolysis system,” BioResources, 6(1), pp. 34–47. [12]
Baños, R., Manzano-Agugliaro, F., Montoya, F. G., Gil, C., Alcayde, A., and
Gómez, J., 2011, “Optimization methods applied to renewable and sustainable energy: A review,” Renew. Sustain. Energy Rev., 15(4), pp. 1753–1766. [13]
de Lourdes Bravo, M., Naim, M. M., and Potter, A., 2012, “Key issues of the
upstream segment of biofuels supply chain: a qualitative analysis,” Logist. Res., 5(12), pp. 21–31. [14]
Mafakheri, F., and Nasiri, F., 2014, “Modeling of biomass-to-energy supply
chain operations: Applications, challenges and research directions,” Energy Policy, 67, pp. 116–126. [15]
De Meyer, A., Cattrysse, D., Rasinmäki, J., and Van Orshoven, J., 2014,
“Methods to optimise the design and management of biomass-for-bioenergy supply chains: A review,” Renew. Sustain. Energy Rev., 31, pp. 657–670. [16]
Steele, P., Puettmann, M. E., Kanthi Penmetsa, V., and Cooper, J. E., 2012,
“Life-cycle assessment of pyrolysis bio-oil production,” For. Prod. J., 62(4), p. 326. [17]
Ruiz-Femenia, R., Guillén-Gosálbez, G., Jiménez, L., and Caballero, J. A.,
2013, “Multi-objective optimization of environmentally conscious chemical supply chains under demand uncertainty,” Chem. Eng. Sci., 95, pp. 1–11. [18]
Sharma, B., Ingalls, R. G., Jones, C. L., and Khanchi, A., 2013, “Biomass
supply chain design and analysis: Basis, overview, modeling, challenges, and future,” Renew. Sustain. Energy Rev., 24, pp. 608–627.
126 [19]
Mobini, M., Sowlati, T., and Sokhansanj, S., 2011, “Forest biomass supply
logistics for a power plant using the discrete-event simulation approach,” Appl. Energy, 88(4), pp. 1241–1250. [20]
Akgul, O., Zamboni, A., Bezzo, F., Shah, N., and Papageorgiou, L. G., 2010,
“Optimization-based approaches for bioethanol supply chains,” Ind. Eng. Chem. Res., 50(9), pp. 4927–4938. [21]
An, H., Wilhelm, W. E., and Searcy, S. W., 2011, “A mathematical model to
design a lignocellulosic biofuel supply chain system with a case study based on a region in Central Texas,” Bioresour. Technol., 102(17), pp. 7860–7870. [22]
Andersen, F., Iturmendi, F., Espinosa, S., and Diaz, M. S., 2012, “Optimal
design and planning of biodiesel supply chain with land competition,” Comput. Chem. Eng., 47, pp. 170–182. [23]
Geijzendorffer, I. R., Annevelink, E., Elbersen, B. S., Smidt, R. A., and Mol,
R. M. de, 2008, “Application of a GIS-BIOLOCO tool for the design and assessment of biomass delivery chains.” [24]
Mas, M. D., Giarola, S., Zamboni, A., and Bezzo, F., 2010, “Capacity planning
and financial optimization of the bioethanol supply chain under price uncertainty,” Computer Aided Chemical Engineering, S. Pierucci and G. Buzzi Ferraris, ed., Elsevier, pp. 97–102. [25]
Zamora Cristales, R. A., 2013, “Economic optimization of forest biomass
processing and transport.” [26]
Aden, A., Ruth, M., Ibsen, K., Jechura, J., Neeves, K., Sheehan, J., and Wallace,
B., 2002, Lignocellulosic Biomass to Ethanol Process Design and Economics Utilizing Co-current Dilute Acid Prehydrolysis and Enzymatic Hydrolysis for Corn Stover, National Renewable Energy Laboratory. [27]
Phillips, S., and Eggeman, T. J., 2007, Thermochemical ethanol via indirect
gasification and mixed alcohol synthesis of lignocellulosic biomass, Citeseer. [28]
Badger, P. C., and Fransham, P., 2006, “Use of mobile fast pyrolysis plants to
densify biomass and reduce biomass handling costs—A preliminary assessment,” Biomass Bioenergy, 30(4), pp. 321–325.
127 [29]
Kersten, S., and Garcia-Perez, M., 2013, “Recent developments in fast
pyrolysis of ligno-cellulosic materials,” Curr. Opin. Biotechnol., 24(3), pp. 414–420. [30]
ROI, 2003, “Renewable Oil International LLC. Company” [Online]. Available:
http://www.renewableoil.com/Pages/default.aspx. [Accessed: 05-May-2015]. [31]
Polagye, B. L., Hodgson, K. T., and Malte, P. C., 2007, “An economic analysis
of bio-energy options using thinnings from overstocked forests,” Biomass Bioenergy, 31(2–3), pp. 105–125. [32]
Ringer, M., Putsche, V., and Scahil, J., 2006, Large-Scale Pyrolysis Oil
Production: A Technology Assessment and Economic Analysis. Golden (CO): National Renewable Energy Laboratory; 2006 Nov. Report No, NREL/TP-510-37779. Contract No.: DE-AC36-99-GO10337. [33]
Zamora-Cristales, R., and Sessions, J., 2015, “Are double trailers cost effective
for transporting forest biomass on steep terrain?,” Calif. Agric., 69(3), pp. 177–183. [34]
Mirkouei, A., and Haapala, K. R., 2014, “Integration of Machine Learning and
Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process,” 24th International Conference on Flexible Automation and Intelligent Manufacturing (FAIM), May 20-23, 2014, San Antonio, Texas, Flexible Automation and Intelligent Manufacturing. [35]
Cantú-Paz, E., 2003, Genetic and Evolutionary Computation--GECCO 2003:
Genetic and Evolutionary Computation Conference, Chicago, IL, USA, July 12-16, 2003 : Proceedings, Springer Science & Business Media. [36]
Ihaka, R., and Gentleman, R., 1996, “R: a language for data analysis and
graphics,” J. Comput. Graph. Stat., 5(3), pp. 299–314. [37]
Sorenson, C. B., 2010, “A Comparative Financial Analysis of Fast Pyrolysis
Plants in Southwest Oregon,” The University of Montana Missoula, MT. [38]
Flint, B. R., 2013, “Analysis and operational considerations of biomass
extraction on steep terrain in western Oregon.” [39]
Brinker, R. W., Miller, D., Stokes, B. J., and Lanford, B. L., 1989, “Machine
rates for selected forest harvesting machines.” [40]
BLS, 2015, “Producer Price Index (PPI), Bureau of Labor Statistics, United
States Department of Labor.”
128 [41]
Yang, Z., Kumar, A., and Huhnke, R. L., 2015, “Review of recent developments
to improve storage and transportation stability of bio-oil,” Renew. Sustain. Energy Rev., 50, pp. 859–870. [42]
Gurobi Optimization, Inc., 2014, Gurobi Optimizer Reference Manual.
[43]
Van Rossum, G., 2007, “Python Programming Language.,” USENIX Annual
Technical Conference. [44]
ODF, 2015, “State of Oregon: Oregon Department of Forestry - Home”
[Online]. Available: http://www.oregon.gov/odf/Pages/index.aspx. [Accessed: 12-Oct2015].
129
Computational Codes function [ ] = ObjectiveFunctionJMSE( ) %UNTITLED Summary of this function goes here % Detailed explanation goes here global i j k l s t; PeriodIJ(:,:,1)=xlsread('Data.xlsx','XIJ'); Fh= 111255;%fixed harvesting- $/smh Vh= 118032;%variable harvesting- $/pmh Lh= 76464;%labor harvesting - $/smh Fc= 74178;%fixed collection Vc= 101513;%variable collection Lc= 65520;%labor collection Fp= 146482;%fixed grinding - $/h Vp= 508500;%variable grinding Lp= 51705;%labor grinding Fm= 600762;%fixed mobile Vm= 89082;%variable mobile Lm= 375357;% Fl= 4314095;% Vl= 1578218;% Ll= 1155334;% Fsmass= 80798;% Fsoil= 170078;% Fstrns= 24633;%$/h Vstrns= 129520;%$/h Lstrns= 47360;%$/h % Fmtrns= 12;%$/h % Vmtrns= 65; % Lmtrns= 24; Fbtrns= 47536; Vbtrns= 233420;%147620 Lbtrns= 47360; Fktrns= 44074;% Vktrns= 145332;% Lktrns= 47360;% % CAPs=17; % CAPm=25; % CAPb=31; % CAPtnk=25; % URh=1;%0.73; % URc=1;%0.623; % URp=1;%0.68; % % URm=0.88; % % URl=0.9; % URstrns=1;%0.7; % % URmtrns=0.7; % URbtrns=1;%0.7; % URktrns=1;%0.7; PRh=60000; % =25ton*2400hr/yr PRp=37500; % =25ton*1500hr/yr PRmob=1/16425; %=50*328.5 PRfix=1/73000; %=200*365 PRs=28000; %=2000*14 % PRm=1/4200;
130 PRb=1/50000; %=2000*25 PRtnk=1/46000; %=2000*23 C1=(Fh+Lh+Vh); %harvesting C2=(Fc+Lc+Vc);%collecting C3=(Fp+Lp+Vp);%processing C4=(Fm+Lm+Vm);%mobile C5=(Fl+Ll+Vl);%Fixed C6=(Fstrns+Lstrns+Vstrns);%small truck % C7=(Fmtrns+Lmtrns+(Vmtrns*URmtrns));%medium truck C7=(Fstrns+Lstrns+Vstrns);%small truck C8=(Fbtrns+Lbtrns+Vbtrns);%big truck C9=(Fktrns+Lktrns+Vktrns);%tanker
fileID=fopen('ObjectiveFunctionJMSE.txt','w'); fprintf(fileID,'%s\n','# Objective Function');
%C1 fprintf(fileID,'%s ',strcat('m.setObjective(')); Temp=''; for J=1:j JJ=num2str(J); for T=1:t for I=1:i II=num2str(I); TT=num2str(T); C1=num2str(C1); Temp=strcat(Temp,strcat(C1,' * ', num2str((PeriodIJ(I,J,T))/PRh),' + ')); if I==i && J==j && T==t end end end end fprintf(fileID,'%s\n',Temp);
Temp=''; for J=1:j JJ=num2str(J); for T=1:t for I=1:i II=num2str(I); TT=num2str(T); C2=num2str(C2); Temp=strcat(Temp,strcat(C2,' * ', num2str((PeriodIJ(I,J,T))/PRh),' + ')); if I==i && J==j && T==t end end end
131 end fprintf(fileID,'%s\n',Temp);
Temp=''; for J=1:j JJ=num2str(J); for T=1:t for I=1:i II=num2str(I); TT=num2str(T); C3=num2str(C3); Temp=strcat(Temp,strcat(C3,' * ', num2str((PeriodIJ(I,J,T))/PRp),' + ')); if I==i && J==j && T==t end end end end fprintf(fileID,'%s\n',Temp);
Temp=''; for T=1:t TT=num2str(T); for K=1:k KK=num2str(K); for J=1:j JJ=num2str(J); C4=num2str(C4); PRmob=num2str(PRmob); Temp=strcat(Temp,strcat(C4,'*XJK', JJ, '_', KK, '_', TT,' * ',PRmob,' + ')); end end fprintf(fileID,'%s\n',Temp); end
Temp = ' '; for T=1:t TT=num2str(T); for L=1:l LL=num2str(L); for J=1:j JJ=num2str(J); C5=num2str(C5); PRfix=num2str(PRfix); Temp=strcat(Temp,strcat(C5,'*XJL', JJ, '_', LL, '_', TT,' * ',PRfix,' + ')); end end
132 fprintf(fileID,'%s\n',Temp); end
%C6 for L=1:l LL=num2str(L); for T=1:t TT=num2str(T); Fsmass=num2str(Fsmass); fprintf(fileID,'%s \n', strcat(Fsmass,' * ', 'BJL',LL,'_',TT,' + ')); end end
%C7 for S=1:s SS=num2str(S); for T=1:t TT=num2str(T); Fsoil=num2str(Fsoil); fprintf(fileID,'%s \n', strcat(Fsoil,' * ', 'BS',SS,'_',TT,' + ')); end end Temp=''; for J=1:j JJ=num2str(J); for T=1:t for I=1:i II=num2str(I); TT=num2str(T); C6=num2str(C6); Temp=strcat(Temp,strcat(C6,' * ', num2str((PeriodIJ(I,J,T))/PRs),' + ')); if I==i && J==j && T==t end end end end fprintf(fileID,'%s\n',Temp); Temp = ' '; for T=1:t TT=num2str(T); for K=1:k KK=num2str(K); for J=1:j JJ=num2str(J); C7=num2str(C7);
133 PRs=num2str(PRs); Temp=strcat(Temp,strcat(C7,'*XJK', JJ, '_', KK, '_', TT,' * 1/',PRs,' + ')); end end end fprintf(fileID,'%s\n',Temp);
Temp = ' '; for T=1:t TT=num2str(T); for L=1:l LL=num2str(L); for J=1:j JJ=num2str(J); C8=num2str(C8); PRb=num2str(PRb); Temp=strcat(Temp,strcat(C8,'*XJL', JJ, '_', LL, '_', TT,' * ',PRb,' + ')); end end end fprintf(fileID,'%s\n',Temp);
Temp=''; for T=1:t TT=num2str(T); for S=1:s SS=num2str(S); for K=1:k KK=num2str(K); C9=num2str(C9); PRtnk=num2str(PRtnk); Temp=strcat(Temp,strcat(C9,'*YKS', KK, '_', SS, '_', TT,' * ',PRtnk,' + ')); end end end fprintf(fileID,'%s\n',Temp(1:end-1));
fprintf(fileID,'%s\n',', GRB.MINIMIZE)'); fprintf(fileID,'%s\n','# Optimize'); fprintf(fileID,'%s\n','m.optimize()'); fprintf(fileID,'%s\n','m.objVal'); fprintf(fileID,'%s','m.printAttr(','"x"',')'); end
134
CHAPTER 4: EVOLUTIONARY DECISION MAKING FOR BIOMASS-BASED ENERGY SUPPLY CHAINS
To be Submitted to Environmental Science and Technology
135
Chapter 4:
Evolutionary Decision Making for Biomass-based Energy Supply Chains
Abstract Bioenergy sources (e.g., biomass-based energy) have been introduced as a means to address environmental, energy security, and human health challenges attendant with conventional energy sources (e.g., fossil-based energy) over the past few decades. Societal and government interest in bioenergy has put additional scrutiny on feedstock supply and logistics, systems analysis, and system integration, as well as cross-cutting sustainability. Moreover, the uncertainties of externalities represent the key challenges in bioenergy supply chains and lead to reduced robustness and competitiveness. This study aims to develop a multi-criteria decision making method capable of improving sustainability performance in biomass-based energy supply chains. Thus, the impacts of a mixed supply chain, when uses mixed-mode bio-refineries and mixed-pathway transportation, can be explored along with assessing the impact of real-world uncertainties. The method developed couples economic and environmental impact analyses. The economic analysis employs a support vector machine method to predict the pattern of uncertainty parameters and a stochastic optimization model to incorporate uncertainties (i.e., quality and accessibility) into the model. The model minimizes the total annual cost (i.e., fixed, variable, and labor costs) of a biomass-based energy supply chain network by using a genetic algorithm. The environmental impact analysis employs a life cycle assessment method to determine global warming potential for the proposed mixed supply chain with the assistance of two software tools (SimaPro 8 and
136 GREET 2015). A case study for the Pacific Northwest is used to demonstrate the application of this method and to verify the accuracy of method and models. The results indicate that the proposed mixed supply chain can improve sustainability performance of traditional supply chain, by reducing the annual cost (by ~ 24 %) and environmental impacts (by ~ 5 %).
Introduction Sustainability has attracted increased attention of the industrial, government, and academic research communities. Since being highlighted as a global concern three decades ago by the Brundtland Commission: “Sustainable energy requires meeting today’s energy needs without compromising future generation’s ability to meet their energy needs” [1]. For instance, concern over fossil fuel consumption and its direct relationship with greenhouse gas (GHG) emissions and environmental impacts have simulated intensive research in alternative energy technology development. Renewable energy has been suggested as an environmentally friendly form of energy [2]. For instance, biomass is a renewable energy resource that can be harvested and replenished indefinitely. However, as an emerging energy technology, its sustainability performance should be evaluated by considering the economic, environmental, and social aspects across the biomass-based energy supply chain (BESC). Decision makers (e.g., policy makers and investors) can take steps to avoid the impacts of fossil fuelbased GHG emissions by promoting the use of biomass-based energy sources (e.g., biofuel, bio-oil, and biochar).
137 Designing a robust and sustainable BESC network requires addressing the uncertainties of externalities, such as supply, logistics, quality, delivery, production, and price. Several methods have been introduced in the literature to incorporate uncertainties in supply chain (SC) optimization, such as analytical methods (e.g., stochastic programming and Markov decision process) and simulation methods (e.g., discrete event simulation and Monte Carlo).
The overarching objective of the research presented here is to ensure energy security in the dynamic and competitive energy market by producing competitive bioenergy from forest biomass resources. In particular, this study aims to explore the feasibility of biomass-based energy production, using the proposed mixed SC instead of a conventional SC. Conventional BESCs have three main segments: upstream, midstream, and downstream (Fig. 4.1). The focus of this study is on upstream and midstream segments of BESC, which involve collection, transportation, pre-treatment, conversion, and short term storage. The conversion process converts biomass into denser energy carriers (e.g., bio-oil and biochar) that ease handling, transportation, and storage.
Study Scope Collection & Staging
Transportation & Storage
Upstream
Pre-treatment
Conversion
Midstream
Distribution
End-use
Downstream
Figure 4.1. Conventional biomass-based energy supply chain entities
138
An evolutionary decision making method is developed to simultaneously evaluate the objectives of cost and environmental impact along with assessing the role of network uncertainties and the mixed SC in supporting broader biomass-based energy commercialization. The proposed mixed SC includes mixed-mode bio-refineries (i.e., mobile and fixed) and mixed-pathway transportation (i.e., new and traditional) (Fig. 4.2). The proposed decision making method integrates two phases: stochastic optimization modeling to minimize cost and life cycle assessment to determine associated global warming potential.
Traditional Pathway ?
? Question marks indicate location and asset decision points
Biomass Transferring
Grinding
? Fixed Bio-refining
Collection
?
Bio-oil Transferring
New Pathway
Mobile Bio-refining
Figure 4.2. A mixed biomass-to-bio-oil supply chain (mobile bio-refinery image courtesy Phillip C. Badger; fixed bio-refinery image courtesy UPM Lappeenranta Bio-refinery, Finland)
The stochastic optimization model aims to explore the commercial feasibility of biooil production by minimizing the annual BESC cost, which includes collection, grinding, pyrolysis, and short-term storage. The model incorporates supply
139 uncertainties in the upstream BESC segment by using two stochastic constraints, as other objectives, to assess the effects of uncertainties on bio-oil economic feasibility, as well as other constraints. Moreover, the model has uncertainty parameters that are estimated through machine learning technique (i.e., support vector machine). A metaheuristic computational algorithm is developed to solve the stochastic model using a genetic algorithm as an artificial intelligence technique with assistance of the MATLAB computational language.
The life cycle assessment (LCA) method is used to assess environmental impacts, i.e., global warming potential (GWP) throughout the bio-products’ (i.e., bio-oil and biochar) life cycle. The LCA study applies a cradle-to-grave system boundary, including the upstream (collection, staging, and grinding), midstream (drying and pyrolysis), and downstream (distribution and combustion) BESC segments. The LCA method includes three steps: Goal and scope definition, life cycle inventory (LCI), and life cycle impact assessment (LCIA). Commercial LCA software packages, i.e., SimaPro 8 and GREET 2015 [3], [4] are used to estimate emission factors for biomass collection (including collection, staging, and size reduction), transportation (using truck and tanker), and bio-oil production. Prior studies (e.g., Steele et al. 2011 [2] and Johnson et al. 2010 [5]) were used to determine emission factors bio-oil combustion. Finally, environmental impact assessment, using GWP, was conducted using actual data for a case study in the Pacific Northwest.
140 Background Over the past few decades, renewable energy sources have been introduced as a promising replacement for conventional energy, due to potential environmental and social benefits. According to the United States (U.S.) Department of Energy, around 10% of total U.S. energy consumption is provided from renewable energy resources and half of this renewable energy demand is provided through biomass [2]. Biomass is an energy resource produced from natural materials, such as forest resources, energy crops, algae, agricultural residues, and wastes (i.e., urban waste, wood waste, and food waste) [5]. Renewable energy (e.g., biomass-based energy) has several benefits in comparison with conventional energy sources (e.g., petroleum and coal), which include [6]: 1) reduced dependency on imported energy, 2) reduced GHG emissions, 3) improved forest management, 4) reduced poverty potential, especially in rural areas, 5) improved social resources (e.g., water quality), 6) promotion of carbon cycling and biodiversity, and 7) improved human health.
Biomass feedstocks and wastes are distributed across the U.S. and represent its largest renewable energy resource. The predicted available biomass was 602-1,009 million dry tons per year by 2022, the available amount of biomass can meet over one-third of current transportation fuel demands [5]. Thus, biomass can play a key role in the energy industry due to its abundance and low cost. Biomass resources can be classified as lignocellulose, triglycerides, amorphous sugars, and starches [7]. In 2015, the U.S. Department of Agriculture, in collaboration with the U.S. Department of Energy, announced nearly nine million dollars in funding through the Biomass Research and
141 Development Initiative, which would aim to mitigate dependence on foreign oil by supporting the development of biomass-based energy [8]. The main topics of interest included [9] 1) feedstock supply and logistics, 2) thermochemical production processes, 3) integrated bio-refineries, 4) system analysis and integration, and 5) crosscutting sustainability, all of which are in focus here.
As mentioned above, the use of biomass-based energy has the potential to mitigate GHG emissions from fossil-based energy and related impacts. GHG emissions are often expressed as CO2 equivalent emissions, and include several substances, e.g., carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O). Environmentally responsible bio-resource stewardship supports mitigation of GHG emissions, as well as soil degradation, poor water quality, and biodiversity loss [10]. In fact, renewable energy sources, especially biomass, can be carbon neutral. The carbon released into the environment during forest biomass combustion is nearly equal to the carbon absorbed from atmosphere during the growth life of the tree (via photosynthesis) [2].
The use of biomass is inherently limited due to economic factors, such as low energy density and high collection and transportation cost, as well as land use competition; which is why only 45% of available biomass is currently used [5]. The transition from conventional energy to biomass-based energy solutions presents numerous challenges in the traditional and competitive energy market. Ongoing research and innovation is pursuing technological and research breakthroughs for bio-refinery capital cost reduction and process scale-up for commercialization. Most forest harvest residues
142 (FHRs) are underutilized and wasted due to logistical challenges (e.g., high collection, transportation, and storage costs), and low market values [7]. Challenges have also been reported in the literature due to the nature of the BESC [11]. A major BESC challenge is network uncertainties, including uncertainty in supply, logistics, production and yield, and demand and distribution. In order to overcome the existing BESC challenges, development of robust and reliable SCs that incorporate knowledge of the sources of uncertainty is essential. In particular, developing a sustainable and optimal BESC is an expedient short-term solution to tradeoff between cost and environmental impacts.
Optimal SC planning aids decision makers (e.g., investors in energy industry and policy makers in government) to ensure the efficiency and effectiveness of the material and information flow. In the competitive energy market, focus should be on minimizing total cost (or maximizing profit) and environmental impacts to achieve robust, sustainable, and reliable SCs. The major decision making approaches in SC management are strategic (long-term decisions of five years or more), tactical (medium-term decisions of 6-12 months), and operational (short-term decisions that occur daily/weekly). This study focuses on a tactical decision making approach, to support sourcing (e.g., collection site) and logistical decisions (e.g., facility location) in timely and cost effective manner. Some of the decisions are selected at the beginning of SC planning (e.g., raw material type and production technologies) because these decisions will not be changed in a short- or medium-term period. Biomass sourcing decisions play a key role in BESC development through efficient collection, transporting, and pre-treating of biomass to reduce costs and environmental impacts.
143
Since uncertainties affect SC performance, decision makers must incorporate them into their decisions. Stochastic SC modeling is a quantitative method for developing optimization models that can manage uncertainty sources [11]. In contrast to deterministic optimization models that have known parameters, stochastic optimization models incorporate unknown model parameters to consider real-world uncertainties. Stochastic modeling is an analytical method for decision making under a stochastic environment [5], and takes advantage of probability distributions and historical data. One of the challenges of stochastic programming is designing a reliable computational algorithm to effectively solve the model. The existing methods to incorporate uncertainties in SCs include analytical methods (e.g., stochastic mathematical programming and Markov decision process) and simulation methods (e.g., Monte Carlo and discrete event simulation). More details about modeling uncertainties in SCs and stochastic programming optimization in bioenergy SCs have been previously reported [11], [12].
Since the existing uncertainties add complexities to decision making in energy industries, a method for managing uncertainty sources is vital during BESC development [12]. Without considering uncertainty parameters, the results of SC models may be suboptimal or even infeasible, and, subsequently, will be unreliable [12]. The main uncertainty sources in upstream and midstream BESC segments are limited to 1) biomass supply (e.g., collection and staging), 2) logistics (e.g., transportation and storage), 3) pre-treatment (e.g., drying and size reduction), and 4)
144 manufacturing processes (e.g., conversion technology and production processes). This study explores the effects of uncertainty in biomass supply by using a stochastic optimization model and genetic algorithm (GA) technique. GA is an evolutionary search heuristic, which has several evolutionary phases (e.g., initialization, mutation, crossover, and tournament selection) to find an optimal solution.
Uncertainty in biomass supply exists due to the characteristics of this energy resource. Low energy density of forest-based biomass, high moisture content, and variations in quality and availability represent supply uncertainties in upstream forest biomass SCs. These factors, especially biomass quality, affect the net profit of bioenergy production in different seasons. Seasonality has huge impact on ensuring a continuous supply at a desired moisture content since underutilized forest biomass may not be obtainable throughout the year, thus decision makers need to consider multiple sources of biomass to ensure supply chain robustness. Logistics uncertainties include facility location, transportation lead time (i.e., transferring biomass in timely and cost effective manner), and transportation cost. Pre-treatment uncertainties include equipment costs (e.g., grinder, chipper, and dryer) and unexpected equipment breakdowns. Manufacturing process uncertainties include production yield and machine breakdowns. A detailed overview of upstream and midstream BESC segments has been provided previously [7]. A brief overview is provided below for the technologies relevant to this study.
This work strives to address the mentioned gap to evaluate the competitive potential of bio-oil. The mobile bio-refinery is a trailer-mounted unit that uses pyrolysis technology
145 to convert underutilized forest biomass to bio-products [7]. The transportable biorefinery operates using bio-char (as heat energy) in place of fossil fuels (e.g., diesel and gasoline) or electricity [13]. The major benefit of a mobile bio-refinery is its ability to produce higher energy density, intermediate products (e.g., bio-oil and biochar) in close proximity to raw materials (e.g., forests or farms) [14]. This alleviates transferring low energy density biomass to centralized bio-refineries, which can improve economic and environmental performance of bio-oil production by reducing transportation, handling, and storage operations [15].
Pyrolysis conversion technology is thermochemical decomposition of biomass at higher temperature (300-550°C) in the absence of oxygen (Fig. 4.3). Three types of pyrolysis have been introduced in the literature: slow, intermediate, and fast processes. Each type has a specific reaction time and temperature, and production yield. The main products include pyrolysis oil (bio-oil), pyrolysis char (bio-char), and non-condensable gases (syngas), of which bio-oil and bio-char are target products.
Heat from Syngas
Biomass
Size Reduction
Product
Drying
Pyrolysis Reactor
Heat from Biochar
Solid Separation
Gas/Liquid Separation
Biochar
Bio-oil
Pre-processing Process
Figure 4.3. Pyrolysis conversion process principles
146 Bio-oil is mainly produced from forest biomass through a pyrolysis process that condenses a mixture of water and oxygenated carbon. Bio-oil includes carbon, hydrogen, oxygen, nitrogen, and ash, which are 50, 6, 41, 0.17, and 0.92 percent per weight, respectively. Bio-oil has an approximate density of 1.2 kg per liter, and 18 MJ/kg higher heating value (HHV) energy content [2]. Major applications of bio-oil include upgrading to transportation fuel, use in chemical production, and combustion in boilers, engines, and turbines [16]. Bio-oil has also been used to produce electricity, but it requires industrial advancements to be commercialized.
Prior techno-economic studies of bio-oil SCs indicate that the interest has been growing in bioenergy industry, especially due to the potential for near carbon neutrality, or no net carbon release into the environment, in comparison with heating oil and fuel oil No. 6. The results of prior studies indicate that biomass moisture content, biomass chip size, and bio-refinery size have direct impacts on economic and environmental aspects of bio-oil SCs [7], [17]. LCA studies of BESC indicate that there is a strong dependency of results on system boundary, allocation method, and functional unit [18]. For instance, a system that maximizes the GHG emission reductions per unit of energy is not able to achieve the highest GHG reduction per unit of biomass. Steele et al. reported that CO2 and steam emissions make up the majority of emissions released across all bio-oil life-cycle stages, accounting for 67% and 32% of the total emissions by mass, respectively [2]. Additionally, they reported that 1 MJ of bio-oil production releases a total of 0.19 kg CO2 equivalent cradle-to-gate emissions [2].
147 Prior studies inconsistently addressed the triple bottom line (i.e., economic, environmental, and social) in BESCs [19]. No studies have been found that simultaneously evaluate the objectives of cost and environmental impact along with assessing the role of network uncertainties and mixed-mode bio-refineries (i.e., mobile and fixed) to support broader bioenergy commercialization. In the proposed mixed SC, the locations of mobile bio-refineries have been selected close to forest collection sites to reduce the number of truck trips needed to transport biomass feedstocks. Fewer truck trips not only reduce fuel consumption, but also mitigates GHG emissions and can aid bioenergy commercialization by lowering transportation costs. Traditionally, to reduce the number of truck trips, different types of trucks have been considered for in-forest roads and highways. For instance, higher-capacity, double-trailer trucks can be used on main highways, while in-forest road travel is restricted to single-trailer trucks, which have more flexibility (e.g., in tight curves) and can turn around when the road is narrow.
Since utilization of bioenergy aims to accommodate environmental pressures, new bioenergy technologies need to be assessed in terms of environmental impacts. The LCA method can be used to evaluate the environmental impacts of the SC system and assist in defining more environmentally responsible networks. This study applies an LCA method to quantify resource consumption (e.g., biomass and fossil fuel), as well as emissions, to determine environmental impacts in terms of GWP for a mixed BESC. Impacts are then compared to those for a traditional BESC. The methodology developed and a case study undertaken as a demonstration are described in the next section.
148 Methodology As discussed above, various methods have been developed to incorporate uncertainties and environmental impacts in SC management and logistics planning. There are limited studies in the BESC domain, however, that have considered uncertainties and life cycle assessment in quantitative analysis techniques. The methodology here presents a multicriteria decision making method, including two phases for economic analysis and environmental impact analysis (Fig. 4.4). Phase 1 integrates two quantitative analyses (i.e., support vector machine and stochastic modeling) into a decision support system to assess the role of uncertainty sources in BESCs. Phase 2 applies the LCA method, which includes the following steps: 1) defining the goal and scope to set the context of study, 2) compiling a life cycle inventory (LCI) to quantify materials and energy flows in the system, and 3) conducting a life cycle impact assessment (LCIA) to translate LCI data into environmental impact metrics. Phases 1 and 2 are described in more detail below.
Phase 2: Environmental Impact Analysis
Phase 1: Economic Analysis Support Vector Machine
Stochastic Optimization
Life Cycle Goal & Scope Life Cycle Impact Inventory Analysis Definition Assessment
Multi-criteria Decision Making
Figure 4.4. Decision support system for sustainable bioenergy production
149 Economic Analysis (Phase 1) The support vector machine (SVM) method is applied in this study to provide a learning algorithm for pattern recognition of uncertainty parameters (i.e., average biomass availability and average biomass quality in this study). SVM is a supervised method in machine learning, which uses historical data to predict the future trend of parameters of interests. The solution found using SVM is always unique and globally optimal equivalent to solving a linear constrained quadratic programming problem [20], [21]. Quality rate is defined based on ash content (%wt. ash) and moisture content (%wt. H2O). Ash content is the fraction of solid waste remaining after complete combustion process of biomass. It's also known as a percentage weight of the dry base or as a percentage weight of the as received material. High ash content of the biomass generally indicates a reduced heating value. The ash content of forest biomass ranges from 0.08 to 2.3%. In this study, the accessibility rate depends on the available amount of biomass and distance to the staging site.
A dataset of 50 collection sites provides the quality rate, calorific value, available amount of biomass, and distance to the staging site for historical data. In a real case, dataset would be obtained from the land manager, but it is randomly generated for this study. The outputs define the quality rate and accessibility rate for each collection site. Table 4.1 shows an example evaluation of several collection sites, and includes the identified key criteria. These data form the training set, which contains input and output information.
150 A training set along with a testing set are used to train and test the SVM. The provided dataset contains values for 40 sites as a training dataset and 10 sites as a testing dataset. Table 4.2 shows SVM can find the pattern and the weights from the training information in the training phase [22]. Support vectors recognize this pattern of the data, which define the weight for each criterion.
Table 4.1. Example of training set for quality and accessibility of harvesting sites Site No. 1 2 3
Moisture Content (% H2O)
0.90 1.20 0.85
55 60 50
1300 1500 1000
0.50 0.50 0.70 ?
Accessibility Rate Output
9 5 2
0.60 0.80 0.80
?
…
57 60 52 ?
Distance To Staging Site (Mile)
…
?
0.60 0.40 0.90
Available Amount of Biomass (Metric Tons) 1100 1400 1200
…
0.95 1.10 0.90
Quality Rate Output
…
…
…
… 38 39 40 Weight
Ash Content (%)
8 4 9 ?
0.70 0.90 0.50 ?
Table 4.2. Identified weights from the support vector machine method (R results) Site No. Weight
Ash Content (%)
Moisture Content (% H2O)
0.20
2.43
Quality Rate Output 36.34
Available Amount of Biomass (Metric Tons) 3.47
Distance To Staging Site (Mile)
Accessibility Rate Output
2.83
29.51
The weights found in the training phase can be used to determine the outputs (i.e., quality and accessibility rates) for the testing set (Table 4.3). These can then be compared them with actual outputs to check the failure rate. The quality and accessibility rates of sites can be predicted after defining the weights for each criterion and used to classify the qualified and unqualified collection sites, as well as finding the
151 average quality and accessibility rates and apply them in stochastic model (quality and accessibility constraints).
Table 4.3. Example of testing set of harvesting site
41 42 43
Ash Content (%) 0.85 1.20 1.1
Moisture Content (% H2O) 50 59 60
Quality Rate Output ? ? ?
Available Amount of Biomass (Metric Tons) 900 1200 1300
Distance To Staging Site (Mile) 2 10 9
Accessibility Rate Output ? ? ?
…
…
…
…
…
…
…
Site No.
50
0.90
60
?
600
11
?
After recognizing the trend of uncertainty parameters, a stochastic optimization model is formulated to optimize the total cost of BESC over a year time horizon. The stochastic optimization model is applied in this phase as an analytical method to explore the commercial feasibility of bio-oil production. The objective function (Eq. 1) to minimize the total annual cost (TC) of BESC, includes harvesting, collection, transportation, pre-treatment, conversion, and short-term storage. The model encompasses two other objectives that are defined as stochastic constraints (Eqs. 2-3) to assess the role of uncertainties. Biomass quality and accessibility are highly influential and variable uncertainties in biomass supply - considering and incorporating these parameters is essential for establishing sustainable and reliable bioenergy. The model considers the following other constraints: capacity (Eqs. 4-6), harvesting base (number of harvesting sites) as specified by decision makers (Eq. 7), production yield (Eq 8), conservation flows (Eqs. 9-11), the annual available biomass to be processed (Eq. 12), and non-negativity, binary, and integer constraints to guarantee a feasible
152 solution (Eqs. 13-15). Notations of model indices, parameters, and variables are provided in the nomenclature, below. 𝑀𝑖𝑛 𝑇𝐶 = ∑ ∑ ∑ 𝐹𝑐 ∗ 𝐵𝑖𝑗 + (𝐿𝑐 + 𝑉𝑐 ) ∗ 𝑖
𝑗
𝑡
𝑋𝑖𝑗𝑡 𝑈𝑐
+ ∑ ∑ ∑ 𝐹𝑝 ∗ 𝐵𝑖𝑗 + (𝐿𝑝 + 𝑉𝑝 ) ∗ 𝑖
𝑗
𝑡
+ ∑ ∑ ∑(𝐹𝑚 + 𝐿𝑚 + 𝑉𝑚 ) ∗ 𝑗
𝑘
𝑡
𝑖
𝑗
𝑡
(4.1)
𝑋𝑖𝑗𝑡 𝑈𝑝
𝑋𝑗𝑘𝑡 𝑋𝑗𝑙𝑡 + ∑ ∑ ∑(𝐹𝑓 + 𝐿𝑓 + 𝑉𝑓 ) ∗ 𝑈𝑚 𝑈𝑓 𝑗
𝑙
𝑡
𝑗
𝑘
𝑡
𝑋𝑖𝑗𝑡 𝑋𝑗𝑘𝑡 + ∑ ∑ ∑(𝐹𝑠𝑡 + 𝐿𝑠𝑡 + 𝑉𝑠𝑡 ) ∗ + ∑ ∑ ∑(𝐹𝑠𝑡 + 𝐿𝑠𝑡 + 𝑉𝑠𝑡 ) ∗ + 𝑈𝑠𝑡 𝑈𝑠𝑡 𝑋𝑗𝑙𝑡 𝑌𝑘𝑠𝑡 ∑ ∑ ∑(𝐹𝑑𝑡 + 𝐿𝑑𝑡 + 𝑉𝑑𝑡 ) ∗ + ∑ ∑ ∑(𝐹𝑡𝑡 + 𝐿𝑡𝑡 + 𝑉𝑡𝑡 ) ∗ 𝑈𝑑𝑡 𝑈𝑡𝑡 𝑗
𝑙
𝑡
𝑘
𝑠
𝑡
St: 1⁄2 2 ∑ ∑ 𝐵𝑖𝑗𝑡 ∗ 𝜇𝑖𝑗𝑡 + (∑ ∑ 𝐵𝑖𝑗𝑡 ∗ 𝜎𝑖𝑗𝑡 ) 𝑖∈𝐼 𝑡∈𝑇
∀𝑖 ∈ 𝐼, ∀𝑡 ∈ 𝑇 𝛼 ∈ [0,1]
(4.2)
∗ 𝜑 −1 (1
∀𝑖 ∈ 𝐼, ∀𝑡 ∈ 𝑇 𝛽 ∈ [0,1]
(4.3)
𝑖∈𝐼 𝑡∈𝑇
− 𝛼) ≤ 𝑄𝑡
1⁄2
2 ∑ ∑ 𝐵𝑖𝑗𝑡 ∗ 𝜇𝑖𝑗𝑡 + (∑ ∑ 𝐵𝑖𝑗𝑡 ∗ 𝜎𝑖𝑗𝑡 ) 𝑖∈𝐼 𝑡∈𝑇
∗ 𝜑 −1 (1
𝑖∈𝐼 𝑡∈𝑇
− 𝛽) ≤ 𝐴𝑡
𝑋𝑖𝑗𝑡 ≤ 𝐶𝑎𝑝𝑖 ∗ 𝐵𝑖𝑗𝑡
∀𝑖 ∈ 𝐼, ∀𝑗 ∈ 𝐽, ∀𝑡 ∈ 𝑇 (4.4)
∑ ∑ 𝑋𝑗𝑘𝑡 ≤ 𝐶𝑎𝑝𝑚
∀𝑗 ∈ 𝐽 , ∀𝑡 ∈ 𝑇
(4.5)
∀𝑗 ∈ 𝐽 , ∀𝑡 ∈ 𝑇
(4.6)
∀𝑖 ∈ 𝐼, ∀𝑡 ∈ 𝑇
(4.7)
𝑃𝑌 = Weight * 𝑄𝑡
∀𝑡 ∈ 𝑇
(4.8)
∑ ∑ 𝑋𝑖𝑗𝑡 − ∑ ∑ 𝑋𝑗𝑙𝑡 − ∑ ∑ 𝑋𝑗𝑘𝑡 = 0
∀𝑖 ∈ 𝐼, ∀𝑙 ∈ 𝐿, ∀𝑡 ∈𝑇
(4.9)
∀𝑗 ∈ 𝐽, ∀𝑠 ∈ 𝑆, ∀𝑡 ∈𝑇
(4.10)
𝑗∈𝐽 𝑡∈𝑇
∑ ∑ 𝑋𝑗𝑙𝑡 ≤ 𝐶𝑎𝑝𝑓 𝑗∈𝐽 𝑡∈𝑇
∑ ∑ 𝐵𝑖𝑗𝑡 ≤ 𝑁𝑡 𝑖∈𝐼 𝑡∈𝑇
𝑖∈𝐼 𝑡∈𝑇
𝑙∈𝐿 𝑡∈𝑇
𝑗∈𝑗 𝑡∈𝑇
𝑃𝑌 ∗ ∑ ∑ 𝑋𝑗𝑘𝑡 − ∑ ∑ 𝑌𝑘𝑠𝑡 = 0 𝑗∈𝑗 𝑡∈𝑇
𝑠∈𝑆 𝑡∈𝑇
153 𝑃𝑌 ∗ ∑ ∑ 𝑋𝑗𝑙𝑡 − ∑ ∑ 𝑌𝑙𝑠𝑡 = 0 𝑗∈𝐽 𝑡∈𝑇
𝑠∈𝑆 𝑡∈𝑇
∑ ∑ ∑ 𝑋𝑖𝑗𝑡 ≥ 𝜃𝑡 𝑖∈𝐼 𝑗∈𝐽 𝑡∈𝑇
∀𝑗 ∈ 𝐽, ∀𝑠 ∈ 𝑆, ∀𝑡 ∈𝑇
(4.11)
∀𝑖 ∈ 𝐼, ∀𝑗 ∈ 𝐽, ∀𝑡 ∈ 𝑇 (4.12)
𝑋𝑖𝑗𝑡 , 𝑋𝑗𝑘𝑡 , 𝑋𝑗𝑙𝑡 ≥ 0
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖, 𝑗, 𝑘, 𝑙, 𝑎𝑛𝑑 𝑡 (4.13)
𝐵𝑖𝑗𝑡 = {0, 1}
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖, 𝑗 𝑎𝑛𝑑 𝑡
(4.14)
𝑌𝑘𝑠𝑡 , 𝑌𝑙𝑠𝑡 are integer
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑘, 𝑙, 𝑠, 𝑎𝑛𝑑 𝑡
(4.15)
Environmental Impact Analysis (Phase 2) This section presents the environmental impact analysis approach for the production of bio-oil from biomass, including supply chain (e.g., collection, transportation, storage, and production) effects. The scope of the study includes three life cycle stages: 1) upstream processing (i.e., biomass collection, staging, pre-processing, and transportation), 2) midstream processing (i.e., pre-treatment and conversion to bio-oil), and 3) downstream processing (i.e., distribution and combustion). Thus, the LCA study applies a cradle-to-grave system boundary for the bio-oil SC. The functional unit selected for the study is 1 gallon (3.78 liters) of bio-oil. The LCA is conducted using data and information from prior studies and LCA software (GREET 2015 and SimaPro 8). GREET 2015 provided data required for exploring collection and transportation environmental impacts. SimaPro 8 provided data for bio-oil production and bio-oil combustion environmental impacts. The life cycle impact assessment method selected is Global Warming Potential (GWP), a single indicator method that assesses the total GHG emissions. The GWP emission factor (in kg CO2 equivalent (eq.)) is calculated using a 20-year time horizon, where the emission rates for CO2, CH4, and N2O are RCO2
154 (1 kg CO2 eq./kg CO2), RCH4 (56 kg CO2 eq./kg CH4), and RN2O (280 kg CO2 eq./kg N2O), respectively [20].
LCIAs are conducted and compared for the traditional and new (mixed) supply chain pathway for biomass-to-bio-oil production. In the long term, biomass-based energy products can reduce GHG emissions because the released CO2 can be recaptured as part of the natural process; the forest absorbs carbon from the atmosphere and convert it to biomass. The CO2 released during fossil-based energy combustion (e.g., in equipment and trucks) is not part of the natural process. Equipment type and use characteristics are dependent upon the type of biomass utilized. The upstream forest biomass SC typically uses a harvester, forwarder, grinder, and loader. The process inputs are biomass- and fossil-based energy (mainly diesel fuel) and lubricants, and the outputs are ground biomass (chips) and GHG emissions from equipment operation. The GWP emission factor for biomass collection (EFup), which includes biomass collection and grinding, in kg CO2 eq. per metric ton of biomass, and the environmental impacts (GWP) of the upstream segment (Gup) is calculated using Eqs. 16 and 17. Variables are defined in the Nomenclature section.
EFup=RCO2*EFupCO2+RCH4*EFupCH4+RN2O*EFupN2O
(4.16)
Gup=Available Biomass*EFup
(4.17)
The next step is to transfer the chips to the bio-refinery using different types of trucks with various capacities. The inputs include chips and transportation fuel, and the
155 outputs are transferred chips and GHG emissions from fuel combustion. Factors such as biomass quality and moisture content have a direct effect on GWP. Transferring high quality biomass with lower moisture content can reduce the number of truck trips to produce the same amount of bio-oil and, consequently, reduce environmental impacts. Travel distance and truck weight also directly affect fuel consumption. A decision making method is applied using a Geographical Information System (GIS)-based approach to define the shortest path between sites (e.g., collection site to a transportable bio-refinery or fixed bio-refinery location). The GWP of biomass transportation (Gmass) is quantified using GREET 2015. The emission factor (EFmass) and GWP of transferring biomass (Gmass), are calculated using Eqs. 18 and 19.
EFmass=RCO2*EFmassCO2+RCH4*EFmassCH4+RN2O*EFmassN2O
(4.18)
Gmass=Processed Biomass*EFmass*Distance
(4.19)
In the midstream SC segment, biomass moisture content and particle size determine the type of pre-treatment equipment and conversion technology. The input is feedstock and the outputs are biochar, bio-oil, and GHG emissions. The emissions include steam released and biogenic GHG emissions during the chip drying and production process, respectively. Biogenic CO2 emissions are a part of the carbon cycle that encompasses photosynthesis, whereby atmospheric CO2 is uptake by other growing plants and trees. The heat for feedstock drying and pyrolysis in the new, mixed pathway is provided by the biochar and syngas produced during pyrolysis. Biochar and syngas are funneled directly to the furnace (i.e., self-generated and closed-loop process).
156
Biomass particle size and required process temperature play a key role in determining the pyrolysis type (e.g., fast, intermediate, or slow). Fast pyrolysis has a higher production yield 350-500°C, but requires a high process temperature and small feedstock particle size (0.3-0.8 mm) [23]. Reducing the particles to the proper size requires more fossil-based energy (grinding). The emission factor (EFpro) and GWP of bio-oil production process (Gpro), using mobile or fixed bio-refinery are calculated using Eqs. 20 and 21.
EFpro=RCO2*EFproCO2+RCH4*EFproCH4+RN2O*EFproN2O
(4.20)
Gpro=Produced Bio-oil*EFpro
(4.21)
The produced bio-oil will be transported to the distribution center by tanker truck, using diesel fuel as an input. Tanker truck trips and distance from the bio-refinery to the distribution center are two major factors impacting fuel consumption. The tanker truck capacity has a direct relationship with the number of trips. The GWP of bio-oil transportation (Goil) is quantified using GREET 2015. The emission factor (EFoil) and GWP of transferring bio-oil (Goil), are calculated using Eqs. 22 and 23.
EFoil=RCO2*EFoilCO2+RCH4*EFoilCH4+RN2O*EFoilN2O
(4.22)
EIoil= Produced Bio-oil*EFoil*Distance
(4.23)
157 In this study, combustion is considered as the last life cycle phase for bio-oil. The emission factor (EFcomb) and GWP (Gcomb) of bio-oil combustion are calculated using Eqs. 24 and 25.
EFcomb=RCO2*EFoilCO2+RCH4*EFoilCH4+RN2O*EFoilN2O
(4.24)
Gcomb=Produced Bio-oil * EFcomb
(4.25)
From the foregoing, GWP can be obtained from the various biomass-to-bio-oil SC activities. Table 4.4 presents GHG emission rates from life cycle inventory databases in SimaPro 8 and GREET 2015 for biomass collection, bio-oil production, and bio-oil production. It can be seen that emissions of CO2 are greater than other GHG emissions for each phase of the bio-oil life cycle.
Table 4.4. Cradle-to-grave global warming potential for bio-oil production and consumption over 20-year time period [2], [3] Substance
Biomass Collection (kg CO2 eq. per metric ton of biomass)
Bio-Oil Production (kg CO2 eq. per metric ton of bio-oil)
1.21 E+1 6.12 E-1 4.48 E-2 1.27 E+1
7.12E+02 0.00E+00 5.85E+02 1.30E+03
CO2 (Biogenic and Fossil) CH4 N2O Total
Bio-Oil Combustion (kg CO2 eq. per metric ton of bio-oil) 2.57E+03 5.65E+00 1.55E+03 4.13E+03
Bio-oil combustion releases CO2 emissions from biogenic substances, however, which are not counted as emissions [24].
158 Application of the Method A case study for biomass-based energy production is conducted to demonstrate the application of the proposed method in the Pacific Northwest, USA. The study extends prior work, which investigated supply chain cost minimization through a multi-criteria decision making for sustainable bio-oil production [25]. A GIS-based approach is applied to define the shortest path between sites (e.g., collection site to a transportable bio-refinery or fixed bio-refinery location) [15]. Therefore, transportable bio-refineries are located close to collection sites to address existing logistical challenges (e.g., low bulk density). In addition, different road types are considered between collection sites and fixed bio-refineries, which along with the processing modes, impacts the types of trucks used in the supply chain. The actual data about the forest zones, available amounts and types of biomass, and locations of collection sites were obtained from the State of Oregon Department of Forestry (ODF) [26] to reasonably demonstrate the decision making method, mathematical models, and life cycle assessment.
The Base Case in the study considers twenty high-potential collection sites in three ODF forest districts, which are Forest Grove, Astoria, and Tillamook. The main biomass types available are Douglas Fir, Western Hemlock, and Red Alder. The required equipment in the Base Case includes forwarders, grinders, single trailer trucks, double-trailer trucks, tanker trucks, and mobile bio-refinery units, fixed bio-refineries. Due to the high cost of using a lowboy to move the grinder and loader, the grinder is placed near the mobile bio-refinery or collection sites. In this study, the Base Case includes five staging sites, two mobile bio-refineries, and a fixed bio-refinery with
159 storage capacity. Staging sites are considered to take advantage of assembling loads and using the maximum allowable legal weight. Fixed, variable, and labor costs are calculated after Brinker et al. [25]. The U.S. Producer Price Index is used to adjust cost values for inflation to 2016. To complete the analysis of the simulated supply chain (mobile bio-refineries are not currently in use by industry), the following assumptions were made:
1) The capacities of mobile and fixed bio-refineries are 50 and 200 dry metric tons per day, respectively. 2) The annual scheduled production of a mobile bio-refinery is 329 days (12 hours per day) [13]. 3) The annual scheduled production of a fixed bio-refinery is 365 days (24 hours per day). 4) At least 20,000 metric tons of forest biomass (θ) is available from the twenty collection sites over a one-year time horizon. 5) The type of truck for each route is known. 6) The time horizon is one year. 7) Green wood has a 50% moisture content [2]. 8) The higher heating value of bio-oil is 17.6 MJ/kg [27]. 9) The total distance from the storage facility to the end use is assumed as 150 miles (~241 km) roundtrip. 10) The production yield for biomass-to-bio-oil conversion is 50%. 11) The quality rate (qij) and the accessibility rate (𝛼 ij) for each site is known.
160 12) The quality rate is assumed to range from 0-5% for FHR of 50-60% moisture content. 13) The accessibility rate ranges from 0-5% for FHR with collection costs of $15-25 per dry metric ton. 14) The effective lifetime of a transportable and a fixed bio-refinery unit are assumed to be ten and twenty years, respectively.
The roundtrip distances between the staging sites, and mobile and fixed bio-refineries are defined with the assistance of GIS using a shortest path method. Table 4.5 presents roundtrip distances, numbers of tractor-trailer and tanker truck trips, and truck capacities for the upstream and midstream segments of the mixed SC, e.g., from collection sites to the mobile bio-refinery, and then to storage near the fixed biorefinery.
Table 4.5. Base case transportation details Distance Truck Tanker Truck Capacity (Miles) Trips Trips (Metric Tons) 29,412 1,471 14 C to MB 54,945 549 23 MB to FB 29,412 1,471 14 C to S 86,957 870 25 S to FB C: Collection, MB/FB: Mobile/Fixed bio-refinery, S: Staging site Pathway
GWP (Mg CO2 eq.) 230.46 257.90 230.46 864.57
The mean (μ) and variance (σ2) of qij and 𝛼 ij are obtained from simulated database for each staging site (Table 4.6). The probabilities of the quality rate (α), average acceptable quality rate (Q), accessibility rate (β), and the average acceptable accessibility rate (A) are calculated using the SVM method and simulated data. Since the probabilities are between 0 and 1, the α and β values are between 0 and 1. For
161 instance, a particular site may produce biomass with an average moisture content of 55%, the quality rate is 60%, while the variance is 4% (obtained from simulated data). Staging sites with forest biomass below the acceptable rates (A and Q) will not be considered by the algorithm because they do not meet the quality and accessibility constraints in the stochastic model.
Table 4.6. Collection site attributes Quality Rate, (𝐪𝐢𝐣 ) – N (𝝁, 𝝈𝟐 ) N (0.025, 0.040) N (0.022, 0.025)
Accessibility Rate, (𝐚𝐢𝐣 ) – N (μ, 𝝈𝟐 ) N (0.034, 0.050) N (0.030, 0.035)
1200
N (0.019, 0.020)
N (0.022, 0.040)
900 600
N (0.035, 0.035) N (0.030, 0.057)
…
…
Site 3 Site 49 Site 50
…
Available Biomass (metric tons) 1100 1400
…
Harvesting Site Site 1 Site 2
N (0.015, 0.020) N (0.040, 0.025)
A short-term storage facility is provided near the mobile bio-refinery and associated costs are considered in the bio-refinery costs. The base number of collection sites (N) is 20 in this case study, meaning the decision makers want to select 20 collection sites out of 50 collection sites with the highest biomass quality (i.e., high net calorific value or moisture content) and accessibility (i.e., biomass collection cost and available amount of biomass). The results of the Phase 1 (economic analysis) and Phase 2 (environmental impact analysis) for the mixed biomass-to-bio-oil supply chain are reported in the next section.
Results and Discussion As discussed in Phase 1, a mathematical optimization model was developed to assess the potential for bio-oil commercialization, using SVM and stochastic optimization
162 model. The obtained weight of each criterion via SVM can help decision makers to find the rate of quality (α) and accessibility (β), along with an average rate of quality (Q) and accessibility (A). In this study, Q and A are predicted at 75% and 70%, with the assistance of R programming language. The stochastic optimization model developed was solved using a GA within MATLAB [8]. An optimal solution was found after 300 iterations in under six seconds using a system configured with a Windows 7, 64-bit Operation System, Intel Core i7 processor, and 8GB RAM. The solution indicates that all available biomass would be processed in Sites 1-20, producing 10,000 metric tons (2,201,433 gallons) of bio-oil over a one-year time horizon. The annual cost is predicted as $2,387,595, resulting in a cost of $0.286/liter ($1.08/gallon). The optimal solution for the proposed mixed SC network indicated that the available amount of biomass would be processed using two transportable bio-refineries.
The proposed mathematical model is constituted as an NP-hard problem. Thus, as the number of sites increases, decision making will become increasingly difficult (even impossible) without a proper computational algorithm. The major concern when developing an NP-hard problem is how to solve it and represent the solutions. For instance, if the number of collection sites is 20, the number of solution combinations (either feasible or infeasible) will be 1,048,576 (220), which would lend itself to a heuristic or metaheuristic approach. Both approaches resulted in identical solutions in this case, verifying the GA approach.
163 Both the traditional and new pathways use similar equipment for biomass collection, bio-refinery processing, and bio-oil combustion. Further, the GHG emissions resulting from these activities has similar values in both pathways. For instance, both fixed and mobile bio-refineries employ pyrolysis process technology for bio-oil production, and the GHG emissions factor is predicted to be 1.3 Mg CO2 eq. per metric ton of bio-oil for both. Consequently, the environmental impact (GWP) of a bio-refinery producing 10,000 metric tons bio-oil from 20,000 metric tons biomass is 13,000 Mg CO2 eq. Additionally, the GHG emissions factor for biomass collection and bio-oil combustion in both pathways is predicted as 12.7 kg CO2 eq. per metric ton of biomass and 4.13 Mg CO2 eq. per metric ton of bio-oil.
Figure 4.5 indicates the GWP for each life cycle stage associated with bio-oil production and consumption. The CO2 eq. absorption value is assumed to be equal to the CO2 eq. released into atmosphere through bio-oil combustion. Since it is assumed biochar is used as an energy source, the majority of bio-refinery processing GHG emissions and GWP are negligible. The key difference between the traditional and new pathway results from truck and truck-tanker trips between the collection sites and the fixed bio-refinery (also the bio-oil storage location). The total GHG emissions in the Base Case using the truck-tanker pathway is 488.35 Mg CO2 eq. (0.22 kg CO2 eq. per gallon of bio-oil) and truck-truck pathway is 1,095,037 kg CO2 eq. (0.50 kg CO2 eq. per gallon of bio-oil).
164 The GHG emissions from bio-oil combustion are classified as biogenic, or part of the natural cycle, and are absorbed by new, growing biomass. Particulates emissions are negligible over the all life cycle stages. GHG emissions from biomass, grinding, and transportation result from fossil fuel combustion, and are not considered as part of the natural carbon cycle. The net cradle-to-grave GHG emissions for bio-oil (i.e., biomass collection, grinding, transportation, bio-refinery processing, distribution, bio-oil combustion, and CO2 eq. absorption) is predicted as 13,712 Mg CO2 eq. (6.22 kg CO2 eq. per gallon of bio-oil) using a mobile bio-refinery and truck-tanker pathway, and 14,319 Mg CO2 eq. (6.50 kg CO2 eq. per liter of bio-oil) using fixed bio-refinery and truck-truck pathway. The new pathway reduces GWP by 606 Mg CO2 eq. per year (028 kg CO2 eq. per liter). The result indicates GWP reduces by about 4% through the use of the transportable bio-refinery and truck-tanker transportation pathway.
Global Warming Potential (kg CO2 eq. per gallon)
25 20 Mobile Refinery Processing
15
Fixed Refinery Processing 10
Truck-Truck Pathway
5
Truck-Tanker Pathway
0
Bio-oil Combustion Biomass Collection
-5
CO2 absorption -10 -15 -20 New Pathway Traditional Pathway
Figure 4.5. Cradle to grave global warming potential for bio-oil production and consumption
165 Sensitivity Analysis Based on the structure of the optimization model, several parameters have significant effects on the economic and environmental performance of the system, which can be examined using sensitivity analysis. The purpose of sensitivity analysis is to explore the effect of variables (e.g., binary, continuous, and integer variables) and right hand side parameters (e.g., capacity) on the optimization results. This section presents a sensitivity analysis to assess the effect of major parameters, such as available amount biomass and mobile bio-refinery cost. Apart from the Base Case, two different cases are presented below.
Effect of Mobile Bio-refinery Cost The mobile bio-refinery cost is the major cost driver of bio-oil production in the Base Case. Thus, exploring the effect of this attribute is essential to assess the commercial feasibility for bioenergy production. In the first alternative case (Case 1), the biorefinery costs (i.e., fixed, variable, and labor costs) of the mobile bio-refinery are reduced by 50%. In the second case (Case 2), the bio-refinery costs of the mobile biorefinery are increased by 50%. Table 4.7 reports the optimum results for each case, as well as the results without considering the mobile bio-refinery in the Base Case.
Table 4.7. Effect of mobile bio-refinery cost on the overall annual cost Cases Base Case Case 0a Case 1 (-50% cost) Case 2 (+50% cost) a Base Case without mobile bio-refinery
Annual Overall Cost ($) 2,387,595 2,957,327 1,739,071 2,957,327
Annual Cost of Bio-refinery ($/yr) Fixed Cost Variable Labor Cost Cost 600,762 89,082 375,357 4,314,095 1,578,218 1,155,334 300,381 44,541 187,678 4,314,095 1,578,218 1,155,334
166
The results show that changes in the mobile bio-refinery cost directly impact the total cost. The overall annual cost decreased in Case 1 by approximately $648,524 (27%) and in Case 2, the annual cost increased by $569,732 (23%). Case 1 indicates that the optimal supply chain would consist of two mobile bio-refineries. In Case 2, the fixed bio-refinery would be responsible for processing all forest biomass due to higher production, transportation, and storage costs of using the mobile bio-refinery compared with the fixed bio-refinery (Figure 4.6). Although the bio-refinery costs are higher in Case 2, the increased processing rate of using the fixed bio-refinery reduces total annual cost. While transportation cost of the traditional pathway is higher than new pathway, the combination of single- and double-trailer trucks in the traditional pathway reduces the transportation costs compared to using only single trailer trucks. Since the amount of processed biomass and the number and location of bio-refineries remains constant, the GWP remains the same for in both cases. Global Warming Potential 16
3.0
14
12
2.5
10 2.0 8 1.5 6 1.0
4
0.5
2
0.0
Cradle to Grave Global Warming Potential (Gg CO2 eq.)
Annual Cost ($ million)
Annual Cost 3.5
0 Base Case
Case 1
Case 2
Figure 4.6. Effect of mobile bio-refinery cost on environmental and economic measures
167 Effect of Available Amount of Forest Biomass The amount of available biomass in the Base Case is based on the remaining nonmerchantable products at the roadside, which is not the only potential source of biomass for bio-oil production at the sites considered. Other sources include slash, branches, tops, and breakage, equating to an additional 30,000 metric tons, which are mainly burned due to high collection costs. The effect of the available amount of forest biomass is also investigated due to the importance of this attribute in the proposed method. In Phase 1, the available amount of biomass is the main parameter, along with biomass quality. Therefore, changing the amount of available biomass affects the total annual cost of simulated network, due to the direct impacts of these parameters on each entity in the BESCs. As explained above, the available amount of biomass is assumed to be 20,000 metric tons in the Base Case. In Case 3, the available amount is decreased by 50%. In Case 4, the available amount is increased by 50%. Changing the amount of biomass has a direct impact on collection, pre-processing, transportation, and biorefinery costs. Table 4.8 reports the predicted annual cost for each case.
Table 4.8. Effect of available amount of forest biomass on the overall annual cost Cases Base Case Case 3 (-50% amount of biomass) Case 4 (+50% amount of biomass)
Overall Cost ($) 2,387,595 1,684,017 4,388,222
Amount of Available Biomass (Metric Ton) 20,000 10,000 30,000
The total supply chain cost is found to change directly, but nonlinearly with the amount of available biomass processed. In Cases 3 and 4, the annual cost of the network increased by $703,578 (29%) and $2,000,627 (84%), respectively. The optimal supply chains utilize one mobile bio-refinery in Case 3, and two mobile bio-refineries in Case
168 4. The effect of biomass availability on the various components of supply chain costs are shown in Figure 4.7. This added amount of biomass would increase the environmental impacts from biomass collection, size reduction, and transport. Case 3 investigates the effect of decreasing the amount of available biomass. The results indicate that GWP is decreased by 6,791 and 7,086 Mg CO2 eq. for the new and traditional pathways, respectively, compared to the Base Case. The mixed SC reduced GWP by about 4.3% compared to the traditional SC, in Case 3. In Case 4, the results indicate that GWP is increased by 6,790 and 7,085 Mg CO2 eq. for the new and traditional pathways, respectively, compared to the Base Case. The mixed SC reduced GWP by about 4% compared to the traditional SC, in Case 3. Since the upstream activities (e.g., biomass collection and transportation) mainly use fossil-based energy sources, the emissions released during these activities negatively affect environmental performance. Annual Cost
Global Warming Potential
25
Annual Cost ($ million)
4.5 4.0
20
3.5 3.0
15
2.5 2.0
10
1.5
1.0
5
0.5 0.0
Cradle to Grave Global Warming Potential (Gg CO2 eq.)
5.0
0 Base Case
Case 3
Case 4
Figure 4.7. Effect of the available amount of forest biomass on environmental and economic measures
169 Conclusions Growing concerns over increasing energy consumption indicate that alternative energy sources are essential to reduce environmental impacts, improve energy independence, and aid rural economic development. For these reasons, BESC sources (e.g., biofuels and bio-oils) represent a promising replacement for conventional energy (e.g., coal and diesel). Prior studies have reported that biomass will play a key role due to its availability, low cost, and abundance. BESC can reduce the environmental impacts in comparison with fossil fuel use, especially due to its potential to achieve carbon neutrality. However, a large fraction of biomass resources have been underutilized for bioenergy production due to logistical challenges (e.g., high collection, transportation, and storage cost). Therefore, advanced methods are needed to develop an effective, efficient, and commercially viable bioenergy industry for achieving reliable supply and affordability and addressing the existing challenges in bioenergy production processes and systems.
This research study investigated the costs and environmental impacts of bio-oil production, using an evolutionary decision making method. The two-phase decision making method represents a pioneering approach to incorporate uncertainty parameters (i.e., quality and accessibility) in the upstream and midstream B2BSC segments. The economic analysis (Phase 1) employs a support vector machine technique (as a machine learning technique) to predict the uncertainty parameters trend and stochastic optimization modeling (as an operations research approach) to minimize the total annual cost of BESC. The environmental impact analysis (Phase 2) employs life cycle
170 assessment methods to predict global warming potential over 20-year time horizon by considering cradle-to-grave system boundary of bio-oil SC.
Bio-oil produced from sustainably managed biomass could reduce costs and environmental impacts of transportation fuels and other applications that rely on oil feedstocks. Bio-oil quality and cost are two major barriers to commercialization however. Thus, this research aims to promote sustainability of the bio-oil supply chain (SC) through a mixed SC that employs mobile and fixed bio-refineries. The traditional pathway employs a fixed bio-refinery and a truck-truck transportation approach, including double-trailer trucks to reduce truck trips by assembling loads to take advantage of the maximum allowable legal weight based on the number of axles and axle spacing. The new, mixed pathway employs mobile bio-refineries and truck-tanker transportation approach, including trailer trucks and tankers for transferring biomass and bio-oil.
The mixed SC network is presented to compare the production costs and cradle-tograve environmental impacts of traditional and new SC pathways by considering all life cycle stages, i.e., collection, grinding, transportation, production, distribution, and combustion. The analysis demonstrates the benefits of the mixed bio-oil SC and the role of mobile bio-refinery and truck-tanker pathway. The results of environmental impact analysis indicate that forest biomass collection, grinding, and transportation are the main activities in bio-oil SCs that release fossil carbon emissions. In this research, the majority of required energy for bio-oil production is sourced from biochar and
171 syngas. Thus, the resulting GHG emissions are classified as biogenic, which are offset by absorption by trees in the forest. The impact (GWP) of each life cycle phase is evaluated by quantifying the energy consumption and GHG emissions, as well as analyzing resource use. The results illustrate that the substitution of the traditional with the new, mixed SC pathway can reduce GHG emissions by 4.3% in the cases studied.
This research shows that locating mobile bio-refineries close to the collection and staging sites can reduce costs and environmental impacts of processing low energy density biomass by reducing the number of truck trips and associated fuel consumption. Therefore, the mixed SC represents a promising approach to meet cross-cutting environmental and economic goals. By taking advantage of ongoing work in supply chain and process technology development, as well as mathematical modeling and optimization, viable commercial approaches will emerge to support sustainable bioenergy. In addition to technological approaches, future work should address the influence of social aspects on the cross-cutting sustainability of bioenergy. This industry has the potential to positively impact current and future generations through economic development and improved environmental conditions.
Acknowledgment The authors gratefully acknowledge Michael Wilson and Robert Nall (Oregon Department of Forestry), Phillip C. Badger (Renewable Oil International LLC), and UPM Biofuels, Finland for providing input to support this research.
172 Annex A: Nomenclature Indices 𝑖 𝑗 𝑘 𝑙 𝑡 𝑠 𝑐 𝑝 𝑚 𝑓 𝑠𝑡 𝑑𝑡 𝑡𝑡 𝑀
Set of collection sites Set of staging sites Set of mobile bio-refinery sites Set of fixed (non- mobile) bio-refinery sites Set of time periods Staging Collection Pre-processing Mobile bio-refinery Fixed bio-refinery Single-trailer truck Double-trailer truck Tanker truck Large positive constant (Big M)
Parameters Average accessibility rate 𝐴 Average quality rate 𝑄 Quality rate of processed biomass 𝛼 Accessibility rate of processed biomass 𝛽 Annual fixed cost for a defined site (e.g., collection, staging, refinery, 𝐹𝑠𝑖𝑡𝑒 or storage) ($) Annual variable cost for a defined site (e.g., collection, staging, 𝑉𝑠𝑖𝑡𝑒 refinery, or storage) ($) Annual labor cost for a defined site (e.g., collection, staging, refinery, 𝐿𝑠𝑖𝑡𝑒 or storage) ($) Annual fixed cost of a defined truck (e.g., single-trailer, double-trailer, 𝐹𝑡𝑟𝑢𝑐𝑘 or tanker trucks) ($) Annual variable cost of a defined truck (e.g., single-trailer, double𝑉𝑡𝑟𝑢𝑐𝑘 trailer, or tanker trucks) ($) Annual labor cost of a defined truck (e.g., single-trailer, double-trailer, 𝐿𝑡𝑟𝑢𝑐𝑘 or tanker trucks) ($) Annual utilization rate of a forwarder (metric tons) 𝑈𝑐 Annual utilization rate of a grinder (metric tons) 𝑈𝑝 Annual processing rate of a mobile bio-refinery (metric tons) 𝑈𝑚 Annual processing rate of a fixed bio-refinery (metric tons) 𝑈𝑓 Annual utilization rate of a single-trailer truck (metric tons) 𝑈𝑠𝑡 Annual utilization rate of a double-trailer truck (metric tons) 𝑈𝑑𝑡 Annual utilization rate of a tanker truck (metric tons) 𝑈𝑡𝑡
173 𝐶𝑎𝑝𝑖 𝐶𝑎𝑝𝑚 𝐶𝑎𝑝𝑙 𝜃
Annual capacity of a collection site (metric tons) Annual capacity of a mobile bio-refinery (metric tons) Annual capacity of a fixed bio-refinery (metric tons) Annual available amount of biomass (metric tons) Percentage yield of converting biomass to bio-oil PY Base number of collection sites 𝑁 Total emission factor of biomass transportation EFmass Total emission factor of bio-oil combustion EFcomb EFoilCO2 CO2 emission factor of bio-oil combustion EFoilCH4 CH4 emission factor of bio-oil combustion EFoilN2O N2O emission factor of bio-oil combustion Total emission factor of bio-oil transportation EFoil Total mission factor of production process EFpro EFproCO2 CO2 emission factor of production process EFproCH4 CH4 emission factor of production process EFproN2O N2O emission factor of production process Total emission factor of upstream activities EFup EFupCO2 CO2 emission factor of upstream activities EFupCH4 CH4 emission factor of upstream activities EFupN2O N2O emission factor of upstream activities GWP of biomass transportation Gmass GWP of bio-oil transportation Goil GWP of bio-oil combustion Gcomb GWP of production process Gpro GWP of upstream activities Gup CO2 emission rate RCO2 CH4 emission rate RCH4 N2O emission rate RN2O Continuous Variables Amount of biomass transported from site i to site j at time t 𝑋𝑖𝑗𝑡 Amount of biomass transported from site i to site k at time t 𝑋𝑖𝑘𝑡 Amount of biomass transported from site j to site l at time t 𝑋𝑗𝑙𝑡 Integer Variables Amount of bio-oil transported from site k to site s at time t 𝑌𝑘𝑠𝑡 Amount of bio-oil transported from site l to site s at time t 𝑌𝑙𝑠𝑡 Binary Variables Binary variable for transportation from site i to site j at time t 𝐵𝑖𝑗𝑡
174 References [1]
G. H. Brundtland, Our Common Future, World Commission on Environment
and Development (WCED). Oxford University Press, 1987. [2]
P. Steele, M. E. Puettmann, V. Kanthi Penmetsa, and J. E. Cooper, “Life-cycle
assessment of pyrolysis bio-oil production,” For. Prod. J., vol. 62, no. 4, p. 326, 2012. [3]
“SimaPro - World’s Leading LCA Software Package | PRé Sustainability.”
[Online]. Available: http://www.pre-sustainability.com/simapro. [Accessed: 21-Jan2015]. [4]
Argonne National Laboratory, GREET LCA. 2015.
[5]
R. D. Perlack, L. M. Eaton, A. F. Turhollow Jr, M. H. Langholtz, C. C. Brandt,
M. E. Downing, R. L. Graham, L. L. Wright, J. M. Kavkewitz, A. M. Shamey, and others, “US billion-ton update: biomass supply for a bioenergy and bioproducts industry,” 2011. [6]
U.S.
EIA,
“Annual
Energy
Outlook
2015
-
Energy
Information
Administration,” 2015. [Online]. Available: http://www.eia.gov/forecasts/aeo/. [Accessed: 26-Jan-2016]. [7]
A. Mirkouei, K. R. Haapala, J. Sessions, and G. S. Murthy, “A Review and
Future Directions in Techno-Economic Modeling and Optimization of Upstream Forest Biomass to Bio-oil Supply Chains,” Renew. Sustain. Energy Rev. Revis., 2015. [8]
USDA, “View Opportunity | GRANTS.GOV,” 2015. [Online]. Available:
http://www.grants.gov/view-opportunity.html?oppId=274814.
[Accessed:
29-Jan-
2016]. [9]
V. Reed, “U.S. Department of Energy Biomass Program | Department of
Energy,” 2012. [Online]. Available: http://energy.gov/eere/bioenergy/downloads/usdepartment-energy-biomass-program. [Accessed: 07-Jan-2016]. [10]
B. Lattimore, C. T. Smith, B. D. Titus, I. Stupak, and G. Egnell, “Environmental
factors in woodfuel production: Opportunities, risks, and criteria and indicators for sustainable practices,” Biomass Bioenergy, vol. 33, no. 10, pp. 1321–1342, 2009. [11]
I. Awudu and J. Zhang, “Uncertainties and sustainability concepts in biofuel
supply chain management: A review,” Renew. Sustain. Energy Rev., vol. 16, no. 2, pp. 1359–1368, Feb. 2012.
175 [12]
N. Shabani and T. Sowlati, “A hybrid multi-stage stochastic programming-
robust optimization model for maximizing the supply chain of a forest-based biomass power plant considering uncertainties,” J. Clean. Prod., vol. 112, pp. 3285–3293, 2016. [13]
C. B. Sorenson, “A Comparative Financial Analysis of Fast Pyrolysis Plants in
Southwest Oregon,” The University of Montana Missoula, MT, 2010. [14]
A. Mirkouei, K. R. Haapala, J. Sessions, and G. S. Murthy, “Multi-criteria
Decision Making for Sustainable Bio-Oil Production using a Mixed Supply Chain,” J. Clean. Prod. Be Submitt., 2015. [15]
A. Mirkouei, P. Mirzaie, K. R. Haapala, J. Sessions, and G. S. Murthy,
“Reducing the Cost and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains,” J. Clean. Prod., 2015. [16]
Union of Concerned Scientists, “The Promise of Biomass Clean Power and Fuel
- If Handled Right,” 2012. [Online]. [17]
A. Mirkouei, K. R. Haapala, J. Sessions, and G. S. Murthy, “Evolutionary
Optimization of Bioenergy Supply Chain Cost with Uncertain Forest Biomass Quality and Availability,” in Proceedings of the IIE-ISERC, May 21-24, Anaheim, California, USA, 2016. [18]
P. Thornley, P. Gilbert, S. Shackley, and J. Hammond, “Maximizing the
greenhouse gas reductions from biomass: The role of life cycle assessment,” Biomass Bioenergy, vol. 81, pp. 35–43, 2015. [19]
S. Eisenbarth and K. Van Treuren, “Sustainable and responsible design from a
christian worldview,” Sci. Eng. Ethics, vol. 10, no. 2, pp. 423–429, 2004. [20]
N. Cristianini and J. Shawe-Taylor, An introduction to support vector machines
and other kernel-based learning methods. Cambridge University Press, 2000. [21]
J. Shawe-Taylor and N. Cristianini, Kernel methods for pattern analysis.
Cambridge University Press, 2004. [22]
A. Karatzoglou, D. Meyer, and K. Hornik, “Support vector machines in R,”
2005. [23]
S. Septien, S. Valin, C. Dupont, M. Peyrot, and S. Salvador, “Effect of particle
size and temperature on woody biomass fast pyrolysis at high temperature (1000–1400 C),” Fuel, vol. 97, pp. 202–210, 2012.
176 [24]
L. Johnson, B. Lippke, and E. Oneil, “Modeling Biomass Collection and Woods
Processing Life-Cycle Analysis,” For. Prod. J., vol. 62, no. 4, pp. 258–272, 2012. [25]
R. W. Brinker, D. Miller, B. J. Stokes, and B. L. Lanford, “Machine rates for
selected forest harvesting machines,” Circ. 296 Revis. Ala. Agric Exp Stn. Auburn Univ. 32 Pp, 2002. [26]
ODF, “State of Oregon: Oregon Department of Forestry - Home,” 2015.
[Online]. Available: http://www.oregon.gov/odf/Pages/index.aspx. [Accessed: 12-Oct2015]. [27]
Dynamotive, “Dynamotive Energy Systems Corporation, Canadian BioOil
Plant: Summary (USD).” 2009.
177
Computational Codes clc; clear; close all; global NFE Penalty nSite nQuarter T; %% Assign Problem Parameters Assigndata; %% Problem Definition CostFunction=@Fitness; nVar=nSite;
% Cost Function
% Number of Decision Variables
%% GA Parameter MaxIt=100; nPop=100;
% Maximum Number of Iterations % Population Size
pCrossover=0.7; % Parents (Offsprings) Population Size Ratio nCrossover=round(pCrossover*nPop/2)*2; pMutation=0.1; % Mutants Population Size Ratio nMutation=round(pMutation*nPop); TournamentSelectionSize=3; Penalty=1000; %% Initialization *generate an initial random population of bit strings* NFE=0; individual.Position1=[]; individual.Position2=[]; individual.Sol=[]; individual.Cost=[]; pop=repmat(individual,nPop,1); for i=1:nPop for j=1:nQuarter pop(i).Position1{j}=randi([0 1],1,nSite(j)); if sum(pop(i).Position1{j})>T(j) t=sum(pop(i).Position1{j})-T; a=find(pop(i).Position1{j}==1);
178 a=a(randperm(numel(a))); a=a(1:t); pop(i).Position1{j}(a)=0; end pop(i).Position2{j}=rand(1,nSite(j)); end [pop(i).Cost, pop(i).Sol]=CostFunction(pop(i).Position1,pop(i).Position2); end % Sort Population Costs=[pop.Cost]; [Costs, SortOrder]=sort(Costs); pop=pop(SortOrder); WorstCost=Costs(end); BestSol=[]; BestCost=zeros(MaxIt,1); MeanCost=zeros(MaxIt,1); nfe=zeros(MaxIt,1); %% GA Main Loop for it=1:MaxIt % Crossover pop2=repmat(individual,nCrossover/2,2); for k=1:nCrossover/2 i1=TournamentSelection(pop,TournamentSelectionSize); i2=TournamentSelection(pop,TournamentSelectionSize); [pop2(k,1).Position1, pop2(k,2).Position1]=Crossover1(pop(i1).Position1,pop(i2).Position1) ; [pop2(k,1).Position2, pop2(k,2).Position2]=Crossover2(pop(i1).Position2,pop(i2).Position2) ; [pop2(k,1).Cost, pop2(k,1).Sol]=CostFunction(pop2(k,1).Position1,pop2(k,1).Position2) ; [pop2(k,2).Cost, pop2(k,2).Sol]=CostFunction(pop2(k,2).Position1,pop2(k,2).Position2) ; end pop2=pop2(:);
% Mutation pop3=repmat(individual,nMutation,1); for k=1:nMutation
179 i1=TournamentSelection(pop,TournamentSelectionSize); pop3(k).Position1=Mutate1(pop(i1).Position1); pop3(k).Position2=Mutate2(pop(i1).Position2); [pop3(k).Cost, pop3(k).Sol]=CostFunction(pop3(k).Position1,pop3(k).Position2); end % Merge Populations pop=[pop pop2 pop3]; %#ok % Sort Population Costs=[pop.Cost]; [Costs, SortOrder]=sort(Costs); pop=pop(SortOrder); WorstCost=max(WorstCost,Costs(end)); % Delete Extra Individuals pop=pop(1:nPop); Costs=Costs(1:nPop); % Save Results BestSol=pop(1); BestCost(it)=Costs(1); MeanCost(it)=mean(Costs); nfe(it)=NFE; % Show Information disp(['Iteration ' num2str(it) ': ' ... 'Best Cost = ' num2str(BestCost(it)) ' , ' ... 'Mean Cost = ' num2str(MeanCost(it))]); end %% Results figure; subplot(2,1,1); semilogy(BestCost,'r','LineWidth',2); hold on; semilogy(MeanCost,'b:','LineWidth',2); xlabel('Generation (Iteration)'); legend('Best Costs','Mean Costs'); subplot(2,1,2); semilogy(nfe,BestCost,'r','LineWidth',2); hold on; semilogy(nfe,MeanCost,'b:','LineWidth',2); xlabel('Function Evaluations'); % fitness func. uses func. eval. to calcul. a value of worth for individual and compare with each other
180 legend('Best Costs','Mean Costs'); xlim([0 nfe(end)]);
181
CHAPTER 5: SUMMARY AND CONCLUSIONS
182
Chapter 5:
Summary and Conclusions
Summary Recent growing interest in reducing greenhouse gas (GHG) emissions requires the application of effective energy solutions, such as the utilization of renewable resources. Biomass represents a promising renewable resource for bioenergy, since it has the potential to reduce GHG emissions from various industry sectors. In spite of the potential benefits, use of biomass is limited due to logistical challenges of collection and transport to bio-refineries. This study proposes a forest biomass-to-bio-oil mixed supply chain network to reduce the total cost of bioenergy production and GHG emissions compared to a conventional bioenergy supply chain. The mixed supply chain includes mixed-mode bio-refineries and mixed-pathway transportation. This research integrates knowledge of the renewable energy production, operations research, sustainability, and supply chain management disciplines to evaluate economic and environmental targets of bioenergy production with the use of the multi-criteria decision making approach.
The presented approach includes qualitative and quantitative methods to address the existing challenges and gaps in the bioenergy manufacturing system. The qualitative method employs decision tree analysis to classify the potential biomass harvesting sites, considering biomass quality and availability. The quantitative method proposes mathematical models to optimize the upstream and midstream biomass-to-bioenergy supply chain cost, using mixed bio-refinery modes (transportable and fixed) and
183 transportation pathways (traditional and new). Quantitative methods include machine learning techniques such as support vector machine; operations research techniques, such as deterministic optimization; stochastic optimization; artificial intelligence, such as genetic algorithms, and life cycle assessment. A mathematical model was formulated with the aim of minimizing the total costs of a supply chain network in the PNW. Life cycle assessment was then conducted for this case study SC with the assistance of available life cycle inventory data for a biomass-to-bio-oil supply chain. Impact assessment, on a global warming potential (GWP) basis, is conducted with the assistance of databases within SimaPro 8 and GREET 2015 software packages. Sensitivity analysis for the case investigated indicates that using the mixed supply chain can reduce the total cost by 32% and GHG emissions by 6% compared to the traditional supply chain.
Conclusions The objective of this research is to support engineering decision making for costeffective substitution of fossil-based with biomass-based energy sources to improve energy security and overcome related environmental challenges. Forest biomass are underutilized and wasted due to logistical challenges such as low market value, high collection and transportation costs, and immature production technologies. This research investigates the effects of converting biomass to bioenergy at forest, using mobile bio-refinery, and transportation of high energy density product (bio-oil) instead of low energy density product (forest biomass).
184 The role of mobile (transportable) bioenergy production technology provides a benefit to the B2BSC network. Overall production costs can be reduced when implementing mobile units. In general, by exploring the method, it was found that the use of mobile refinery is more suitable when travel time, travel distance, and unit transportation costs will increase. The optimal integration of mobile and fixed bio-refineries can improve the robustness and reduce the overall cost of bioenergy production. The benefits of this combination would be true for meeting not only economic targets, but also environmental targets, since transportation is a key source of environmental impacts in the bioenergy production.
Contributions Fava et al. (1991) and Azapagic (1999) took initiative to integrate the mathematical optimization model and life cycle assessment (LCA) and conduct a thorough analysis to simultaneously optimize both economic and environmental performances [1–3]. A few studies have attempted to integrate the LCA and mathematical model for designing sustainable supply chains [4–10]. To the best of our knowledge, this is the first research that develops a multi-criteria decision making method to examine biomass-tobioenergy SCs. This method is capable of integrating pre-treatment processes with an upstream segment of forest biomass energy SCs. The proposed decision making method encompasses qualitative (decision tree analysis) and quantitative method (optimization modeling with life cycle assessment) for developing sustainable and optimal biomass-based energy supply chain (BESC). Additionally, this study uses the
185 actual data in the case study (Base Case) to demonstrate the application and verify the methods.
The presented study focuses on upstream and midstream segments of biomass-tobioenergy supply chain to provide the contributions to the research community:
Systems integration: the proposed mixed supply chain includes mixed-mode biorefinery and mixed-pathway transportation.
Cross-cutting sustainability: the proposed decision making methods address sustainability challenges, such as assessing commercialization and environmental impacts.
System analysis: this research analyzes the upstream and midstream segments of biomass-based energy SC system to better understand the major drivers.
Integrated bio-refineries: this research considers different bio-refinery modes (e.g., mobile, transportable, and fixed) to develop an optimal and sustainable biomass energy SC.
Thermochemical: this study applies techno-economic analysis by evaluating pyrolysis process to produce bio-oil and bio-char and explores the economic and environmental impacts.
Feedstock supply and logistics: the key focus of this study is on the upstream segment because it has direct impacts on the production process.
186 The proposed methods and models have applications in various sectors: renewable energy sector (other type of biomass), conventional energy sector (petroleum-based), industrial sector (chemicals and bio-chemicals), transportation sector, and domestic sector, as well as various disciplines: operations research (optimization), sustainability (economic, environmental, and social), manufacturing systems and processes, supply chain, and reverse supply chain.
Opportunities for Future Research Potential benefits of this work include substitute for fossil-based energy, reduction of GHG emissions, improvement of the economics of forest management activities, creation of new jobs in the forest and bioenergy sectors, promotion of economic development in rural areas, and humanitarian engineering.
Designing a robust and sustainable BESC network requires addressing the uncertainties of externalities, such as supply, logistics, quality, delivery, production, and price. Several methods have been introduced in the literature to incorporate uncertainties in supply chain optimization, such as analytical methods (stochastic optimization and Markov decision process) and simulation methods (discrete event simulation and Monte Carlo). While many studies predominantly explored the complete SC (i.e., upstream, midstream, and downstream) to tackle sustainability issues, there is a dearth of literature for detailed analysis of each individual segment of the SC. Exploring and addressing the gaps in each segment can raise the awareness of decision makers and, subsequently, aid in identifying alternative ways of making business more robust and
187 sustainable. In summary, this research reveals some of the gaps in research related to the upstream and midstream segment of forest biomass-to-bioenergy SCs. Specifically, the following potential paths for a future research are defined:
Development of a biomass SC optimization model for integrating pretreatment processes into the upstream segment of forest biomass-to-bioenergy SCs
Exploration of pretreatment processes to identify economic, environmental, technological, and political challenges and barriers to implementation
Development of a techno-economic assessment method to evaluate the investment factors by comparing different system designs and assumptions
Development and implementation of novel pretreatment processes that could be adopted by industry into practice (e.g., transportable bio-refinery facilities)
Exploration of the state-of-the-art within other disciplines to integrate adopted methods and approaches (e.g., mathematical analysis and experimental design)
Exploration of issues related to the existing metrics and measures for optimization based on triple bottom line sustainability.
Further, there is a need to develop a detailed research plan for each of the paths proposed. Investigations into biomass-to-bioenergy SCs are increasing due to internal and external stakeholders’ needs, which include growing demand for bioenergy and for reduction of economic and environmental impacts.
188
189
Appendices
190
191
Appendix A: Reducing the Cost and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains
Journal of Cleaner Production 113 (2016) 495e507
192
Contents lists available at ScienceDirect
Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro
Reducing the cost and environmental impact of integrated fixed and mobile bio-oil refinery supply chains Amin Mirkouei a, *, Pantea Mirzaie a, Karl R. Haapala a, John Sessions b, Ganti S. Murthy c a
School of Mechanical, Industrial and Manufacturing Engineering, Oregon State University, Corvallis, OR 97331, USA Department of Forest Engineering, Resources and Management, College of Forestry, Oregon State University, Corvallis, OR 97331, USA c Department of Biological and Ecological Engineering, College of Agriculture, Oregon State University, Corvallis, OR 97331, USA b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 August 2015 Received in revised form 9 November 2015 Accepted 10 November 2015 Available online 2 December 2015
The call to decrease reliance on fossil fuels to reduce impacts on the environment and improve energy independence has created new duties and responsibilities within society. As one approach, mobile bio-oil refineries have been developed to facilitate the production of bio-oil near the source of underutilized forest harvest residues. These mobile refineries are expected to improve the robustness of woody biomass to bio-oil supply chains by reducing overall supply chain costs and environmental impacts. The use of mobile refineries in combination with large-scale non-mobile refineries, however, must be examined to better understand the potential economic and environmental benefits and drawbacks of such a supply chain. The research presented herein develops a mathematical model capable of helping decision makers in determining the optimal combination and location of fixed refineries and mobile refineries for a known quantity of woody biomass and a given set of harvesting locations by considering capital, operational, and transportation costs. A hypothetical case for northwest Oregon, USA is applied to verify the mathematical model. The supply chain environmental impacts are assessed by considering the carbon footprint of the transportation activities and the bio-refinery infrastructure. The results indicate that the use of a mobile refinery along with a fixed refinery is more suitable when transportation costs and distances increase. It is also found that the capital intensity of mobile refineries can influence the importance of their role. High capital cost can be detrimental to their application within a mixed-mode bio-oil supply chain. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Supply chain Bio-oil Woody biomass Environmental impacts Mobile refinery Carbon footprint
1. Introduction Due to the increasing cost of fossil fuels, their impacts on the environment, and an increasing call for energy independence, many countries are focusing on technology and policy mechanisms for substituting fossil sources with renewables, e.g., biomass, wind, and hydro, to achieve improved energy sustainability performance. According to the U.S. Energy Information Administration, the fossil fuel share of total energy use will decrease from 82% in 2011 to 78% in 2040, while the renewables share, including bio-fuels, is expected to increase from 9% to 13% in the same time period (USEIA, 2014). Bio-oil can be produced from forest harvest residues (FHR). Bio-oil from woody biomass (WB) is a potential energy source in
* Corresponding author. E-mail addresses:
[email protected], (A. Mirkouei). http://dx.doi.org/10.1016/j.jclepro.2015.11.023 0959-6526/© 2015 Elsevier Ltd. All rights reserved.
[email protected]
terms of reduced social, environmental, and economic impacts since it is a by-product of harvesting, represents a fire hazard, impedes planting of seedlings, harbors rodents that eat seedlings, and its combustion is considered carbon neutral (Page-Dumroese et al., 2009; Steele et al., 2012), particularly when compared to conventional disposal methods using fire. The main applications of bio-oil include combustion in engines, turbines, and boilers, as well as production of chemicals, transportation fuels, and hydrogen (Czernik and Bridgwater, 2004; Bridgwater, 2012; Kersten and Garcia-Perez, 2013). Optimization of biomass to bio-oil supply chains (SCs), however, is required to assist industry in supplying the market with a product cognizant of the three domains of sustainability, i.e., economic, environmental, and social (Mullaney et al., 2002). The research herein is motivated by the premise that bio-oil can be considered an economically and technically feasible alternative for fossil fuel-based applications. Current bio-oil output is not sufficient to meet societal demand, however, due to high costs of
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
FHR transportation and the scarcity of bio-oil refineries (Searcy et al., 2007). A recent comprehensive study concluded that the conversion process itself, in addition to the SC barriers, is inhibiting commercialization of bio-refineries (Sharma et al., 2013). The application of small scale and transportable bio-oil refineries has been investigated in recent years to more economically produce greater quantities of bio-oil. Mobile refineries (MRs) have been fabricated to be placed in the forest to produce bio-oil from FHR (Badger and Fransham, 2006). It is posited that if MRs can be utilized in a SC along with fixed (non-mobile) refineries (FRs), the limitations of producing a sufficient volume of bio-oil can be overcome, and this source of renewable energy can be deployed more effectively and efficiently. To improve the robustness of bio-oil SC networks and to be able to respond to rapid growth in fuel consumption, different SC schemes consisting of combinations of current technologies and techniques are required. As such, mixed-mode bio-oil SCs, including the utilization of both FRs and MRs, can be part of the solution to meet increasing consumer fuel demands, while mitigating the impacts associated with fossil fuels. Previous research has proposed approaches to improving biomass to bio-fuel SCs, either by applying new techniques and methodologies for different entities or by focusing on the whole SC from a system perspective. The effect of integrating MRs along with FRs within a single SC to produce bio-oil from WB has remained unanswered. The next section reviews prior research to highlight the limitations of current approaches. 1.1. Literature review To understand the nature of bio-energy, it is important to introduce the different types of biomass that may be used in conversion centers (Frombo et al., 2009). Agricultural, municipal, and forest harvest and agricultural products can be used to produce biofuels, e.g., bio-ethanol, bio-methanol, bio-diesel and bio-oil. However, since the supply and cost of agricultural products, such as corn, are uncertain, recent studies have focused on utilizing biomass residues, such as FHR, to produce bio-energy (Aden et al., 2002). The USDA Forest Service suggested that approximately 73 million acres of national forest land in the United States have an excessive amount of forest biomass, which can be considered a reliable source for producing bio-oil (Troyer et al., 2003). Bio-oil can be produced from degradable biomass using the fast pyrolysis process. FHR is the most common biomass source used in fast pyrolysis (Mullaney et al., 2002). The use of low quality wood chips created from residues remaining after thinning and harvesting processes at roadside in the forest has a positive effect of creating job opportunities and improving the overall economy (Mullaney et al., 2002). The high costs of collection and transportation of FHR are considered a barrier to its use as compared to other forms of biomass. Since bio-oil (18e21 MJ/kg) has higher energy density than green whole tree chips (8.53 MJ/kg), transport of bio-oil is energetically more advantageous compared to raw biomass transport (Badger and Fransham, 2006). According to Badger (2002), the high moisture content levels and low energy density of WB, compared to fuel in liquid form, are the key drivers for high transportation cost. An analytical SC model developed by Allen et al. (1998) found that 20e50% of the overall cost of SC can be attributed to transportation activities. Several studies have investigated the trade-offs between process operation and transportation costs. Polagye et al. (2007) and Granatstein et al. (2009) argued that the biofuel production using mobile and/or transportable facilities is more costly than fixed and/ or relocatable facilities. Conversely, You and Wang (2011) and Yue et al. (2014) reported that a combination of pretreatment
193
facilities and upgrading facilities within a region of 60e135 km2 has lower cost than a fixed facility alone. Therefore, further study is needed to determine the competitiveness of mobile and transportable facility. A wheel-mounted and transportable refinery capable of processing WB at rate of 13.6 metric dry tons (15 US dry tons) per day has been fabricated and reported in the literature (Badger et al., 2010; Mirkouei and Haapala, 2015). The unit can be placed next to in-forest collection areas and produce bio-oil through the fast pyrolysis process. Due to the novel nature of MRs, previous studies focusing on improving the biomass SC network through mathematical models have failed to address a key question: How many MRs and FRs should work simultaneously in an integrated SC to process a specific amount of WB in an optimal manner? As Fig. 1 illustrates, the bio-fuel SC consists of feedstock production (WB), collection and transportation, processing and storage, and delivery to the end user or to storage. An optimal SC is dependent on critical decision making with regard to each of these SC entities, as well as at the system level, to ensure all goals and needs are met. The major goal is to meet the bio-oil demand of the region by processing FHR left at roadside after log making in selected forests. A critical need is to determine optimal SC costs, which can be used to define the lowest cost transportation and capital infrastructure alternatives for the system. The issues of harvesting and collection, storage location and layout, and delivery have been investigated in previous studies. Collection and harvesting methods have been analyzed to create an in-depth solution to avoid biomass logistics pitfalls. The cost analysis of harvesting switch grass in round bales and an economic study comparing round and square bales of switch grass have been performed by Cundiff and Marsh (1996), respectively. The impact of covered on-field storage on the delivery cost, intermediate storage scenarios, and adjacent storage layout on the overall SC cost have been analyzed (Cundiff et al., 1997; Papadopoulos and Katsigiannis, 2002; Tatsiopoulos and Tolis, 2003). In most cases, low-cost storage layouts have been identified. Rentizelas et al. (2009) addressed a key research gap in the storage and collection stage by proposing three types of storage layouts for agricultural biomass, i.e., an adjacent warehouse with drying capability, metal roof storage, and ambient storage without drying infrastructure. Handling and storage of WB is different than agricultural biomass, since WB needs to be chipped after collection to be ready for the conversion process. Thus, the decision of implementing centralized or decentralized chipper equipment has been studied by Gronalt and Rauch (2007) through a heuristic approach. In addition to research focusing on the impact of individual stages on the whole SC, other studies have developed various methods and approaches to improve the biomass SC from a systems perspective. Sandia National Laboratories provided a dynamic model that considers various types of biomass to meet the national demand for cellulosic ethanol (West et al., 2009). Zhang et al. (2013) proposed a mixed integer linear programming (MILP) model to minimize the overall cost of producing ethanol from switch grass while considering all stages in the SC network. Another MILP model was developed by Zhu et al. (2011) to assess restrictions on harvesting seasons and scattered geographical distribution on the €rheden (1989) preoverall system performance. Eriksson and Bjo sented a linear programming (LP) model to minimize the transportation cost of pellet fuel from numerous supply sites to a central heating plant. However, this model failed to guide decision makers about whether to use more harvesting area or sawmills to meet demand. This drawback was addressed by Gunnarsson et al. (2004) by proposing a large and comprehensive MILP model for a forest fuel network.
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
194
Fig. 1. Schematic biomass to bio-oil supply chain network.
1.2. Motivation Previous mathematical models are limited in their ability to answer the question posed above in two key aspects: 1) WB has not been considered as a feedstock to produce bio-oil and 2) the role of a MR has not been addressed when accompanied with an FR in the SC. The goal of the mathematical model presented in this study, therefore, is to determine the number of FRs and MRs that are needed to process a known amount of FHR while minimizing overall SC cost. The cost optimization model developed below focuses on transportation costs and the capital and operational costs of mobile and fixed bio-oil refineries. Subsequently, the environmental impact, in terms of carbon footprint (mass of CO2 equivalent emissions), is assessed for various alternatives. Truck travel time is an important factor to consider in answering where the MRs should be located and which harvesting area will be served by each refinery in fixed and mobile bio-oil refinery decision making. Geographical information system (GIS)-based approaches have been widely used in prior research to address transportation and location-allocation problems. Muttiah et al. (1996) implemented GIS in a decision support system to identify waste disposal sites, for example. A public facility location was also identified by Yeh and Chow (1996) by integrating a GIS approach with a locationallocation model. In the model presented below, truck travel distance is considered. In the research reported herein, a combination of FRs and MRs that minimizes cost is identified by developing and applying an MILP mathematical model. GIS software is used to calculate the shortest path between harvesting areas and refineries. Next, carbon footprint (CF) analysis is undertaken to estimate the environmental impact of the resulting SC. Finally, the effect of this decision making approach on the design of a hypothetical bio-oil SC in northwest Oregon, USA is explored, and the effect of key factors on the selected economic and environmental measures is analyzed through sensitivity analysis. 2. Methodology The MILP model presented in this research is developed to minimize WB to bio-oil SC cost. Operation of each fixed and mobile refinery in a predetermined population to define the SC network is facilitated using a binary variable in the mathematical model. The major cost elements of the model include transportation cost, refinery operational and capital costs, and storage cost. Since the focus of the model is on the optimal number of fixed and mobile refineries needed to process a known amount of WB, the model considers the supply side, rather than the demand
side of the SC, i.e., it is assumed that all bio-oil will be utilized in existing markets. Sections 2.1 and 2.2 explain the formulation of the objective function and constraints in greater detail, while Section 2.3 describes the approach for evaluating environmental impacts. 2.1. Model objective function The objective of the proposed MILP model is to minimize the overall cost of a mixed-mode biomass SC by considering two major cost elements. First, the transportation activities involved in the network consist of delivering WB from the forest to the mobile or FR using both in-forest and main roads, as well as delivering bio-oil from the MR to the bio-oil storage at an FR location. Second, costs account for establishing and operating the FRs and MRs, and for the bio-oil and biomass storage facilities adjacent to the FRs. The objective function (Z) to be minimized is represented in Eq. (1).
Min Z ¼ C1 þ C2 þ C3 þ C4 þ C5 þ C6 þ C7 þ C8 þ C9
(1)
Each of the terms (C1eC9) in Eq. (1) is defined in sequence below. The model nomenclature is reported in Annex A. The cost (C1) of WB transportation from each harvesting area to the selected MR for time period (t) is calculated (Eq. (2)) as the summation of the product of Dijt, the shortest distance between the harvesting area (i) and the MR location (j), Cs, operational cost of an in-forest tractor trailer (small tractor trailer), and nijt, the required number of in-forest tractor trailer trips for each route.
C1 ¼
XXX i
j
Dijt $Cs $nijt
(2)
t
The cost (C2) of in-forest transportation of WB from each harvesting areas to the main road junction (k) is calculated (Eq. (3)) as the summation of the product of Dikt, the shortest distance from the harvesting area to the main road, Cs, operational cost of an in-forest tractor trailer (single trailer), and nikt, required number of in-forest tractor trailer trips for each route.
C2 ¼
XXX i
k
Dikt $Cs $nikt
(3)
t
The cost (C3) of on-road (highway) transportation of WB from each main road junction to the FR (l), is calculated (Eq. (4)) as the summation of the product of Dklt, the shortest distance from the highway to the FR, C2s, operational cost of an on-highway tractor trailer (truck tractor with two trailers), and nklt, the required number of on-highway tractor trailer trips to haul the biomass transferred from the in-forest tractor trailers for each route. Travel
195
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
cost on a specific road class is considered directly proportional to distance.
C3 ¼
XXX
nijt Xijt CAP s
ci 2I; cj 2J; ct 2T
(11)
(4)
nikt Xikt=CAP
ci 2I; ck 2K; ct 2T
(12)
In-forest trucks have a smaller capacity but greater maneuverability, as reflected in Equations (2)e(4) (Schroeder et al., 2007). The cost (C4) of transporting bio-oil from a mobile to an FR is calculated (Eq. (5)) as the summation of the product of Djlt, the shortest distance from the MR to the FR, Ctnk, operational cost of a bio-oil tanker, and njlt, the required number of bio-oil tankers trips to transport produced bio-oil for each route.
nklt Xklt=CAP
ck 2K; cl 2L; ct 2T
(13)
cj 2J; cl 2L; ct 2T
(14)
k
C4 ¼
l
XXX j
l
Dklt $C2s $nklt
t
Djlt $Ctnk $njlt
(5)
t
Equations (6)e(9) calculate the annual cost of the FRs (C5), the MRs (C6), and associated bio-oil (C7) and biomass (C8) storage facilities. In this model, it is assumed that the bio-oil produced by an MR will be held in storage adjacent to the FR. These formulations are defined as the products of Cfix, Cmob, Cst_oil, and Cst_mass, which are the annualized capital and operational costs of the FR, MR, biooil storage facility, and biomass storage facility, respectively, and the related binary variables (i.e., a, b, g, and d) indicating whether entity is in operation. Operational costs for each refinery consist of electricity, grinding or chipping, chemical supplies, and natural gas, which are dependent on the size of the refinery.
C5 ¼
XX l
C6 ¼
C7 ¼
C8 ¼
(7)
glct $Csto il
(8)
dlct $Cstm ass
(9)
t
XXX k
l
. njlt Yjlt CAP tnk
Equations (15) and (16) ensure that only a specific amount of WB (determined by percentage yield, S) can be transformed to bio-oil. Percentage yield is highly dependent on the moisture content; as moisture content of WB increases, the percentage yield will decrease. Equation (17) ensures the conservation-of-flow constraints for any flow in and out of node k (forest and main road junction).
S
XX
Xijt
i2I t2T
S
X X
XX
Yjlt ¼ 0
ci 2I; cl 2L; ct 2T
(15)
l2L t2T
Xklt
k2K t2T
X X X
X
Ylct ¼ 0
ck 2K; ct 2T
(16)
t2T
Xikt
i2I k2K t2T
X XX
Xklt ¼ 0
ci 2I; cl 2L; ct 2T
k2K l2L t2T
(17) Equations (18) and (19) ensure that the flow of WB from harvesting areas is possible to the fixed and/or mobile refineries, respectively, when the selected refinery is operating.
X
Ylct M$alt
ct 2T
(18)
XX
Xijt M$bjt
ci 2I; ct 2T
(19)
i2I t2T
The holding cost (C9) of WB is calculated (Eq. (10)) as the summation of the product of Cin, the inventory cost of storing WB, and Xklt, the amount of WB that will be stored in each biomass storage facility.
C9 ¼
b
t2T
t
XX l
bjt $Cmob
t
XX l
(6)
t
XX j
alt $Cfix
m
Cin $Xklt
(10)
t
Since the fast pyrolysis processing technology is assumed to be the same for each type of refinery, the processing costs (e.g., collection and grinding costs) are not considered in the formulation of the objective function. Recent research has modeled economies of scale of bio-oil production, but uncertainties remain regarding scalability of pyrolysis reactors (Arbogast et al., 2012).
2.2. Model constraints The constraints applied in optimizing the objective function presented in the previous section are defined below. Equations (11)e(14) ensure that the number of in-forest tractor-trailer trips (for transport to MRs or to main road junctions), on-highway tractor-trailer trips, and bio-oil tanker truck trips, respectively, have sufficient capacity to transport the WB and bio-oil.
Equations (20) and (21) ensure that the bio-oil produced in a MR and biomass transported from harvesting areas respectively will be transported to storage adjacent to an FR if the refinery is operating. This constraint also allows bio-oil and biomass to be distributed among different FRs to optimize the flow of these products in the SC.
XX
Yjlt þ
j2J t2T
X
Ylct M$glct
cj 2J; ct 2T
(20)
t2T
X X
Xklt M$dlct
ck 2K; ct 2T
(21)
k2K t2T
Equations (22) and (23) ensure that the amount of bio-oil produced from each MR and/or FR does not exceed the capacity of the selected refinery.
XX
Yjlt CAPmob
cl 2L; ct 2T
(22)
l2L t2T
X
Ylct CAPfix
ct 2T
(23)
t2T
Equation (24) ensures that the amount of WB held at the fixed facility does not exceed the storage capacity, while Eq. (25) determines the amount of bio-oil stored in the bio-oil storage facility during a specific period of time (t). Equation (26) ensures that the
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
available amount of WB (q) in each harvesting area is transported to either an MR or an FR.
X X
Xklt CAPstm ass
ck 2K; ct 2T
(24)
196
emissions from WB collection, transportation, and refinery processing. 3. Application of the model
k2K t2T
XX
Yjlt þ
j2J t2T
Ylct CAPsto il
cj 2J; ct 2T
(25)
t2T
XXX i2I j2J t2T
q
X
Xijt þ
X X X
Xikt
i2I k2K t2T
ci 2I; cj 2J; ck 2K; ct 2T
(26)
Equations (27)e(30) are the binary constraints, Equation (31) the integer constraint, and Equation (32) the non-negativity constraint, respectively, and are applied to ensure the solution is feasible.
alt bjt
1 0
if fixed refinery is working; otherwise
(27)
1 0
if mobile refinery is working; otherwise
(28)
1 0
if biooil storage is working; otherwise
(29)
1 0
if biomass storage is working; otherwise
(30)
glct dlct
A hypothetical case for WB to bio-oil processing is presented below for a four-county region located in northwest Oregon. Actual harvesting data and available information in the literature have been used to assess the reasonableness of the mathematical model presented in the previous section.
nijt ; nikt ; nklt ; and njlt are integers
(31)
Xijt ; Xikt ; Xklt ; and Yjlt 0
(32)
In this section, the formulations of the objective function and constraints required for calculating the optimal number of FRs and MRs have been described. Next, Section 3 focuses on applying the mathematical model to a hypothetical case for specific region of northwest Oregon using realistic data. 2.3. Environmental impacts evaluation Currently, the economics of bioenergy utilization indicate that the applied technologies are not promising compared to conventional (fossil-based) energy, while environmental assessments have demonstrated bioenergy products (e.g., bio-oil, bio-fuel, and bio-jet fuel) can improve air and water quality. Therefore, utilization of bioenergy is expected to increase to accommodate environmental needs and pressures (McKendry, 2002), and proposed bioenergy technologies will need to be evaluated with regard to environmental impacts. Several criteria have been considered to evaluate environmental impacts of bioenergy production systems, e.g., net energy yield and CF. Net energy yield analysis has been used to measure the energy efficiency and sustainability of bioenergy from biomass (Schmer et al., 2008). CF analysis refers to the total amount of greenhouse gases (GHGs) emitted throughout the bioenergy production system, measured in mass of CO2 equivalent (kg CO2 eq.). Transportation contributes a significant amount of CF in biomass to bioenergy SCs (Lam et al., 2010). Since transforming WB to bio-oil at bio-refineries is novel, there is little literature focusing on environmental assessment of the existing production technologies and SCs. The environmental impact assessment method applied herein estimates the CF, including CO2, CH4, and N2O
3.1. Background and assumptions Data for WB amount and harvesting area locations have been provided by the State of Oregon Department of Forestry (ODF, 2015). Data for three forest districts, i.e., Astoria, Tillamook, and Forest Grove obtained from ODF include timber sale harvest data, a GIS layer for 49 harvesting areas, and forest road and highway GIS layers. The GIS layers containing spatial information for counties located in Oregon are from the U.S. Bureau of Land Management (US BLM, 2013). Non-merchantable products remaining at roadside in the forest after harvesting are considered as FHR, and defined as products that lack sufficient quality to be used for pulpwood. According to ODF, red alder, western hemlock, and Douglas-fir tree species are the primary timber species. Moisture content is assumed constant (45% MC, wet basis) for all FHR. The transport of FHR is assumed to be mass-limited. The three forest districts are scattered across four counties, i.e., Tillamook, Clatsop, Columbia, and Washington counties. No large scale bio-oil refineries are operating in Oregon, thus, it was decided that potential fixed bio-oil refinery sites would be located at the center of each county. Fig. 2 illustrates the distribution of harvesting areas and the initial location of FRs across the region. The potential locations for MRs were decided based on the geographical distribution of harvesting areas scattered across the three forest districts. Therefore, one MR is located among the harvesting areas that are geographically closest to each other. Consequently, 16 MR locations were selected throughout the region to serve the harvest areas (Fig. 3). Biomass from each harvest area was assumed to be processed by the nearest MR. Since the time horizon applied in this study is one year, the capital and operational costs of FRs and MRs are dependent to two major factors: the expected operational life and the period of operation in a specific location. Due to its mobility, the latter factor is more appropriate for MRs. To determine annualized cost, the expected life of the MR was assumed to be 10 years (Page-Dumroese et al., 2009); the depreciated life has been considered same for the FR. 3.1.1. Description of mobile and fixed refineries The capacity, operational cost, and capital cost of an MR were calculated with respect to the data developed by Mullaney et al. (2002) and Page-Dumroese et al. (2009). The annual operational cost for an MR was estimated using an empirical model (Eq. (33)), which is based on the linear interpolation of operational cost per year (Cop_mob) for different capacities (metric tons per day) of fixed refineries CAPfix.
Cop_mob ¼ 7454:2$CAP fix þ 70; 527
(33)
The operational cost for FRs has been extracted from work by Mullaney et al. (2002). Attributes of both refineries considered in this application are summarized in Table 1. The biomass storage cost was obtained from a study developed by Rentizelas et al. (2009) for a covered storage facility without
197
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
Fig. 2. Distribution of harvesting areas and potential locations of fixed refineries.
Fig. 3. Distribution of mobile refineries among harvesting areas.
Table 1 Summary of mobile and fixed refinery attributes. Attributes
Mobile
Fixed
Refinery size (metric tons per day, tpd) Biomass (45% MC) processing rate (metric tons per year) Bio-oil production capacity (thousand liters per year) Refinery capital cost (US $) Expected operational/depreciated life (years) Refinery operational cost (US $ per year) Biomass storage facility capital cost (US $) Bio-oil storage facility capital cost (US $) Bio-oil product yield (%)
13.6 4950 1650 1,584,890 10 196,320 e e 50
363 132,000 43,930 15,396,680 20 5,212,610 732,550 1,516,870 50
198
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
drying infrastructure, which has a capital cost of $156/m2, adjusted for inflation to 2014 (BLS, 2015). It is assumed to have a square footprint (28 m per side) and a height of 6 m. Due to lack of information and for simplicity the flow of biomass is considered as just-in-time, thus, the transformation process begins as the batches of woody residue arrive at the refineries. As a result, inventory (holding) cost is not considered in this application. For bio-oil storage cost, two steel tanks, each with the capacity of about 28,600 m3 (180,000 bbl.), are selected to be located next to an FR (B2-Consultants LLC, 2013). This capacity is assumed to accommodate the oil production of the FR, as well as that of MRs located throughout the surrounding region. Bio-oil from MRs is brought by the tankers to the storage facility locations prior to shipment, thus this storage is not limited only to the production of the harvest areas presented in this study. The processing cost per metric ton for both the FRs and MRs have been assumed to be same in this study. The annual cost of an FR (Cfix) is calculated (Eq. (34)) as the summation of the capital cost of the refinery (Ccap_fix) divided by its expected operational life (Kfix) and the prorated operational cost, which is based on the operational cost (Cop_fix), the amount of biomass processed, and the annual processing rate (Rann_fix). The annual cost of an MR (Cmob is calculated (Eq. (35)) by considering its prorated cost based on its annual processing rate (Rann_mob), assuming the ability to serve multiple harvest areas throughout the year, the amount of biomass processed, operational costs (Cop_mob), and capital cost (Ccap_mob) divided by the expected operational life of the refinery (Kmob). The annual cost of a biomass storage facility (Cst_mass) is calculated (Eq. (36)) divided by the total capital and operational cost of storage (TCst_mass) by the expected operational life of the biomass storage facility (Kst_mass). The annual cost of a bio-oil storage facility (Cst_oil) is calculated (Eq. (37)) by dividing the total capital and operational cost of storage (TCst_oil) by the expected operational life of the bio-oil storage facility (Kst_oil).
, Ccap_fix X X X ¼ þ Xklt * Cop_fix Rann_fix Kfix t
Sharp Ridge Forest Roads
Forest RoadHighway Junction
Highway
Tillamook County Fig. 4. Shortest path between the Sharp Ridge harvesting area and the fixed refinery located in Tillamook County.
their limited capacity. For longer distance transport, a trailer can be attached to double capacity. In the case of the additional trailer, the hook-lift truck shuttles bins between a hook-up point along an improved road and the in-forest site. This study used a hook-lift truck with capacity of 13.6 metric tons (15 U.S. tons) for hauling biomass on in-forest roads. If on-highway hauling is used, an additional trailer is added at the highway road junction, doubling capacity. The cost of hook-up and unhook-up cost is $120 per round trip. The capacity of tanker transport is assumed to be 18.2 metric tons (20 U.S. tons). The truck operating cost of $3.26 per mile ($2.02 per km) is adjusted for inflation to 2014, based on the value $2.98 per mile ($1.85 per km) reported by Mason et al. (2008), which includes wages and fixed and variable costs.
(35)
3.1.4. Calculation of carbon footprint According to the U.S. Environmental Protection Agency (U.S. EPA), the CF for heavy and medium duty trucks is 0.298 kg CO2 eq. per US ton-mile (0.328 kg CO2 eq. per metric ton-mile) (U.S. EPA, 2008). According to Steele et al. (2012) the cradle-to-grave environmental impact of producing bio-oil is 0.0323 kg CO2 eq. per MJ of bio-oil, and includes collection of biomass and the pyrolysis process.
Cst_mass ¼ TCst_mass =Kst_mass
(36)
4. Results and discussion
Cst_oil ¼ TCst_oil Kst_oil
(37)
The results of the SC cost model are presented in this section for a base case and several alternative scenarios. Also, the environmental impact of SC alternatives is evaluated by assessing their CF, a key environmental performance indicator for bio-based energy SCs. The alternative scenarios are developed to assess the effect of model factors on predicted economic and environmental measures. The results for each scenario are also discussed.
Cfix
k
(34)
l
2 3 , XXX Ccap_mob 6 7 ¼ þ Cop_mob $4 Xijt R 5 Kmob ann_mob t
Cmob
i
j
3.1.2. Calculation of travel distance The shortest path between harvesting areas and refineries and between the FRs and MRs is calculated by developing a road network in GIS software (ArcGIS 10). The road GIS layer provides two main categories of roads (forest roads and highways); all forest roads feed into highways. Fig. 4 depicts the path between one harvesting area (Sharp Ridge) and the FR located in Tillamook County. The distance from the nearest highway location to the refinery is assumed to be negligible. 3.1.3. Truck operating cost The operating cost per mile varies with respect to the type of terrain traversed, as well as truck size. Straight frame hook-lift trucks can be used for biomass transport for in-forest roads, but they are primarily used for short distance delivery of WB due to
4.1. Computational results The mathematical model presented in Section 2 is solved through an optimization solver (Gurobi 5.6.3) along with Python 2.7 (Gurobi Optimization Inc., 2014; Van Rossum, 2007). The model has 2127 constraints, 28 binary variables, 1088 integer variables, and 1092 continuous non-negative variables. Gurobi solves the problem to optimality after 6786 simplex iterations and 9.12 s. For the case presented, with 16 potential mobile and four FR locations, the results indicate that the optimal SC would consist of 14 MRs that are located within all four counties. The MRs would be
199
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
responsible for processing WB from all harvesting areas except Modified Green, Grease Alder, and Mombo Combo, which are in Clatsop County. The MR would process almost all of the WB available in the harvesting areas served. The total production for the region is estimated to be about 4,537,205 L (4537.2 m3 or 1,198,733 gal.) of bio-oil. The optimal annual cost for the base case is calculated as $1,004,443 or $0.221/L ($0.84/gal.) by considering the transportation costs and annualized capital and operational costs for each facility. Table 2 reports key parameter values for the optimal solution for this base case. Table 3 represents the estimated CF for the considered biomass collection, transportation, FRs and MRs for this case. 4.2. Sensitivity analysis According to the framework of the model presented in Section 2, several factors can have a crucial effect on the economic and environmental performance of the SC. The purpose of the sensitivity analysis is to evaluate the effect of major factors on the base case results presented in Section 4.1, i.e., MR capital cost, operational life, and operational cost, available WB, and storage facility cost. Two cases, in addition to the base case, are considered for each factor and the cost and carbon footprint results for each case are compared (Table 4). 4.2.1. Effect of mobile refinery capital cost The effects of MR capital cost and operational life are investigated in this section. In the first alternative scenario (Case 1), the capital cost of an MR has been decreased by 50% and in the second scenario (Case 2), the operational life of the refinery is increased by 10 years in addition to the capital cost reduction. These scenarios reflect MR technology improvement. According to the modeling results (Fig. 5), the optimal number and location of the FR and MR would remain the same compared to the base case with the changes explored for capital cost. The amount of WB processed by the MR increases in Case 1 and Case 2. The overall cost decreased by about $167,711 (17%), in Case 1, and by about $250,933 (30%) in Case 2. The CF remains constant in both cases since the location and number of refineries, along with the quantity of WB processed by each remains constant. The model indicates that the optimal location and environmental impact appear to be independent respectively to changes in the capital cost and operational life of the MRs. 4.2.2. Effect of mobile refinery operational cost The effect of the operational cost of the MR on the overall SC cost is investigated by developing two scenarios. In the first (Case 3), the operational cost is increased by 50% and in the second scenario (Case 4), this cost is decreased by 50%. The optimal number and location of FRs and MRs have changed in Case 3. The amount of WB processed by each refinery remains constant. The overall cost of the system, however, changes directly as a function of the operational
Table 3 Environmental impact of transportation activities and refinery infrastructure. Supply chain entity
Carbon footprint (Mg CO2 eq.)
Transportation Fixed refinery processing Mobile refinery processing Biomass collection Total
168.5 e 4372.8 102.2 4643.5
cost of the MR (Fig. 6). Hence, the annual cost in Case 3 is increased by about $65,527 (6.5%) and the annual cost in Case 4 is reduced by $216,887 (21.6%). This analysis demonstrates that changes to the operational cost of the MR appear to have effect on the optimal SC configuration. In Case 3, the results indicate that the optimal SC would consist of an FR rather than MRs in the base case and the FRs would be responsible for processing WB from all harvesting areas. Conversely, in Case 4, the results of the analysis indicate that 14 MRs would be responsible for processing WB. 4.2.3. Effect of available woody biomass The amount of WB considered in the base case was based on non-merchantable products remaining at roadside after thinning and final harvest processes. This source of woody residue is not the only feedstock source for the bio-oil process. Branches, tops, and breakage, not at roadside, can be another source of WB (forest slash), much of which would be piled and burned. According to ODF, roughly 2268e2722 metric tons (2500e3000 U.S. tons) are burned in each state forest zone. This amount is greater for private forests, equating to about 11,340e11,794 metric tons (12,500e13,000 U.S. tons) for each forest zone in this study. In the first scenario (Case 5), the state forest slash is added to the nonmerchantable product to determine the mass of WB. It is assumed that 8165 metric tons (9000 U.S. tons) of slash and 2722 metric tons (3000 U.S. tons) slash per forest zone, are uniformly distributed among the 49 harvesting areas. In the second scenario (Case 6), privately-owned forest slash is added to the amount used in Case 5. This amount is calculated as 29,937 metric tons (33,000 U.S. tons) total and uniformly distributed among the 49 harvesting areas. Collection cost of slash is not considered. The results indicate that MRs would not be used in Case 5 and Case 6. With increased biomass, the annual cost increases by about $103,260 (10%) in Case 5 (Fig. 7). This indicates increasing the volume of biomass leads to an increase in transportation costs, which are offset by elimination of capital and operating costs of the MR. In Case 6, the cost is increased by about $229,004 (23%), due to the additional trips necessary for WB transport. Commensurately, the CF increases by 1,476,205 kg CO2 equivalent (32%) in Case 5 and by 6,367,912 kg CO2 equivalent (137%) in Case 6. 4.2.4. Effect of storage facility cost Two scenarios have been developed to evaluate the effect of storage facility capital costs on modeling results. In the first
Table 2 Parameter details for the optimal solution of the mathematical model for the base case. Parameter description
Name
Value
Woody biomass transferred to the mobile refinery (metric tons) Woody biomass transferred to the fixed refinery (metric tons) Number of in-forest truck trips Number of highway truck trips Number of bio-oil tanker trips between mobile and fixed refineries Bio-oil produced by the mobile refinery (metric tons) Bio-oil produced by the fixed refinery (metric tons)
Xijt Xikt, Xklt nijt nklt njlt Yjlt Ylct
10,886 0 806 0 304 5443 0
200
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507 Table 4 Summary description of each case scenario.
1 2 3 4 5 6 7 8 9 10
Objective
Description
Effect Effect Effect Effect Effect Effect Effect Effect Effect Effect
Capital cost of a mobile refinery has been decreased by 50% Operational life of the refinery is increased by 10 years Operational cost is increased by 50% Operational cost is decreased by 50% 7407 metric tons biomass added 27,158 metric tons added Biomass and bio-oil storage cost is increased by 50% Biomass and bio-oil storage cost is reduced by 50% MRs are located 10 miles (16 km) farther away MRs are located 30 miles (48.2 km) farther away
of of of of of of of of of of
mobile refinery capital cost mobile refinery capital cost mobile refinery operational cost mobile refinery operational cost available woody biomass available woody biomass storage facility cost storage facility cost mobile refinery locations mobile refinery locations
Storage Cost
TransportaƟon Cost
TransportaƟon Cost
TransportaƟon CF
TransportaƟon CF
Fixed Refinery Processing CF
Mobile Refinery Processing CF
Biomass CollecƟon CF
Mobile Refinery Processing CF
Biomass CollecƟon CF
5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0
1.0 0.8 0.6
0.4 0.2 0.0 Base Case
Case 1
12.0
1.2
10.0
1.0
8.0
0.8 6.0 0.6 4.0
0.4
2.0
0.2
0.0
0.0 Base Case
Fig. 5. Effect of mobile refinery capital cost on environmental and economic measures.
Case 5
Case 6
Fig. 7. Effect of available woody biomass on environmental and economic measures.
Storage Cost
TransportaƟon Cost
Mobile Refinery Processing Cost
Storage Cost
TransportaƟon CF
Fixed Refinery Processing CF
TransportaƟon Cost
TransportaƟon CF
Mobile Refinery Processing CF
Biomass CollecƟon CF
Mobile Refinery Processing CF
Biomass CollecƟon CF
1.0 0.8 0.6 0.4 0.2 0.0 Base Case
Case 3
1.2
Carbon Footprint (Gg CO2 eq.)
5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0
1.2
Annual Cost ($ million)
1.4
Case 2
Case 4
Fig. 6. Effect of mobile refinery operational cost on environmental and economic measures.
scenario (Case 7), the cost of establishing biomass and bio-oil storage is increased by 50% and in the second scenario (Case 8) these costs are reduced by 50%. The results show no change in the optimal combination refineries from the base case, nor in the amount of WB processed by the MR, and consequently, the CF (Fig. 8). The overall cost of the system, however, has a direct correlation with facility capital cost. The overall cost is increased by about $75,843 (7.5%) in Case 7 and is decreased by the same percentage in Case 8 ($75,843). The results show that the optimal
Annual Cost ($ million)
Annual Cost ($ million)
1.2
Annual Cost ($ million)
Storage Cost
Carbon Footprint (Gg CO2 eq.)
Mobile Refinery Processing Cost
Carbon Footprint (Gg CO2 eq.)
Case Case Case Case Case Case Case Case Case Case
5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0
1.0 0.8 0.6 0.4 0.2 0.0 Base Case
Case 7
Carbon Footprint (Gg CO2 eq.)
Case
Case 8
Fig. 8. Effect of storage cost on environmental and economic measures.
solution is not sensitive to the changes in the bio-oil and biomass storage costs. 4.2.5. Effect of mobile refinery locations The impact of the distance of the MR from the forest, a primary factor in decision making, is investigated in this section. In the first scenario (Case 9), MRs are located 10 miles (16 km) farther than the original location. In the second scenario (Case 10), MRs are located at a distance of 30 miles (48.2 km) from the original locations, while holding all other parameters constant. The latter is farther than
201
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
Mobile Refinery Processing Cost
Storage Cost
TransportaƟon Cost
TransportaƟon CF
Fixed Refinery Processing CF
Mobile Refinery Processing CF
Biomass CollecƟon CF
Annual Cost ($ million)
5.0 1 4.0
0.8
3.0
0.6
0.4
2.0
0.2
1.0
0
0.0 Base Case
Case 9
Case 10
Carbon Footprint (Gg CO2 eq.)
Fixed Refinery Processing Cost
Fig. 9. Effect of mobile refinery location on environmental and economic measures.
would be realistic, but is used to evaluate sensitivity of the model. The overall SC cost and CF are found to be directly related to the distance (Fig. 9). The optimal number of the MRs remains constant in Case 9. When the distance increases to 30 miles (16 km), however, the results indicate that MRs would not be used and all WB would be processed with a single FR. Also, the total quantity of WB processed remains unchanged. The annual cost is increased by about $26,587 (2.6%) in Case 9 and by about $65,571 (6.5%) in Case 10. CF is increased by about 52,929 kg CO2 eq. in Case 9 and is increased by 1677 kg CO2 equivalent in Case 10. 5. Summary An MILP mathematical model was developed to obtain the optimal combination of fixed and mobile bio-oil refinery facilities to serve a given forested region. The optimization model was demonstrated using an optimization solver (Gurobi 5.6.3) along with Python 2.7 to evaluate a set of three forest districts comprising 49 harvesting areas in northwestern Oregon. For a base case considering 16 mobile and four FR locations, the solver identified 12 MR locations. The amount of WB processed in the base case at the 12 selected refinery sites was 10,886 metric tons. The overall annual cost of the system, consisting of transportation, refinery (capital and operational), and storage costs, was calculated to be $1,004,443 or $0.221/L ($0.84/gal.) of bio-oil, adjusted for inflation to 2014. The environmental impact assessment estimated the CF of transportation and production of bio-oil. Transportation CF was determined using the U.S. EPA method (U.S. EPA, 2008), and CF of biomass collection and pyrolysis processing was based on work by Steele et al. (2012). The total annual CF for the base case was calculated to be about 4.643 106 Gg CO2 eq., or 1.02 kg CO2 eq./L. According to U.S. EIA (2015), the wholesale price of heating oil averaged $0.76/L ($2.91/gal.) in 2014, which is approximately 3.5 times higher than the calculated cost of bio-oil presented in this study. Additionally, the CF of crude oil production is 0.095 kg CO2 eq./MJ (Unnasch et al., 2009), which is about three times higher than for bio-oil production. Sensitivity analysis on modeling parameters was performed to estimate the effect of each factor on the results, including the SC configuration and associated cost and CF. The results of the sensitivity analysis indicated that operational cost of the MR, amount of WB, and transportation-related factors appeared to have a significant impact on the number and location of the facilities. As would
be expected, the overall cost of the system was found to be directly related to the capital cost of the storage facilities and the operational cost of the MR, i.e., as these costs increase, the overall cost of the system increases. Two scenarios were developed to examine the effect of changes in the MR capital cost to evaluate the impact of technology improvement. The results indicated that changes in MR capital cost had no impact on the optimal solution for the cases evaluated. Additionally, the location of the fixed facility was investigated to evaluate the effect of distance from the harvesting area on cost and CF modeling results. Table 5 indicates the total distances between each entity in base case, Case 3, Case 5, Case 6, Case 9, and Case 10. The analysis found that as distance between harvesting area and MR increases, the transportation cost will increase due to transferring more WB than bio-oil. Thus, adapting the role of new MR technology in the SC is more critical under long distance scenarios (e.g., increased in-forest truck capacities would aid in decreasing MR-related transportation costs). Further, the effect of distance on the amount of WB processed by the MR confirms the hypothesis that transportation activities greatly influence sustainability performance of the WB SC, measured by cost and CF. Another factor examined in the sensitivity analysis was the available amount of WB. The available amount of WB was first increased by adding state in-forest slash to the non-merchantable product mass used in the base case (Case 5) and, in Case 6, by adding private and state forest slash to the base amount. The results showed that an increase in the volume of available WB led to an increase in transportation costs and CF. The solution also eliminated the MR and accepted higher transportation costs over the increased cost of MRs. Comparisons between the base case scenario and other scenarios in the sensitivity analysis are summarized in Table 6. 6. Conclusions Earlier studies focused on upstream biomass to bioenergy SC cost considered harvesting/collection, logistics, and storage, while pretreatment was not investigated. Mirkouei et al. (under review) reviewed the literature in detail by examining the conventional upstream segment biomass-to-bioenergy SCs structure. The work presented here addresses two limitations of modeling and optimization of bio-energy SCs. First, while previous studies have focused on optimizing the SCs for producing other types of forest fuels, the cost model developed is the first reported that focuses on SC optimization for the use of in-forest WB for bio-oil production. Second, the mathematical model developed in this study is the first reported that optimizes an integrated SC for a combination of fixed and mobile bio-oil refineries. Finally, it is the first known study to simultaneously investigate economic and environmental impact measures for the bio-oil SC as indicators of sustainability performance. In spite of the advances made in this work, there are limitations in the model that should be addressed by future research to Table 5 Sum of round trip distances between each entity for different case scenarios. Forest to mobile Mobile refinery to Forest to main Main road to refinery (km) storage facility road (km) fixed refinery (km) (km) Base Case Case Case Case Case a
casea 3 5 6 9 10
29,914 e e e 56,510 e
25,691 e e e 403,823 e
e 33,114 56,846 97,934 e 33,050
e 36,257 80,778 136,050 e 36,268
Other cases not reported in the table are the same as the base case.
202
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507 Table 6 Summary comparisons between different case scenarios. Case
Total biomass (thousands of metric tons)
Total bio-oil (thousands of liters)
Total carbon footprint (Mg CO2 eq.)
Total cost (thousands of dollars)
Unit cost ($/L)
Unit carbon footprint (kg CO2 eq./L)
Base case Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8 Case 9 Case 10
10.89 10.89 10.89 10.89 10.89 18.29 38.04 10.89 10.89 10.89 10.89
4536 4536 4536 4536 4536 7622 15,851 4536 4536 4536 4536
4643 4643 4643 4645 4643 6270 11,661 4643 4643 4701 4645
1004.4 836.7 753.5 1069.9 787.5 1107.7 1233.4 1080.2 928.6 1031.0 1070.0
0.22 0.18 0.17 0.24 0.17 0.15 0.08 0.24 0.20 0.23 0.24
1.02 1.02 1.02 1.02 1.02 0.82 0.74 1.02 1.02 1.04 1.02
improve the accuracy and relevance of results in the design of biomass SCs. The model was implemented in a non-dynamic case study, considering a one-year time period, which neglected the impact of dynamic changes, e.g., scheduling impacts due to weather, demand, and downtime, and the availability of trucks and processing equipment. Since timber harvesting varies by season, the time of year can have a significant impact on the overall cost, e.g., additional storage and inventory cost or more transportation activities in certain months of the year. Also, since distance-finding was done manually using GIS data to determine forest road and highway distances individually from two map layers, only routes from each harvest area to the nearest MRs were determined. Due to the manual method, MR locations were fixed, which inhibits optimization of refinery locations. Thus, if the MRs were not operating or had reached capacity, biomass would be transported to an FR, as opposed to the next nearest MR. Another limitation is related to the assumption of mass-limited biomass transport, which may be the case for high-density, non-merchantable product, but FHRs such as tops and limbs can widely vary in density, causing transport to be volume-limited. Thus, WB density can have a profound impact on the required number of trips from the harvesting area to the refinery. The generalized mathematical model and assessment approach developed for evaluating the economic and environmental performance of bio-oil production SCs is complex, and the application to a relatively small geographic region demonstrated herein required many assumptions to simplify analysis and facilitate optimization using integer programming and a solver package. Evaluation of the system, however, was able to demonstrate insights into the factors that influence the SC configuration and associated costs and CFs. It is critical that mathematical approaches, such as that explored here, continue to be developed so informed system-level decisions can be made that will advance the development and implementation of robust and sustainable energy systems. Acknowledgments The authors would like to thank Michael Wilson and Rob Nall of the Oregon Department of Forestry for their helpful contributions to this research. Annex A. Nomenclature
Indices c fix i
set of storage sites fixed bio-refinery set of harvesting sites
in j k l mob s st_mass st_oil t tnk
inventory set of mobile sites set of main-road junctions set of fixed refinery sites mobile bio-refinery small tractor-trailer biomass storage bio-oil storage time period tanker trucks
Parameters Ccap_fix capital cost of fixed refinery (US $) Ccap_mob capital cost of mobile refinery (US $) annual cost of a fixed refinery (including capital and Cfix operational costs) (US $) annual inventory cost of woody biomass ($/metric ton) Cin annual cost of a mobile refinery (including capital and Cmob operational cost) (US $) operational cost of fixed refinery (US $) Cop_fix Cop_mob operational cost of mobile refinery (US $) transportation cost of small in-forest tractor-trailer to Cs mobile refinery (US $/mile) transportation cost of on-highway tractor-trailer to fixed C2s refinery (US $/mile) Cst_mass annual cost of biomass storage at a fixed refinery (US $) annual cost of bio-oil storage at a fixed refinery (US $) Cst_oil transportation cost of tanker trailer to fixed refinery (US Ctnk $/mile) capacity of small tractor-trailer with two trailers (metric CAP2s ton) capacity of large tractor-trailer (metric ton) CAPb capacity of fixed refinery (metric ton) CAPfix CAPmob capacity of mobile refinery (metric ton) capacity of medium tractor-trailer (metric ton) CAPm capacity of small tractor-trailer (metric ton) CAPs CAPst_mass capacity of biomass storage facility (metric ton) CAPst_oil capacity of bio-oil storage facility (metric ton) capacity of tanker truck (metric ton) CAPtnk distance between a harvest site and a mobile refinery Dijt (miles) distance between a harvest site and a main road junction Dikt (miles) distance between a main road junction and a fixed Dklt refinery (miles) distance between a mobile refinery and a fixed refinery Djlt (miles) expected operational life of the fixed refinery (year) Kfix expected operational life of the mobile refinery (year) Kmob
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507
Kst_mass Kst_oil M Rann_fix Rann_mob S TCst_mass TCst_oil
q
expected operational life of the biomass storage facility (year) expected operational life of the bio-oil storage facility (year) large positive constant (Big M) annual processing rate of fixed refinery (metric ton) annual processing rate of mobile refinery (metric ton) percentage yield total capital and operational costs of biomass storage facility (US $) total capital and operational costs of bio-oil storage facility (US $) the available amount of woody biomass in all harvesting sites (metric ton)
Continuous variables amount of biomass from site i to site j at time t (metric Xijt ton) amount of biomass from site i to site k at time t (metric Xikt ton) amount of biomass from site k to site l at time t (metric Xklt ton) amount of bio-oil from site j to site l at time t (metric ton) Yjlt amount of bio-oil from site l to site c at time t (metric ton) Ylct Binary variables alt binary bjt binary glct binary dlct binary
variable variable variable variable
for for for for
fixed refinery at time t mobile refinery at time t bio-oil storage at time t biomass storage at time t
Integer variables number of in-forest tractor-trailer trips to transfer woody nijt biomass form harvesting area to nearest mobile refinery at time t number of in-forest tractor-trailer trips to transfer woody nikt biomass from harvesting area to nearest main road junction at time t number of highway tractor-trailer trips to transfer woody nklt biomass from main road junction to nearest fixed refinery at time t number of tanker trailer trips to transfer bio-oil from njlt mobile refinery to fixed refinery at time t References Aden, A., Ruth, M., Ibsen, K., Jechura, J., Neeves, K., Sheehan, J., Wallace, B., 2002. Lignocellulosic Biomass to Ethanol Process Design and Economics Utilizing CoCurrent Dilute Acid Prehydrolysis and Enzymatic Hydrolysis for Corn Stover. National Renewable Energy Laboratory. Allen, J., Browne, M., Hunter, A., Boyd, J., Palmer, H., 1998. Logistics management and costs of biomass fuel supply. Int. J. Phys. Distrib. Logist. Manag. 28, 463e477. Arbogast, S., Bellman, D., Paynter, J.D., Wykowski, J., 2012. Advanced bio-fuels from pyrolysis oil: the impact of economies of scale and use of existing logistic and processing capabilities. Fuel Process. Technol. 104, 121e127. http://dx.doi.org/ 10.1016/j.fuproc.2012.04.036. B2-Consultants LLC, 2013. Costs for Fixed Roof Oil Storage Tanks w/tdn 30 D 10 W. Badger, P., Badger, S., Puettmann, M., Steele, P., Cooper, J., 2010. Techno-economic analysis: preliminary assessment of pyrolysis oil production costs and material energy balance associated with a transportable fast pyrolysis system. BioResources 6, 34e47. Badger, P.C., 2002. Processing Cost Analysis for Biomass Feedstocks. DOE/EERE. Badger, P.C., Fransham, P., 2006. Use of mobile fast pyrolysis plants to densify biomass and reduce biomass handling costsda preliminary assessment. Biomass Bioenergy 30, 321e325. http://dx.doi.org/10.1016/j.biombioe.2005. 07.011. BLS, 2015. Producer Price Index (PPI). Bureau of Labor Statistics, United States Department of Labor. Bridgwater, A.V., 2012. Review of fast pyrolysis of biomass and product upgrading. Biomass Bioenergy 38, 68e94. http://dx.doi.org/10.1016/j.biombioe.2011.01
203
.048. Overcoming Barriers to Bioenergy: Outcomes of the Bioenergy Network of Excellence 2003e2009. Cundiff, J.S., Dias, N., Sherali, H.D., 1997. A linear programming approach for designing a herbaceous biomass delivery system. Bioresour. Technol. 59, 47e55. http://dx.doi.org/10.1016/S0960-8524(96)00129-0. Cundiff, J.S., Marsh, L.S., 1996. Harvest and storage costs for bales of switchgrass in the southeastern United States. Bioresour. Technol. 56, 95e101. Czernik, S., Bridgwater, A.V., 2004. Overview of applications of biomass fast pyrolysis oil. Energy Fuels 18, 590e598. http://dx.doi.org/10.1021/ef034067u. € rheden, R., 1989. Optimal storing, transport and processing for a Eriksson, L.O., Bjo forest-fuel supplier. Eur. J. Oper. Res. 43, 26e33. Frombo, F., Minciardi, R., Robba, M., Rosso, F., Sacile, R., 2009. Planning woody biomass logistics for energy production: a strategic decision model. Biomass Bioenergy 33, 372e383. Granatstein, D., Kruger, C.E., Collins, H., Galinato, S., Garcia-Perez, M., Yoder, J., 2009. Use of Biochar from the Pyrolysis of Waste Organic Material as a Soil Amendment. Final project report. Center for Sustaining Agriculture and Natural Resources, Washington State University, Wenatchee, WA. Gronalt, M., Rauch, P., 2007. Designing a regional forest fuel supply network. Biomass Bioenergy 31, 393e402. €nnqvist, M., Lundgren, J.T., 2004. Supply chain modelling of Gunnarsson, H., Ro forest fuel. Eur. J. Oper. Res. 158, 103e123. Gurobi Optimization, Inc, 2014. Gurobi Optimizer Reference Manual. Kersten, S., Garcia-Perez, M., 2013. Recent developments in fast pyrolysis of lignocellulosic materials. Curr. Opin. Biotechnol. 24, 414e420. Lam, H.L., Varbanov, P., Klemes, J., 2010. Minimising carbon footprint of regional biomass supply chains. Resour. Conserv. Recycl. 54, 303e309. http://dx.doi.org/ 10.1016/j.resconrec.2009.03.009. Mason, C.L., Casavant, K., Lippke, B., Nguyen, D., Jessup, E., 2008. The Washington Log Trucking Industry: Costs and Safety Analysis. University of Washington and Washington State University. http://www.ruraltech.org/pubs/reports/2008/log_ trucks/index.aspd. McKendry, P., 2002. Energy production from biomass (part 2): conversion technologies. Bioresour. Technol. 83, 47e54. Mirkouei, A., Haapala, K.R., 2015. A network model to optimize upstream and midstream biomass-to-bioenergy supply chain costs. In: ASME 2015 International Manufacturing Science and Engineering Conference (MSEC), MSEC20159355. Charlotte, NC. Mirkouei, A., Haapala, K.R., Sessions, J., Murthy, G.S., 2015. A review and future directions in techno-economic modeling and optimization of upstream forest biomass to bio-oil supply chains. Renew. Sustain. Energy Rev. (under review). Mullaney, H., Farag, I., LaClaire, C., Barrett, C., 2002. Technical, Environmental and Economic Feasibility of Bio-Oil in New Hampshire's North Country. UNH Project. Muttiah, R.S., Engel, B.A., Jones, D.D., 1996. Waste disposal site selection using GISbased simulated annealing. Comput. Geosci. 22, 1013e1017. ODF, 2015. State of Oregon: Oregon Department of Forestry. http://www.oregon. gov/odf/Pages/index.aspx (accessed 10.12.15.). Page-Dumroese, D., Coleman, M., Jones, G., Venn, T., Dumroese, R.K., Anderson, N., Chung, W., Loeffler, D., Archuleta, J., Kimsey, M., et al., 2009. Portable in-woods pyrolysis: using forest biomass to reduce forest fuels, increase soil productivity, and sequester carbon. In: Proceedings of the 2009 North American Biochar Conference. Http://cees. Colorado. Edu/biochar_production. Html (accessed 05.01.10.). Papadopoulos, D.P., Katsigiannis, P.A., 2002. Biomass energy surveying and technoeconomic assessment of suitable CHP system installations. Biomass Bioenergy 22, 105e124. http://dx.doi.org/10.1016/S0961-9534(01)00064-2. Polagye, B.L., Hodgson, K.T., Malte, P.C., 2007. An economic analysis of bio-energy options using thinnings from overstocked forests. Biomass Bioenergy 31, 105e125. http://dx.doi.org/10.1016/j.biombioe.2006.02.005. Rentizelas, A.A., Tolis, A.J., Tatsiopoulos, I.P., 2009. Logistics issues of biomass: the storage problem and the multi-biomass supply chain. Renew. Sustain. Energy Rev. 13, 887e894. http://dx.doi.org/10.1016/j.rser.2008.01.003. Schmer, M.R., Vogel, K.P., Mitchell, R.B., Perrin, R.K., 2008. Net energy of cellulosic ethanol from switchgrass. Proc. Natl. Acad. Sci. 105, 464e469. Schroeder, R., Jackson, B., Ashton, S., Hubbard, W., Biles, L., Mayfield, C., Ashton, S., 2007. Biomass Transportation and Delivery. Searcy, E., Flynn, P., Ghafoori, E., Kumar, A., 2007. The relative cost of biomass energy transport. In: Applied Biochemistry and Biotechnology, pp. 639e652. Sharma, B., Ingalls, R.G., Jones, C.L., Khanchi, A., 2013. Biomass supply chain design and analysis: basis, overview, modeling, challenges, and future. Renew. Sustain. Energy Rev. 24, 608e627. http://dx.doi.org/10.1016/j.rser.2013.03.049. Steele, P., Puettmann, M.E., Kanthi Penmetsa, V., Cooper, J.E., 2012. Life-cycle assessment of pyrolysis bio-oil production. For. Prod. J. 62, 326. Tatsiopoulos, I.P., Tolis, A.J., 2003. Economic aspects of the cottonestalk biomass logistics and comparison of supply chain methods. Biomass Bioenergy 24, 199e214. http://dx.doi.org/10.1016/S0961-9534(02)00115-0. Troyer, J., Mann, Rex, King, Michael, Berry, John, Hitchcock, Neal, Frye, Steve, Artley, Don, DuFault, Art, Keller, Paul, 2003. Large Fire Cost Reduction Action Plan. USDA Forest Service, USDI, and the National Association of State Foresters. Unnasch, S., Wiesenberg, R., Tarka Sanchez, S., Brandt, A., Mueller, S., Plevin, R., 2009. Assessment of Life Cycle GHG Emissions Associated with Petroleum Fuels. Life Cycle Associates Report LCA-6004-3P. Prepared for New Fuels Alliance. US BLM, 2013. Downloadable Data and Map Services.
A. Mirkouei et al. / Journal of Cleaner Production 113 (2016) 495e507 U.S. EIA, 2015. U.S. Weekly Heating Oil and Propane Prices (OctobereMarch) [WWW Document]. http://www.eia.gov/dnav/pet/pet_pri_wfr_dcus_nus_w. htm (accessed 10.30.15.). USEIA, 2014. Annual Energy Outlook 2014 with Projections to 2040 (No. DOE/EIA0383 (2014)). Energy Information Administration, United States Department of Energy, Washington, D.C., USA. U.S. EPA, 2008. Climate Leaders: Greenhouse Gas Inventory Protocol Core Module Guidance, Direct Emissions from Mobile Combustion Sources (No. EPA430-K08-004). Van Rossum, G., 2007. Python programming language. In: USENIX Annual Technical Conference. West, T., Dunphy-Guzman, K., Sun, A., Malczunski, L., Reichmuth, D., Larson, R., Ellison, J., Taylor, R., Tidwell, V., Klebanoff, L., Hough, P., Lutz, A., Shaddix, C., Brinkman, N., Wheeler, C., O'Toole, D., 2009. Feasibility, Economics, and Environmental Impact of Producing 90 Billion Gallons of Ethanol per year by 2030 (No. Paper 86). US Department of Energy Publications. Sandia National Laboratories.
204
Yeh, A.G.-O., Chow, M.H., 1996. An integrated GIS and location-allocation approach to public facilities planningdan example of open space planning. Comput. Environ. Urban Syst. 20, 339e350. http://dx.doi.org/10.1016/S0198-9715(97) 00010-0. Decision Support Systems in Urban Planning. You, F., Wang, B., 2011. Life cycle optimization of biomass-to-liquid supply chains with distributedecentralized processing networks. Ind. Eng. Chem. Res. 50, 10102e10127. Yue, D., You, F., Snyder, S.W., 2014. Biomass-to-bioenergy and biofuel supply chain optimization: overview, key issues and challenges. Comput. Chem. Eng. 66, 36e56. http://dx.doi.org/10.1016/j.compchemeng.2013.11.016. Selected papers from ESCAPE-23 (European Symposium on Computer Aided Process Engineering e 23), 9e12 June 2013, Lappeenranta, Finland. Zhang, J., Osmani, A., Awudu, I., Gonela, V., 2013. An integrated optimization model for switchgrass-based bioethanol supply chain. Appl. Energy 102, 1205e1217. http://dx.doi.org/10.1016/j.apenergy.2012.06.054. Zhu, X., Li, X., Yao, Q., Chen, Y., 2011. Challenges and models in supporting logistics system design for dedicated-biomass-based bioenergy industry. Bioresour. Technol. 102, 1344e1351. http://dx.doi.org/10.1016/j.biortech.2010.08.122.
205
Appendix B: A Network Model to Optimize Upstream and Midstream Biomass-to-Bioenergy Supply Chain Costs
Proceedings of the ASME 2015 Manufacturing Science and Engineering Conference MSEC2015 June 8-12, 2015 in Charlotte, North Carolina, USA
MSEC2015-9355 A NETWORK MODEL TO OPTIMIZE UPSTREAM AND MIDSTREAM BIOMASS-TOBIOENERGY SUPPLY CHAIN COSTS Amin Mirkouei School of Mechanical, Industrial and Manufacturing Engineering Oregon State University Corvallis, Oregon, USA
[email protected]
Karl R. Haapala School of Mechanical, Industrial and Manufacturing Engineering Oregon State University Corvallis, Oregon, USA
[email protected]
ABSTRACT Growing awareness and concern within society over the use of and reliance on fossil fuels has stimulated research efforts in identifying, developing, and selecting alternative energy sources and energy technologies. Bioenergy represents a promising replacement for conventional energy, due to reduced environmental impacts and broad applicability. Sustainable energy challenges, however, require innovative manufacturing technologies and practices to mitigate energy and material consumption. This research aims to facilitate sustainable production of bioenergy from forest biomass and to promote deployment of novel processing equipment (mobile bio-refinery units). The study integrates knowledge from the renewable energy production and supply chain management disciplines to evaluate economic targets of bioenergy production with use of qualitative and quantitative techniques. The decision support system method employs two phases: (1) classification of potential biomass harvesting sites via decision tree analysis and (2) optimization of the supply network through a mixed integer linear programming model that minimizes the costs of upstream and midstream supply chain segments. While mobile units are shown to reduce biomass-to-bioenergy supply chain costs, production and deployment of the units is limited due to undeveloped bioenergy supply chains and quality uncertainty. It is reiterated that future research must address process-related and systemic issues in pursuit of sustainable energy technology development.
INTRODUCTION Research into modeling and analysis of renewable energy production technologies and systems has increased over the last few decades due to related positive economic, environmental, and social impacts, e.g., access to clean energy and growth of
new labor markets. Investigators have paid much attention to approaches that meet cost targets and reduce carbon footprint e.g., emissions of carbon dioxide, methane, and nitrous oxides [1]. Bioenergy is renewable – biomass can be harvested and replenished in a relatively short timeframe. However, renewable energy sources exhibit a high-energy balance, the ratio of production energy to the energy released during consumption, compared to other sources, such as fossil fuels. According to the United States Energy Information Administration [2], over 80% of U.S. energy is provided through conventional fuels such as gas, coal, and petroleum (Figure 1). Renewable energy sources provide 8% to 10% of total U.S. energy consumption, and half of this renewable energy demand is provided through biomass. Thus, biomass plays a key role in the renewable energy industry and attention should be placed on optimizing the production, conversion, and use of biomass in producing bioenergy [2], [3]. The manufacturing research community has initiated several efforts investigating improved biomass processing methods and biomass-to-bioenergy supply chain optimization, as introduced briefly below. Solar Geothermal Wind Hydroelectric Biomass
1% 5% 9%
Renewable Nuclear Electric Power Coal Natural Gas Petroleum
8% 9%
Renewable Energy Sources
35% 50%
All Energy Sources
21% 25% 37% 0%
20%
40%
60%
80%
100%
Figure 1. SOURCES OF ENERGY IN THE U.S. [2]
V002T05A011-1
Copyright © 2015 by ASME
207 Zhang et al. [4] proposed a new pelleting method to increase the density of cellulosic biomass, which has several benefits such as reducing transportation and storage costs. Since location is a significant transportation cost driver, Zhang et al. [5] introduced a two-stage methodology to identify the best location for biofuel production based on multiple attributes. They considered several variables, such as fuel price, transportation distance, and feedstock availability. Zhang et al. [6] pursued size reduction strategies for converting biomass to bioenergy to understand the effects of particle size and biomass crystallinity in cellulosic biofuel production. In addition, Cui et al. [7] addressed a new approach for growing algae on solid carriers that can lead to meeting the cost targets for algae biofuels. High volume algae production for biofuels is expensive through existing methods. A successful bioenergy industry would benefit society by reducing energy costs and reducing the imports of fossil fuel as well as improving energy security and health benefits [8]. The development and deployment of bioenergy sources is limited, however, due to supply uncertainties and lack of flexible, robust, and cost-effective production technologies and systems [9], [10] The development of production technologies, production systems, and supply chains must occur in concert to realize the expected economic, environmental, and social benefits of an established bioenergy industry. Several challenges must be overcome before achieving an efficient, optimized system that will ensure viability and sustainability of the bioenergy network at all decision levels. A key challenge is minimization of bioenergy production costs. High costs of biomass feedstock supply inhibit the development of a strong bioenergy market [4]. Thus, a hybrid method is needed that will simultaneously minimize associated costs and overcome barriers associated with competitive markets for the upstream and midstream segments of biomass-to-bioenergy supply chains (B2BSC). However, in the investigation here, environmental and social aspects are not addressed directly. Optimal and sustainable B2BSC management requires the following: (1) choice of highly productive biomass, (2) use of advanced coordination of biomass supply (e.g., collection and transportation) at different decision levels (e.g., strategic, tactical, and operational), and (3) use of efficient conversion facilities that can mitigate bioenergy production costs [11], [12]. In a broad sense, supply chain management considers all logistics network activities, which includes challenges of integrating energy processing and production in B2BSCs. Figure
Biomass
Harvesting
Collection
2 indicates each upstream and midstream entity in the B2BSC network, as well as interrelationships and interdependencies among the entities. A motivation of this research is to examine the substitution of conventional energy sources with renewable energy sources through informed technology development, which will lead to improvements of social, environmental, and economic aspects of energy production, supply, and use. Strategic decision making examines internal and external perspectives from diverse disciplines. Thus, this study develops a detailed strategic decision model of B2BSC costs to satisfy conflicting objectives of decision makers. Ongoing work is developing a strategic design model of B2BSC environmental impacts. Both will form an integrated, multi-objective sustainability assessment model. The models can be utilized to consider several decision making objectives as follows: • Biomass source location and collection methods, • Transportation modes and routing, and • Production technology methods and locations. In pursuing these objectives, research must identify methods that can contend with optimization aspects and overcome challenges of traditional approaches discussed above. The approach herein utilizes multi-criteria decision making to explore quantitative and qualitative aspects of B2BSC to assist industry decision makers in determining optimal solutions for minimizing production costs. First, related prior work is reviewed. The decision-making methodology is then presented and demonstrated using a case study. Finally, conclusions and opportunities for future research are discussed.
BACKGROUND The academic and industrial communities have been intensively investigating novel uses of bioenergy sources for several decades. Investigators have placed recent scrutiny on the production logistics for relevant bioenergy products, including forest biomass [13], [14]. Research has mainly focused on developing quantitative models to represent existing systems and addressing related logistics issues, such as transportation and production planning [1], [15]. The results of this work have led to the establishment of various concepts in the domain of traditional bioenergy sources and the integration of key contributions across a broader range of areas including
Transportation
Upstream
Processing
Conversion
Bioenergy Products
Midstream
Figure 2. UPSTREAM AND MIDSTREAM BIOMASS-TO-BIONERGY SUPPLY CHAIN ENTITIES
V002T05A011-2
Copyright © 2015 by ASME
208 production planning, operations management, and supply chain decision making. The need for further investigation is not only through empirical studies, but also through the development of conceptual studies that can integrate methods and tools for improving B2BSC performance [9], [16]. B2BSC development is supported by academia, government and industry [14], [17]–[19]. Several studies highlight the need to integrate supply chain processes with either conventional energy sources (e.g., gas and diesel) or bioenergy sources (e.g., biogas and biodiesel). Several methods and metrics have been classified by researchers to enhance the performance of bioenergy processing. In understanding the effect of biomass on production, conversion, and use, it is important to evaluate the type or types of biomass to be considered. The major energy-use sources are solid biomass (e.g., crops and straw), wet biomass (e.g., manure and organic waste), oil seeds (e.g., sunflower and rapeseed), and plant materials (e.g., sugar and starch). Products of converted biomass include heat and power, oils and fuels, and chemicals, each of which has specific properties that impact effectiveness in desired applications [18]. For example, biopower can be used for heat and power generation and has environmental benefits over conventional power sources, while biofuels are a potential option for transportation systems that can have lower environmental impacts than conventional fuels. The advantages of using biomass to produce bioenergy have been identified as [8], [19]: (1) excellent availability across the world, (2) potential of using agricultural and forestry wastes and byproducts, (3) potential of reducing fire hazards, and (4) the carbon-neutral nature of biomass combustion. Identified drawbacks include [20]: (1) high transportation costs and environmental impacts due to high moisture content (low energy density), (2) large storage area requirements due to biomass properties (low material density), and (3) high production costs because of a low number of existing bio-refineries (long transportation distances). Biomass supply is generally coordinated through purchases of feedstocks from suppliers or through harvesting, collection, and transportation by the company from the forest or field. Many research efforts in bioenergy supply chain management have applied mathematical programming models that represent the existing systems through the objective function and real world constraints to provide the optimal outcomes [15]– [17]. In general, mathematical programming is categorized as four distinct techniques: (1) linear programming, (2) integer programming, (3) mixed integer linear programming, and (4) non-linear programming [14]. Among these techniques, mixed integer linear programming (MILP) is the most common optimization method because it has capabilities to be applied in different types of decision problems. For bioenergy supply chain optimization, the MILP approach makes it feasible to consider integer variables representing discrete units of equipment. While the economic objective is most frequently addressed among the mathematical programming models [14], the models have differences in the specific characteristics, such as objective
function, decision variables, and constraints. Different types of objective functions are considered at the strategic decision level for economic objective problems, e.g., minimize total cost [21], maximize overall profits [22], maximize net present value [23], maximize revenue [24], minimize risk on investment [25], and minimize transportation costs [26]. Transportation optimization problems consider decision variables such as transportation mode, capacity, and routing. Common alternative biomass transport modes include roll-off container trucks, hook-lift container trucks, dump trucks, and logging trucks. Each of these modes is dependent on the circumstance and collection approach in the forest or field [27]. Due to equipment limitations, collection mode can have a significant effect on biomass availability and collection costs. Similarly, different production technologies can be applied to produce bioenergy from biomass, which impact costs and environmental impacts. In general, energy production pathways are classified as thermochemical and biochemical [28], [29]. Since these processes have different characteristics, investigators need to consider issues such as type of biomass and preferred bio-products when developing optimization and sustainability analyses. Biochemical conversion can, for example, convert corn stover to ethanol using chemical processing. Thermochemical conversion technology is explored in this research, and involves gasification and pyrolysis based conversions to convert biomass to bio-oil [15], [30]. Gasification converts biomass into a syngas, tar, and a solid char at high temperature in absence of combustion. Fast pyrolysis converts biomass to bio-oil and biochar at temperatures between 250 and 550°C in the absence of oxygen [31]. To address several of the disadvantages mentioned above (e.g., low material and energy density), mobile bio-refinery units have been developed, including those that use a thermochemical conversion technology called fast pyrolysis [32]. The mobile bio-refineries are a truck-mounted unit that can travel to farms and/or forests, where the biomass feedstock is available. The mobile bio-refinery enables conversion of low-energy density biomass to high-energy density intermediate bio-products [33]. The main advantages of mobile bio-refineries are reported as [21], [24]: • Reduced capital and operational costs of establishing a bioenergy production system, • Reduced land area requirements for biomass storage, and • Reduced transportation costs due to transferring of denser bioenergy (e.g., bio-oil) instead of unprocessed biomass. Since their initial development, however, these units have not been adopted by industry, and their benefit from a cost and environmental impact perspective is unclear. In particular, the advantages of using mobile refining technology in place of centralized, or fixed, bio-refineries are likely situationally dependent on biomass type, transportation distances, time of the year, policies and regulations, and numerous other factors. The authors [34] previously proposed an integrated decision making method to evaluate the potential suppliers and select the best supplier to purchase biomass for bioenergy production
V002T05A011-3
Copyright © 2015 by ASME
209 through an optimization goal programming model, which addressed three main objectives: cost, quality, and delivery time. In addition, a cost model was developed and applied mixedmode (mobile and fixed units) biomass processing and bio-oil supply in northwest Oregon, which demonstrated the potential cost benefits of mobile units [3]. In the methodology presented below, this approach is extended to consider feedstock supply quality, multiple process outputs, and multi-criteria decision making. This research effort integrates both qualitative and quantitative techniques to assess criteria relevant to decision making, an approach which has not been previously reported in literature. It also considers the effect of harvesting and collection activities on the overall cost of a mixed-mode (mobile and fixed refining) biomass supply chain. Finally, most prior research has calculated the shortest path distance between harvesting sites and refineries to optimize cost. Rather than distance, the mathematical model herein considers travel time, which is more representative of actual accumulation of transportation costs in forest biomass supply chains due to low travel speed.
METHODOLOGY As discussed above, various approaches have been proposed in literature to design optimal biomass processing and bioenergy production networks. There are limited frameworks, however, that have combined both qualitative and quantitative techniques. The methodology reported here presents a decision method that integrates qualitative (Decision Tree Analysis) and quantitative (Mathematical Programming) decision support systems (Figure 3) and addresses the role of mobile bio-refinery unit in this network. The methodology proceeds in two phases to accomplish B2BSC network optimization. In Phase 1, the harvesting areas are classified according to the availability and quality of biomass feedstocks. In Phase 2, a mathematical optimization model is developed based on transportation time within the system. Since transportation is the main cost driver in B2BSC, transporting biomass to a central processing facility may not be cost effective. Thus, using mobile unit is a potential alternative to fixed refineries when the travel time from the forest to the refinery is high or if there is a low volume of biomass. It is crucial to note here that, due to the transport speeds, travel time is more indicative of transportation costs in this case than distances traveled. The phases of the methodology are described in more detail below.
Phase 1. Classification of Harvesting Area A decision tree (DT) analysis is proposed to evaluate and select the potential biomass harvesting areas based on availability and quality of biomass feedstock. An optimal decision tree has a low misclassification error rate in supervised classification learning and is a common method in data mining [35]. Thus, a DT analysis is developed and applied to identify the capacity and potential of each harvesting site. DT analysis is conducted with the assistance of R language [36].
Qualitative Decision Tree Analysis
Quantitative Mathematical Programming
Decision Support System Figure 3. MULTI-CRITERIA DECISION MAKING Phase 2. Mathematical Optimization Model A mathematical model is developed as a quantitative method to assess the feasibility and commercialization of mobile biorefinery unit for bioenergy production within the B2BSC network. A mixed integer linear programming (MILP) model is formulated in this study, with the aim of minimizing the capital and operational costs of B2BSC. A key component of the model is travel time, which is an important factor for selection of harvesting sites, choice of mobile bio-refinery locations, and consequent reduction of overall costs. Understanding truck travel time can assist in answering two key research questions: 1) Which sites will be harvested? 2) Where will mobile bio-refinery units be located? The mathematical optimization model takes into account the overall costs of the combination of mobile and fixed bio-refinery units. The final products of mobile bio-refinery units are bio-oil, which will be transferred to a fixed bio-refinery, and bio-char, which will be transferred to a distribution center. While the focus of this study is on the upstream and midstream segments of the B2BSC network, the downstream segment, which includes distribution and demand activities, will be considered in future work. Objective Function. The objective function (TC) addresses the total annual costs and includes the capital and operational costs of harvesting, collection, and processing, inventory, mobile bio-refineries, fixed bio-refineries, and transportation vehicles (Eq. 1). The model nomenclature is addressed in Annex A. Since transportation has significant impacts on the total costs, the model considers four truck types to address maneuverability on various forest roads and main roads: 1) Small trucks transport biomass from harvesting sites to collection sites, 2) Medium trucks transport biomass from collection sites to mobile refinery units, as well as transporting biomass and bio-oil to main roads, 3) Large trucks transport biomass from main roads to fixed refineries, and 4) Tanker trucks transport bio-oil from mobile units to fixed refineries.
V002T05A011-4
Copyright © 2015 by ASME
210 𝑀𝑀𝑀𝑀𝑀𝑀 𝑇𝑇𝑇𝑇 = � � ��𝐹𝐹ℎ 𝐵𝐵𝑖𝑖𝑖𝑖𝑖𝑖 + 𝑉𝑉ℎ 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 � 𝑖𝑖
𝑗𝑗
6) Ensures the number of truck-trailers is sufficient in terms of capacity of each truck to transport the biomass (Eq. 11-14) 7) Integer, non-negativity, and binary constraints to guarantee that the solution is feasible (Eq. 15-17).
(1)
𝑡𝑡
+ � � ��𝐹𝐹𝑐𝑐 𝐵𝐵𝑖𝑖𝑖𝑖𝑖𝑖 + 𝑉𝑉𝑐𝑐 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 � 𝑖𝑖
𝑗𝑗
𝑡𝑡
𝑗𝑗
𝑘𝑘
𝑡𝑡
𝑗𝑗
𝑘𝑘
𝑡𝑡
𝑗𝑗
𝑘𝑘
𝑡𝑡
𝑙𝑙
𝑡𝑡
𝑗𝑗
𝑖𝑖
𝑗𝑗
𝑡𝑡
𝑗𝑗
𝑘𝑘
𝑡𝑡
𝑗𝑗
𝑙𝑙
𝑡𝑡
𝑘𝑘
𝑙𝑙
𝑡𝑡
+ � � ��𝐹𝐹𝑝𝑝 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 + 𝑉𝑉𝑝𝑝 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 �
� 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 − � 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 − � 𝑊𝑊𝑗𝑗𝑗𝑗𝑗𝑗 = 0 𝑖𝑖∈𝐼𝐼
+ � � ��𝐹𝐹𝑤𝑤 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 + 𝑉𝑉𝑤𝑤 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 �
𝑀𝑀𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 ≥ 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗
𝑀𝑀𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 ≥ 𝑊𝑊𝑗𝑗𝑗𝑗𝑗𝑗
+ � � � ��𝐹𝐹𝑙𝑙 𝐵𝐵𝑙𝑙𝑙𝑙 + 𝑉𝑉𝑙𝑙 (𝑊𝑊𝑗𝑗𝑗𝑗𝑗𝑗 + 𝑍𝑍𝑘𝑘𝑘𝑘𝑘𝑘 )� + � � � �𝐹𝐹𝐹𝐹𝑠𝑠 𝐵𝐵𝑖𝑖𝑖𝑖𝑖𝑖 + 𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 𝑉𝑉𝑉𝑉𝑠𝑠
𝑙𝑙∈𝐿𝐿
𝑀𝑀𝐵𝐵𝑖𝑖𝑖𝑖𝑖𝑖 ≥ 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖
+ � � ��𝐹𝐹𝑚𝑚 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 + 𝑉𝑉𝑚𝑚 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 � 𝑘𝑘
𝑘𝑘∈𝐾𝐾
𝑀𝑀𝐵𝐵𝑘𝑘𝑘𝑘𝑘𝑘 ≥ 𝑍𝑍𝑘𝑘𝑘𝑘𝑘𝑘
𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 � 𝐶𝐶𝐶𝐶𝐶𝐶𝑠𝑠
𝑆𝑆 ∗ � 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 − � 𝑍𝑍𝑘𝑘𝑘𝑘𝑘𝑘 = 0
𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 + � � � �𝐹𝐹𝐹𝐹𝑚𝑚 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 + 𝑇𝑇𝑗𝑗𝑗𝑗𝑗𝑗 𝑉𝑉𝑉𝑉𝑚𝑚 � 𝐶𝐶𝐶𝐶𝐶𝐶𝑚𝑚
𝑗𝑗∈𝐽𝐽
𝑗𝑗∈𝐽𝐽 𝑘𝑘∈𝐾𝐾
𝑍𝑍𝑘𝑘𝑘𝑘𝑘𝑘 + � � � �𝐹𝐹𝐹𝐹𝑘𝑘 𝐵𝐵𝑘𝑘𝑘𝑘𝑘𝑘 + 𝑇𝑇𝑘𝑘𝑘𝑘𝑘𝑘 𝑉𝑉𝑉𝑉𝑘𝑘 � 𝐶𝐶𝐶𝐶𝐶𝐶𝑘𝑘
Model Constraints. The following constraints are applied in solving the optimization model: 1) Ensures the conservation-of-flow constraints for any flows in and out of node j (Eq. 2) 2) Transportation arc triggers to allow transport of material between collection sites (Eq. 3-6) 3) Percentage yield to ensure that a specific amount of the biomass can be transformed to bio-products (i.e., bio-oil and bio-char) (Eq. 7) 4) Capacity constraints to ensure that the amount of biomass and bio-products do not exceed the capacity of storage sites (Eq. 8-9) 5) Ensures the amount of transported biomass has to be equal to the sum of the available amount of biomass in each harvesting site (Eq. 10)
Mobile Refinery
Collection Facility Harvesting Sites
(3)
∀𝑗𝑗 ∈ 𝐽𝐽 , ∀𝑙𝑙 ∈ 𝐿𝐿 , ∀𝑡𝑡 ∈ 𝑇𝑇
(5)
∀𝑘𝑘 ∈ 𝐾𝐾 , ∀𝑡𝑡 ∈ 𝑇𝑇
(7)
∀𝑡𝑡 ∈ 𝑇𝑇
(9)
∀𝑗𝑗 ∈ 𝐽𝐽 , ∀𝑘𝑘 ∈ 𝐾𝐾 , ∀𝑡𝑡 ∈ 𝑇𝑇
(4)
∀𝑘𝑘 ∈ 𝐾𝐾 , ∀𝑙𝑙 ∈ 𝐿𝐿 , ∀𝑡𝑡 ∈ 𝑇𝑇
(6)
� � 𝑍𝑍𝑘𝑘𝑘𝑘𝑘𝑘 + � � 𝑊𝑊𝑗𝑗𝑗𝑗𝑗𝑗 ≤ 𝐶𝐶𝐶𝐶𝐶𝐶𝑙𝑙
𝑘𝑘∈𝐾𝐾 𝑙𝑙∈𝐿𝐿
𝑗𝑗∈𝐽𝐽 𝑙𝑙∈𝐿𝐿
� � 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑡𝑡 𝑖𝑖∈𝐼𝐼 𝑗𝑗∈𝐽𝐽
(2)
∀𝑖𝑖 ∈ 𝐼𝐼 , ∀𝑗𝑗 ∈ 𝐽𝐽 , ∀𝑡𝑡 ∈ 𝑇𝑇
𝑙𝑙∈𝐿𝐿
� � 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 ≤ 𝐶𝐶𝐶𝐶𝐶𝐶𝑚𝑚𝑚𝑚𝑚𝑚
𝑊𝑊𝑗𝑗𝑗𝑗𝑗𝑗 + � � � �𝐹𝐹𝐹𝐹𝑏𝑏 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 + 𝑇𝑇𝑗𝑗𝑗𝑗𝑗𝑗 𝑉𝑉𝑉𝑉𝑏𝑏 � 𝐶𝐶𝐶𝐶𝐶𝐶𝑏𝑏
∀𝑗𝑗 ∈ 𝐽𝐽 , ∀𝑡𝑡 ∈ 𝑇𝑇
∀𝑡𝑡 ∈ 𝑇𝑇
(8)
∀𝑡𝑡 ∈ 𝑇𝑇
(10)
𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖 ≥ 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 /𝐶𝐶𝐶𝐶𝐶𝐶𝑠𝑠
∀𝑖𝑖 ∈ 𝐼𝐼 , ∀𝑗𝑗 ∈ 𝐽𝐽 , ∀𝑡𝑡 ∈ 𝑇𝑇
(11)
𝑛𝑛𝑗𝑗𝑗𝑗𝑗𝑗 ≥ 𝑊𝑊𝑗𝑗𝑗𝑗𝑗𝑗 /𝐶𝐶𝐶𝐶𝐶𝐶𝑏𝑏
∀𝑗𝑗 ∈ 𝐽𝐽 , ∀𝑙𝑙 ∈ 𝐿𝐿 , ∀𝑡𝑡 ∈ 𝑇𝑇
(13)
𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑎𝑎𝑎𝑎 𝑖𝑖, 𝑗𝑗, 𝑘𝑘, 𝑙𝑙 𝑎𝑎𝑎𝑎𝑎𝑎 𝑡𝑡
(15)
𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑎𝑎𝑎𝑎 𝑖𝑖, 𝑗𝑗, 𝑘𝑘, 𝑙𝑙 𝑎𝑎𝑎𝑎𝑎𝑎 𝑡𝑡
(17)
𝑛𝑛𝑗𝑗𝑗𝑗𝑗𝑗 ≥ 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 /𝐶𝐶𝐶𝐶𝐶𝐶𝑚𝑚 𝑛𝑛𝑘𝑘𝑘𝑘𝑘𝑘 ≥ 𝑍𝑍𝑘𝑘𝑘𝑘𝑘𝑘 /𝐶𝐶𝐶𝐶𝐶𝐶𝑘𝑘
𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 , 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 , 𝑍𝑍𝑗𝑗𝑗𝑗𝑗𝑗 , 𝑊𝑊𝑘𝑘𝑘𝑘𝑘𝑘 ≥ 0
𝐵𝐵𝑖𝑖𝑖𝑖𝑖𝑖 , 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 , 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 , 𝐵𝐵𝑘𝑘𝑘𝑘𝑘𝑘 , 𝐵𝐵𝑙𝑙𝑙𝑙 = {0, 1} 𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖 , 𝑛𝑛𝑗𝑗𝑗𝑗𝑗𝑗 , 𝑛𝑛𝑗𝑗𝑗𝑗𝑗𝑗 , 𝑛𝑛𝑘𝑘𝑘𝑘𝑘𝑘 ≥ 0
Tanker
∀𝑗𝑗 ∈ 𝐽𝐽 , ∀𝑘𝑘 ∈ 𝐾𝐾 , ∀𝑡𝑡 ∈ 𝑇𝑇
(12)
∀𝑘𝑘 ∈ 𝐾𝐾 , ∀𝑙𝑙 ∈ 𝐿𝐿 , ∀𝑡𝑡 ∈ 𝑇𝑇
(14)
𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎𝑎𝑎𝑎𝑎 𝑖𝑖, 𝑗𝑗, 𝑘𝑘, 𝑙𝑙 𝑎𝑎𝑎𝑎𝑎𝑎 𝑡𝑡
(16)
Fixed Bio-refinery
Truck
Figure 4. A HYPOTHETICAL MIXED-MODE BIOMASS-TO-BIOENERGY PRODUCTION NETWORK
V002T05A011-5
Copyright © 2015 by ASME
211
4th Priority
2nd Priority
3rd Priority
1th Priority
Figure 5. CLASSIFICATION TREE FOR PROVIDING BIOMASS FROM POTENTIAL SITES (R RESULTS)
CASE STUDY A hypothetical case study is conducted to confirm the validity of the model and demonstrate the application to biomass-to-bioenergy supply chain management. Figure 4 indicates a mixed-mode network of mobile and fixed biorefineries for bioenergy production. Different road types are considered between harvesting sites and fixed bio-refineries. Based on the model, the potential locations for a mobile unit were selected from among the harvest areas close to collection sites. It is assumed that the mobile unit is stationed in a single location for the entire year. In this hypothetical case, business decision makers have a desire to reduce bioenergy production costs. A randomly generated dataset is used to provide volume and quality values for 100 harvesting sites containing different generations of forest biomass. Figure 5 indicates the platform of classification tree analysis for the first four priorities (first four potential harvesting sites), which starts with evaluating biomass quality levels and then evaluates the available biomass volume for each harvesting site. With this information, each site is classified into one of the four priority levels. It can be seen that harvesting sites with volume greater than 45 tons and quality greater than 67% (on an arbitrary scale that represents measures such as moisture content, particle size, grit contamination, and wood quality) are classified as a first priority. A random sample is chosen from the dataset, rather than stratifying by all items. Half of the data is used as a training set and the other half is the testing set. Besides considering availability and quality, Phase 1 can be expanded using this approach to define the appropriate priority for harvest areas using other criteria, e.g., distance and seasonality.
data development in prior research [3], [27], [37]. Table 1 depicts sizes of mobile and fixed bio-refineries, as well as related capital and operational costs. The effective lifetime of a bio-refinery is assumed to be ten years [3]. Table 1. SELECTED BIO-REFINERY ATTRIBUTES [27] Refinery Size (tpd) Capital Cost ($/year) Operational Cost ($/year)
Mobile Unit 15 147,200 182,340
Fixed Plant 400 1,430,000 3,052,272
tpd = tons per day
Table 2 presents capital and operational costs of different types of trucks, which are needed since different types of roads (in-forest roads and highways) accommodate different vehicle types. Small trucks have more flexibility and transportability on in-forest roads, while large trucks have higher travel speeds on the highway. The effective life of a truck is assumed as ten years. Table 2. TRUCK COSTS BASED ON CAPACITY [19] Capital Costs ($/year) Operational Costs ($/h)
Small 4,220.00 63.40
Medium 7,572.00 71.14
Large 12,834.00 88.40
Since speed is the distance covered per unit of time, truck travel time is calculated from the speed and shortest path between sites. The shortest path is calculated from geographic information system (GIS) software (Arc GIS 10.2.2) by developing a road network for northwest Oregon. Table 3 illustrates the average truck speed on different roads (loaded or unloaded) [27]. Table 3. TRAVEL SPEED ON DIFFERENT ROADS [19]
Modeling Overview From DT analysis, it is determined that the hypothetical network contains eight high potential harvesting sites, as well as two collection sites, a mobile bio-refinery unit, and a fixed biorefinery. The information and data are collected with respect to
Paved roads Gravel roads Dirt roads
Average Truck Speed (kmh) 70 (43 mph) 15 (9 mph) 10 (6 mph)
kmh = kilometers per hour; mph = miles per hour
V002T05A011-6
Copyright © 2015 by ASME
212 Capital and operational costs of harvesting, collection, and processing activities in the upstream segment of the B2BSC are presented in Table 4.
analysis evaluates the effect of two major factors: mobile unit location and fixed bio-refinery location. Apart from the base case, two different cases are presented below.
Table 4. COSTS OF UPSTREAM ENTITIES [19], [26] Collect 145,680 594,180
Process 334,595 592,410
Assumptions The current model formulation assumes the following: • The available amount of biomass is known • The type of truck for each type of route is known • The quality of biomass is unknown • The time horizon is one year • The loading and unloading costs at the forest road and main road junction are ignored.
Computational Results The mathematical optimization model was solved through an optimization solver (Gurobi 5.6.3) along with Python 2.7 [38], [39]. The model has 41 decision variables (which include 14 binary variables and 27 integer variables) and 46 constraints. The problem was solved after 33 iterations in 0.13 seconds, and the annual cost in this hypothetical network was predicted as $33,244,553 from the objective function. The model results for the presented hypothetical case include two collection sites, one mobile unit, and one fixed refinery, and indicate that available biomass would be processed in both the mobile unit and fixed plant. The predicted annual cost without considering the mobile unit was slightly higher, $35,132,235. The mobile unit allows the overall cost to be reduced by over $1,887,682 in this case. Since the throughput of the mobile unit was assumed to be 15 tons per day, the annual throughput is 5000 ton per year. In the case here, the amount of forest biomass to be processed is 13000 tons, with the remaining 8000 tons processed by the fixed facility. The analysis illustrates that the role of new mobile bioenergy production technology provides a benefit to the B2BSC network. In particular, the results demonstrate that overall production costs can be minimized when implementing mobile unit. In general, by exploring the model, it was found that the use of mobile unit is more suitable when the following criteria increase: travel time, travel distance, and unit transportation costs.
Sensitivity Analysis Based on the structure of the optimization model, several factors are seen to have a significant effect on the overall cost of the system, which can be examined using sensitivity analysis. The main purpose of sensitivity analysis is to assess the effect of basic variables, non-basic variables, and right hand side parameters on the optimization results. The presented sensitivity
Effect of Mobile Unit Location The distance between each entity has an effect on travel time and the overall cost of the proposed hypothetical network. The effect of travel time between collection sites and mobile biorefinery unit was investigated in two different cases. In the base case, the average travel time is 60 min. between collection sites and the mobile unit and 180 min. from the mobile unit to the fixed plant. In the first alternative case (Case 1), the location of the mobile unit was changed such that average travel time from collection sites is reduced by 50% (30 min.), and the average travel time between the mobile unit and the fixed plant was increased to 210 min. In the second case (Case 2), average travel time was increased to 90 min. from each collection site to each mobile unit, and the average travel time from each mobile unit to the fixed plant was decreased to 150 min. Table 5 reports the optimization results for each case. Table 5. EFFECT OF MOBILE UNIT LOCATION ON THE OVERALL ANNUAL COST Cases Base Case Case 1 Case 2
Overall Cost ($/year) 33,244,553 32,681,124 33,806,911
Travel Time (min.) Collection to Mobile Unit to Mobile Unit Fixed Plant 60 180 30 210 90 150
The results show that increasing and decreasing the travel time impacts the annual cost. The overall cost decreased in Case 1 by approximately $563,429 (1.7%) and in Case 2, the annual cost increased by $562,358 (1.7%). Figure 6 demonstrates the comparison of overall annual cost among these cases. 34 Overall Cost ( Millions $/year)
Capital Costs ($/year) Operational Costs ($/year)
Harvest 91,300 283,260
34 34 33 33 33 33 33 32 32 32 Base Case
Case 1
Case 2
Figure 6. EFFECT OF MOBILE UNIT LOCATION ON OVERALL ANNUAL COST
V002T05A011-7
Copyright © 2015 by ASME
213 Effect of Fixed Plant Locations The effect of fixed plant location is also investigated. In Case 3, the average travel time between the mobile unit and the fixed plant is decreased by 50%. In Case 4, the travel time is increased by 50%. Changing the travel time to the fixed plant impacts whether the mobile unit is used in the network. Table 6 reports the predicted annual cost when considering the use of the mobile unit and when not considering the use of the mobile unit for each case. Table 6. EFFECT OF FIXED PLANT LOCATION AND USE OF MOBILE UNIT ON THE OVERALL ANNUAL COST Cases Base Case Case 3 Case 4
Overall Cost With Mobile Unit ($/year) 33,244,553 30,137,996 36,350,039
Overall Cost Without Mobile Unit ($/year) 35,132,235 30,819,748 39,446,412
Travel Time (Mobile to Fixed) 180 90 270
Overall Cost (Millions $/year)
Cost is found to change directly when the travel time increases or decreases. In Case 3, the annual cost of a network with the mobile unit decreased by $3,106,557 (~9.5%) and without the mobile unit increased by $4,312,487 (~12%). Due to increasing the travel time in Case 4, the costs increased by $3,105,486 (~9.5%) and $4,314,177 (~12%) with the mobile unit and without the mobile unit, respectively. Figure 7 shows the effect of fixed refinery plant location on the annual cost of the hypothetical network. 50 40 30 20 10 0 Base Case
Case 3
Case 4
Overall Cost With Mobile Unit ($/year) Overall Cost Without Mobile Unit ($/year)
Figure 7. EFFECT OF FIXED REFINERY PLANT LOCATION ON OVERALL ANNUAL COST
CONCLUSIONS Growing energy demand and related concerns about energy security and climate change indicate that society requires new alternative energy sources. Studies dealing with biomass-tobioenergy production report that biomass will occupy a significant fraction of total electricity, heat, and transportation energy sources in the future. Subsequently, bioenergy technology development will play a strong role in energy supply, and represents a key future market for manufacturing industry. However, current market readiness has led to uncertainties and
undeveloped supply chains that inhibit the development of a sustainable and robust bioenergy market. The proposed method presents a decision support system for biomass-to-bioenergy supply chain (B2BSC) optimization, which includes two phases: classification tree analysis as a qualitative technique and mathematical programming model as a quantitative technique. The qualitative phase employs a decision tree (DT) analysis to classify the potential harvesting sites based on defined criteria (quality and volume) that have high impacts on bioenergy production. The quantitative phase formulates a mixed integer linear programming (MILP) model to optimize the cost of upstream and midstream segments of biomass processing and bioenergy production. The model considers both mobile and non-mobile (fixed) bio-refineries for a hypothetical case representing a realistic B2BSC network. This approach can be used by decision makers to design optimal bioenergy production systems and evaluate the applicability of mobile production units. From the above method and application, it was found that integration of qualitative and quantitative methods offers a promising approach to supplementing existing methods for B2BSC operation and management through the use of mobile bio-refinery technology. Since transportation is one of the main cost drivers in B2BSC networks, the model was intended to investigate the commercialization and feasibility of using mobile bio-refinery units to facilitate the production of bioenergy near the unprocessed biomass. The results of the presented method have shown that the optimal integration of mobile and fixed biorefineries can improve the robustness and reduce the overall cost of bioenergy supply chains. The benefits of this combination would be particularly true for meeting not only economic targets but also environmental targets, as transportation is a key source of environmental impacts in the conversion of biomass to bioenergy products. Future work should address assumptions of the model, explore the effect of environmental impacts, and consider the downstream segment (end use) of the B2BSC network. Future work should also evaluate the proposed method with actual B2BSC system data to assess its robustness and accuracy. This future research will further highlight the potential for new technology development, as well as the development of methodological decision-making approaches for the design of optimal and sustainable technologies and systems for biomassto-bioenergy production.
ACKNOWLEDGEMENTS The authors wish to acknowledge Ms. Rana Azghandi and Mr. Kamyar Raoufi for their assistance in this research, as well as the valuable input of Dr. John Sessions.
REFERENCES [1]
F. You and B. Wang, “Life Cycle Optimization of Biomass-to-Liquid Supply Chains with Distributed–
V002T05A011-8
Copyright © 2015 by ASME
214
[2]
[3] [4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
Centralized Processing Networks,” Ind. Eng. Chem. Res., vol. 50, no. 17, pp. 10102–10127, 2011. USEIA, “Annual energy outlook 2011 with projections to 2035,” Energy Information Administration, United States Department of Energy, Washington D.C., USA, DOE/EIA-0383 (2011), 2011. P. Mirzaie, “A supply chain model for optimizing fixed and mobile bio-oil refineries on a regional scale,” 2013. P. F. Zhang, Z. J. Pei, D. H. Wang, X. R. Wu, W. L. Cong, M. Zhang, and T. Deines, “Ultrasonic vibration-assisted pelleting of cellulosic biomass for biofuel manufacturing,” J. Manuf. Sci. Eng., vol. 133, no. 1, p. 011012, 2011. F. Zhang, D. M. Johnson, and J. W. Sutherland, “A GISbased method for identifying the optimal location for a facility to convert forest biomass to biofuel,” Biomass Bioenergy, vol. 35, no. 9, pp. 3951–3961, Oct. 2011. M. Zhang, X. Song, P. Zhang, Z. J. Pei, T. W. Deines, and D. Wang, “Size reduction of cellulosic biomass in biofuel manufacturing: separating the confounding effects of particle size and biomass crystallinity,” J. Manuf. Sci. Eng., vol. 135, no. 2, p. 021006, 2013. Y. Cui, W. W. Yuan, and Z. Pei, “Effects of Carrier Material and Design on Microalgae Attachment for Biofuel Manufacturing: A Literature Review,” pp. 525– 540, Jan. 2010. Food and Agriculture Organization of the United Nations, “State of the world’s forests,” 2013. [Online]. Available: http://www.fao.org/docrep/005/Y7581E/Y7581E00.HTM . [Accessed: 17-Oct-2014]. I. Awudu and J. Zhang, “Uncertainties and sustainability concepts in biofuel supply chain management: A review,” Renew. Sustain. Energy Rev., vol. 16, no. 2, pp. 1359– 1368, Feb. 2012. P. Badger, S. Badger, M. Puettmann, P. Steele, and J. Cooper, “Techno-Economic Analysis: Preliminary Assessment of Pyrolysis Oil Production Costs and Material Energy Balance Associated with a Transportable Fast Pyrolysis System,” BioResources, vol. 6, no. 1, pp. 34–47, 2010. R. Baños, F. Manzano-Agugliaro, F. G. Montoya, C. Gil, A. Alcayde, and J. Gómez, “Optimization methods applied to renewable and sustainable energy: A review,” Renew. Sustain. Energy Rev., vol. 15, no. 4, pp. 1753–1766, May 2011. M. de Lourdes Bravo, M. M. Naim, and A. Potter, “Key issues of the upstream segment of biofuels supply chain: a qualitative analysis,” Logist. Res., vol. 5, no. 1–2, pp. 21– 31, 2012. F. Mafakheri and F. Nasiri, “Modeling of biomass-toenergy supply chain operations: Applications, challenges and research directions,” Energy Policy, vol. 67, pp. 116– 126, 2014. A. De Meyer, D. Cattrysse, J. Rasinmäki, and J. Van Orshoven, “Methods to optimise the design and management of biomass-for-bioenergy supply chains: A
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
review,” Renew. Sustain. Energy Rev., vol. 31, pp. 657– 670, 2014. F. You, L. Tao, D. J. Graziano, and S. W. Snyder, “Optimal design of sustainable cellulosic biofuel supply chains: Multiobjective optimization coupled with life cycle assessment and input–output analysis,” AIChE J., vol. 58, no. 4, pp. 1157–1180, 2012. R. Jamshidi, S. M. T. Fatemi Ghomi, and B. Karimi, “Multi-objective green supply chain optimization with a new hybrid memetic algorithm using the Taguchi method,” Sci. Iran., vol. 19, no. 6, pp. 1876–1886, 2012. R. Ruiz-Femenia, G. Guillén-Gosálbez, L. Jiménez, and J. A. Caballero, “Multi-objective optimization of environmentally conscious chemical supply chains under demand uncertainty,” Chem. Eng. Sci., vol. 95, pp. 1–11, 2013. B. Sharma, R. G. Ingalls, C. L. Jones, and A. Khanchi, “Biomass supply chain design and analysis: Basis, overview, modeling, challenges, and future,” Renew. Sustain. Energy Rev., vol. 24, pp. 608–627, Aug. 2013. P. Steele, M. E. Puettmann, V. Kanthi Penmetsa, and J. E. Cooper, “Life-cycle assessment of pyrolysis bio-oil production,” For. Prod. J., vol. 62, no. 4, p. 326, 2012. M. Mobini, T. Sowlati, and S. Sokhansanj, “Forest biomass supply logistics for a power plant using the discrete-event simulation approach,” Appl. Energy, vol. 88, no. 4, pp. 1241–1250, Apr. 2011. O. Akgul, A. Zamboni, F. Bezzo, N. Shah, and L. G. Papageorgiou, “Optimization-based approaches for bioethanol supply chains,” Ind. Eng. Chem. Res., vol. 50, no. 9, pp. 4927–4938, 2010. H. An, W. E. Wilhelm, and S. W. Searcy, “A mathematical model to design a lignocellulosic biofuel supply chain system with a case study based on a region in Central Texas,” Bioresour. Technol., vol. 102, no. 17, pp. 7860– 7870, Sep. 2011. F. Andersen, F. Iturmendi, S. Espinosa, and M. S. Diaz, “Optimal design and planning of biodiesel supply chain with land competition,” Comput. Chem. Eng., vol. 47, pp. 170–182, Dec. 2012. I. R. Geijzendorffer, E. Annevelink, B. S. Elbersen, R. A. Smidt, and R. M. de Mol, “Application of a GISBIOLOCO tool for the design and assessment of biomass delivery chains,” 2008. M. D. Mas, S. Giarola, A. Zamboni, and F. Bezzo, “Capacity planning and financial optimization of the bioethanol supply chain under price uncertainty,” in Computer Aided Chemical Engineering, vol. Volume 28, S. Pierucci and G. Buzzi Ferraris, Ed. Elsevier, 2010, pp. 97–102. B. Aksoy, H. Cullinan, D. Webster, K. Gue, S. Sukumaran, M. Eden, and N. Sammons, “Woody biomass and mill waste utilization opportunities in Alabama: Transportation cost minimization, optimum facility location, economic feasibility, and impact,” Environ. Prog. Sustain. Energy, vol. 30, no. 4, pp. 720–732, Dec. 2011.
V002T05A011-9
Copyright © 2015 by ASME
215 [27] R. A. Zamora Cristales, “Economic optimization of forest biomass processing and transport,” 2013. [28] U.S. Environmental Protection Agency and National Renewable Laboratory, “State Bioenergy Primer: Information and Resources for States on Issues, Opportunities, and Options for Advancing Bioenergy" http://www.epa.gov/statelocalclimate/documents/pdf/bioe nergy.pdf#page=2&zoom=auto,-193,550.” 15-Sep-2990. [29] A. Aden, M. Ruth, K. Ibsen, J. Jechura, K. Neeves, J. Sheehan, and B. Wallace, Lignocellulosic Biomass to Ethanol Process Design and Economics Utilizing Cocurrent Dilute Acid Prehydrolysis and Enzymatic Hydrolysis for Corn Stover. National Renewable Energy Laboratory, 2002. [30] S. Phillips and T. J. Eggeman, Thermochemical ethanol via indirect gasification and mixed alcohol synthesis of lignocellulosic biomass, vol. 112. Citeseer, 2007. [31] S. Kersten and M. Garcia-Perez, “Recent developments in fast pyrolysis of ligno-cellulosic materials,” Curr. Opin. Biotechnol., vol. 24, no. 3, pp. 414–420, 2013. [32] “ROI 2003.” [Online]. Available: http://www.renewableoil.com/Pages/default.aspx. [Accessed: 17-Oct-2014].
Annex A: nomenclature Indices 𝑖𝑖 𝑗𝑗 𝑘𝑘 𝑙𝑙 𝑡𝑡 ℎ 𝑐𝑐 𝑝𝑝 𝑤𝑤 𝑚𝑚𝑚𝑚𝑚𝑚 𝑙𝑙 𝑠𝑠 𝑚𝑚 𝑏𝑏 𝑘𝑘
Set of harvesting sites Set of collection sites Set of mobile bio-refinery sites Set of non-mobile bio-refinery sites Set of time periods Harvesting Collection Processing Inventory Mobile bio-refinery Non-mobile bio-refinery Small truck Medium truck Big truck Tanker
Parameters Capital cost harvesting site 𝐹𝐹ℎ Operational cost harvesting site 𝑉𝑉ℎ Capital cost collection site 𝐹𝐹𝑐𝑐 Operational cost collection site 𝑉𝑉𝑐𝑐 𝐹𝐹𝑝𝑝 Capital cost processing site 𝑉𝑉𝑝𝑝 Operational cost processing site Capital cost inventory 𝐹𝐹𝑤𝑤 Operational cost inventory 𝑉𝑉𝑤𝑤 Capital cost mobile bio-refinery 𝐹𝐹𝑚𝑚 Operational cost mobile bio-refinery 𝑉𝑉𝑚𝑚
[33] P. C. Badger and P. Fransham, “Use of mobile fast pyrolysis plants to densify biomass and reduce biomass handling costs—A preliminary assessment,” Biomass Bioenergy, vol. 30, no. 4, pp. 321–325, Apr. 2006. [34] A. Mirkouei and K. R. Haapala, “Integration of Machine Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process.” Flexible Automation and Intelligent Manufacturing, 2014. [35] E. Cantú-Paz, Genetic and Evolutionary Computation-GECCO 2003: Genetic and Evolutionary Computation Conference, Chicago, IL, USA, July 12-16, 2003 : Proceedings. Springer Science & Business Media, 2003. [36] R. Ihaka and R. Gentleman, “R: a language for data analysis and graphics,” J. Comput. Graph. Stat., vol. 5, no. 3, pp. 299–314, 1996. [37] B. R. Flint, “Analysis and operational considerations of biomass extraction on steep terrain in western Oregon,” 2013. [38] Gurobi Optimization, Inc., Gurobi Optimizer Reference Manual. 2014. [39] G. Van Rossum and others, “Python Programming Language.,” in USENIX Annual Technical Conference, 2007. 𝐹𝐹𝑙𝑙 𝑉𝑉𝑙𝑙 𝐹𝐹𝐹𝐹𝑠𝑠 𝑉𝑉𝑉𝑉𝑠𝑠 𝐹𝐹𝐹𝐹𝑚𝑚 𝑉𝑉𝑉𝑉𝑚𝑚 𝐹𝐹𝐹𝐹𝑏𝑏 𝑉𝑉𝑉𝑉𝑏𝑏 𝐹𝐹𝐹𝐹𝑘𝑘 𝑉𝑉𝑉𝑉𝑘𝑘 𝐶𝐶𝐶𝐶𝐶𝐶𝑠𝑠 𝐶𝐶𝐶𝐶𝐶𝐶𝑚𝑚 𝐶𝐶𝐶𝐶𝐶𝐶𝑏𝑏 𝐶𝐶𝐶𝐶𝐶𝐶𝑘𝑘 𝐶𝐶𝐶𝐶𝐶𝐶𝑚𝑚𝑚𝑚𝑚𝑚 𝐶𝐶𝐶𝐶𝐶𝐶𝑙𝑙 𝛼𝛼 𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 𝑇𝑇𝑗𝑗𝑗𝑗𝑗𝑗 𝑇𝑇𝑗𝑗𝑗𝑗𝑗𝑗 𝑇𝑇𝑘𝑘𝑘𝑘𝑘𝑘 S
Capital cost non-mobile bio-refinery Operational cost non-mobile bio-refinery Capital cost of small truck Operational cost of small truck Capital cost of medium truck Operational cost of medium truck Capital cost of big truck Operational cost of big truck Capital cost of tanker truck Operational cost of tanker truck Capacity of small truck Capacity of medium truck Capacity of big truck Capacity of tanker truck Capacity of mobile bio-refinery Capacity of non-mobile bio-refinery Available amount of biomass Time from site i to site j at time t Time from site j to site k at time t Time from site j to site 1 at time t Time from site k to site 1 at time t Percentage yield of converting biomass to bio-oil
Continuous variables 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 Amount of biomass from site i to site j at time t 𝑌𝑌𝑗𝑗𝑗𝑗𝑗𝑗 Amount of biomass from site j to site k at time t Amount of biomass from site k to site l at time t 𝑍𝑍𝑘𝑘𝑘𝑘𝑘𝑘 𝑊𝑊𝑗𝑗𝑗𝑗𝑗𝑗 Amount of biomass from site j to site l at time t 𝑛𝑛𝑖𝑖𝑖𝑖𝑖𝑖 Number of truck from site i to site j at time t 𝑛𝑛𝑗𝑗𝑗𝑗𝑗𝑗 Number of truck from site j to site k at time t 𝑛𝑛𝑗𝑗𝑗𝑗𝑗𝑗 Number of truck from site j to site l at time t V002T05A011-10
Copyright © 2015 by ASME
216 𝑛𝑛𝑘𝑘𝑘𝑘𝑘𝑘
Number of truck from site k to site l at time t
Binary Variables Binary variable for transportation from site i to site 𝐵𝐵𝑖𝑖𝑖𝑖𝑖𝑖 j at time t Binary variable for transportation from site j to site 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 k at time t
𝐵𝐵𝑘𝑘𝑘𝑘𝑘𝑘 𝐵𝐵𝑗𝑗𝑗𝑗𝑗𝑗 𝐵𝐵𝑙𝑙𝑙𝑙
V002T05A011-11
Binary variable for transportation from site k to site l at time t Binary variable for transportation from site j to site l at time t Binary variable for site l at time t
Copyright © 2015 by ASME
217
Appendix C: Evolutionary Optimization of Bioenergy Supply Chain Cost with Uncertain Forest Biomass Quality and Availability
Proceedings of the 2016 Industrial and Systems Engineering Research Conference H. Yang, Z. Kong, and MD Sarder, eds.
Evolutionary Optimization of Bioenergy Supply Chain Cost with Uncertain Forest Biomass Quality and Availability Amin Mirkouei1*, Karl R. Haapala1, Ganti S. Murthy2, and John Sessions3 1School of Mechanical, Industrial, and Manufacturing Engineering 2Department of Biological and Ecological Engineering 3Department of Forest Engineering, Resources and Management Oregon State University, Corvallis, OR, 97331 Abstract Over the past few decades, bioenergy sources (e.g., bio-oil, biochar, and biofuels) have been introduced as a means to address environmental, energy security, and human health challenges attendant with fossil-based energy. Government and societal interest in bioenergy has put additional scrutiny on feedstock supply and logistics, systems analysis and integration, and cross-cutting sustainability. One of the key challenges in the biomass-to-bioenergy supply chain (B2BSC) is the uncertainties of externalities (e.g., supply, transportation, logistics, production, demand, and price) that can inhibit environmental performance and reduce competitiveness and robustness. The study presented herein aims to develop a multi-criteria decision making method capable of helping investigators to fill key gaps in B2BSC decision support systems. The methodology employs two quantitative methods: 1) the support vector machine method is used to predict the pattern of uncertainty parameters, and 2) a stochastic programming method is used to assess the role of a transportable bio-refinery and the impact of real-world uncertainties in B2BSCs. The developed stochastic optimization model considers upstream and midstream costs (i.e., fixed, variable, and labor costs) of a B2BSC and includes two stochastic constraints (i.e., quality and availability) to incorporate uncertainties into the model. The model can minimize the total cost of a B2BSC network by using a genetic algorithm approach. The results for a case from Oregon, USA indicate that incorporating quality and availability uncertainty in the model can aid in harvesting site selection to use high quality biomass and reduce collection cost.
Keywords Biomass, Bioenergy, Supply Chain, Uncertainty, Genetic Algorithm
1. Introduction According to the U.S. Energy Information Administration, over one-tenth of total U.S. energy consumption is produced from renewable energy resources [1]. Recent interest in renewable energy due to the potential environmental and social benefits has stimulated research efforts globally and in the United States. Bioenergy is a form of renewable energy that has several benefits in comparison with conventional energy sources (e.g., coal and petroleum), which include [2]: 1) reduced dependency on imported energy, 2) reduced greenhouse gas (GHG) emissions, 3) improved forest management, 4) reduced poverty potential, especially in rural areas, 5) improved social resources (e.g., water quality), 6) promotion of carbon cycling and biodiversity, and 7) improved human health. In addition to bioenergy, major U.S. renewable energy sources include biomass, geothermal, hydroelectric power, solar, and wind [1]. Since some of these sources are concentrated geographically, their growth is limited in the United States. For instance, geothermal generation is limited to the western United States. On the other hand, biomass feedstocks and wastes are distributed across the U.S. and represent the largest renewable energy source (602-1009 million dry tons per year by 2022). Thus, biomass can play a key role in the energy industry. Biomass resources can be classified as lignocellulose, triglycerides, amorphous sugars, and starch [2]. Lignocellulosic biomass is low-cost, readily-available, and fast-growing. In 2015, the U.S. Department of Agriculture, in collaboration with the U.S. Department of Energy, announced nearly nine million dollars in funding through the Biomass Research and Development Initiative (BRDI), which would “reduce the nation’s dependence on foreign oil by supporting the development of bioenergy feedstocks, biofuels, and bio-based products” [3]. The main topics of interest include [4]: 1) feedstock supply and logistics, 2) thermochemical and biochemical production processes, 3) integrated biorefineries, 4) system analysis and integration, and 5) cross-cutting sustainability.
*Corresponding author. E-mail address:
[email protected]; Tel: +1-541-602-3488.
218
Mirkouei, Haapala, Murthy, and Sessions
219 The transition from conventional energy to bioenergy-based solutions presents a numerous challenges in the traditional and competitive energy market. Ongoing research and innovation is pursuing technological and research breakthroughs for bio-refinery capital cost reduction and process scale-up for commercialization. Most forest harvest residues (FHRs) are underutilized due to logistical challenges, such as high collection, transportation, and storage costs, as well as low market values [2]. Challenges have also been reported in the literature due to the nature of the bioenergy supply chain (SC) [2,5]. A major drawback of bioenergy is network uncertainties within the biomass-tobioenergy supply chain (B2BSC), including uncertainty in supply, logistics, production and yield, and demand and distribution. In order to overcome the existing challenges and fill the gaps in B2BSC, development of robust and reliable SCs that incorporates knowledge of the sources of uncertainty is crucial. The overarching objective of the research presented here is to ensure energy security in the dynamic and competitive energy market by producing a competitive bioenergy from forest-based biomass resources. B2BSCs have three main segments: upstream, midstream, and downstream (Fig. 1). The focus of this study is on upstream and midstream segments of B2BSC, which contain harvesting, collection, transportation, pre-treatment, conversion, and short term storage. Conversion process converts biomass into denser energy carriers (e.g., bio-oil and biochar) that ease handling, transportation, and storing. A multi-criteria decision making method is developed to simultaneously evaluate the objectives of cost, quality, and availability along with assessing the role of network uncertainties and transportable bio-refineries to support broader bioenergy commercialization. The proposed decision making method integrates two quantitative methods: support vector machine method and stochastic optimization model. Developing a sustainable and optimal B2BSC is an expedient short-term solution to trade off between cost and biomass quality and availability. Optimal SC planning aids decision makers (e.g., investors in energy industry and policy makers in government) to ensure the efficiency and effectiveness of the material and information flow. In the competitive energy market, B2BSCs should focus on minimizing total cost or maximizing profit through a decision making approach to achieve robust, sustainable, and reliable SCs. The major decision making approaches in SC management are strategic (long-term decisions – five years or more), tactical (medium-term decisions – 6-12 months), and operational (short-term decisions – daily/weekly). This study focuses on a tactical decision making approach in timely and cost effective manner, such as sourcing (e.g., harvesting site) and logistical (e.g., facility location) decisions. Some of the decisions are selected at the beginning of SC planning (e.g., raw material type and production technologies) because these decisions will not be changed in a short- or medium-term period. Biomass sourcing decisions play a key role in B2BSC development through efficiently harvesting, transporting, and pre-treating biomass in order to reduce the cost and environmental impacts. Since uncertainties affect the performance of SCs, decision makers must incorporate them in all decisions. Stochastic programming is a quantitative method for developing optimization models that can manage uncertainty sources [5]. In contrast with deterministic optimization models that have known parameters, stochastic optimization models incorporate unknown model parameters to consider real-world uncertainties. Stochastic programming is an analytical method for decision making under a stochastic environment [5], and takes advantage of probability distributions and historical data. One of the challenges of stochastic programming is designing a reliable computational algorithm to effectively solve the model. The existing methods to incorporate uncertainties in SCs include analytical methods (e.g., stochastic mathematical programming and Markov decision process) and simulation methods (e.g., Monte Carlo and discrete event simulation). More details about modeling uncertainties in SCs and stochastic programming optimization in bioenergy SCs have been previously reported [5,6].
Figure 1. Forest-based biomass-to-bioenergy supply chain entities Uncertainty in biomass supply exists due to the characteristics of this energy resource. Low energy density of forestbased biomass, high moisture content, and variations in quality and availability represent supply uncertainties in upstream forest biomass SCs. These parameters, especially biomass quality, affect the net profit of bioenergy production in different seasons. Seasonality has huge impact on ensuring a continuous supply since underutilized FHR
Mirkouei, Haapala, Murthy, and Sessions
220 may not be obtainable throughout the year, thus decision makers need to consider multiple sources of biomass. Logistics uncertainties include facility location and transportation lead time (i.e., transferring biomass in timely and cost effective manner). Pre-treatment uncertainties include equipment costs (e.g., grinder, chipper, and dryer) and unexpected equipment breakdowns. Manufacturing process uncertainties include production yield and machine breakdowns. A detailed overview of upstream and midstream B2BSC segments has been provided previously [2]. Since the existing uncertainties add complexities to decision making in energy industries, a method for managing uncertainty sources is vital during bioenergy SC development [7]. Without considering uncertainty parameters, the results of SC models may be suboptimal or even infeasible, and, subsequently, will be unreliable [7]. The main uncertainty sources in upstream and midstream B2BSC segments are limited to 1) biomass supply (e.g., harvesting and collection), 2) logistics (e.g., transportation and storage), 3) pre-treatment (e.g., drying and size reduction), and 4) manufacturing processes (e.g., conversion technology and production processes). This study explores the effects of uncertainty by using a stochastic optimization model and genetic algorithm technique. The proposed methodology and an actual case study are described in the next section.
2. Methodology As discussed above, various methods have been developed to incorporate uncertainties in SC management and logistics planning. There are limited studies in the B2BSC domain, however, that have considered uncertainties in quantitative analysis techniques. The methodology here presents a multi-criteria decision making method that integrates two quantitative analyses (i.e., support vector machine and stochastic programming) into a decision support system to assess the role of uncertainty sources in B2BSCs. In Phase 1, the support vector machine (SVM) method is applied to provide a learning algorithm for pattern recognition of uncertainty parameters. SVM is a supervised method in machine learning, which uses historical data to predict the future trend of several parameters (i.e., acceptable biomass availability failure rate and acceptable biomass quality failure rate in this study) [8]. Quality failure rate is defined based on biomass net calorific value (MJ/kg) and moisture content (% H2O). The availability failure rate depends on the biomass collection cost at each harvesting site. In Phase 2 (the phase reported herein), the stochastic programming method is developed to optimize the total cost of biomass-to-bio-oil SC over a quarter-year time horizon. 2.1. Stochastic Optimization Model A stochastic optimization model is used in this phase as an analytical method to explore the commercial feasibility of bio-oil production. The objective function (Eq. 1) to minimize the total cost (TC) of B2BSC over the quarter-year, includes harvesting, collection, transportation, pre-treatment, conversion, and short-term storage. The model encompasses two other objectives that are defined as stochastic constraints (Eqs. 2-3) to assess the role of uncertainties. Biomass quality and availability are highly influential and variable uncertainties in biomass supply - considering and incorporating these parameters is essential for establishing sustainable and reliable bioenergy. The model considers the following other constraints: capacity (Eqs. 4-6), harvesting base (number of harvesting sites) as specified by decision makers (Eq. 7), annual available biomass to be processed (Eq. 8), and non-negativity, binary, and integer constraints to guarantee a feasible solution (Eqs. 9-11). Notations of model indices, parameters, and variables are provided in the nomenclature, below. Min TC = ∑ ∑ 𝐹ℎ ∗ 𝐵𝑖𝑗 + (𝐿ℎ + 𝑉ℎ ) ∗ 𝑖
𝑗
𝑖
+ ∑ ∑ 𝐹𝑝 ∗ 𝐵𝑖𝑗 + (𝐿𝑝 + 𝑉𝑝 ) ∗ 𝑖
𝑋𝑖𝑗 𝑋𝑖𝑗 + ∑ ∑ 𝐹𝑐 ∗ 𝐵𝑖𝑗 + (𝐿𝑐 + 𝑉𝑐 ) ∗ 𝑃𝑅ℎ 𝑃𝑅𝑐
𝑗
(1)
𝑗
𝑋𝑖𝑗 𝑋𝑖𝑗 + ∑ ∑ 𝐷𝑖𝑗 ∗ 𝐶𝑖𝑗 ∗ 𝑛𝑖𝑗 + ∑ ∑ 𝐹𝑚 ∗ 𝐵𝑖𝑗 + (𝐿𝑚 + 𝑉𝑚 ) ∗ 𝑃𝑅𝑝 𝑃𝑅𝑚 𝑖
𝑗
𝑖
𝑗
St: 1⁄2
∑ 𝐵𝑖𝑗 ∗ 𝜇𝑖𝑗 + (∑ 𝐵𝑖𝑗 ∗ 𝑖∈𝐼
𝑖∈𝐼
∑ 𝐵𝑖𝑗 ∗ 𝜇𝑖𝑗 + (∑ 𝐵𝑖𝑗 ∗ 𝑖∈𝐼
𝜎𝑖𝑗2 )
𝑖∈𝐼
∗ 𝜑−1 (1 − 𝛼) ≤ 𝑄
∀𝑖 ∈ 𝐼 𝛼 ∈ [0,1]
(2)
∗ 𝜑−1 (1 − 𝛽) ≤ 𝐴
∀𝑖 ∈ 𝐼 𝛽 ∈ [0,1]
(3)
1⁄2
𝜎𝑖𝑗2 )
Mirkouei, Haapala, Murthy, and Sessions
221 𝑋𝑖𝑗 ≤ 𝐶𝑎𝑝𝑖 ∗ 𝐵𝑖𝑗
∀𝑖 ∈ 𝐼, ∀𝑗 ∈ 𝐽
(4)
𝑛𝑖𝑗 ≥ 𝑋𝑖𝑗 /𝐶𝑎𝑝𝑡
∀𝑖 ∈ 𝐼, ∀𝑗 ∈ 𝐽
(5)
∀𝑖 ∈ 𝐼 ∀𝑖 ∈ 𝐼
∑ 𝑋𝑖𝑗 ≤ 𝐶𝑎𝑝𝑚 𝑖∈𝐼
∑ 𝐵𝑖𝑗 ≤ 𝑁 𝑖∈𝐼
∀𝑖 ∈ 𝐼, ∀𝑗 ∈ 𝐽
(8)
𝑋𝑖𝑗 ≥ 0
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 𝑎𝑛𝑑 𝑗
(9)
(6)
𝐵𝑖𝑗 = {0, 1}
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 𝑎𝑛𝑑 𝑗
(10)
(7)
𝑛𝑖𝑗 is integer
𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 𝑎𝑛𝑑 𝑗
(11)
𝑖∈𝐼 𝑗∈𝐽
Nomenclature Indices 𝑐 ℎ 𝑖 𝑗 𝑚 𝑝
Collection Harvesting Set of harvesting sites Set of transportable bio-refinery sites Transportable bio-refinery Pre-treatment
Parameters 𝐴 𝑎𝑖𝑗 𝐶𝑖𝑗 𝐶𝑎𝑝𝑖 𝐶𝑎𝑝𝑚 𝐶𝑎𝑝𝑡 𝐷𝑖𝑗 𝐹𝑐 𝐹ℎ 𝐹𝑚 𝐹𝑝 𝐿𝑐 𝐿ℎ 𝐿𝑚
∑ ∑ 𝑋𝑖𝑗 ≥ 𝜃
Maximum acceptable biomass availability failure rate Availability failure rate of transferred biomass from site i to bio-refinery site j Transportation cost of in-forest trailer-truck to transportable bio-refinery (US $/mile) Available amount of biomass in harvesting site i (metric tons) Annual capacity of a transportable bio-refinery unit (metric tons) Capacity of a trailer-truck (metric tons) Distance between harvesting site and transportable biorefinery unit (miles) Annual fixed cost for a collection ($) Annual fixed cost for a harvesting ($) Annual fixed cost for a transportable bio-refinery ($) Annual fixed cost for a pre-treatment ($) Annual labor cost for a collection ($) Annual labor cost for a harvesting ($) Annual labor cost for a transportable bio-refinery ($)
𝐿𝑝 𝑁 𝑃𝑅𝑐 𝑃𝑅ℎ 𝑃𝑅𝑚 𝑃𝑅𝑝 𝑄 𝑞𝑖𝑗 𝑉𝑐 𝑉ℎ 𝑉𝑚 𝑉𝑝 𝜃 𝜇𝑖𝑗 𝜎𝑖𝑗2 𝛼 𝛽
Annual labor cost for a pre-treatment ($) Base number of harvesting Annual utilization rate of a forwarder (metric tons) Annual utilization rate of a harvester (metric tons) Annual treatment rate of a transportable bio-refinery (metric tons) Annual utilization rate of a grinder (metric tons) Maximum acceptable biomass quality failure rate Quality failure rate of transferred biomass from harvesting site i to bio-refinery site j Annual variable cost for a collection ($) Annual variable cost for a harvesting ($) Annual variable cost for a transportable bio-refinery ($) Annual variable cost for a pre-treatment ($) Annual available amount of biomass (metric tons) Mean of the original normal distribution Variance of the original normal distribution Probability of the quality constraint Probability of the availability constraint
Continuous Variables 𝑋𝑖𝑗
Amount of biomass transported from site i to site j
Integer Variables 𝑛𝑖𝑗
Number of in-forest truck trips to transfer forest biomass form site i to nearest site j
Binary Variables 𝐵𝑖𝑗
Binary variable for biomass transportation from site i to site j
2.2. Case Study A case study is used to demonstrate the proposed method. The Base Case in the study considers five high-potential harvesting sites in Tillamook County (Oregon) using information from the Oregon Department of Forestry (ODF) and prior research [9,10]. Table 1 indicates the main attributes of each harvesting site. The main FHR types available are Douglas fir and red alder. Fixed, variable, and labor costs are calculated after Brinker et al. [11]. The U.S. Producer Price Index has been used to adjust cost values for inflation to 2015 [12]. Phase 2 study demonstration assumes: 1) production yield for biomass-to-bio-oil conversion is 50%, 2) at least 2500 metric tons of FHR (θ) is available in the five harvesting sites over the quarter-year time horizon, 3) the quality failure rate (qij) and the availability failure rate (aij) for each site is known, 4) the maximum acceptable quality failure rate (Q=25%) and the maximum acceptable availability failure rate (A=30%) were found in Phase 1 using the SVM method, 5) the quality failure rate has a linear relationship with moisture content, and 6) the availability failure rate has a linear relationship with collection cost. The mean (𝜇) and variance (𝜎 2 ) of qij and aij would be obtained from historical data for each harvesting site (assumed values are shown in Table 1). The quality failure rate is assumed to range from 0-5% for FHR of 50-60% moisture content, and availability failure rate ranges from 0-5% for FHR with collection costs of $15-25 per dry metric ton. Thus, since it is known that Site 1 has an average moisture content of 55%, the mean quality failure rate is 2.5%, while the variance is 4% (obtained from historical data). A harvesting site with FHR above the maximum failure rates (A and Q) will be considered by the algorithm, but the stochastic constraints must be met for all harvesting sites. The required equipment in the Base Case includes a harvester, a forwarder, trailer trucks (18.2 metric tons), a grinder, and a transportable bio-refinery unit (50 dry metric tons per day refinery size). Since the available amount of biomass
Mirkouei, Haapala, Murthy, and Sessions
222 is low in this study, the grinder is placed near the transportable bio-refinery due to the high cost of using a lowboy to move the grinder and loader. Transportable bio-refineries are located close to harvesting and collection sites to address existing logistical challenges (e.g., low bulk density). A transportable bio-refinery uses pyrolysis to convert biomass to intermediate bio-products of higher density than FHR [13,14]. The effective lifetime of a transportable bio-refinery unit is assumed to be ten years and annual production is assumed as 16,425 metric tons (329 scheduled operation days -12 hours per day), or 4,100 metric tons per quarter [13]. A short-term storage facility is provided near the transportable bio-refinery and associated costs are considered in the bio-refinery costs. The base number of harvesting sites (𝑁) is four in this case study, meaning the decision makers want to select four of the five harvesting sites with the highest biomass quality (i.e., high net calorific value or moisture content) and availability (i.e., biomass collection cost). The probability of the quality failure (1-𝛼) and availability failure (1-𝛽) constraints are 23% and 25%, respectively; 𝛼 and 𝛽 values are between 0 and 1, since the probabilities are between 0 and 1. For instance, quality failure is above 23% for a particular site, the algorithm will not consider it as a potential harvesting site. Table 1. Harvesting site attributes Harvesting Site Site 1 Site 2 Site 3 Site 4 Site 5
Available FHR (metric tons) 1160 460 450 530 580
Distance from site to bio-refinery (miles) 35.0 12.0 6.0 5.5 26.0
Quality Failure Rate, (𝐪𝐢𝐣 ) – N (𝝁, 𝝈𝟐 ) N (0.025, 0.040) N (0.022, 0.025) N (0.019, 0.020) N (0.035, 0.035) N (0.030, 0.057)
Availability Failure Rate, (𝐚𝐢𝐣 ) – N (μ, 𝝈𝟐 ) N (0.034, 0.050) N (0.030, 0.035) N (0.022, 0.040) N (0.015, 0.020) N (0.040, 0.025)
3. Results and Discussion The stochastic optimization model developed was solved using a genetic algorithm (GA) within MATLAB [8]. GA is an evolutionary search heuristic. It used several evolutionary phases (e.g., initialization, mutation, crossover, and tournament selection) to find an optimal solution. The solution for the case described indicates the model has 15 decision variables (five-binary, five-integer, and five-continuous) and 30 constraints. The exhaustive solution has 32 (25) feasible and infeasible solutions. An optimal solution was found after 100 iterations in under five seconds using a system configured with Windows 7, 64-bit Operation System, Intel Core i7 processor, and 8GB RAM. The solution indicates that all available FHR would be processed Sites 1-4, producing 1,250 metric tons (275,573 gallons) of biooil over the quarter-year time horizon. Site 5 is not selected due to the long distance between the harvesting site and bio-refinery. The quarter-year cost is predicted as $376,229, resulting in a cost of $0.36/liter ($1.36/gallon). In addition to GA, an exhaustive search was conducted using Microsoft Excel without considering stochastic constraints. Since the size (number of decision variables) of the Base Case is small, the exhaustive search was used to compare the results and verify the computational algorithm and solution. The proposed mathematical model is constituted as an NP-hard problem. Thus, as the number of sites increases, decision making will become increasingly difficult (even impossible) without a proper computational algorithm. The major concern when developing an NPhard problem is how to solve it and represent the solutions. For instance, if the number of harvesting sites changes to 20, the number of solution combinations (either feasible or infeasible) will be 1,048,576 (2^20), which would lend itself to a heuristic or metaheuristic approach. Both approaches resulted in identical solutions.
4. Conclusion Due to numerous real-world uncertainties, advanced methods are needed to develop an effective, efficient, and commercially viable bioenergy industry. In order to assist biomass-to-bioenergy supply chain (B2BSC) development, dynamic, multi-criteria decision making is crucial. Future methods should incorporate uncertainty parameters to address real-world challenges, using analytical, simulation, or hybrid methods. This study provides a brief review of the decision making methods and modeling uncertainties in the bioenergy supply chain management and logistics. A proposed multi-criteria decision making method presented herein represents a pioneering approach to incorporate two uncertainty parameters (i.e., quality and availability) in the upstream and midstream B2BSC segments. The approach also consider the use of a small, transportable bio-refinery in place of a large, centralized bio-refinery. The decision making approach includes two quantitative methods, support vector machine (Phase 1) and stochastic programming (Phase 2), to assess the role of integrating transportable bio-refinery technology in addressing uncertainties of bio-oil production economics. Phase 1 helps decision makers to quantify uncertainty parameters and use them as inputs in the Phase 2 stochastic optimization model. Phase 2 is used to evaluate the effect of biomass
Mirkouei, Haapala, Murthy, and Sessions
223 quality and availability on bio-oil cost. The stochastic model considers harvesting, collection, transportation, pretreatment, conversion, and short-term storage activities. The uncertainty in biomass quality and availability is formulated using probability constraints (with a normal distribution) in the optimization model. Since the model has two stochastic constraints, this model is constituted as NP-hard problem. Due to the difficulty in solving NP-hard problems, a genetic algorithm approach was applied as a metaheuristic technique with the assistance of MATLAB. The results of proposed model for a case study using data from Oregon, USA demonstrate that the proposed multicriteria decision making method offers a promising approach to incorporate uncertainties in logistics and SC planning. In particular, the results indicate that incorporating quality and availability uncertainty in the model can aid in selecting appropriate harvesting sites to collect highest quality biomass and reduce collection costs. This will provide more reliable, robust, and sustainable B2BSC networks. Future research should expand the SC network by considering the downstream segment, as well as considering the production of other bio-products, such as bio-char, which can lead to additional revenue streams and environmental impact reductions (e.g., carbon pollution).
Acknowledgement The authors wish to acknowledge Michael Wilson and Robert Nall of the Oregon Department of Forestry for providing data to support this research.
References [1] [2]
[3] [4]
[5] [6] [7]
[8]
[9]
[10] [11] [12] [13]
[14]
DOE/EIA, 2015, “Monthly Energy Review - Energy Information Administration” [Online]. Available: http://www.eia.gov/totalenergy/data/monthly/. [Accessed: 11-Jan-2016]. Mirkouei, A., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015, “A Review and Future Directions in Techno-Economic Modeling and Optimization of Upstream Forest Biomass to Bio-oil Supply Chains,” Renewable & Sustainable Energy Reviews, in revision. USDA, 2015, “View Opportunity | GRANTS.GOV” [Online]. Available: http://www.grants.gov/viewopportunity.html?oppId=274814. [Accessed: 29-Jan-2016]. Reed, V., 2012, “U.S. Department of Energy Biomass Program | Department of Energy” [Online]. Available: http://energy.gov/eere/bioenergy/downloads/us-department-energy-biomass-program. [Accessed: 07-Jan2016]. Awudu, I., and Zhang, J., 2012, “Uncertainties and sustainability concepts in biofuel supply chain management: A review,” Renewable and Sustainable Energy Reviews, 16(2), pp. 1359–1368. Shabani, N., Sowlati, T., Ouhimmou, M., and Rönnqvist, M., 2014, “Tactical supply chain planning for a forest biomass power plant under supply uncertainty,” Energy, 78, pp. 346–355. Shabani, N., and Sowlati, T., 2016, “A hybrid multi-stage stochastic programming-robust optimization model for maximizing the supply chain of a forest-based biomass power plant considering uncertainties,” Journal of Cleaner Production, 112, pp. 3285–3293. Mirkouei, A., and Haapala, K. R., 2014, “Integration of Machine Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process,” 24th International Conference on Flexible Automation and Intelligent Manufacturing (FAIM), May 20-23, 2014, San Antonio, Texas, Flexible Automation and Intelligent Manufacturing. Mirkouei, A., Mirzaie, P., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015, “Reducing the Cost and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains,” Journal of Cleaner Production, doi:10.1016/j.jclepro.2015.11.023. ODF, 2015, “State of Oregon: Oregon Department of Forestry - Home” [Online]. Available: http://www.oregon.gov/odf/Pages/index.aspx. [Accessed: 12-Oct-2015]. Brinker, R. W., Miller, D., Stokes, B. J., and Lanford, B. L., 2002, “Machine rates for selected forest harvesting machines,” Circular 296 (revised). Alabama Agric. Exp. Station, Auburn University; 32 pp. BLS, 2015, “Producer Price Index (PPI), Bureau of Labor Statistics, United States Department of Labor” [Online]. Available: http://www.bls.gov/ppi/. [Accessed: 30-Oct-2015]. Mirkouei, A., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015, “Multi-criteria Decision Making for Sustainable Bio-Oil Production using a Mixed Supply Chain,” ASME Journal of Manufacturing Science & Engineering, in review. Mirkouei, A., and Haapala, K. R., 2015, “A Network Model to Optimize Upstream and Midstream Biomass-toBioenergy Supply Chain Costs,” ASME 2015 International Manufacturing Science and Engineering Conference (MSEC), MSEC2015-9355, June 8-12, 2015, Charlotte, NC.
224
Appendix D: Reducing Greenhouse Gas Emissions for Sustainable Bio-Oil Production Using a Mixed Supply Chain
ASME 2016 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference IDETC2016 August 21-24, 2016, Charlotte, North Carolina, USA
IDETC2016-59262 REDUCING GREENHOUSE GAS EMISSIONS FOR SUSTAINABLE BIO-OIL PRODUCTION USING A MIXED SUPPLY CHAIN Amin Mirkouei School of Mechanical, Industrial and Manufacturing Engineering Oregon State University Corvallis, Oregon 97331
[email protected]
Karl R. Haapala School of Mechanical, Industrial and Manufacturing Engineering Oregon State University Corvallis, Oregon 97331
[email protected]
John Sessions Department of Forest Engineering Resources and Management Oregon State University Corvallis, Oregon 97331
[email protected]
Ganti S. Murthy Department of Biological and Ecological Engineering Oregon State University Corvallis, Oregon 97331
[email protected]
ABSTRACT Recent growing interest in reducing greenhouse gas (GHG) emissions requires the application of effective energy solutions, such as the utilization of renewable resources. Biomass represents a promising renewable resource for bioenergy, since it has the potential to reduce GHG emissions from various industry sectors. In spite of the potential benefits, biomass is limited due to logistical challenges of collection and transport to bio-refineries. This study proposes a forest biomass-to-biooil mixed supply chain network to reduce the GHG emissions compared to a conventional bioenergy supply chain. The mixed supply chain includes mix-mode bio-refineries and mixedpathway transportation. Life cycle assessment is conducted for a case study in the Pacific Northwest with the assistance of available life cycle inventory data for biomass-to-bio-oil supply chain. Impact assessment, on a global warming potential (GWP) basis, is conducted with the assistance of databases within SimaPro 8 software. Sensitivity analysis for the case investigated indicates that using the mixed supply chain can reduce GHG emissions by 2-5% compared to the traditional supply chain. INTRODUCTION Sustainability has attracted the attention of investigators in both industrial and academic communities. Provision of
sustainable energy requires meeting today’s energy needs without compromising future generation’s ability to meet their energy needs. Fossil fuel consumption has a direct relationship with GHG emissions and environmental impacts. Therefore, renewable energy has been suggested as an environmentally friendly form of energy [1]. For instance, biomass is an energy resource that can be harvested and replenished indefinitely. However, its sustainability performance should be evaluated by considering the economic, environmental, and social aspects across the bioenergy supply chain (SC). Policy makers and other decision makers can take steps to avoid the impacts of carbon emissions by promoting the use of bioenergy sources (e.g., biofuel, bio-oil, and biochar). Therefore, sustainability analysis can be applied to mitigate the negative effects (e.g., market volatility) of biomass-based energy SCs. The United States Energy Information Administration (U.S. EIA) estimates biomass energy generation will increase 3.1% per year on average through 2030 [2]. After 2030, new dedicated biomass bio-refineries are expected to account for most of the growth in renewable energy from biomass resources. According to the American Petroleum Institute [3], the U.S. imported nearly $1 billion of crude oil each day in 2014, with 40% for the production of gasoline. Bioenergy has been suggested as part of a comprehensive climate strategy to cut the use of oil in half by 2030, in addition to phasing out the
Copyright © 2016 by ASME
226 use of coal [4]. Cofiring using high efficiency coal boilers, for example, is an option to convert biomass to electricity cleanly and at low cost. Therefore, special attention should be placed on supporting research and commercialization of bioenergy sources, such as underutilized forest harvest residue (FHR) to balance the cost, energy, and environmental tradeoffs. Bioenergy used in this manner can support several national priorities including [5]: 1) reducing dependence on imported energy and increasing energy security, 2) promoting the use of domestic, sustainable, and reliable energy resources, 3) improving economic growth by establishing an advanced bio-industry, and 4) maximizing greenhouse gas (GHG) reductions by reducing energy production and consumption. The use of bioenergy has the potential to mitigate GHG emissions from fossil energy and related impacts, e.g., global warming potential (GWP). GHG emissions are expressed as CO2 equivalent emissions, and include several substances, e.g., carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O). Environmentally responsible bio-resource stewardship supports mitigation of GHG emissions, as well as soil degradation, poor water quality, and biodiversity loss [6]. Renewable energy sources, especially biomass, can be carbon neutral. The carbon released into the environment during forest biomass combustion is nearly equal to the carbon absorbed from atmosphere during the growth life of the tree (through photosynthesis) [1]. Biomass resources can benefit bioenergy and biochemical production, waste management, and soil amendment with the use of biochar. Forest-based biomass is one of the lowest cost and most abundant resources used to produce second and third generation bioenergy [7]. First generation bioenergy sources have benefits (e.g., ability to be blended with fossil-based energy and GHG mitigation) that can mitigate dependence on foreign energy. The main concern of first generation bioenergy is sourcing of feedstocks, which include food crops (e.g., sugar cane, wheat, and maize). Recent analyses indicate most first generation bioenergy sources have negative effects on land use, biodiversity, and food prices due to competition with food crops [8]. Therefore, forest biomass can play a key role in decarbonizing the environment through the production of biooil, bio-char, biofuels, and renewable chemicals, as well as benefiting rural economies with new bioenergy industry [7]. The forest biomass-to-bioenergy SC represents a key challenge to bioenergy feasibility. The U.S. EPA Renewable Fuel Standard program, mandating the production of 36 billion gallons (136 billion liters) of renewable fuel by 2022, has raised serious concerns in terms of long-term effects of removing FHRs on soil productivity, water quality, and habitat [9]. However, maintaining forests and farms properly can improve the environment for future preparations (e.g., planting). FHR is of low energy density and highly dispersed, resulting in high costs for collection and transport. In addition, some processing technologies remain at the development stage. The SC includes
three major segments: upstream, midstream, and downstream (Fig. 1). The major cost driver in forest biomass-to-bioenergy SCs is the production process, including pretreatment (e.g., drying and pyrolysis) and upgrading process (e.g., hydroprocessing) for biofuels. Therefore, advances in pre- and post-conversion technologies (e.g., pyrolysis and hyrdroprocessing) are needed to reduce the total cost of bioenergy production [10,11]. After the bio-oil production process (conversion), biomass supply and logistics entities (e.g., harvesting, collection, transportation, and storage) represent the next highest barriers to commercialization, due to the economic and environmental challenges [10]. Thus, these entities are of key focus in this work. The overarching goal of the research that encompasses this work is to support engineering decision making for costeffective substitution of fossil-based with bio-based energy sources to improve energy security and overcome related environmental challenges. In particular, this work aims to assess the environmental impacts (in terms of GWP) of a proposed cost-optimal forest biomass-to-bio-oil mixed supply chain, and to evaluate the reduction of GHG emissions compared to a more traditional supply chain. The biomass-to-bio-oil mixed SC includes mix-mode (transportable and fixed) bio-refineries and mixed-pathway (truck-tanker and truck-truck) transportation, which are explained in greater detail below. The analysis approach is demonstrated using a case from the Pacific Northwest to determine the material and energy flows. Life cycle assessment (LCA) is applied to quantify the GHG
FIGURE 1. BIOMASS-TO-BIOENERGY SUPPLY CHAIN SEGMENTS AND ENTITIES
Copyright © 2016 by ASME
227 emissions associated with each SC entity. Environmental impact assessment is conducted to estimate energy use, GHG emissions, and GWP. The results provide tangible information about environmental impacts of bio-oil production and consumption, which can be useful for government policy makers, investors, and industrial and academic researchers and practitioners.
gasoline) or electricity [14]. The major benefit of a transportable bio-refinery is its ability to produce higher energy density, intermediate products (e.g., bio-oil and biochar) in close proximity to raw materials (e.g., forests or farms) [11]. This alleviates transferring low energy density biomass to centralized bio-refineries, which can improve economic and environmental performance of bio-oil production by reducing transportation, handling, and storage operations [10].
BACKGROUND Biomass can be classified as energy crops, agriculture residue, forest residue, and wastes resources (e.g., urban wood wastes) [12]. The predicted available forest and agricultural biomass resources was about 473 million dry tons in 2012, and is expected increase to nearly 1.1 billion dry tons by 2030 [12]. This amount of biomass can meet over one-third of current transportation fuel demands. The use of biomass is inherently limited due to economic factors (e.g., low energy density and high collection/transportation cost) and land use competition, which is why only 45% of available biomass is currently used [12]. Prior studies have inconsistently addressed the triple bottom line (i.e., economic, environmental, and social) in biomass-based energy SCs [13]. No studies have been found that apply an LCA approach to estimate environmental impacts of a biomass-to-bio-oil SC using a transportable bio-refinery unit (Fig. 2).
Pyrolysis conversion technology is a thermochemical decomposition of biomass at higher temperature (300-550°C) in the absence of oxygen (Fig. 3). Three types of pyrolysis have been introduced in the literature, which are slow, intermediate, and fast process. Each type has specific length of reaction time and temperature, and production yield. The main products of pyrolysis process include pyrolysis oil (bio-oil), pyrolysis char (bio-char), and non-condensable gases (syngas), of which biooil and bio-char are target products. Bio-oil is mainly produced from forest biomass through pyrolysis process that can condense a mixture of water and oxygenated carbon. Bio-oil includes carbon, hydrogen, oxygen, nitrogen, and ash, which are almost 50, 6, 41, 0.17, and 0.92 percent per weight, respectively. Additionally, bio-oil has an approximate density of 1.2 kg per liter, 18 MJ/kg higher heating value energy content [1]. The major applications of bio-oil include: upgrading to transportation fuel, using in chemical production, and combustion in boilers, engines, and turbines [4]. Bio-oil has also been used to produce electricity, but it requires industrial advancements to commercialize this product.
The transportable bio-refinery is a trailer-mounted unit that uses pyrolysis technology to convert underutilized FHR to bioproducts [7]. The transportable bio-refinery operates using biochar (as heat energy) in place of fossil fuels (e.g., diesel and
Traditional Pathway ?
Grinding Biomass Transferring ? Fixed Bio-refining
Collection
?
Transportable Bio-refining Harvesting
New Pathway
Bio-oil Transferring
? Question marks indicate location and asset decision points
FIGURE 2. A MIXED BIOMASS-TO-BIO-OIL SUPPLY CHAIN (TRANSPORTABLE BIO-REFINERY IMAGE COURTESY PHILLIP C. BADGER; FIXED BIO-REFINERY IMAGE COURTESY UPM LAPPEENRANTA BIOREFINERY, FINLAND)
Copyright © 2016 by ASME
228 Biomass Product
Pre-procesing
Pyrolysis Process
Size Reduction Drying
Heat from Biochar and Syngas
Gas/Liquid Separation
Solid Separation Pyrolysis Reactor Heat from Biochar and Syngas
Biochar
Bio-oil
FIGURE 3. PYROLYSIS CONVERSION PROCESS PRINCIPLES [2] Prior techno-economic analyses of bio-oil SCs indicate that the interest has been growing in this field, especially due to near carbon neutrality, or no net carbon release into the environment, in comparison with heating oil [1]. The results of prior studies indicate that biomass moisture content, biomass chip size, and bio-refinery size have direct impacts on economic and environmental aspects of bio-oil SCs [7,15,16]. The results of LCA studies on biomass-to-bioenergy SC indicate that there is a strong dependency on system boundary, allocation method, and functional unit [17]. For instance, a system that maximizes the GHG emission reductions per unit of energy is not able to achieve the highest GHG reduction per unit of biomass. Steele et al. reported that CO2 and steam emissions are the first and second highest emissions released across all bio-oil life-cycle stages, accounting for 67% and 32% of the total emissions by mass, respectively [1]. Additionally, Steele et al. reported that 1 MJ of bio-oil production releases a total of 0.19 kg CO2 cradleto-gate emissions [1]. In the mixed SC, the locations of transportable biorefineries have been selected close to forest harvesting sites to reduce the number of truck trips to transport biomass feedstocks. Reduced fuel consumption not only mitigates GHG emissions, but can aid bioenergy commercialization by minimizing transportation costs. Traditionally, to reduce the truck trips, different types of trucks have been considered for in-forest roads and highways. For instance, higher-capacity, double-trailer trucks are used on main highways, while in-forest road travel is restricted to single-trailer trucks, which have more flexibility in limited areas (e.g., tight curves) and can turn arounds when the roadway is narrow. The LCA method can be used to evaluate environmental impacts of the SC system to assist in defining more environmentally responsible networks. Since utilization of bioenergy aims to accommodate environmental pressures, new bioenergy technology need to be assessed in terms of environmental impacts. This study applies an LCA method to
quantify resource consumption (e.g., biomass and fossil fuel), as well as emissions, to determine GWP for a biomass-to-bio-oil mixed SC. This is then compared to a traditional biomass-tobio-oil SC. The proposed methodology and an actual case study are described in the next section. LIFE CYCLE ASSESSMENT METHOD The life cycle assessment (LCA) method includes: 1) defining the goal and scope to set the context of study, 2) life cycle inventory (LCI) to quantify materials and energy flows in the system, and 3) life cycle impact assessment (LCIA) to translate LCI data into environmental impact metrics. The application of LCA to the biomass-to-bioenergy mixed supply chain is described in the following sections. Goal and Scope This study aims to evaluate the environmental impacts for the production of bio-oil from FHR, including supply chain (transportation) effects. The scope of the study includes three life cycle stages: 1) upstream processing (i.e., biomass harvesting, collection, pre-processing, and transportation), 2) midstream processing (i.e., pre-treatment and conversion to bio-oil), and 3) downstream processing (i.e., distribution and combustion). Thus, the LCA study applies a cradle-to-grave system boundary to the bio-oil SC. The functional unit selected for the study is 1 gallon (3.78 liters) of bio-oil. The LCA is conducted using data and information from prior studies, including databases within a commercial LCA software package, SimaPro 8 [1,18,19]. The impact assessment method selected is Global Warming Potential (GWP), a single indicator method that assesses the total GHG emissions. The GWP emission factor (in kg CO2 equivalent (eq.)) is calculated using a 20-year time horizon, where the coefficients for CO2, CH4, and N2O are RCO2 (1 kg CO2 eq./kg CO2), RCH4 (56 kg CO2 eq./kg CH4), and RN2O (280 kg CO2 eq./kg N2O), respectively [20]. Thus, over 20 years, CH4 can trap an estimated 56 times more heat than an equivalent mass of CO2. Determination of Life Cycle Global Warming Potential Life cycle impact assessments are conducted and compared for the traditional and new (mixed) supply chain pathway for biomass-to-bio-oil production. In the long term, biomass-based energy products reduce GHG emissions because the released CO2 is part of the natural process and forest absorbs it from the atmosphere and store in the wood. While the released CO2 after fossil-based energy combustion is not part of the natural process. Equipment type and use characteristics are dependent upon the type of biomass utilized. The upstream forest biomass SC typically uses a harvester, forwarder, grinder, and loader. The process inputs are biomass and fossil-based energy (mainly
Copyright © 2016 by ASME
229 diesel fuel) and lubricants, and the outputs are ground biomass (chips) and air emissions from equipment operation. The GWP emission factor of biomass collection (EFup), which includes biomass harvesting, collection, and grinding, in kg CO2 eq. per metric ton of available biomass, and the environmental impacts (GWP) of the upstream segment (EIup), in kg CO2 eq., is calculated using Eqs. 1 and 2. Variables are defined in the Nomenclature section. EFup=RCO2*EFupCO2+RCH4*EFupCH4+RN2O*EFupN2O (1) EIup=Available Biomass*EFup
(2)
The next step is to transfer the chips to the bio-refinery using different types of trucks with various capacities. The inputs include chips and transportation fuel, and the outputs are transferred chips and air emissions from fuel combustion. Factors such as biomass quality and moisture content have a direct effect on GWP. Transferring high quality biomass with lower moisture content can reduce the number of truck trips (TrT) and, consequently, environmental impacts. Travel distance and truck weight also directly increase fuel consumption. A decision making method is applied, using a Geographical Information System (GIS)-based approach to define the shortest path between sites (e.g., collection site to a transportable bio-refinery or fixed bio-refinery location). The environmental impacts of biomass transportation (EImass) are quantified using Eqs. 3 and 4. The emission factor of transferring biomass (EFmass), using chip vans (a small type of truck) is equal to 0.403 kg CO2 eq. per ton-mile [21]. TrT=Amount of Transferred Biomass / Truck Capacity (3) EImass=TrT*W*Distance*EFmass
(4)
In the midstream segment, moisture content and particle size determine the type of pre-treatment equipment and conversion technology. The input is feedstock and the outputs are biochar, bio-oil, and air emissions. The air emissions include steam released and biogenic GHG emissions during the chip drying and production process, respectively. The heat for feedstock drying and pyrolysis in the new, mixed pathway is provided by the biochar and syngas produced during pyrolysis. Biochar and syngas are funneled directly to the furnace (i.e., self-generated and closed-loop process), and used as furnace inputs for biomass drying and conversion. Biomass particle size and required process temperature play a key role in determining the pyrolysis type (e.g., fast, intermediate, or slow). Fast pyrolysis has a higher production yield, but requires a high process temperature and small feedstock particle size (0.3-0.8 mm) [22]. Reducing the particles to the proper size requires more fossil-based energy. The emission factor (EFpro), in kg CO2 eq. per metric ton of bio-oil, and environmental impacts of
bio-oil production process (EIpro), using transportable or fixed bio-refinery are calculated using Eqs. 5 and 6. EFpro=RCO2*EFproCO2+RCH4*EFproCH4+RN2O*EFproN2O (5) EIpro=Produced Bio-oil*EFpro
(6)
The produced bio-oil will be transported to the distribution center by tanker truck, using diesel fuel as an input. Tanker truck trips (TaT) and distance from the bio-refinery to the distribution center are two major factors impacting fuel consumption. The tanker truck capacity has a direct relationship with the number of trips. In this study, the capacity is assumed to be 18 metric tons [10]. The environmental impacts of bio-oil transportation (EIoil) is quantified using Eqs. 7 and 8. The emission factor of transferring bio-oil (EFoil) using a tanker truck is assumed to be 0.403 kg CO2 eq. per ton-mile [21]. TaT=Amount of Produced Bio-oil / Tanker Capacity (7) EIoil=TaT*W*Distance*EFoil
(8)
In this study, combustion is considered as the last life cycle phase for bio-oil. The emission factor (EFcomb) and environmental impact (EIcomb) of bio-oil combustion are calculated using Eqs. 9 and 10. EFcomb=RCO2*EFoilCO2+RCH4*EFoilCH4+RN2O*EFoilN2O (9) EIcomb=Produced Bio-oil * EFcomb
(10)
From the foregoing, GWP can be obtained from the various biomass-to-bio-oil SC activities, i.e., harvesting, collection, grinding, transportation, drying, production, and consumption. Table 1 presents GWP values from life cycle inventory databases in SimaPro 8 and from prior studies for bio-oil production [1,23]. It can be seen that emissions of CO2 are greater than other GHG emissions over the bio-oil life cycle. Bio-oil combustion releases CO2 emissions from biogenic substances, however, which are not counted as emissions [18]. Biogenic CO2 emissions are a part of the carbon cycle that encompasses photosynthesis, whereby atmospheric CO2 is uptaken by other growing plants and trees. BIOMASS TO BIO-OIL MIXED SUPPLY CHAIN CASE A case study is presented to demonstrate the application of the method and to compare the results for the proposed mixed SC to a more traditional SC. In this case, decision makers have desire to reduce the GHG emissions compared to the traditional SC approach. The Base Case considered has 20 harvesting sites, five collection (staging) sites, two transportable bio-refineries, and one fixed bio-refinery with a storage facility.
Copyright © 2016 by ASME
230 TABLE 1. CRADLE-TO-GRAVE GLOBAL WARMING POTENTIAL FOR BIO-OIL PRODUCTION AND CONSUMPTION OVER 20-YEAR TIME PERIOD [1,23] Substance Biomass Collection Bio-Oil Production Bio-Oil Combustion (kg CO2 eq. per metric (kg CO2 eq. per (kg CO2 eq. per metric ton of biomass) metric ton of bio-oil) ton of bio-oil) CO2 (Biogenic and Fossil) 11.4 712 2,570 CH4 0.830 0.00 5.65 N 2O 58.5 585 1,550 Total 70.8 1,300 4,130 Staging sites are considered to take advantage of assembling loads and using the maximum allowable legal weight. The study extends prior work, which investigated supply chain cost minimization through a multi-criteria decision making for sustainable bio-oil production [11]. To complete the analysis of the hypothetical supply chain (transportable bio-refineries are not currently in use), several assumptions were made. The capacities of transportable and fixed bio-refineries are 50 and 200 dry metric tons, respectively. The annual scheduled production of a transportable bio-refinery is 329 days (12 hours per day), producing 1.8 million gallons of bio-oil [14]. The annual scheduled production of a fixed biorefinery is 365 days (24 hours per day), resulting in 8 million gallons of bio-oil, annually. A minimum of 20 thousand dry metric tons of biomass is to be processed. The selected harvesting sites are located in three forest districts in Oregon, USA: Tillamook, Astoria, and Forest Grove. Data about biomass types and availability were obtained from the Oregon Department of Forestry. The main types of biomass available in these forest districts are Douglas fir, red alder, and western hemlock. The following assumptions were also made: • The available amount of biomass is known • The type of truck for each route is known • The time horizon is one year • Green wood has a 50% moisture content [1] • Intermediate pyrolysis technology used in the transportable bio-refinery has 50% production yield [11] • The higher heating value of bio-oil is 17.6 MJ/kg [25] • The total distance from the storage facility to the end use is assumed as 150 miles (~241 km) roundtrip The roundtrip distances between the collection sites, and transportable and fixed bio-refineries are defined with the assistance of GIS (geographic information system) using a shortest path method. Table 2 presents roundtrip distances, numbers of tractor-trailer and tanker truck trips, and truck capacities for the upstream and midstream segments of the mixed SC, e.g., from harvesting sites to the transportable biorefinery, and then to storage near the fixed bio-refinery.
In an earlier study, a mathematical optimization model was developed to assess the bio-oil commercialization [11]. Using the same assumptions, an optimal solution for the proposed mixed SC network indicated that available amount of biomass would be processed using two transportable bio-refineries. The total annual production volume was predicted to be about 2.2 million gallons of bio-oil. The total annual cost of production was predicted to be about $2.4 million, resulting in a cost of $1.10 per gallon ($0.29 per liter). The results for the environmental impact analysis are reported in the next section. RESULTS AND DISCUSSION Both the traditional and new pathways use similar equipment for biomass collection, bio-refinery processing, and bio-oil combustion. Further, the GHG emissions resulting from these activities have similar values in both pathways. For instance, both fixed and transportable bio-refineries employ pyrolysis process technology for bio-oil production, and the total emission factor is predicted to be 1,297 kg CO2 eq. per metric ton of bio-oil for both. Consequently, the environmental impact (GWP) of a bio-refinery producing 10,000 metric tons bio-oil from 20,000 metric tons biomass is 12,970,000 kg CO2 eq. Additionally, the total emission factor for biomass collection and bio-oil combustion in both pathways are predicted as 71 kg CO2 eq. per metric ton of biomass and 4,125 kg CO2 eq. per metric ton of bio-oil. Figure 4 indicates the GWP for each life cycle stage associated with bio-oil production and consumption. The CO2 eq. absorption value is assumed to be equal to the CO2 eq. released into atmosphere through bio-oil combustion. Since it is assumed to use biochar as an energy source, the majority of biorefinery processing GHG emissions are negligible. The key TABLE 2. BASE CASE TRANSPORTATION DETAILS Distance Truck Tanker Truck Capacity Pathway (miles) Trips Trips (metric tons) H to TB 29,412 1,471 14 TB to FB 54,945 549 18 H to S 29,412 1,471 14 S to FB 86,957 870 23 H: harvesting, TB/FB: transportable/fixed bio-refinery, S: staging site
Copyright © 2016 by ASME
231 Global Warming Potential (kg CO2 eq. per gallon)
difference between the traditional and new pathway results from 25and truck-tanker trips between the harvesting sites and the truck fixed bio-refinery (also the bio-oil storage location). The total 20 emission in the Base Case using the truck-tanker pathway is Mobile Bio-refinery Processing liter of bio-oil) and 783,500 kg CO2 eq. (0.090 kg CO2 eq. per 15 Bio-refinery eq. (0.140 kg CO2 eq. truck-truck pathway is 1,148,000 kg CO2Fixed Processing per 10 liter of bio-oil). Truck-Truck Pathway
The GHG emissions from bio-oil combustion are classified 5 Pathway as biogenic, or part of the natural cycle,Truck-Tanker and are absorbed by new,0 growing biomass. Particulates emissions are negligible Bio-oil Combustion over the all life cycle stages. GHG emissions from biomass -5 collection, grinding, and transportation result from fossil fuel Biomass Collection combustion, and are not considered as part of the natural carbon -10 cycle. The net cradle to grave GHG emissions for bio-oil (i.e., CO2 Absorption biomass harvesting, collection, grinding, transportation, bio-15 Net Emissions refinery processing, distribution, bio-oil combustion, and CO2 eq.-20 absorption) is predicted as 15,170,000 kg CO2 eq. (1.82 kg New Pathway Traditional Pathway
FIGURE 4. CRADLE TO GRAVE GLOBAL WARMING POTENTIAL FOR BIO-OIL PRODUCTION AND CONSUMPTION CO2 eq. per liter of bio-oil) using a transportable bio-refinery and truck-tanker pathway, and 15,534,000 kg CO2 eq. (1.86 kg CO2 eq. per liter of bio-oil) using fixed bio-refinery and trucktruck pathway. The new pathway reduces GHG emissions by 365,000 kg CO2 eq. The result indicates GHG emissions reduce by about 2.3% through the use of the transportable bio-refinery and truck-tanker transportation pathway.
Effect of Number of Truck Trips The number of truck trips has a direct effect on environmental impacts, since more truck trips will result in more fuel use and more emissions. Truck capacities and biomass moisture content can reduce or increase the number of truck trips. In Case 2, the number of tanker trips between the transportable and fixed bio-refinery is increased by 100%. Also, the number of tuck trips between the staging site and fixed biorefinery is increased by 100%. The results show an increase of 497,800 kg and 868,700 kg CO2 eq. for the new and traditional pathways, respectively, compared to the Base Case. There is about 4.6% non-biogenic GHG emissions reduction by using the mixed SC, rather than the traditional SC, in Case 2. Effect of the Available Amount of Biomass The amount of available biomass in the Base Case is based on the remaining non-merchantable products at the roadside, which is not the only potential source of biomass for bio-oil production at the sites considered. Other sources include slash, branches, tops, and breakage, equating to an additional 30,000 metric tons, which are mainly burned due to high collection costs. This added amount of biomass would increase the environmental impacts from biomass collection, size reduction, and transport. Case 3 investigates the effect of increasing the amount of available biomass. The results indicate that GWP is increased by 1,105,000 and 1,662,000 kg CO2 eq. for the new and traditional pathways, respectively, compared to the Base Case. The mixed SC reduced GHG emissions by about 4.8% compared to the traditional SC, in Case 3.
SENSITIVITY ANALYSIS Travel distance, truck trips, available amount of biomass are the major factors that can have crucial effects on the environmental and economic performance of the bioenergy SC. Therefore, another case, in addition to Base Case, is considered to evaluate the effects of each factor on the proposed SC.
Since the upstream activities (e.g., biomass collection and transportation) mainly use fossil-based energy sources, the emissions released during these activities negatively affect environmental performance. Figure 5 indicates the results of these activities in each case study investigated in sensitivity analysis.
Effect of Travel Distance Travel distance is a primary factor in biomass-to-bioenergy SC decision making. The location of the staging site and biorefinery determines distances and influences environmental impacts. In Case 1, the effect of the fixed bio-refinery location is investigated. The distance between the transportable and fixed bio-refinery is increased by 50%, and the distance between staging site and fixed bio-refinery is increased by 50%. The results indicate that GWP increases by about 248,900 and 434,400 kg CO2 eq. for the new and traditional pathways, respectively, compared to the Base Case results. Total GWP is about 3.5% lower when using the new, mixed SC, as opposed to the traditional SC, in Case 1. Non-biogenic GHG emissions from biomass and bio-oil transport have the highest impacts.
CONCLUSION AND FUTURE WORK Growing concerns about increasing energy consumption indicate that alternative energy sources are essential to improve energy independence, aid rural economic development, and reduce environmental impacts. For these reasons, biomassbased energy resources represent a promising replacement for fossil-based energy. Prior studies have reported that biomass will play a key role due to its availability, low cost, and abundance. Biomass based energy can reduce the environmental impacts in comparison with fossil fuel use, especially due to its potential to achieve carbon neutrality. However, a large fraction of biomass resources have been underutilized for bioenergy production. Therefore, methods for achieving reliable supply and affordability are crucial to address the existing challenges in bioenergy production processes and systems.
Copyright © 2016 by ASME
232 Global Warming Potential (kg CO2 eq. per gallon) Case 3
Case 2
Case 1 Trailer Truck Tanker Truck Biomass Collection
Base Case 0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
FIGURE 5. LIFE CYCLE IMPACTS OF COLLECTION AND TRANSPORTATION OF BIOMASS AND BIO-OIL IN MIXED SUPPLY CHAIN This study investigated the environmental impacts of biooil production. Bio-oil produced from sustainably managed biomass could decrease environmental impacts of transportation fuels and other applications that rely on oil feedstocks. Bio-oil quality and cost are two major barriers to commercialization. This study is part of a larger effort that aims to promote sustainability of the bio-oil supply chain (SC) through a mixed SC that employs transportable and fixed bio-refineries. The traditional pathway employs a fixed bio-refinery and a trucktruck transportation approach, including double-trailer trucks to reduce truck trips by assembling loads to take advantage of the maximum allowable legal weight based on the number of axles and axle spacing. The new, mixed pathway employs a transportable bio-refineries and truck-tanker transportation approach, including trailer trucks and tankers for transferring biomass and bio-oil. The mixed SC network is presented to compare the cradle-to-grave environmental impacts of traditional and new SC pathways by considering all life cycle stages, i.e., harvesting, collection, grinding, transportation, production, distribution, and combustion. The analysis demonstrates the benefits of the mixed bio-oil SC and the role of transportable bio-refinery and truck-tanker pathway. The results indicate that forest biomass harvesting, collection, grinding, and transportation are the main activities in bio-oil SCs that release fossil carbon emissions. In this study, the majority of required energy for bio-oil production is sourced from biochar and syngas. Thus, the resulting GHG emissions are classified as biogenic, which are absorbed by trees in the forest and released to atmosphere from biomass combustion. The impact (GWP) of each life cycle phase is evaluated by quantifying the energy consumption and GHG emissions, as well as analyzing resource use. The results illustrate that the substitution of the traditional with the new, mixed SC pathway can reduce GHG emissions by 2-5% for the cases studied. This research shows that locating transportable biorefineries close to the harvesting and collection sites can reduce
costs and environmental impacts of processing low energy density biomass by reducing the number of truck trips and fuel consumption. Therefore, the mixed SC represents a promising approach meet cross-cutting environmental and economic goals. By taking advantage of ongoing work in supply chain and process technology development, as well as mathematical modeling and optimization, viable commercial approaches will emerge to support sustainable bioenergy. In addition to technological approaches, future work should address the influence of social aspects on the cross-cutting sustainability of bioenergy. This industry has the potential to positively impact current and future generations through economic development and improved environmental conditions. NOMENCLATURE EFmass Total emission factor of biomass transportation EFcomb Total emission factor of bio-oil combustion EFoilCO2 CO2 emission factor of bio-oil combustion EFoilCH4 CH4 emission factor of bio-oil combustion EFoilN2O N2O emission factor of bio-oil combustion EFoil Total emission factor of bio-oil transportation EFpro Total mission factor of production process EFproCO2 CO2 emission factor of production process EFproCH4 CH4 emission factor of production process EFproN2O N2O emission factor of production process EFup Total emission factor of upstream activities EFupCO2 CO2 emission factor of upstream activities EFupCH4 CH4 emission factor of upstream activities EFupN2O N2O emission factor of upstream activities EImass Environmental impacts of biomass transportation EIoil Environmental impacts of bio-oil transportation EIcomb Environmental impacts of bio-oil combustion EIpro Environmental impacts of production process EIup Environmental impacts of upstream activities RCO2 CO2 coefficient RCH4 CH4 coefficient RN2O N2O coefficient TrT Truck trips TaT Tanker truck trips W Average truck weight in round trip ACKNOWLEDGMENTS The authors wish to express gratitude to Michael Wilson and Robert Nall of the Oregon Department of Forestry, Phillip C. Badger (Renewable Oil International), and UPM Lappeenranta Biorefinery, Finland for providing valuable input to support this research. REFERENCES [1] Steele, P., Puettmann, M. E., Kanthi Penmetsa, V., and Cooper, J. E., 2012, “Life-cycle assessment of pyrolysis bio-oil production,” Forest Products Journal, 62(4), p. 326. [2] U.S. EIA, 2015, Annual Energy Outlook 2015, U.S. Energy Information Administration.
Copyright © 2016 by ASME
233 [3]
[4] [5] [6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
U.S. EIA, 2014, “Oil: Crude and Petroleum Products Energy Explained, Your Guide To Understanding Energy.” Union of Concerned Scientists, 2012, “The Promise of Biomass Clean Power and Fuel - If Handled Right.” US DOE, 2014, Bioenergy Walkthrough - Bioenergy Technology Office. Lattimore, B., Smith, C. T., Titus, B. D., Stupak, I., and Egnell, G., 2009, “Environmental factors in woodfuel production: Opportunities, risks, and criteria and indicators for sustainable practices,” Biomass and bioenergy, 33(10), pp. 1321–1342. Mirkouei, A., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015, “A Review and Future Directions in TechnoEconomic Modeling and Optimization of Upstream Forest Biomass to Bio-oil Supply Chains,” Renewable and Sustainable Energy Reviews, in revision. Naik, S. N., Goud, V. V., Rout, P. K., and Dalai, A. K., 2010, “Production of first and second generation biofuels: a comprehensive review,” Renewable and Sustainable Energy Reviews, 14(2), pp. 578–597. US EPA, 2015, “Program Overview for Renewable Fuel Standard Program” [Online]. Available: http://www.epa.gov/renewable-fuel-standardprogram/program-overview-renewable-fuel-standardprogram. [Accessed: 17-Jan-2016]. Mirkouei, A., Mirzaie, P., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015, “Reducing the Cost and Environmental Impact of Integrated Fixed and Mobile Bio-Oil Refinery Supply Chains,” Journal of Cleaner Production, doi:10.1016/j.jclepro.2015.11.023. Mirkouei, A., Haapala, K. R., Sessions, J., and Murthy, G. S., 2015, “Multi-criteria Decision Making for Sustainable Bio-Oil Production using a Mixed Supply Chain,” ASME Journal of Manufacturing Science & Engineering, in review. Perlack, R. D., Eaton, L. M., Turhollow Jr, A. F., Langholtz, M. H., Brandt, C. C., Downing, M. E., Graham, R. L., Wright, L. L., Kavkewitz, J. M., Shamey, A. M., and others, 2011, “US billion-ton update: biomass supply for a bioenergy and bioproducts industry.” Eisenbarth, S., and Van Treuren, K., 2004, “Sustainable and responsible design from a christian worldview,” Science and Engineering Ethics, 10(2), pp. 423–429.
[14] Sorenson, C. B., 2010, “A Comparative Financial Analysis of Fast Pyrolysis Plants in Southwest Oregon,” The University of Montana Missoula, MT. [15] Mirkouei, A., and Haapala, K. R., 2015, “A Network Model to Optimize Upstream and Midstream Biomass-toBioenergy Supply Chain Costs,” ASME 2015 International Manufacturing Science and Engineering Conference (MSEC), MSEC2015-9355, June 8-12, 2015, Charlotte, NC. [16] Mirkouei, A., Haapala, K. R., Sessions, J., and Murthy, G. S., 2016, “Evolutionary Optimization of Bioenergy Supply Chain Cost with Uncertain Forest Biomass Quality and Availability,” Proceedings of the IIE/ISERC, May 21-24, Anaheim, California, USA, in review. [17] Thornley, P., Gilbert, P., Shackley, S., and Hammond, J., 2015, “Maximizing the greenhouse gas reductions from biomass: The role of life cycle assessment,” Biomass and Bioenergy, 81, pp. 35–43. [18] Johnson, L., Lippke, B., and Oneil, E., 2012, “Modeling Biomass Collection and Woods Processing Life-Cycle Analysis*,” Forest Products Journal, 62(4), pp. 258–272. [19] Ringer, M., Putsche, V., and Scahil, J., 2006, Large-Scale Pyrolysis Oil Production: A Technology Assessment and Economic Analysis. Golden (CO): National Renewable Energy Laboratory; 2006 Nov. Report No, NREL/TP510-37779. Contract No.: DE-AC36-99-GO10337. [20] United Nations, 2014, “Global Warming Potentials.” [21] Leal Filho, W., 2010, Universities and Climate Change: Introducing Climate Change to University Programmes, Springer Science & Business Media. [22] Septien, S., Valin, S., Dupont, C., Peyrot, M., and Salvador, S., 2012, “Effect of particle size and temperature on woody biomass fast pyrolysis at high temperature (1000–1400 C),” Fuel, 97, pp. 202–210. [23] “SimaPro - World’s Leading LCA Software Package | PRé Sustainability” [Online]. Available: http://www.presustainability.com/simapro. [Accessed: 21-Jan-2015]. [24] ODF, 2015, “State of Oregon: Oregon Department of Forestry Home” [Online]. Available: http://www.oregon.gov/odf/Pages/index.aspx. [Accessed: 12-Oct-2015]. [25] Dynamotive, 2009, “Dynamotive Energy Systems Corporation, Canadian BioOil Plant: Summary (USD).”
Copyright © 2016 by ASME
234
Appendix E: Integration of Machine-Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process
Integration of Machine Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process
Integration of Machine Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process Amin Mirkouei* and Karl R. Haapala School of Mechanical, Industrial and Manufacturing Engineering Oregon State University Corvallis, OR 97331-2600, USA ABSTRACT Recent concerns over the use of and reliance on fossil fuels have stimulated research efforts in identifying, developing, and selecting alternative energy sources. Biofuels represent a promising replacement for conventional fuels for heating and mobility applications, however, variability in the quality and availability of biomass feedstocks greatly affect the utility of biofuels due to the impact on cost and life cycle environmental performance. Thus, methods for mitigating these potential impacts are needed when selecting biomass feedstock suppliers. In the research herein, the selection of the best supplier is investigated for a biomass supply chain (BSC) network by including both qualitative and quantitative factors. Most existing supplier-selection methods consider four steps: (1) Problem formulation, where Decision-Tree Analysis is applied as a qualitative method for defining the type of biomass feedstock materials for biofuel production, (2) Criteria definition, (3) Preevaluation of qualified suppliers, which employs the Support Vector Machine (SVM) method, and (4) Final selection. Integration of machine learning (ML) techniques and a mathematical programming model is undertaken with this method to select the most appropriate feedstock suppliers. It is shown that integrating ML and mathematical programming methods offers a promising approach to supplementing existing supplier selection methods for biomass-to-biofuel supply chains.
1.
INTRODUCTION
In recent decades, biofuel has been recognized as a potential source of energy that could have positive effects on the environment, economy, and society. Biofuel is made from organic material or biomass, grown in fields and forests [1]. A successful biofuel industry would benefit society by reducing energy and fuel costs, while reducing the imports of oil and improving energy security. Biofuel has been proposed as a replacement for conventional liquid fuels because it can reduce life cycle emissions, and their associated impacts, e.g., climate change [2]. Sharma et al. [3] provided a starting point for understanding biomass feedstocks and biofuel production and presented a review of biomass supply chain (BSC) design and modeling, specifically for mathematical programming. BSC design and modeling must account for uncertainties such as seasonality, weather, physical and chemical characteristics, distribution, and supplier agreements [4]–[6]. The BSC includes the biomass feedstock supplier, storage sites, biorefinery sites, pre-treatment facilities, and distribution [7]. Globalization and a competitive environment have lead companies to give more attention to consumer expectations, final price, product quality, and lead times. For a firm to remain competitive and achieve environmental goals, careful supply chain management is critical. Supply chains move material between the source and end-users, and consist of suppliers, manufacturers, distributors, and customers [8]. Supplier selection is a key aspect of supply chain management. Since the 1960s, several methods have been introduced by investigating different stages and characteristics of the supplier selection process. The supplier selection process has been divided into four steps: Problem formulation, criteria definition, pre-evaluation of qualified suppliers, and final selection [9]. Figure 1 indicates the steps of the supplier selection process, based on work by Aissaoui et al. [10]. In general, the two final steps of this process have been a focus of prior research [11]–[13]. In the research herein, the selection of the best supplier is investigated for a biomass supply chain (BSC) network by including both qualitative and quantitative factors by adapting the above supplier selection process. Machine learning (ML) techniques and a mathematical programming model are integrated to select the most appropriate *
Corresponding author: Tel: (541) 602-3488; Email:
[email protected]
235
Flexible Automation and Intelligent Manufacturing, FAIM2014
biomass feedstock suppliers. Most decision tools for problem formulation include qualitative methods that assist experts and decision-makers in carefully identifying the need for a decision and the alternatives that seem to be available [10]. The first step is to identify what purchasers require of the supplier. The Kraljic Portfolio Purchasing Model is often applied to make right decisions in purchasing process [14]. Figure 2 indicates Kraljic’s classification according to the potential Profit Impact and Supply Risk for each item. Problem Formulation Formulation of Criteria Pre-selection of Suppliers
Final Selection
Make versus buy? More or fewer suppliers? Replace current supplier?
Consider relative importance of various criteria to evaluate all suppliers. Develop proposed list of vendors. Approve sellers. Conduct quotation analysis. Perform order allocation. Figure 1: Decision methods in supplier selection [9]
According to Kraljic [14], Profit Impact is high when items add significant value to the company's output (product), for example items may comprise a high proportion of the final product. Supply Risk is high when items have constrained availability, for example due to scarcity, logistics challenges, or government instability. Routine items (low profit impact, low supply risk) are non-critical items produced in standard configuration. The best method of controlling these items is to optimize inventory, and there is no need to consider other attributes. Bottleneck items (low profit impact, high supply risk) are those whose supply involves various risks and problems. In this situation, contract guarantees, supplier control, and strategies to maintain high inventory levels are recommended. Leverage items (high profit impact, low supply risk) include the materials for which the buyer has maneuverability to bargain and readily available alternative products and suppliers. Strategic items (high profit impact, high supply risk) require the most attention since they are critical to ensuring high output value, but are attendant with supply constraints. The company can focus on developing long-term supply relationships, assessing risks regularly, and devising contingency plans to mitigate risk. They may choose to make the item in-house, rather than purchasing it. 2.
BIOMASS FEEDSTOCK SUPPLIER SELECTION PROCESS
Profit Impact
Biomass feedstocks are commonly classified into first (edible crops), second (lignocellulosic and other nonedible sources), and third (algal biomass) generations. The various types of biomass feedstocks can be converted into biofuels [15]. Each generation of feedstocks can be classified according to the Kraljic purchasing model depending upon their supply risks and profit impacts. The BSC includes several phases from harvesting to the arrival of biomass feedstock at the bio-refinery. An objective of this research is to develop a supplier selection process to predict BSC network cost that will account for variation in feedstock quality and variability in feedstock availability over time. The general approach is presented below.
Leverage items
Strategic items
Routine items
Bottleneck items
Supply Risk
Figure 2: Kraljic’s classification matrix for purchasing various items [14]
2.1. PROBLEM FORMULATION For this step, the Decision Tree (DT) method is applied. The DT method is commonly used in machine learning and data mining. DT analysis can take categorical, binary, and numeric value input and output variables, and it can handle missing attributes and outliers well. DT analysis is also good in explaining reasoning for its prediction and therefore gives good insight about the underlying data [16]. Classification tree analysis is a DT approach usually
236
Integration of Machine Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process
used to illustrate the basic idea of machine learning. A classification tree approach is developed next and employed to explain the concepts as a part of the narrative. Classification tree analysis is conducted with the assistance of R.
Figure 3: Classification tree for purchasing three different generations of biomass feedstock (R results)
A randomly generated dataset is used to provide profit impact and supply risk values for 200 suppliers for three different generations of biomass feedstocks. Figure 3 indicates the basic form of classification tree analysis for the three generations, which starts with all items in one group and then identifies characteristics (shown in Figure 2) that best classifies each item into one of the three generations. A random sample is chosen from the dataset, rather than stratifying by all items. Half of the data is used as a training set and the other half is the testing set. In general, DT analysis has a fairly low misclassification error rate. From the figure, for instance, it can be seen that items with supply risk > 0.4 and profit impact > 0.5 are classified as third-generation biomass feedstocks. 2.2. FORMULATION OF CRITERIA Supplier selection is a term used in supply chain management that refers to the process of evaluating and approving potential suppliers by several criteria. Dickson [17] identified 23 different criteria evaluated in supplier selection. Most research related to the evaluation and selection of suppliers has identified price, quality, capacity, and delivery time to be the key criteria considered in supplier selection problems. In this research, each of these factors will be considered (in section 2.3 and 2.4) due to their importance in selecting a proper biomass feedstock suppliers. 2.3. PRE-SELECTION OF POTENTIAL SUPPLIERS In this step, pre-selection is a quantitative method that evaluates and selects qualified suppliers from all suppliers. The Support Vector Machine (SVM) method is a supervised method in machine learning, and uses historical company data as inputs and outputs. SVM can provide a learning technique for pattern recognition which is reasonable in learning theory [18]. In addition, SVM is equivalent to solving a linear constrained quadratic programming problem, so that the SVM solution is always unique and globally optimal [19], [20]. A dataset of 150 suppliers provides an example for the corresponding features in each feature sample. This dataset would typically be obtained from historical data for each company, but is randomly generated for this study. The outputs for each supplier define the set of performance data for that supplier. In supply chain management, it is notably difficult to obtain this type of data, and large datasets are not readily available. Thus, the provided dataset is defined as an example for this step in order to validate the performance of supplier selection process. Table 1 shows an instance of the pre-evaluation of suppliers, including the identified key criteria. These data form the training set, which contains input and output information. The performance of suppliers can be predicted after defining the weights for each criterion. SVM analyzes supplier data to identify any patterns that are then used for classification. SVM then uses the key criteria identified in the second step (Formulation of Criteria) as features to find the weight of each criterion. In literature, these features are called credit indexes, which are quantified from historical company data [21]. The outputs (target variables) are the final supplier credit index values that define the performance of the suppliers. The approach assumes historical data can be obtained for a given the output. It is further assumed if data is not available it can be obtained based on short-term supplier selection experiences [22]. High output scores identify suppliers that have the highest probability of being selected.
237
Flexible Automation and Intelligent Manufacturing, FAIM2014
Table 1: Example of training set for supplier selection
Training data
Price ($) 7 5 10 9
66 61 69 76
Failure rate 0.07 0.11 0.08 0.05 ?
0.89 0.73 0.69 0.8
SVM finds data pattern
…
?
Output
…
?
Tardiness rate 0.9 0.6 0.8 0.8
…
…
?
Capacity
…
…
… Weight
Constant cost ($) 705 791 968 655
History of data
Supplier No. 1 2 3 4
?
The training set along with the testing set are used to train and test the SVM. The provided dataset contains values for 100 suppliers as a training dataset and 50 supplier values as a test dataset. Table 2 shows SVM can find the pattern and the weights from the training information in the training phase [23]. Support vectors recognize this pattern of the data, which define the weight for each criterion. Table 2: Identified weights from the Support Vector Machine (SVM) method (R results) Items Weight
Price ($) 513
Constant cost ($) 569.9
Capacity 481.27
Failure rate 56.67
Tardiness rate 56.12
The weights found in the training phase can be used to predict output (performance) in the test set (Table 3) to classify the qualified and unqualified suppliers.
Capacity
…
…
…
71 66 72 62
Failure rate 0.13 0.07 0.13 0.07
Tardiness rate 0.7 0.6 0.7 0.6
Output
…
Constant cost ($) 777 651 764 991
…
Price ($) 8 5 9 7
…
Supplier No. 1 2 3 4 …
Testing data
Table 3: Example of test set of supplier selection
? ? ? ?
SVM finds performance rate
In general, the SVM learning algorithm provides us with a quantitative tool for supplier evaluation and selection. SVM has been applied for the pre-selection of potential supplier for the following reasons [22]:
Pre-selection of potential suppliers by other methods, such as the Analytic Hierarchy Process, are an involved and prolonged task. Also, even if the company has a wealth of historical data, the training and testing tasks will take a short amount of time to identify the data pattern.
SVM is a non-parametric and objective method, since it focuses on identifying patterns in the data.
SVM can find a solution even if some data are missing.
Unlike statistical methods, SVM can control large-scale problems.
SVM is a supervised method that is more accurate than unsupervised methods, e.g., clustering analysis.
2.4. FINAL SELECTION In the final step, after evaluating and selecting some potential suppliers via SVM methods, a mathematical model is introduced to select the supplier with the aim of minimizing the cost of purchasing biomass feedstocks, considering limitations such as capacity, quality, supply base, lead time, and other limitations when numerous sources of uncertainty exists in the BSC with respect to quality and lead time such as weather uncertainty, low bulk density of biomass feedstocks, and suppliers contracts and government policies [3]. The mentioned criteria (in step 2.2) have been used for selecting the final suppliers. A genetic algorithm has been applied to solve the problem and
238
239
Integration of Machine Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process
a numeric example has been provided, along with sensitivity analysis, in order to validate the performance of the model. The following mathematical method provides a quantitative tool for biomass feedstock supplier selection. 2.4.1.
HYPOTHESES
The amount of demand of each strategic item is definite.
The supply base for each commodity is fixed and is pre-determined.
The failure rate (defect rate) of one random variable is specified with a standard normal distribution and the maximum acceptable defect rate of commodity.
The tardiness rate (delay rate) of one random variable is specified with a standard normal distribution and the maximum acceptable delay rate.
2.4.2.
DEFINITION OF PARAMETERS
Di: Demand of ith biomass feedstock. Yij: The amount of ith biomass feedstock purchased from jth supplier. Xij: Binary variable (0 or 1) indicating selection or non-selection of jth supplier for ith biomass feedstock. Cij: The capacity of jth supplier for supplying ith biomass feedstock. qij: Failure rate of jth supplier for supplying ith biomass feedstock. qia: Maximum acceptable failure rate of ith biomass feedstock. Lij: Tardiness rate of jth supplier for supplying ith biomass feedstock. Lia: Maximum acceptable tardiness rate of ith biomass feedstock. T: Supply base. Pij: The purchase price of ith biomass feedstock from j supplier. Fij: Fixed cost (constant cost) of jth supplier for supplying ith biomass feedstock. TBP: The total budget for purchasing biomass feedstock. 1-A: The probability of evaluating the considered quality, which is specified in advance by the decision maker. 1-B: The probability of evaluating the time between ordering and receiving the considered commodity. 2.4.3.
THE FINAL MATHEMATICAL MODEL
The objective function is obtained by the sum of fixed and variable costs of the commodities from the selected suppliers (Eq. 1). The model considers several constraints: capacity, quality, supplies base, lead-time, demand, total amount of budget, and non-negative limitation, respectively (Eqs. 2-9). 𝑚
𝑖 𝑀𝑖𝑛 ∑𝑛𝑖=1 ∑𝑗=1 (𝑃𝑖𝑗 𝑌𝑖𝑗 + 𝐹𝑖𝑗 𝑋𝑖𝑗 )
(1)
ST: 𝑌𝑖𝑗 ≤ 𝐶𝑖𝑗 𝑋𝑖𝑗 𝑖 = 1 … 𝑛, 𝑗 = 1 … 𝑚𝑖 𝑚
𝑖 ∑𝑗=1 𝑋𝑖𝑗 𝜇𝑖𝑗 + 𝑀∅−1 (1 − 𝐴) ≤ 𝑞𝑖𝑎
𝑚
𝑖 ∑𝑗=1 𝑋𝑖𝑗 ≤ 𝑇 𝑖 = 1 … 𝑛
𝑚
𝑖 ∑𝑗=1 𝑋𝑖𝑗 𝜇𝑖𝑗 + 𝑁∅−1 (1 − 𝐵) ≤ 𝐿𝑖𝑎
𝑚
𝑖 ∑𝑗=1 𝑌𝑖𝑗 ≥ 𝐷𝑖
𝑚
𝑖 ∑𝑛𝑖=1 ∑𝑗=1 𝑃𝑖𝑗 𝑌𝑖𝑗 ≤ 𝑇𝐵𝑃
(2) (3) (4) (5) (6) (7)
𝑋𝑖𝑗 = 0 𝑜𝑟 1 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑖 𝑎𝑛𝑑 𝑗
(8)
𝑌𝑖𝑗 ≥ 0
(9)
𝑖 = 1 … 𝑛, 𝑗 = 1 … 𝑚𝑖
Flexible Automation and Intelligent Manufacturing, FAIM2014
2.4.4.
NUMERICAL EXAMPLE
In this section, the model is investigated through a numerical example using two methods: exhaustive search and a Genetic Algorithm. In the example, the hypothesis is that the management of a company tends to minimize the cost for suppliers in preparing a biomass feedstock. As Table 4 shows, the goal is to select the best supplier when the demand for the commodity is 90 units. Meanwhile, the supply base (number of suppliers) for the first commodity is considered to be three (T=3). The total allocated budget for preparing this biomass feedstock is $150,000. Table 4: Supplier information for biomass feedstock supply chain Items Price ($) Constant cost ($) Supplier No. 1 088 0888 Supplier No. 2 088 058 Supplier No. 3 008 058 Supplier No. 4 088 088 The demand for this commodity is 90 units The maximum number of suppliers is 3 (T=3)
2.4.5.
Capacity (Ton) 08 08 05 55
Failure rate N (0.0040,0.050) N (0.0034,0.055) N (0.0044,0.065) N (0.0040,0.055) q1a =8105 1-A=8105
Tardiness rate N (0.0024,0.065) N (0.0049,0.055) N (0.0040,0.050) N (0.0040,0.050) L1a =810 1-B=810
SOLVING THE MATHEMATICAL MODEL
For the given commodity with four suppliers, there are 16 cases for selecting suppliers; and, as indicated in Table 5, feasibility or infeasibility is determined based on the demand and the amount of the commodity supplied. Table 5: Example of solutions satisfying demand and amount supplied (Constraints 2, 6) Case 1 2
Supplier No. 1 1 1
Supplier No. 2 1 1
1 0
65 0
Feasibility Feasible Feasible …
0 0
Amount (Yij) 290 225 …
0 0
…
0 0
Supplier No. 4 1 0
…
…
…
…
15 16
Supplier No. 3 1 1
Infeasible Infeasible
Among the feasible cases explored using the exhaustive search technique (evaluating and comparing all possible scenarios), the best case is shown in Table 6. It is seen that the first and third suppliers are selected. The lowest cost for two suppliers is equal to $74,750. It is possible to realize the feasibility of the solution based on the limitations of supply capacity and demand of this commodity due to the low number of suppliers. In larger problems, however, this decision will be very difficult because, in addition to the two selection limitations (quality and lead-time), there are stochastic constraints. Therefore, this model is constituted as NP-hard. The Genetic Algorithm (GA) is applied to solve this model, which is a common approach to solving NP-hard models [24]. Table 6 also indicates the GA solution from MATLAB, and shows that the first and the fourth suppliers have been selected with this approach. The best and the least cost for supplying this strategic commodity is $74,900, which is less than the initial solution. Table 6: Comparision of solutions (units purchased and overall cost) using two methods Methods Exhaustive Search
Supplier No. 1
Supplier No. 2
Supplier No. 3
Supplier No. 4
80
0
10
0
Overall Cost $74,750
Genetic Algorithm
79
0
0
11
$74,900
One reason for the difference between these two solutions is due to the stochastic constraints. The exhaustive search satisfies the stochastic constraints to find a feasible solution GA satisfies not only the stochastic constraints, but also selects the supplier (# 4) exhibiting lower variance. To clarify, variance is a measure of uncertainty or, as in this problem, quality. The overall cost of the numerical example obtained by the exhaustive search method and the GA method are very close (within $150, or a 0.1% difference). Figure 4 indicates the evaluation of objective function (cost) for the GA iterations. The top graph shows the cost for each consecutive generation. It shows that the algorithm quickly reaches the optimum value (after 20 generations). The bottom graph is related to the objective function, where the solid line indicates the optimum cost and the dashed line indicates the mean cost in each generation.
240
Integration of Machine Learning and Mathematical Programming Methods into the Biomass Feedstock Supplier Selection Process
Figure 4: Genetic Algorithm cost results, Top: For each generation and Bottom: For function evaluations (MATLAB result)
2.4.6.
SENSITIVITY ANALYSIS
There are many methods for confirming mathematical optimization models. Thus, the supplier selection model can be evaluated using sensitivity analysis for three cases in future work:
Coefficients of basic variables in objective function: Changes in coefficients of basic variables can influence the optimization results.
Coefficients of non-basic variables in objective function: Changes in coefficients of non-basic variables should not influence the optimization results. Thus, if there is an effect methodological errors can be revealed.
Right hand numbers: In the provided model, the maximum acceptable failure rate of the ith biomass feedstock (qia) and the maximum acceptable tardiness rate of the ith biomass feedstock (Lia) are constituted as right hand numbers. The changes of this parameter can make the optimum solution of the problem feasible or infeasible.
3.
CONCLUSIONS
From the above method and application, the evaluation and selection of the best supplier is obtained for a hypothetical biomass feedstock supply chain. Due to numerous biomass supply uncertainties, research to support biomass supply chain (BSC) network design requires the development of advanced methods. Biomass feedstocks greatly affect the BSC due to the impact on cost and life cycle environmental impacts. One of the most applicable approaches to select suppliers is introduced applied in this research. The first step is Problem Formulation, which defines what buyers need to achieve from suppliers. The Kraljic Portfolio Purchasing Model determines the items in four clusters simultaneously to define which kind of supplier can provide these items. The decision tree method is applied to solve this classification model. The second step is criteria definition. While more than 200 factors are introduced in the literature, four of them (price, quality, capacity, delivery time) have been primarily used in prior work. In this effort, each of these four factors was considered due to their importance in selecting a proper supplier. The third step is pre-selection of potential suppliers, which evaluates and selects the qualified supplier from all suppliers. The SVM method is provided to find potential suppliers. In the final step that is significance phase, a mathematical model is proposed with the purpose of selecting the supplier which minimizes the cost of purchasing feedstock. Since the final model has two stochastic constraints, the genetic algorithm is demonstrated to solve this model for the hypothetical example. It is shown that this integrated machine learning and mathematical programming method offers a promising approach to supplementing existing supplier selection methods for biomass-to-biofuel supply chains. Future research must evaluate the robustness of the method, however, based on actual supply chain data. This future work will reveal the potential for agricultural development, technological advancements, and methodological approaches for ensuring viable and sustainable alternatives for conventional liquid fuels and other energy sources.
241
Flexible Automation and Intelligent Manufacturing, FAIM2014
ACKNOWLEDGEMENT The authors wish to acknowledge Ms. Rana Azghandi for her assistance in this research. REFERENCES [1]
“Energy 101: Feedstocks for Biofuels and More,” Energy.gov. [Online]. Available: http://energy.gov/eere/videos/energy101-feedstocks-biofuels-and-more. [Accessed: 04-Jan-2014].
[2]
F. You and B. Wang, “Life cycle optimization of biomass-to-liquid supply chains with distributed–centralized processing networks,” Ind. Eng. Chem. Res., vol. 50, no. 17, pp. 10102–10127, 2011.
[3]
B. Sharma, R. G. Ingalls, C. L. Jones, and A. Khanchi, “Biomass supply chain design and analysis: Basis, overview, modeling, challenges, and future,” Renew. Sustain. Energy Rev., vol. 24, pp. 608–627, Aug. 2013.
[4]
J. S. Cundiff, N. Dias, and H. D. Sherali, “A linear programming approach for designing a herbaceous biomass delivery system,” Bioresour. Technol., vol. 59, no. 1, pp. 47–55, Jan. 1997.
[5]
E. Iakovou, A. Karagiannidis, D. Vlachos, A. Toka, and A. Malamakis, “Wasted Biomass-to-Energy Supply Chain Management: A Critical Synthesis,” Elsevier Ltd, pp. 1860–1870, 2010.
[6]
S. Gold and S. Seuring, “Supply chain and logistics issues of bio-energy production,” J. Clean. Prod., vol. 19, no. 1, pp. 32–42, 2011.
[7]
D. Vlachos, E. Iakovou, A. Karagiannidis, and A. Toka, “A strategic supply chain management model for waste biomass networks,” in 3rd International Conference on Manufacturing Engineering, 2008, pp. 797–804.
[8]
H. Min and G. Zhou, “Supply chain modeling: past, present and future,” Comput. Ind. Eng., vol. 43, no. 1, pp. 231–249, 2002.
[9]
L. De Boer, E. Labro, and P. Morlacchi, “A review of methods supporting supplier selection,” Eur. J. Purch. Supply Manag., vol. 7, no. 2, pp. 75–89, 2001.
[10] N. Aissaoui, M. Haouari, and E. Hassini, “Supplier selection and order lot sizing modeling: A review,” Comput. Oper. Res., vol. 34, no. 12, pp. 3516–3540, 2007. [11] A. Amid, S. H. Ghodsypour, and C. O’brien, “Fuzzy multiobjective linear model for supplier selection in a supply chain,” Int. J. Prod. Econ., vol. 104, no. 2, pp. 394–407, 2006. [12] E. A. Demirtas and Ö. Üstün, “An integrated multiobjective decision making process for supplier selection and order allocation,” Omega, vol. 36, no. 1, pp. 76–90, 2008. [13] G. H. Hong, S. C. Park, D. S. Jang, and H. M. Rho, “An effective supplier selection method for constructing a competitive supply-relationship,” Expert Syst. Appl., vol. 28, no. 4, pp. 629–639, 2005. [14] P. Kraljic, “Purchasing must become supply management,” Harv. Bus. Rev., vol. 61, no. 5, pp. 109–117, 1983. [15] L. G. Papageorgiou, “Supply chain optimisation for the process industries: Advances and opportunities,” Comput. Chem. Eng., vol. 33, no. 12, pp. 1931–1938, 2009. [16] Y. Zhao, R and Data Mining: Examples and Case Studies. Academic Press, 2012. [17] G. W. Dickson, “An analysis of vendor selection systems and decisions,” J. Purch., vol. 2, no. 1, pp. 5–17, 1966. [18] V. N. Vapnik, “An overview of statistical learning theory,” Neural Netw. IEEE Trans. On, vol. 10, no. 5, pp. 988–999, 1999. [19] N. Cristianini and J. Shawe-Taylor, An introduction to support vector machines and other kernel-based learning methods. Cambridge university press, 2000. [20] J. Shawe-Taylor and N. Cristianini, Kernel methods for pattern analysis. Cambridge university press, 2004. [21] H. Guosheng and Z. Guohong, “Comparison on neural networks and support vector machines in suppliers’ selection,” J. Syst. Eng. Electron., vol. 19, no. 2, pp. 316–320, 2008. [22] G. Kim, “Solving support vector machine classification problems and their applications to supplier selection,” 2011. [23] A. Karatzoglou, D. Meyer, and K. Hornik, “Support vector machines in R,” 2005. [24] G. Zhou, H. Min, and M. Gen, “The balanced allocation of customers to multiple distribution centers in the supply chain network: a genetic algorithm approach,” Comput. Ind. Eng., vol. 43, no. 1, pp. 251–261, 2002.
242
