Development of a Genetic Algorithm for Evaluating ...

Journal of Energy and Power Engineering Volume 9, Number 1, January 2015 (Serial Number 86)

David Publishing

David Publishing Company www.davidpublishing.com

Publication Information: Journal of Energy and Power Engineering is published monthly in hard copy (ISSN 1934-8975) and online (ISSN 1934-8983) by David Publishing Company located at 240 Nagle Avenue #15C, New York, NY 10034, USA. Aims and Scope: Journal of Energy and Power Engineering, a monthly professional academic journal, covers all sorts of researches on Thermal Science, Fluid Mechanics, Energy and Environment, Power System and Automation, Power Electronic, High Voltage and Pulse Power, Sustainable Energy as well as other energy issues. Editorial Board Members: Prof. Ramesh K. Agarwal (USA), Prof. Hussain H. Al-Kayiem (Malaysia), Prof. Zohrab Melikyan (Armenia), Prof. Pinakeswar Mahanta (India), Prof. Carlos J. Renedo Estébane (Spain), Prof. Mohamed Ahmed Hassan El-Sayed (Trinidad and Tobago), Prof. Carlos Redondo Gil (Spain), Prof. Roberto Cesar Betini (Brazil), Prof. Rosário Calado (Portugal), Prof. Dr. Ali Hamzeh (Germany), Prof. Noor-E-Alam Ahmed (Australia), Prof. E. Ubong (USA), Prof. Shripad T. Revankar (USA), Prof. Almoataz Youssef Abdelaziz (Egypt), Prof. Guihua Tang (China), Prof. Mohammad Rasul (Australia), Prof. Rene Wamkeue (Canada), Prof. Ya-Ling He (China), Prof. Filippo de Monte (Italy), Prof. Masoud Rokni (Denmark), Prof. Hosni I. Abu-Mulaweh (USA), Prof. Quan Zhang (China), Prof. Peng-Sheng Wei (Taiwan), Prof. Vinod Kumar (India), Prof. Yuan-Kang Wu (Taiwan), Dr. Kaige Wang (USA), Dr. Fude Liu (Hong Kong), Prof. Isa Salman Hasan Qamber (Bahrain). Manuscripts and correspondence are invited for publication. You can submit your papers via Web Submission, or E-mail to [email protected] or [email protected]. Submission guidelines and Web Submission system are available at http://www.davidpublishing.org, www.davidpublishing.com. Editorial Office: 240 Nagle Avenue #15C, New York, NY 10034, USA E-mail: [email protected]; [email protected]; [email protected].

[email protected];

Copyright©2015 by David Publishing Company and individual contributors. All rights reserved. David Publishing Company holds the exclusive copyright of all the contents of this journal. In accordance with the international convention, no part of this journal may be reproduced or transmitted by any media or publishing organs (including various websites) without the written permission of the copyright holder. Otherwise, any conduct would be considered as the violation of the copyright. The contents of this journal are available for any citation. However, all the citations should be clearly indicated with the title of this journal, serial number and the name of the author. Abstracted / Indexed in: Database of EBSCO, Massachusetts, USA Database of Cambridge Science Abstracts (CSA), USA Chinese Database of CEPS, American Federal Computer Library Center (OCLC), USA Ulrich’s Periodicals Directory Summon Serials Solutions Chinese Scientific Journals Database, VIP Corporation, Chongqing, China Chemical Abstracts Service (CAS) ProQuest CNKI Google Scholar

Subscription Information: Price: Print $800 (per year) Online $600 (per year) Print and Online $880 (per year) David Publishing Company Tel: 1-323-984-7526, 001-323-410-1082; Fax: 1-323-984-7374; 323-908-0457 E-mail: [email protected] Digital Cooperative Company: www.bookan.com.cn

D

DAVID PUBLISHING

David Publishing Company www.davidpublishing.com

Journal of Energy and Power Engineering Volume 9, Number 1, January 2015 (Serial Number 86)

Contents Clean and Sustainable Energy 1

Biogas in Brazil: A Governmental Agenda Melissa Cristina Pinto Pires Mathias and João Felippe Cury Marinho Mathias

16

Analysis of Cooling Characteristics in Datacenter Using Outdoor Air Cooling Yoshiyuki Inoue, Hirofumi Hayama, Taro Mori, Koki Kikuta and Noriyuki Toyohara

25

Development and Test of an Experimental Apparatus to Study Thermal-Choking in IDEAL GASES and Self-decomposition in Superheated N2O Patrick Lemieux, Alberto Fara, Pablo Sanchez and William Murray

40

Green Product of Liquid Fuel from Plastic Waste by Pyrolysis at 900 °C Dianta Mustofa Kamal and Fuad Zainuri

45

Numerical Simulation on Ice Shedding Phenomena in Turbomachinery Ryosuke Hayashi and Makoto Yamamoto

54

The Shift from “Grid-Tie” to Partly “Off-Grid” Balint David Olaszi and Jozsef Ladanyi

Power and Electronic System 59

Stability Monitoring and Control of Generation Based on the Synchronized Measurements in Nodes of Its Connection Alexander Fishov, Maria Shiller, Anton Dekhterev and Vladimir Fishov

68

Identifying the Factors for Ensuring Customers Actively Engaged in Smart Grids Linda Hull, Even Bjørnstad, Yvonne Boerakker, Magnus Brolin, Yeoungjin Chae and Duncan Yellen

78

Development of a Genetic Algorithm for Evaluating the Performance of Overhead Power Distribution Lines and Proposing Solutions to Improve Distribution Line Safety Fabio Romero, Alden Antunes, Dario Takahata, André Meffe, Carlos Olivera, Fermando Lange, Hamilton Souza and João Castilho

83

MPPT Control System for PV Generation System with Mismatched Modules Chengyang Huang, Kazutaka Itako, Takeaki Mori and Qiang Ge

91

Application of an Analytical Method to Locate a Mixing Plane in a Supersonic Compressor Emmanuel Benichou and Isabelle Trébinjac

102

Thermal Design of Power Transformers via CFD Ralf Wittmaack

108

Static Analysis of Eolic Blade through Finite Element Method and OOP C++ Tiago Freires, Silvia Caroline and Raimundo Menezes Jr

D

Journal of Energy and Power Engineering 9 (2015) 1-15 doi: 10.17265/1934-8975/2015.01.001

DAVID

PUBLISHING

Biogas in Brazil: A Governmental Agenda Melissa Cristina Pinto Pires Mathias1 and João Felippe Cury Marinho Mathias2 1. Department of Mechanical Engineering, Pontifícia Universidade Católica, Rio de Janeiro 22,451-900, Brazil 2. Institute of Economics, Federal University of Rio de Janeiro, Rio de Janeiro 22,290-240, Brazil Received: March 20, 2014 / Accepted: May 04, 2014 / Published: January 31, 2015. Abstract: Environmental issues have become an important aspect in the energy sector and also in governmental policies all over the world. Renewable energy plays a decisive role in a cleaner energy production with less environmental impacts. In this new world scenario, several countries create incentive programs aiming to increase the renewable energy share in their energy balance. Biogas is an interesting example of a smart and rational use of organic waste material. Nevertheless, international experience shows that its production relays mainly on government incentives. This paper investigates what the Brazilian government role is to introduce and promote the biogas industry in the country. We estimate the production of agricultural waste energy using biogas derived from cattle and swine waste and analyze the Brazilian legal and regulatory framework for renewable energy, focusing on biogas production, transport and sale. The results point to an unexplored potential of the use of cattle and swine waste for biogas production in Brazil. However, this potential can only become a reality if some legal and regulatory issues are solved. Brazilian government agenda has to include not only filling the legal and regulatory blanks but also creating incentives for the use of biogas. Key words: Biogas, renewable energy, policy, Brazil.

1. Introduction 

The relationship between energy and economic development is widely known and discussed. In fact, there is extensive literature that associates economic growth with consumption of electricity and other energy sources and discusses the causal relationship between energy and economic growth. Despite the knowledge about the impact and external factors generated by energy use and production, there is an increasing demand for energy and, in order to meet this growing demand, investments must be made to allow an increase in energy supply, taking into account not only the amount of energy that is generated, but also the impact on society and the economy. Brazil has already produced a large amount of energy from renewable sources, mainly from hydropower plants (81.9% in 2011, according to the

Corresponding author: Melissa Cristina Pinto Pires Mathias, professor, research fields: natural gas, biomethane and regulation. E-mail: [email protected].

Brazilian Energy Balance 2012 [1])1. However, there is enormous potential for the use of other renewable sources of energy, such as solar energy, wind power, and sources generated from organic biomass. In order to encourage the use of renewable energy sources other than the large hydropower plants to produce electricity, in 2002, the Brazilian Government passed a law creating a policy called PROINFA (Incentive Program for Renewable Energy Sources). This policy program aims to push the development of renewable energy (wind, biomass, and small hydro). As a result of PROINFA, there has been a considerable increase in the supply of electricity generated by wind2 1

In 2011, according to Ref. [1], domestic electricity supply by source in Brazil was: 81.9% (hydro); 6.6% (biomass); 0.5% (wind); 4.4% (natural gas); 2.5% (oil products); 2.7% (nuclear); 1.4% (coal and coal products). 2 According to Ref. [1], the production of electricity from wind power reached 2,705 GWh in 2011. This represents a 24.3% increase over the previous year, when it reached 2,177 GWh. According to Ref. [2], a turning point for windpower development in Brazil was the renewable energy incentive program (Proinfa) of the Ministry of Mines and Energy, established in 2002. The installed capacity for wind power rose from 56 GWh to 2,705 GWh between 2002 and 2011.

2

Biogas in Brazil: A Governmental Agenda

and biomass. According to data from the Ministry of Mines and Energy, until December 31, 2012, 19 biomass plants, 41 windpower plants, and 59 SHPs (small hydropower plants) had been founded. The biomass energy projects are associated mainly with sugarcane vinasse. The use of other forms of biomass, such as landfills and animal waste, still has very limited participation in the country’s energy supply [1]. The use of organic biomass has the advantage of not only generating energy (heat and electricity) but also reducing the environmental impact associated with the disposal of this biomass. Therefore, there is a dual environmental advantage: waste reduction and renewable energy generation. Nevertheless, there are great obstacles to the implementation of the use of biomass for power generation in Brazil, mainly the fact that the residue is spread out over a territory of continental proportions. In addition, investments are generally small and decentralized, which minimizes the advantages of credit access. There is also the need to reconcile the interests of distributing companies (aimed at maximizing the number of clients) and those of potential users of the energy generated by organic biomass (who will not only leave the distributing companies but may also become potential suppliers of the future energy surplus). Among the alternative uses for biomass, residue is biogas production, which can be used directly or transformed into electricity. In either case, the use of biomass to produce energy represents distributed generation, as the source of the energy production is close to the consumer. In Brazil, there is a great potential for various types of biogas production, particularly sugarcane vinasse, urban waste (including solid waste from landfills and liquid waste from effluent treatment plants), and cattle and swine waste. Ref. [3] estimates that the potential for electricity generation from different sources of biogas in Brazil is between 1.21% and 1.30% of the installed capacity and concludes that there is a great potential for power

generation from biogas obtained from the anaerobic digestion of organic residue. In fact, given the variety of possibilities for biogas production, their heterogeneous nature, and their geographical dispersion in the territory, each of these possibilities must be analyzed taking into account their peculiarities and characteristics with regard to socioeconomic contribution to the country and to their local area. As previously mentioned, alternative energy sources generally have one element in common: they are considered distributed generation with decentralized, small-scale energy production, unlike the conventional style of energy production, which in Brazil’s case is concentrated in large hydropower plants or in enterprises with great installed capacity. Distributed generation still faces resistance from the industry’s planners and operators, one of the reasons being the complexity of managing the quality and quantity of energy needed to supply large demands with numerous small-scale energy sources scattered throughout the territory [4]. However, there are also benefits to be considered, particularly the avoided cost of enterprises in electricity generation and transmission (or gas transportation), as energy is generated in the location where it is consumed. The process of distributed generation from waste biomass involves the transformation of waste into biogas with the use of biodigesters (for a detailed description of the process that takes place in biodisgesters (see Refs. [5, 6])). This biogas can be used directly or to produce electricity. In both cases, the energy produced can be used for private consumption and the excess can be sold to distributors of natural gas or electricity, respectively. In the case of cattle and swine waste and sugarcane vinasse, given their agricultural nature, one of the by-products of this process is biofertilizer, which makes its use even more advantageous. In fact, the concept of agroenergy is being consolidated in connection with the idea of distributed generation. This consolidation of agroenergy as an officially


recognized and stimulated economic activity gives rise to a new business and a new source of income for rural properties in addition to the revenue generated by traditional agricultural products. Through the structure of prices, periods, and firm long-term contracts with publically regulated official distributors, agroenergy constitutes new perspectives in the field [7]. The development of biogas systems is inserted in this context. Thus, sustainability in the current model of rural production in Brazil becomes viable with the inclusion of agroenergy in rural properties, based on the technology of environmental sanitation through the treatment of residual biomass in biodigesters, which allows the exploitation of the potential for energy generation with the use of the concept of distributed generation [8]. Given the range of possibilities of biogas generation, the present article estimates the potential production derived from cattle and swine waste. The intention is not to exhaust the topic of biogas in Brazil, but to measure the potential production using animal waste and to propose a list of actions that can be implemented by the government to encourage the development of the country’s biogas systems. The work is based on the hypothesis that the country has enormous potential for biogas production and use, but it lacks political tools, such as a more focused legislation that would facilitate the development of biogas systems. It is worth noting that biodigester technology is available, however, the greatest barriers do not seem to be technological but political and regulatory in nature. To provide a glimpse of the potential for biogas production in Brazil, the article analyzes one of several possible sources of biogas: the biomass derived from cattle and swine waste. Swine and confined cattle production is very relevant in the country’s rural areas, and the effective use of the biogas generation potential would contribute to sustainability in these areas. In general terms, the aforementioned activity generates residue that poses a high risk of environmental pollution primarily due to its great amount of nutrients

3

and biodegradable organic materials, which can contaminate the soil and ground water and reach superficial bodies of water. To achieve the proposed objective of assessing the potential of biogas generation from cattle and swine waste and identify the political and regulatory barriers that open a government agenda on the subject, this work is divided into four more sections: the methodological aspects are presented in Section 2, followed by the state of the art of biogas in Brazil and an estimation of the potential for biogas generation from cattle and swine waste; Section 4 lays out the governmental agenda for biogas development in the country; finally, the final considerations are presented.

2. Methodological Aspects Biogas is produced when microorganisms break down organic materials in the absence of oxygen, also called anaerobic digestion. The biogas produced consists of methane (50%-80%), carbon dioxide (20%-50%), and traces of hydrogen sulphide (0-0.4%), for example [6]. The biogas can be used for different energy services, such as heat, CHP (combined heat and power), and vehicle fuel, although the latter requires upgrading, by which most of the carbon dioxide and the hydrogen sulphide are removed. Additional treatment (to achieve local specifications) will also allow injection into the natural gas grid. The various sources of organic biomass are converted using biodigesters and can have different energy uses. Fig. 1 presents a simplified flow diagram of a generic anaerobic digestion plant based on organic feedstocks. Biogas digesters have come to symbolize access to modern energy services in rural areas and are slated to considerably improve health and sanitation, and to yield significant socioeconomic and environmental benefits [5]. In order to identify the possibilities of the use of biogas in Brazil, as well as the political and regulatory challenges for the development of biogas systems in the country, we will conduct a literature review and statistical analysis that will allow the


4

estimation of the potential for biogas generation from cattle and swine waste. The literature review will describe the state of the art of biogas in Brazil and point out the indicators that will be used to estimate the biogas potential. A study of the international experience, notably the cases of China and India, given their similarities with Brazil, can provide information on how the implementation of biogas took place in these countries and give examples that can be adapted and incorporated into the Brazilian reality. The most recent data on the amount of cattle and swine waste in the country are found in the Agricultural and Livestock Census of 2006 (Instituto Brasileiro de Geografia e Estatística [10]). These data are invaluable to the study of the potential for biogas generation from the biomass derived from animal waste. Moreover, they help to corroborate the theory that there is no dissemination of management systems for the treatment of animal waste, clearly indicating the environmental problem caused by agricultural and livestock activities. There is also important data on the incipient purchase and self-production of electricity by rural properties, which confirms the hypothesis of the potential for exploitation by the national agricultural industry. Due to the need to delimit the study, more specific Inputs

data will be used as described below:  Agricultural and livestock management, which refers to the use of animal waste;  Self-generation and purchase of electricity by the rural property, which represents the amount of energy that the rural property produces on site;  Number of heads of cattle and swine, which will be the basis for the calculation of the biogas generation potential. The data on cattle only refer to confined cattle for the simple reason that intensive systems have access to the animals’ waste. The special supplement of the Agricultural and Livestock Census for Family Farming provides the following piece of data [11]:  Participation of family farming in cattle and swine raising, which represents the number of small scale properties in the production of swine and confined cattle. These data provide the dimensions of small and large scale properties and are useful in the development of policies suited to each type of property. With the total number of swine and confined cattle, it is possible to estimate the production of biogas based on waste/animal/day indicators. Therefore, the following additional data are required:  indicators for estimation of solid cattle and swine waste generation; Outputs

Biogas plant Gas treatment

Feedstock-1

Electricity Pretreatment

Energy market

CHP plant

Digestion

Thermal energy

Feedstock-2

Solids separation

Fiber fraction

Compost

(Optional) Liquid treatment

Liquid fertiliser

Fig. 1 Simplified flow diagram of a generic anaerobic digestion plant based on organic feedstocks. Source: Karellas, Boukis & Kontopoulos (2010: 1274) [9].

Application to land


 indicators for conversion of cattle and swine waste into biogas (biomethane). Finally, we will analyze the legal framework related to the production and use of biogas in the country. For that, we will present and discuss the laws, decrees, and regulations related to biogas.

3. In Search of an Alternative Source of Fuel: The Potential of Biogas Derived from Organic Biomass This section aims to show evidence of the potential for biogas generation from cattle and swine waste, starting with a brief state-of-the-art review on biogas in Brazil. 3.1 The State of the Art of Biogas in Brazil Just over 30 years ago, Brazil fostered a program to introduce biodigesters in rural areas; however, it did not bring significant investments for the use of residual biomass. In light of this unfulfilled expectation, official efforts to stimulate waste treatment with biogas generation and energy use gradually decreased. However, in some regions, notably the southern region, farmers continued the process on their own [7]. Despite the negative experience with electricity production from residual biomass, recent incentive programs for renewable sources, such as PROINFA, have been successful, as in the case of wind power [2]. This program was established by Law 10438 of April 26, 2002 and updated by Law 10762 of November 11, 2003. It aims to diversify the national energy grid and find regional solutions with the use of renewable energy sources, through the exploitation of available raw materials and applicable technologies, starting with an increase in the participation of the electricity derived from these sources in the National Interconnected Electrical System (Sistema Elétrico Interligado Nacional—SIN). One of the sources included in the PROINFA program is biomass. However, in the case of biomass,

5

there are only projects related to sugarcane vinasse. In Brazil, there is practically no treatment of cattle and swine waste and it receives no benefits from PROINFA. This calls for an investigation into the reasons behind the non-use of the potential for biogas generation from animal waste. The literature indicates that the biogas production initiatives in Brazil are incipient and isolated. In reality, renewable energies in general are still classified as “alternative”, which renders them inferior to hydropower, and still considered the noblest renewable source [7]. Sector statistics ignore the energy potential of organic residues, if not for the purposes recorded in the distribution of spaces of the so-called alternative energies, then at least for the correct identification of the economic potential that these residues and effluents represent to their generators. One of the most concrete examples of biogas production incentives in Brazil is that of the Itaipu Renewable Energy Platform (http://www.plataformaitaipu.org/energia/biomassa), which has several demonstration units in the state of State of Paraná. Among its success stories is Condomínio Ajuricaba, a cooperative of small-scale cattle and swine producers that generate approximately 1,000 m3/day of biogas. As previously mentioned, the use of biogas as an energy source represent distributed generation. In addition to the advantages related to avoided cost with investments in generation and distribution, distributed generation allows energy consumption (thermal or electrical) along with production. In the case of biogas, there is the possibility of use in energy-intensive industries, such as the grain, brick, cement, tile, stone and other mineral product industries, as well as packing plants, mills, and other agricultural industries. Therefore, these segments can find in biogas actual possibilities of obtaining energy that is custom-made for their high consumption [12]. Even though initiatives for biogas generation from animal waste are isolated, there is a significant


6

Therefore, there is room for the development of biogas systems in Brazil and particularly the Southern Region, where intensive production is very significant and where most of the heads of swine and cattle are concentrated. In addition to the previous argument, there is the fact that only a few properties have adequate treatment for manure, as shown in Table 3. Not only is there a small number of properties with waste treatment, but most of them use treatment in open tanks. Treatment in biodigesters was insignificant in 2006. A simple data analysis shows that there is room to adopt policies that allow the treatment of animal waste with simultaneous generation of biogas and biofertilizer. Ref. [13] clearly summarizes the importance of animal waste treatment: “The employment of anaerobic digestion technology for waste treatment is possible and desirable given that it contributes to environmental conservation, makes modern production systems viable, and optimizes the enterprise’s cost/benefit ratio, etc.. In the same way, rational use of raw material and correct waste management optimize productive systems to achieve a harmonious coexistence between man and the environment”. In fact, generation of biogas from the anaerobic digestion of biomass is a technology that can produce sustainable energy and also reduce the environmental

potential for it in Brazil’s rural areas, particularly in cattle and swine farms. The Southern Region has characteristics that are very favorable to the development of biogas systems, given that it holds a large part of the cattle and swine production. Data from the IBGE Agricultural and Livestock Census show that biogas production is practically non-existent in rural areas and those rural properties depend on the electricity purchase from distributors. It is also evident that there is no adequate treatment of animal waste, which leads to a significant environmental problem. Table 1 shows the electricity purchased by rural properties and obtained by assignment, and Table 2 shows the total electricity produced in rural properties. In 2006, the Agricultural and Livestock Census counted 5,175,489 agricultural and livestock establishments, of which 68.1% or 3,526,330 units obtained electricity from at least one source. Electricity purchase from distributors is presented in 3,258,676 properties (about 92.4% of the total with energy); electricity obtained by assignment is presented in 7.7%; and electricity generated in the property is presented in 2.1% or 75,457 properties. When the data from Tables 1 and 2 are compared, it is possible to see that a very small number of establishments generate their own electricity, most of which used solar energy or burned fossil fuels in 2006. Table 1

Electricity used by establishments, Brazil and Southern Region, 2006.

Brazil and Southern Region Brazil Southern Region State of Paraná State of Santa Catarina State of Rio Grande do Sul

Purchased 3,258,676 812,468 270,084 175,379 367,005

Obtained by assignment 270,293 33,144 19,667 3,217 10,260

Source: IBGE, Agricultural and Livestock Census of 2006. Table 2 Total electricity used and generated within the establishment, according to source, Brazil and Southern Region, 2006. Brazil and Southern Region Brazil Southern Region State of Paraná State of Santa Catarina State of Rio Grande do Sul

Solar energy 32,217 308 106 56 146

Wind power 273 28 7 5 16

Source: IBGE, Agricultural and Livestock Census of 2006.

Hydropower 7,072 796 287 143 366

Burning fuels 30,669 464 126 50 288

Other source 6,321 538 232 101 205

Biogas in Brazil: A Governmental Agenda Table 3

7

Treatment of manure per establishment, Brazil and Southern Region, 2006.

Brazil and Southern Region Brazil Southern Region State of Paraná State of Santa Catarina State of Rio Grande do Sul

Total properties 5,175,489 1,006,181 371,051 193,663 441,467

Treatment in anaerobic lagoon 3,269 1,618 490 529 599

Treatment in open tanks 131,232 82,609 13,036 28,016 41,557

Treatment in biodigester 2,387 1,223 393 490 340

Treatment with composting 31,849 21,379 6,271 7,823 7,285

Treatment elsewhere 27,197 7,877 3,043 1,478 3,356

Source: Mathias (2014) [14].

risks associated with manure and waste management [15]. The first conclusion drawn from the analysis of the previous tables (Tables 1-3) is that, if the deficiencies of Brazil’s rural areas were addressed with biogas systems, there could be immediate benefits from an economic perspective (at the very least energy generation for private consumption and biofertilizers) and from an environmental perspective (animal waste treatment). The potential for biogas generation from animal waste is shown in the next section. 3.2 Potential for Biogas Generation from Animal Waste In this section, we estimate the potential for the generation of biogas derived from cattle and swine waste. As highlighted in Section 2, the methodology used to obtain this estimate is based on data from the most recent Agricultural and Livestock Census (2006) published by IBGE, which shows the structural data of Brazilian agriculture and livestock production. The information needed to obtain the estimates for animal waste and, consequently, biogas production refers to the total heads of swine and cattle. In the case of swine, the information of interest is the total number of heads and, in the case of cattle, the number of confined animals, as the objective is to obtain biogas from dry animal waste, which is not possible in extensive cattle farming. The data from the Agricultural and Livestock Census [10] included in Table 4 shows the number of swine in the country in 2006, which exceeded 31.1 million heads, more than half of them (16.7 million) concentrated in the Southern Region. Although the number of heads of cattle is far greater (nearly 200

million), only confined animals can be considered for the potential of waste generation, which exceeded 4 million heads in 2006 including a little over 600 thousand heads in the Southern Region. With the number shown in Table 4 and the estimates of daily production of dry material from swine and cattle waste, it is possible to calculate the potential for waste production in tons/day. Considering that swine produce 2.3 kg to 2.5 kg of dry waste per day and that cattle produce 10 kg to 15 kg per day [3], it is possible to obtain the values expressed in Table 5. The extremes are both scenarios considered in this study, with Scenario 1 being the lowest and Scenario 2 being the highest. Based on daily waste production, we can calculate the potential for biogas generation in Brazil and particularly in the Southern Region. The indicator for conversion of animal waste into biogas, more precisely methane gas3 is provided by Ref. [16]: for beef cattle, 40 m3 of methane gas per ton of dry material and, for swine, 350 m3 of methane gas per ton of dry material4. Thus, we have the estimate for the potential for methane gas production (Table 6). 3

The typical composition of biogas is predominantly CH4 (methane gas), which represents between 55% and 75% of biogas. Another important gas that is generated is CO2, with a participation of 25% to 45% in biogas [9]. 4 The data from Ref. [16] are close to those seen in international experience. Ref. [9] provides an indicator of 362.5 m3 of CH4 per ton of dry material for swine. When measured in m3/animal/day, Ref. [17] provides an indicator of 1.43 m3/animal/day for swine and 0.32 m3/animal/day for cattle. However, this refers to the indicator for biogas production and not specifically methane gas. In that case, the data from Ref. [17] are similar to the data from Ref. [18], which show indicators for biogas production from dry material from cattle and swine (1.40 m3/animal/day) in Brazil.


8 Table 4

Number of heads of swine and confined cattle, Brazil and Southern Region, 2006.

Region and states

Swine Number of establishments Number of heads

Confined cattle Number of establishments Confined animals

Brazil Southern Region State of Paraná State of Santa Catarina State of Rio Grande do Sul

1,496,107 451,870 135,477 82,324 234,069

20,864 5,750 2,633 1,299 1,818

31,189,339 16,750,420 4,569,275 6,569,714 5,611,431

4,049,210 603,153 366,577 77,104 159,472

Source: IBGE (2007). Table 5

Potential production of swine and cattle waste, Brazil and Southern Region, 2006 (tons/day).

Region and states Brazil Southern Region State of Paraná State of Santa Catarina State of Rio Grande do Sul

Scenario 1 71,735 38,526 10,509 15,110 12,906

Swine Scenario 2 77,973 41,876 11,423 16,424 14,029

Scenario 1 40,492 6,032 3,666 771 1,595

Confined cattle Scenario 2 60,738 9,047 5,499 1,157 2,392

Source: Prepared by author based on data from IBGE (2007). Table 6

Potential for methane gas production. Brazil and Southern Region: 2006 (m3/day).

Region and States Brazil Southern Region State of Paraná State of Santa Catarina State of Rio Grande do Sul

Scenario 1

Swine Scenario 2

Scenario 1

Confined cattle Scenario 2

25,107,418 13,484,088 3,678,266 5,288,620 4,517,202

27,290,672 14,656,618 3,998,116 5,748,500 4,910,002

1,619,684 241,261 146,631 30,842 63,789

2,429,526 361,892 219,946 46,262 95,683

Source: Prepared by author based on data from IBGE (2007).

The data in Table 6 are very representative, given that in 2006 the country imported 26.8 million m3/day of natural gas (95% from Bolivia and 5% from Argentina). In other words, if all of the swine and cattle waste in Brazil was treated in biodigesters, the potential for gas generation would meet the country’s importation needs. In Brazil, particularly the Southern Region, there is a strong presence of small family farms, as seen in the data analysis of the Agricultural and Livestock Census of 2006. Based on the census, IBGE conducted a study on Family Agriculture in the country. The Institute used the concept of Family Agriculture defined by Law 11 326 of July 24, 2006. According to the law, rural family units must meet the following criteria simultaneously: the area of the rural

establishment must not exceed four modules for tax purposes; the labor employed in the economic activities must be predominantly from the family; the family income must come predominantly from these activities; and the establishment must be managed by the family [11]. Table 7 shows the participation of Family Agriculture in cattle and swine raising in the Southern Region in 2006. From Table 7, it can be concluded that:  The number of rural establishments is composed mainly of family units (between 81% and 90%);  In the case of cattle, with the exception of the state of Santa Catarina, most of the herd belongs to non-family units;

Biogas in Brazil: A Governmental Agenda Table 7

9

Participation of family agriculture in cattle and swine raising, Southern Region, 2006.

State State of Paraná State of Santa Catarina State of Rio Grande do Sul

Cattle Swine Cattle Swine Cattle

Family agriculture Establ. Heads (number) 171,618 3,161,405 115,252 2,840,213 129,254 2,038,705 73,715 4,370,999 283,768 4,063,020

Non-family agriculture Establ. Heads (number) 39,748 5,892,396 20,225 1,729,062 18,084 1,087,297 8,609 2,198,715 46,133 7,121,228

% family agriculture Establ. Heads (number) 81% 35% 85% 62% 88% 65% 90% 67% 86% 36%

Swine

209,282

24,787

89%

Activity

3,942,427

1,669,004

70%

Source: Adapted from IBGE (2009).

 In the case of swine in all three southern states, most of the herd (between 62% and 70%) belongs to family units. Therefore, it can be concluded that the development of biogas systems, particularly small scale systems, can be a favorable strategy for local sustainable development. However, there are various challenges to be overcome before biogas can be produced on a large scale and not only in isolated local properties.

4. Challenges to Biogas Development This section aims to present the pathways to the development of biogas in Brazil. First, some lessons will be learned from international experience that can contribute to the Brazilian experience. Next, some legal, political, and regulatory aspects are pointed out to show the need for a list of topics to be discussed and included in the governmental agenda. 4.1 Lessons from International Experiences: The Cases of China and India Biogas production from animal waste is particularly useful in countries with swine and cattle herds and where the possible sites for residue use are geographically dispersed [19]. The reason for that is that locally produced biogas can be used in the farms themselves, whether for electricity generation for local supply (avoiding investments in the expansion of energy distribution networks to remote areas), for generation of thermal energy (useful in countries with harsh winters) or for drying grain (in farms with

simultaneous cattle raising and production of foods that require thermal processes). If such farms are already connected to distribution networks of electricity or natural gas, the excess energy (electricity or methane, as long as specified) could be injected into the networks to increase the country’s energy supply and reduce its dependence of possible energy importation and delaying the need for investment in energy generation and network expansion. Given the similarities in the size of their territories and the large number of heads of cattle and swine, two of the most important developing countries that successfully use biogas systems can share their experiences and provide examples for Brazil to follow. These countries are China and India and, according to Ref. [20], they dominate the best technologies for the use of biodigesters. The main objective of the China is to obtain biofertilizers for food production, whereas, India seeks to reduce the energy deficit. The biodigester models are distinct: the Chinese is simpler and cheaper and the Indian is more sophisticated and technical to make better use of biogas production. The development of the technology of biogas in China and India is based on animal management, especially, swine and cattle raising [17]. Ref. [17] presents a history of biogas and assesses its future in developing countries, particularly China and India. According to the authors, starting in the 1970s, China promoted the use of biogas in all rural residences in the country. To cite an example, in 2007, there were 26.5 million biogas plants in China, mostly

10


family systems producing 6-10 m3 of gas a day. Ref. [15] also presents an overview of China’s biogas industry. However, they also emphasize how the management systems offer no adequate sanitation to prevent pollution. As a way of stimulating energy production from renewable sources, the Chinese government passed the “Renewable Energy Law” and provided incentives for biogas production in 2006. This shows that a country with ample reserves of hydrocarbons, particularly coal and more recently non-conventional natural gas, also has an interest in the use of biogas and other alternative energies. As in China’s case, India, with its vast territory and widely dispersed rural properties, granted government subsidies for the construction of 4 million family biogas plants between 1999 and 2007. Since the early 1980s, the country has run a project known as the NPBD (National Project on Biogas Development), which provides funding and training to the various development programs proposed by the government. These government subsidies for the development of family biodigesters covered 30%-100% of the total price of equipment between 1980 and 1990 [17]. Therefore, given their vast territories and the characteristics of their rural production, China and India are ideal countries for distributed generation of energy, particularly biogas production. Undoubtedly, international experience suggests that the development of biogas systems requires a set of focused political measures with strong government participation, particularly with regard to the legal framework and the financial incentives provided 5 . 5

Besides China and India, there are strong government incentives for biogas development in Europe, with Germany being the most emblematic example. The German case is particularly interesting for the analysis of an efficient policy for the development of alternative energy sources, especially the implementation of biogas production. In 2000, the country passed the Renewable Energy Source Act, which represents the foremost legal framework for the production of renewable energies, including biogas. Following the publication of this act, there was significant expansion in the production and consumption of renewable sources of energy in the country. According to Ref. [20], biogas production, in particular, showed extraordinary growth, with an eightfold increase

Another topic highlighted in international experience is the incentive for the development of small biogas plants in rural areas [19]. However, there are many political and legal obstacles to biogas development in Brazil that warrant a governmental agenda on the issue. 4.2 Legal and Political Difficulties: A Governmental Agenda Ref. [7] believed that the legal conditions for bioenergy development had been given and all that needed was to stimulate its use. According to these authors, the legal conditions were given in two parts: legislation and policy programs. They are the following:  Decree 5163 of July 30, 2004. This decree regulates electricity trading, the process of granting concessions and authorizations for electricity generation, and other provisions.  ANEEL (National Electricity Agency) Normative Resolution 167 of October 10, 2005. This regulation defines the conditions for the trade of energy derived from distributed generation.  Amendment to Law 9648/98 (Law 10438/02 included changes to article 11, § 4, regarding the system for subrogation of the cost of FFC (Fossil Fuel Consumption), which underwent several revisions and is currently regulated by ANEEL Resolution 146/05). This amendment allowed the transfer of FFC benefits, not only to SHPs as already prescribed by law, but also to wind, solar, and biomass sources implemented in isolated electrical systems to substitute thermal generation using petroleum by-products according to current and future demand, which simultaneously meets the process of universal access and introduction of alternative sources into the energy grid. between 2000 and 2008 boosted by the strong development of biogas units in agricultural properties scattered throughout the country. In Germany, the primary purpose of biogas was electricity generation to replace the power originally derived from burning fossil fuels. In 2008, 15.1% of the total electricity consumption in Germany came from renewable sources, with 1.3% of the electricity coming from biogas.


 PROINFA. This is a policy program for renewable energy generation. It was created by Law 10438/02, which was a key step in shaping the legal framework for electrification efforts in Brazil6. However, in September 2011, Brazil passed the Biofuels Law (Law 12.490/11), which can also be considered a new legal framework for biogas. This law made new contributions, but still left gaps that needed to be addressed to allow actual development of the activities related to the biogas system. According to Ref. [19], when it comes to biogas generation and distribution, there are still political and regulatory hurdles to be overcome. This reference points out that among other rulings, the Biofuels Law changes Law 9478/97 includes in “Principles and Objectives of the National Energy Policy”: (1) the guarantee that biofuels will be supplied throughout the national territory; (2) incentives for electricity generation from biomass and by-products of biofuel production, given that it is clean, renewable, and complementary to hydropower; (3) the promotion of the country’s competitiveness in the international biofuel markets; (4) the attraction of investments in infrastructure for transport and storage of biofuels; (5) the promotion of renewable energy research and development; and (6) the mitigation of greenhouse gas emissions and pollutants in the energy and transport sectors, including the use of biofuels. The same law ruled that the National Council for Energy Policy must “define the strategy and the policy of economic and technological development of the oil, natural gas, fluid hydrocarbon, and biofuel industry, as well as its supply chain”. It also defines the “Biofuel Industry” as a “set of economic activities related to the production, importation, exportation, transfer, 6

Law 10438 obliges concessionaires and licensees to provide “universal electricity service coverage,” without financial contribution by the new consumers toward initial investments (which are to be fully recovered through tariffs). However, the law was not a pure “rural electrification norm”, as it covered a series of (competing) policy goals, namely rural access, power generation from alternative national resources (renewable energies, natural gas, and coal), social tariffs, and “emergency generation” [21].

11

transport, storage, trade, distribution, compliance assessment, and quality certification of biofuels” and “Biofuel Production” as a “set of industrial operations for transforming plant or animal renewable biomass into fuel”. From the Law, it can be inferred that biogas derived from animal waste is classified as biofuel. According to natural gas Law 11909/09, however, the biogas extracted from animal waste cannot be classified as natural gas (despite their identical chemical composition) because the latter only applies to gas derived from petroleum or gas reservoirs. Nevertheless, biogas can be classified as biofuel if regulated as such by the National Agency of Petroleum, Natural Gas and Biofuels (ANP (Agência Nacional de Petróleo, Gás Natural e Biocombustíveis)). Law 12490/11 ruled that it is the ANP’s responsibility to “regulate and authorize activities related to biofuel production, importation, exportation, storage, stockpiling, transportation, transfer, distribution, resale, and trade, as well as compliance assessment and quality certification, inspecting them directly or in association with Union, State, Federal District or Municipal agencies”. However, the Federal Constitution rules that it falls to the States to exploit the local piped gas services, regardless of source or composition. Thus, if the biogas derived from animal waste cannot be considered “natural gas”, its local movement (within a Federal State) is state-regulated and not federally regulated. Nonetheless, there is legal provision that biofuel trade must be authorized by the ANP. In other words, the ANP is responsible for authorizing trade, but the state regulators are responsible for controlling the movement and sale to final buyers. Furthermore, when biogas is used to generate electricity, the activity is also regulated by ANEEL. Thus, a complex set of legal rules apply to biogas, from its production to its use. Table 8 summarizes the legal framework that applies to the production, trade, and use of biogas and biogas-generated electricity. This legal and normative framework assigns a


12

number of duties and responsibilities to different public agencies of the energy sector, which are summarized in Table 9. The analysis of the legal framework and the duties assigned to the different public agencies leads to the conclusion that this framework was developed in a hermetic fashion and did not consider the specificities of the biofuel industry. The different legal documents overlap duties, while also leaving gaps that need to be filled. One of the main juxtapositions is the role of regulating the direct use and trade of biogas. It is unclear whether it is a responsibility of the federal regulatory agency (ANP) or the state regulators. There Table 8

is legal basis for both interpretations. One of the main gaps is the definition of biogas itself, which is not found in any of the normative frameworks provided. The first topic on the governmental agenda for biogas is the clear definition of the duties of the State agencies regarding the production, movement, and use of biogas derived from animal waste, so that its development will not run into legal or bureaucratic matters that hinder the construction of an enterprise that could bring environmental and energy benefits to its area. Even without changes to the legal framework, it is fundamental to coordinate the public agencies in order to allow the development of biogas enterprises.

Legal framework for biogas development in Brazil.

Law/resolution/program

Description Brazilian government policy aimed at promoting the expansion of distributed power generation Law 10438 (PROINFA) through renewable sources, and diversifying primary sources of electricity, thus improving the long-term supplying conditions of the national system. Amendment to the previous law (Law 10438) to guarantee funds for anticipating deadlines Law 10762 defined by regulatory agencies and to restrict the exemption of financial contribution to new consumers with load up to 50 kW. It regulates electricity trade and the process of granting concessions and licenses to generate Decree 5163 electricity, among other provisions. ANEEL normative resolution 167 It defines the conditions for trading energy derived from distributed generation. It authorizes the Program for Distributed Generation with Environmental Sanitation proposed by Authorizing resolution 1482 the Paraná State Energy Company (Companhia Paranaense de Energia—COPEL) as a pilot project for implementing low-voltage distributed generation. It sets the requirements for authorization to exploit and change the installed capacity of thermal ANEEL normative resolution 390 power plants and other alternative energy sources, describes procedures for registering generation plants with reduced installed capacity, and makes other provisions. Law 11909/09 It provides the definition of “natural gas”, which excludes biogas and prevents it from being (Natural Gas Law) injected into the natural gas grid. Law 12490/11 It deals with biofuels in general, but does not mention biogas explicitly. It rules that the ANP is (Biofuel Law) responsible for regulating the entire biofuel industry. Source: prepared by the author based on Refs. [7, 19, 22]. Table 9

Public institutions of the energy sector and their duties.

Agency National Council for Energy Policy (Conselho Nacional de Política Energética—CNPE) Ministry of Mines and Energy (Ministério de Minas e Energia—MME) National Electricity Agency (Agência Nacional de Energia Elétrica—ANEEL) National Agency of Petroleum, Natural Gas and Biofuels (Agência Nacional do Petróleo, Gás Natural e Biocombustíveis—ANP) Statute Regulation Agencies Energy Research Company (Empresa de Pesquisa Energética—EPE) Source: prepared by author.

Legal responsibility To propose an energy policy that takes into account the rational use of the country’s energy resources, among other aspects; to define the strategy and policies for the economic and technological development of the biofuel industry. To implement the energy policies proposed by the CNPE. To regulate electricity generation from biogas; to define the rules for the injection of biogas surplus into the grid (sale to distributors). To regulate and authorize activities related to biofuel production, importation, exportation, storage, stockpiling, transport, transfer, distribution, resale, and trade. To regulate the “local piped gas services” provided by the LDC (Local Distribution Companies). To provide services in the area of studies and research destined to subsidize planning in the energy sector.


To achieve that, each public agency of the energy sector must perform its role as prescribed in the legal framework. Thus, the CNPE should establish guidelines for specific programs, such as those for biofuel use, and propose policies for the use of local resources, which can stimulate local biogas production and use. However, this agency has not had a proactive role in proposing policies. Another important element is the interaction between the different Ministries of State involved in biogas production and use. In order to achieve that, the Ministry of Agricultural Development (focused on small rural properties), the Ministry for the Environment (focused on waste treatment and environmental protection), and the Ministry of Mines and Energy should make a joint effort to allow the CNPE to propose policies that facilitate the inclusion of biogas as an energy source, both for thermal energy and electricity. With regard to planning, the EPE should consider in its agenda the real potential of biogas, taking into account the environmental advantages derived from the proper management of animal waste and from distributed generation (avoided cost of expanding generation and transmission). Of course, the great complexity of Brazil’s electrical system hinders the task of measuring and allocating these benefits, nevertheless, it must be completed. As for regulation, despite juxtapositions in different legal frameworks, it is possible for federal and state regulators to reach an agreement to allow the creation of enterprises that can take advantage of local energy resources. Even without changes to legislation, programs, such as cooperative agreements between regulators, can be used in specific cases to secure private investments in biogas. After the technological and bureaucratic issues are overcome, there is still the need to obtain financing for biogas enterprises. There are government institutions that can be used in this financing, i.e., Bank of Brazil, which has low interest rate loans for small rural

13

enterprises, and the National Bank for Economic and Social Development (Banco Nacional de Desenvolvimento Econômico e Social—BNDES), which can finance investments in medium-size and large rural properties. It must be pointed out, however, that this is only one of the requirements for achieving the investments. The fundamental issue is to find a solution to the legal barriers, primarily through the coordination of the above mentioned agents. Finally, we can add to these agents the Brazilian Biogas Network (Rede Brasil Biogás), a network of technical, scientific, industrial, commercial, and agricultural exchange that revolves around the topic of biogas and the possible energies generated by its use for the purpose of stimulating the flow of knowledge on the topic and having the widest possible outreach through its members. Different spheres of government can interact with the Brazilian Biogas Network to better understand the peculiarities of the industry, thus leading to more focused and specific policies.

5. Conclusions This article sought to demonstrate that, along with Brazil’s enormous biogas generation potential, particularly through the transformation of animal waste, there is an important governmental agenda centered on legal, political, and regulatory challenges. The main elements that compose the government agenda have been presented to assist in the design of policies and actions that will increase distributed generation of biogas and/or electricity derived from biogas, obtained from residual biomass by sanitary and environmentally acceptable means, for private use and sale of possible excess of gas and/or electricity. As previously emphasized in this study, this biogas (or the electricity derived from it) is destined primarily for private use in rural properties with immediate positive results in terms of the environment and the avoided cost of investments in the natural gas and energy sectors. In the country’s southern region, for example, which has the largest production of swine and


14

confined cattle, the consumers of the excess biogas can be businesses located near the generating units. In 2011, the industrial demand for natural gas increased by 8% compared to 2010, particularly in the ceramics (12.9%), pig iron and steel (11.2%), and chemical (6.4%) sectors [1]. Brazil’s white ceramics industry is also concentrated in the region and is a potential consumer of biogas. The article has presented a long list of challenges, the greatest being the development of an energy policy that will allow the consistent introduction of this “alternative” source of energy into the national energy grid, as established by Law 9478/97. This same law states that one of the principles and objectives of the national energy policy is to “increase the participation of biofuels in the national energy grid on an economic, social, and environmental basis”. In fact, the challenges presented in the governmental agenda can be analyzed in two phases. In the first phase, the governmental agenda must create incentives for biogas production aimed at environmental care through the treatment of residual biomass and at energy generation (biogas and electricity) for private consumption. In a potential second phase, with more developed biogas systems and resolved technical and financial issues, it would be possible to encourage the sale of biogas excess to natural gas and/or electricity distribution networks. This phase is apparently more complex and requires more time for implementation. In the academic sphere, a useful tool that should be addressed in relation to biogas is the use of a GIS (Geographical Information System), given that the majority of enterprises represent distributed generation (of gas or electricity). The regional aspect is, therefore, of fundamental importance, and local governments must be mindful of the potential for production and use of this renewable energy source.

References [1]

Brasil. 2013. Brazilian Energy Balance 2012, Empresa de Pesquisa Energética. Accessed November 12, 2013. https://ben.epe.gov.br/downloads/Relatorio_Final_BEN_

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10] [11]

[12] [13]

[14]

2013.pdf. Simas, M., and Pacca, S. 2011. “Windpower Contribution to Sustainable Development in Brazil.” In Proceedings of World Renewable Energy Congress, 2626-33. Salomon, K. R., and Lora, E. S. 2005. “Energetic Potential Estimate for Electric Energy Generation of Different Sources of Biogas in Brazil.” Viçosa, Revista Biomassa & Energia 2 (1): 57-67. Hachisuca, A. M. M., Fernandes, D. M., Silva, F. P., Antunes, M. A., Leonardo, P. C. A. S., Costanzi, R. N., and Akaboci, T. R. V. 2010. Geração distribuída: biomassa residual utilizada como fonte de energia alternativa em unidades de demonstração. Belém, III Simpósio Brasileiro de Sistemas Elétricos. Srinivasan, S. 2008. “Positive Externalities of Domestic Biogas Initiatives: Implications for Financing.” Renewable and Sustainable Energy Reviews 12 (5): 1476-84. Lantz, M., Svenssonb, M., Bjornsson, L., and Borjesson, P. 2007. “The Prospects for an Expansion of Biogas Systems in Sweden—Incentives, Barriers and Potentials.” Energy Policy 35 (3): 1830-43. Bley, J. R. C., Libanio, J. C., Galinkin, M., and Oliveira, M. M. 2009. Agroenergia da biomassa residual: perspectivas energéticas, socioeconômicas e ambientais. FAO. Accessed October 31, 2013. https://www.fao.org.br/download/agroenergia_biomassa_ residual251109.pdf. Fernandes, D. M. 2012. “Biomass and Biogas Pig Farming.” M.Sc. dissertation, Universidade Estadual do Oeste do Paraná. Karellas, S., Boukis, I., and Kontopoulos, G. 2010. “Development of an Investment Decision Tool for Biogas Production from Agricultural Waste.” Renewable and Sustainable Energy Reviews 14 (4): 1273-82. IBGE. 2007. Censo Agropecuário 2006. Brasil, grandes regiões e unidades da federação. Rio de Janeiro: IBGE. IBGE. 2009. Censo Agropecuário 2006. Agricultura Familiar: grandes regiões e unidades da federação. Rio de Janeiro: IBGE. Bley, J. R. C. 2010. Reflexões sobre a economia do biogás. Foz do Iguaçu, Itaipu Binacional. Salomon, K. R. 2007. “Technical-Economic and Environmental Assessment of the Use of the Biogás from Biodigestion Vinasse in Electricity Generating Technologies.” D.Sc. thesis, Universidade Federal de Itajubá, Itajubá. Mathias, J. F. C. M. 2014. “Manure as a Resource: Livestock Waste Management from Anaerobic Digestion, Opportunities and Challenges for Brazil.” In International Food and Agribusiness Management Review, 87-110.

Biogas in Brazil: A Governmental Agenda [15] Jiang, X., Sommer, S. G., and Christensen, K. V. 2011. “A Review of the Biogas Industry in China.” Energy Policy 39 (10): 6073-81. [16] Castanon, N. J. B. 2002. Biogas Produced from Farm Waste. São Paulo: Universidade de São Paulo. [17] Bond, T., and Templeton, M. R. 2011. “History and Future of Domestic Biogas Plants in the Developing World.” Energy for Sustainable Development 15 (4): 347-54. [18] Cervi, R. G., Esperancini, M. S. T., and Bueno, O. C. 2010. “Economic Viability for Electrical Power Generation Using Biogas Produced in Swine Grange.” Engenharia Agrícola 30 (5): 831-44. [19] Mathias, J. F. C. M., and Mathias, M. C. P. P. 2012. “Biogas in Brazil: Opportunities and Challenges.”

15

Presented at the Rio Oil & Gas Expo and Conference, Rio de Janeiro. [20] Coldebella, A. 2006. “Viability of Using Biogas from Cattle and Swine Culture to Generate Electricity and Irrigation.” M.Sc. dissertation, Universidade Estadual do Oeste do Paraná. [21] Ferreira, M., Marques, I. P., and Malico, I. 2012. “Biogas in Portugal: Status and Public Policies in a European Context.” Energy Policy 43 (Apr.): 267-74. [22] Pereira, O. L. S. 2009. “Renewable Energy as a Tool to Assure Continuity of a Low Emission Brazilian Electric Power Sector—Results of an Aggressive Renewable Energy Policy.” In Proceedings of IEEE 2009 Power & Energy Society General Meeting, 1-7.

D


DAVID

PUBLISHING

Analysis of Cooling Characteristics in Datacenter Using Outdoor Air Cooling Yoshiyuki Inoue1, Hirofumi Hayama2, Taro Mori2, Koki Kikuta2 and Noriyuki Toyohara3 1. Graduate School of Engineering, Hokkaido University, Sapporo 060-8628, Japan 2. Faculty of Engineering, Hokkaido University, Sapporo 060-8628, Japan 3. Department of Facilities Design, Taisei Corporation, Tokyo 163-0606, Japan Received: October 20, 2014 / Accepted: November 18, 2014 / Published: January 31, 2015. Abstract: Advancement of the information society has proceeded with the development of information and communication technology, and a demand on a data center has increased. In such a situation, the number of servers is increasing in a data center. Thus, the heat density in a data center is much higher than that of usual offices. And typically, almost 40% of the total power consumption is used for cooling servers in a data center. Thus, cooling effectiveness is one of the most important factors in evaluating the value of the data center. The data center taken up in this paper is located in Ishikari, where is a cold district in Japan. Using the cool outdoor air for cooling servers helps us to cut the power consumption for cooling. This paper first assesses the efficiency of Ishikari data center measuring the temperature of seven parts in a building where the cooling air flowing. Second, this paper describes the most efficient method for the operation and estimates 1.11 of PUE (power usage effectiveness). Key words: Data center, cooling characteristics, outdoor air cooling, PUE. 

Nomenclature N Subscripts θ Hfan Hm Hw V Vfm Vfr Vm VOA Vum Vur Γ θ0 θ0m θ1 θ1d Θ1m

Scale (-) Temperature (°C) Exhaust fan load (W) Heat load (W) Heat loss on wall (W) Air conditioning supply-air volume (m3/s) Effective air conditioning supply-air volume in an equipment(m3/s) Effective air conditioning supply-air volume in a room (m3/s) Equipment ventilation volume (m3/s) Outdoor air introducing volume (m3/s) Ineffective air conditioning supply-air volume in an equipment (m3/s) Ineffective air conditioning supply-air volume in a room (m3/s) Recirculation (-) Air conditioning supply-air temperature (°C) Equipment inlet air temperature (°C) Air conditioning return-air temperature (°C) Exhausted air temperature (°C) Equipment outlet air temperature (°C)

Corresponding author: Yoshiyuki Inoue, graduate student, research field: engineering. E-mail: [email protected].

Θmix ΘOA

Mixing air temperature (°C) Outdoor air temperature (°C)

1. Introduction With the rapid development in information technology in recent years, the energy consumption of datacenters follows the increasing proportion of total energy consumption in society as a whole [1-3]. In normal datacenters, air conditioning systems have to operate year round to treat a large sensible heat load. There are many efficient datacenters located in cold areas that use outdoor air cooling systems [4, 5]. These systems play an important role in reducing the processing loads on servers. Many researchers have studied indoor air conditioning systems. Furihata et al. [6] investigated the temperature of 100 points in sever rooms and found the characteristics which show a performance of datacenters. Hayama [7] proposed a method for estimating the energy consumption of air conditioning systems using the relation between the cooling characteristics and energy consumption.

Analysis of Cooling Characteristics in Datacenter Using Outdoor Air Cooling

However, there is no specific control method for saver rooms. It is important to determine the efficiency and characteristics of an outdoor air cooling system in detail. Therefore, we studied a datacenter located in Ishikari, a cold area in Hokkaido, Japan, which cools the electrical devices by outdoor air. First, we assessed the efficiency of this system by measuring the temperature of each area in the building. We then analyzed the cooling characteristics of this system by using a model for the system. Finally, we estimated the energy consumption and determined a proper management method for more efficient operation.

17

2.3 Temperature Distribution Fig. 5 shows a boxplot of the temperature distribution obtained from the actual measurement. As a reference, there are recommended or allowable conditions in server rooms, as shown in Table 3, which were determined by ASHRAE (American Society of Heating Refrigerating and Airconditioning Engineer) (2011). We can see that the blowout temperatures are relatively stable. They are within the limit of the ASHRAE (2011) [8] recommended temperature range

2. Ishikari Datacenter 2.1 Summary of Facilities The datacenter we assessed is located in Ishikari, Hokkaido, Japan. Fig. 1 shows the building’s facade, Table 1 lists the building specifications, and Fig. 2 illustrates a plan of a server room. In this datacenter, outdoor air comes into a server room from a wall. Fig. 3 shows a section of a server room. To stabilize the airflow volume, every rack has a fan on the surface of their ceiling to provide the necessary amount of cooling air. 2.2 Actual Measurements We measured the temperatures at seven points around a server room from May 2012 to May 2013. Table 2 lists the details of the points and Fig. 4 illustrates where those points are located. Equipment

Table 1

Specification of building.

Items Floor space Outline Setting condition Supply air system Air flow Exhaust air system system Fan Air conditioner Power Exhaust Power fan consumption Number of systems Cooling capacity Chiller Power consumption

Room A Room B 302.1 (m2/room) ASHRAE (2011) Wall Ceiling Exhaust tower FU × 4 AHU × 4 12.6 × 4 (kW) 33.0 × 4 (kW) 1.06 × 16 (kW/room) 1 1,617 (kW) 272 (kW) = 250 (kW) (Chiller) + 55 × 4 (kW) (Cooling tower)

Measured points

11,400 mm 11,400mm

Air conditioning

Fig. 1 The building facade.

26,500 mm 26,500mm

Fig. 2 The plan of the server room.


18

Table 3 Allowable and recommended condition (ASHRAE 2011). Low temperature High temperature Low humidity High humidity Fig. 3 The section of the server rooms. Table 2 Measurement detail. Subject

Point

Place Number

Outdoor

T1

4

Mixing

T2

2

Supply air

T3

3

T4

100

T5

100

Exhausted air T6

100

Return sir

4

Temperature Inlet air Outlet air

T7

Time Term span

1 day

Room A

2012/1/12013/7/31

Room B

Recommended 18 °C 27 °C DP 5.5°C RH 60%, DP15 °C

Allowable 15 °C 32 °C RH 20% RH 80%, DP21 °C

of 18-27 ºC (dew point). However, there is a drop in temperature between the exhaust ports of servers (T5) and return air (T7). This means that there is a large supply of cooling air that does not pass through servers but goes directly to the spaces between the roof and ceilings. We call this phenomenon “short circuit”. Because of the short circuit, there is a small gap in temperatures between supply air and return air. Moreover, there is a rise in temperature between supply air (T3) and the suction ports of servers (T4). This means that there is some cooling air passing through servers that comes back to the server room. We call this phenomenon “recirculation”.

3. Cooling Characteristics 3.1 Analytical Model

Fig. 4 The section of the datacenter. 35 30 25 20 15 10 5 0 ‐5 ‐10 -15 ‐15 35

Fig. 5 Temperature distribution.

T7

T6

T5

T4

T3

T2

T1

35

30 25 20 15 10 5 0 ‐5 ‐10 ‐15

To understand the cooling characteristics, we constructed the model illustrated in Fig. 5. Using this model, we can see how large the short circuit and recirculation are during operation. The meaning of signs is listed in Nomenclature. Table 4 lists the definitional equations of cooling characteristics. These equations explain how cooling air is supplied properly, irrespective of the change in outdoor temperatures. We now discuss the air flow. First, an air conditioner supplies cooling air (temperature = θ0 (°C), volume = V (m3)) to server rooms. Then, the supplied air (V (m3)) separates into effective supplied air in room (Vfr (m3)), which goes above the server racks, and ineffective supplied air (Vur (m3)), which exhausts before reaching the server racks. The temperature of Vfr (m3) rises to θ1d, called the temperature of exhausted air, while Vfr passes through the heating servers (we define the heat load as Hm (m3) and heat loss as Hw (m3)). After that, Vfr


Fig. 5

Model for cooling characteristics.

Table 4 Cooling characteristics. Ventilation flow ratio

 m  VV  m

1m   0 m 1   0

Equipment exhaust heat efficiency

Vm 

V fm V fr



1d   0 1m   0

Room exhaust efficiency

V 

V fr V

r



1   0 1d   0

Exhaust heat efficiency

V 

V fm V

 VrVm

Equipment inlet air temperature ratio

Equipment outlet air temperature ratio

m0 m 

m1m 

0m  0   1   0 V

1m   0 1   0

Exhausted air temperature ratio

Recirculation ratio

m1d 

  V m0 m

1d   0 1   0

Fresh air introduction ratio

q

19

VOA 1   MIX  V 1   OA

Relation equation

  V  m  1

is mixed with Vur (temperature = θ0) and the temperature of Vfr rises to θ1—the temperature of return air. Moreover, the inlet temperature of servers (θ0m (°C)) rises to θ1m (°C), while the amount of ventilation in

servers (Vm) passes through the servers; their heat load is Hm (m3). Then the air flowing through the servers is exhausted. Next, we define the recirculation ratio (γ). This means how much air is coming back to the servers after passing through them. If the ratio is high, it is inefficient. The amount of air coming back to the servers can be expressed as γVm. The supplied air passing through a ceiling chamber (V (m3)) separates into the exhausted air and return air going to the mixing chamber. The return air is mixed with fresh air (temperature = θOA (°C), volume = VOA (m3)), then the temperature becomes θmix (°C). Finally, the mixed air flows into the air conditioner. 3.2 Cooling Characteristics Each characteristic calculated with the obtained temperature data is shown in Fig. 6. Basically, every temperature ratio changes in proper order. However, gaps among the temperature ratios are small, so a small amount of recirculation or short circuit may exist. Until November 2012, the exhausted air temperature ratio was lower than the equipment-exhausted air temperature ratio.

20


Thus, we put a baking sheet on top of the operational server racks so as not to let the air exhaust without passing through the equipment. After we placed the sheet in November 2012, the exhausted air temperature ratio exceeded the equipment-exhausted air temperature ratio. This means the exhausted air temperature became higher than the equipment-exhausted air temperature regardless of the outside temperature. After all, the short circuit problem in equipment—the phenomenon in which supplied air is exhausted without passing through equipment was addressed by using the baking sheet. Also, the recirculation ratio (γ) and equipment-inlet air temperature ratio decreased after November 2012. This

is because the placement rate in the server racks increased. Therefore, there is no relation between using a baking sheet and improvement in γ. Thus, we will have to consider capping the operational parts in server racks to reduce γ. 3.3 Cooling Characteristics Evaluation To quantitatively verify the effect of setting a baking sheet on top of operating servers, we compared the cooling characteristics between June-July 2012 and June-July 2013, as shown in Table 5. The heat-exhaust efficiency (ηV) in 2013 (0.72), was much higher than that in 2012 (0.5). In particular, the room heat-exhaust efficiency (ηVr) improved from 0.57 to 0.71, so the

Room A 3.0 2.5 2.0 1.5 1.0 0.5 0.0

Room B 3.0 2.5 2.0 1.5 1.0 0.5 0.0 5/1

/

6/1

7/1

/

8/1

/

Ventilation flow ratio

換気流量比

Equipment inlet air

temperature ratio 機器吸込み温度差比

Fig. 6

Change of the cooling characteristics.

9/1

/

10/1

/

Room exhaust heat

排熱効率（室） efficiency

Equipment outlet air

temperature ratio 機器吹出し温度差比

11/1

/

12/1

/

1/1

2/1

/

Equipment exhaust

排熱効率（ラック） heat efficiency

Exhausted air tempera排気温度差比 ture ratio

3/1

/

4/1

/

5/1

/

Exhaust heat efficiency 排熱効率 Recirculation ratio

再循環比


Results of cooling characteristics.

June-July 2012 June-July 2013 Room A Room B Room A Room B Ventilation flow ratio 1.14 0.68 0.97 0.88 Room exhaust heat 0.57 0.59 0.71 0.67 efficiency Equipment exhaust 0.88 1.15 1.00 1.13 heat efficiency Exhaust heat 0.50 0.68 0.72 0.76 efficiency Equipment inlet air 2.01 1.49 1.41 1.50 temperature ratio Equipment outlet air 1.78 1.72 1.41 1.50 temperature ratio Exhausted air 0.43 0.55 0.30 0.34 temperature ratio Recirculation ratio 0.43 0.55 0.30 0.34

baking sheet seemed to work well. We can say that there seems to be a small amount of short circuit because the equipment heat-exhaust efficiency (ηVm) also improved. The equipment inlet temperature ratio in 2013 was lower than that in 2012 by 40%. This means that the gap in temperature between supply air (θ0) and equipment-inlet air (θ0m) decreased and the recirculation problem was mitigated. The ventilation flow ratio (κm) decreased from 1.14 to 0.97. The reason of this decrease is that the number of servers decreased and the amount of supply air needed also decreased.

4. Efficient Operation 4.1 Relation between Ventilation Flow Ratio and Power Consumption We estimated power consumption in relation to κm with the method for estimating power consumption established by Futawatari [3]. We set parameters for the placement of a baking sheet, exhaust-air efficiency, and γ against κm and θ0 and analyzed the characteristics of power consumption. We determined the behavior of power consumption and found the best κm. Table 6 lists the setting conditions of the analysis. 4.2. Ideal Ventilation Flow Ratio Fig. 7 shows the estimation of power consumption

Table 6 Analysis condition. Case A Case B Case C Case D Setting equipment inlet 18.0 18.0 27.0 27.0 temperature (°C) Non-sett Non-sett Banking sheet Setting Setting ing ing Analysis time (h) 24 (h) × 365 (day) = 8,760 (h) Heat load (kWh/h) 188 Equipment quantity of 48.2 ventilation (m3/s)

Power consumption (MWh/year)

Table 5

21

1,000 900 800 700 600 500 400 300 200 100 0 0.5

Fan

1.0

Chiller

1.5

Lighting

Total

Fig. 7 Estimation of power consumption against ventilation flow ratio in a year.

against κm in one year for each condition in Table 6. This graph illustrates that there is an exponential relation between power consumption and κm. Thus, to maximize efficiency, it seems best to reduce κm. However, a low κm is detrimental for proper operation. If κm is too low, the circulation ratio increases. A high circulation ratio causes rapid rise in θ0m, which leads to server breakdown. Moreover, when κm is too low, θ0 decreases, which is also detrimental for servers. In this section, we discuss the most efficient κm. To do this, we have to consider two factors: θ0 and γ. 4.1.1 Supply Air Temperature Fig. 8 illustrates the relation between estimated power consumption and θ0 against κm. According to ASHRAE, the lowest θ0 within the allowable range is 15 ºC [1]. Thus, θ0 should be higher than 15 ºC. Then


1,000

20

P attern A

P attern B

P ower consumption (MWh/year)

800

15

600 10

400 5

200

 m  0 .87

 m  0.67 0 30

0 1,000

P attern C

P attern D 25

800

Supply air temperature(℃ )

22

20

600 15

400 10

200

5 0

0 0.5

1.5 0.5

1.0

1.0

1.5

Ventilation flow ratio

Fan

Chiller

Lighting

Total

Supply air temperature

Fig. 8 Power consumption and supply air temperature against ventilation flow ratio in a year.

the best κm is when θ0 is 15 ºC, which is 0.87 in pattern A and 0.67 in pattern B. However, we could not determine the best value, especially in patterns C and D, because even the lowest κm does not result in a θ0 below 15 ºC. The reason is that θ0m is so high that θ0 would not be outside the allowable range. As a whole, we can say that it is better for efficient operation to set a high θ0m. When κm is 1.0, the power consumption is 483 MWh/year in pattern A, 479 MWh/year in pattern B, 261 MWh/year in pattern C, and 249 MWh/year in pattern D. To maximize efficiency and determine the most appropriate κm in patterns C and D, we considered the allowable range of γ. 4.1.2 Consideration of Recirculation Ratio We calculated and led the allowable range of γ by considering the defined equation for cooling characteristics and the allowable range of supply air defined by ASHRAE [1]:

18 

 0 m  1m    27 1 

(1)

The yearly average difference between equipment-exhaust air temperature and θ0m is 4.8 ºC. With this value, we computed the allowable range of γ. (1) θ0m was set to 18 ºC (for patterns A and B):

0    2.2

(2)

(2) θ0m was set to 27 ºC (for patterns A and B):

0    0.6

(3)

Considering these allowable ranges, we show the relation between power consumption and γ with κm in Fig. 9. In patterns A and B, we already obtained the proper values of 0.87 and 0.67. We can confirm that these values are safe by the graph shown as Fig. 8.


1,000

23

1.0

P attern A

P attern B

800

0.5 400 200 0.0 1.0

0 1,000

P attern C

800

P attern D

Recirculation ratio(-)

P ower consumption (MWh/year)

600

600

0.5 400

Ventilation flow ratio [-]

200 0 0.5

m  0.67

Fan

1.50.5 m  0.51

1.0

Chiller

0.0 1.0

Lighting

1.5

Total

Recirculation ratio

Fig. 9 Power consumption and recirculation ratio against ventilation flow ratio in a year. Table 7

Results of Analysis

Pattern

m

A B C D

0.87 0.67 0.67 0.51

Power consumption (MWh/year) 483 349 184 175

PUE 1.29 1.25 1.12 1.11

ICT power consumption (MWh/year) 188 (kWh/h) × 8,760 (h) × 10-3 = 1,646 (MWh/year)

We determined the best κm in patterns C and D, which is 0.67 in pattern C, and 0.51 in pattern D. The results of the analysis are listed in Table 7. We obtained the most effective operation in pattern D. The θ0m was set to 27 ºC, after placing a baking sheet, κm was 0.51, and the annual power consumption was 175 MWh/year. Therefore, PUE (power usage effectiveness) can be estimated as 1.11, which is better than the measured

PUE of 1.20. The PUE is equal to the entire power consumption in the datacenter divided by the power consumption for cooling heated servers. Thus, the PUE 1.11 means that only 11% of all the power is used for cooling equipment.

5. Conclusions We presented the cooling characteristics of a datacenter located in Ishikari, Hokkaido, Japan. This datacenter, which uses an outdoor air cooling system, can take in fresh air to cool servers. We assessed the cooling characteristics based on temperature measurements in the building and server rooms and determined a more effective operation method for cooling servers. We first measured the temperature of seven points in

24


a building through which the cooling air from outside passes. We then found two problems in the server rooms: short circuit and recirculation. To solve these problems, we placed baking sheets on top of server racks that were not in operation. The baking sheets worked well to slow down the supplied air from leaking out into the ceiling chamber. We also constructed a model to determine the cooling characteristics in detail regardless of outdoor temperature. We defined the cooling characteristics as equations. From those equations, we found that there were still short circuit and recirculation. We determined the most efficient operation method. We first set four patterns that differed in the temperature of equipment-inlet air and whether banking sheets were used. We estimated the power consumption in a year against the ventilation flow ratio. From that graph, we found there is an exponential relation between power consumption and ventilation flow ratio. It seemed to be the most appropriate to reduce the ventilation flow ratio to maximize efficiency. However, too low a ventilation flow ratio causes equipment breakdown because the supply-air temperature decreased to below 15 ºC, the lowest temperature of supply air in the allowable range. Also, low ventilation flow ratio increases the recirculation ratio. Under a high recirculation ratio, servers cannot be cooled enough to be safe. Thus, we considered these two factors to determine the most effective operation method. In pattern D and with a ventilation flow of 0.51, we showed that this datacenter can be under control with an annual power consumption of 175 MWh/year for cooling servers and PUE 1.11. However, further efforts are still required to consider the relation between the failure rate of servers and supply-air temperature. As the effective way of

operation is presented in this paper, further efforts are still required to consider the relation between the failure rate of servers and the temperature of supply air.

Acknowledgement We would like to thank to the members in Sakura Internet Corporation for letting us measure the temperature of server rooms in Ishikari Datacenter. Last but not least, this paper was supported by JSPS KAKENHI Grant Number 26289199.

References [1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

ASHRAE. 2011. Thermal Guidelines for Data Processing Environments. Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.. Belady, C. L. 2007. “In the Data Center, Power and Cooling Costs More than the IT Equipment It Supports.” Accessed June 20, 2013. www.electronics-cooling. coml2007/021. Hamilton, J. 2008. “Cost of Power in Large-Scale Data Centers.” Accessed May 14, 2013. http://perspectives.mvdirona.com/2008/11/28/CostOfPow erInLargeScaleDataCenters.aspx. Open Compute Project. 2011. “The Open Compute Server Architecture Specifications.” Accessed July 5, 2013. http://www.opencompute.org. Datacenter. 2011. “Quarterly PUE Benchmark Data.” Accessed June 25, 2013. http://www.google. com/corporate/datacenter/efficiency-measurements.html. Furihata, Y., Hayama, H., Enai, M., and Mori, T. 2003. “Efficient Cooling System for IT Equipment in a Datacenter.” In Proceedings of INTELEC’03 the 25th International Telecommunications Energy Conference, 152-9. Hayama, H. 1997. “Cooling Characteristics of Computer and Consideration of Energy Consumption Measures.” Journal of Architecture, Planning 494: 29-36. ASHRAE. 2011. Thermal Guidelines for Data Processing Environments—Expanded Data Center Classes and Usage Guidance. Whitepaper prepared by ASHRAE Technical Committee (TC) 9.9 Mission Critical Facilities, Technology Spaces, and Electronic Equipment.

D


DAVID

PUBLISHING

Development and Test of an Experimental Apparatus to Study Thermal-Choking in Ideal Gases and Self-decomposition in Superheated N2O Patrick Lemieux1, Alberto Fara2, Pablo Sanchez2 and William Murray3 1. California Polytechnic State University, San Luis Obispo, CA, 93407, USA 2. Politecnico di Torino, Turin 10129, Italy 3. California Polytechnic State University, San Luis Obispo, CA, 93407, USA Received: August 30, 2014 / Accepted: October 10, 2014 / Published: January 31, 2015. Abstract: N2O represents a popular oxidizer for hybrid rocket motors for a variety of reasons, including safety, ease of access and self-pressurization. It is often used as a saturated two-phase fluid in these applications to take advantage of self-pressurization. Recent interest in using this oxidizer in regeneratively cooled engines requires a detailed heat transfer process analysis to the coolant, in order to quantify performance. Since the injection of N2O typically takes place in the two-phase region, our study focuses on heat transfer rates in this region, and extends the region to include superheated vapor. This analysis is critical for these cooling applications, because the exothermic decomposition nature of N2O also means that unchecked heating in the superheated region may result in a runaway reaction in the cooling passages. Furthermore, provided that sufficient heat transfer rates are available, N2O is expected to accelerate in the cooling passages due to Rayleigh flow effects much like those of a calorically perfect gas. The proximity of superheated N2O to its saturated vapor curve, at the conditions studied here, makes the suitability of a perfect gas model questionable, but that benchmarks is still useful. This paper presents the development of an experimental apparatus (a “Rayleigh tube”), specifically designed to study this problem, and test the analytical methods developed to model it. Since we focus on the development of the apparatus, the data presented were uses primarily calorically perfect gas surrogates, but the goal is to apply the apparatus and method to N2O. The design and construction of the Rayleigh tube is presented, along with preliminary results with perfect gases. Finally, we present preliminary results on heated N2O flow. Using a simple model for predicted dry-out point, we investigate where superheating may be expected to occur. We present estimates of critical heating and compare them to the heat required to achieve self-decomposition. Key words: Self-decomposition, N2O, Rayleigh apparatus.

Nomenclature CHF N2O IAF SCR

Critical heat flux Nitrous oxide Inverted annular flow Silicon-controller rectifier

1. Introduction The combination of experimental, analytical and numerical modeling is the best way to obtain heat transfer coefficient data for two-phase flow of fluids, Corresponding author: Patrick Lemieux, professor, research field: mechanical engineering. E-mail: [email protected].

such as N2O, and thereby prevent dangerous self-decomposition excursions [1]. These results and modeling techniques will allow us to develop an appropriate tool for the modeling of N2O cooling in aerospike nozzles. Accordingly, our two main technical objectives in this project were: (1) Develop and characterize an experimental apparatus for tests on N2O that minimizes the uncertainties. The specifications for this apparatus are that it should be simple, flexible and sufficiently rugged to withstand changes in setup and operating conditions, and should be modular and relatively easy to replace or

26

Develo opment and Test T of an Experimental Apparatus A to Study S Therma al-Choking in n IDEAL GASES and Self-decomposiition in Superheated N2O

rebuild, makking it a more appropriatte platform for f a (possibly) deestructive deccomposition sttudy. (2) Devellop models foor the thermo fluid behavioor of its performaance in coooling N2O that characterize c applications. The primaary experimeental apparatuus for this prooject consists off an electrrically-heatedd, instrumennted, straight-pipee flow devicee suitable for Rayleigh, or heat driven, flow w experimennts. Electriccal heating was chosen so that the poower input could be eaasily measured annd controlledd. A conceptt drawing off the apparatus is i shown in Fig. 1. Each set of instrumentattion ports inncludes a thhermocouple for measuring coolant c tempperature, a pressure p tap for measuring static s pressuure, and a thhermocouple for measuring teemperature of o the copper wall of the pipe. p The plan was w to develoop, test and characterize this apparatus with w air (for itts low cost) and helium (as ( a surrogate for f a perfecct gas) befoore using itt to characterize the behaviorr of N2O. This paper discuusses primarily thee use of the apparatus a with air and heliium; a follow-up paper p will disscuss its use with w N2O in more m detail.

Fig. 1 Concept drrawing for elecctrically-heated d, straight-pipee Ray yleigh apparatu us.

2. Design of o Apparattus The basicc dimensionns of the appparatus (thatt is, diameter andd length) are a fundamental function off the maximum flow f rate thaat we felt comfortable c w with testing (prim marily with N2O), and with w length too the critical poinnt, over the full gamut of tests thatt we expected too run. Costt of manuffacturability and materials neecessarily scaaled with lenggth also, so inn the final analysiis, we chose to build an apparatus a withh an internal diam meter of 0.245 inches, andd a total lengtth of 87.50 inchess (i.e., maxim mum L/D for the apparatuus of about 350, high enoughh for most Rayleigh R analyysis, without proohibitively high h heating load demaands, which we arbitrarily capped c at 3 kW, againn in deference too the system cost c and a dessire to operatte on the 220 V ellectrical service available to us). We chhose copper for thhe tube materrial, for ease of o heat transfe fer to

(aa) Rayleigh appparatus with insulation annd frame.

(b) Rayleigh ap pparatus flow tube with eiight sets of instrumentaation ports.

Fig.. 2 Rayleigh apparatus. a

the heated gas. The T apparatuss was to be sub bdivided intoo eveen segments, as shown inn Fig. 1, to quantify q statee chaanges in eitheer Fanno (fricction-driven) or Rayleighh (heaat-driven) moode. Figs. 2-5 illustrate the t layout off insttruments on our o Rayleighh tube and fitttings used too maiintain leak-free operatiion at high h operatingg presssures. Threee measurem ments are made at eachh insttrumentation port: fluid temperature, flluid pressure,, and d copper walll temperaturre. The therm mocouple forr meaasuring coppeer wall tempeerature is at th he same axiall locaation, but is embedded in the pipe wall on thee opp posite side of the pipe.

Develo opment and Test T of an Experimental Apparatus A to Study S Therma al-Choking in n IDEAL GASES and Self-decomposiition in Superheated N2O

277

Fig. 3 A typical t instrum mentation porrt in the Rayyleigh apparatus.

Fig. 4 The Rayleigh tube, shown with on one-half off the calcium silicatte insulation.

To the exxtent possiblee, the Rayleiggh apparatus was designed to contain mosstly off-the-shhelf componeents. Electrical heeat tape withh flexible fibeerglass insulaation was chosen as the heat source, s and thhe most poweerful heat tape fouund is capablee of producinng 78 W/ft of heat tape. Howevver, through a separate seeries of tests, and with appropriate cooling (as expectedd when flowinng a gas in the tuube), we estabblished that we w could obtain at least 424 W/ft W of heat innput along thee pipe by usinng a double-layerr butted wrap w for the t heat tape, t correspondinng to a total heat h flow ratte of just above 3 kW. Subseequent experriments on the complleted Rayleigh appparatus havee shown thatt we can in fact generate a heat load of allmost 4 kW.

3. Mass Fllow Rate Estimates foor N2O For a reppresentative N2O process path during our experimentss, we expect the change in i enthalpy from f entrance intoo to the vaporr dome to the superheated state s to be approoximately 1855 kJ/kg. This correspondds to

Fig. 5 The Rayleeigh flow apparatus installed in the test celll C Poly. at Cal

presssure throttlinng from abouut 5 MPa upsstream of thee app paratus, to 3 MPa M at the tuube inlet, and a continuouss presssure drop thhrough the appparatus varyiing with heatt inpu ut. Given a maximum m off approximately 3 kW off heaat input from the t heat tape uused on our apparatus, a thee max ximum flow rate that takkes N2O to a superheatedd statte at the exit of o the pipe is:: q (1))  0.016 kkg/s = 0.036 lbm/s m   h To T allow heaadroom to reaach decompo osition of thee N2O, O our target m for N2O is 0.02 lbm/ss. This targett flow w rate correspponds to a rellatively low flow f velocityy for a pipe havinng an internaal diameter of o nominallyy 0.25 5 inches. Thiis low flow vvelocity loweers the Fannoo com mponent of the pressuree drop, therreby makingg the Rayleigh com mponent of thhe pressure drrop larger in a relaative sense, annd therefore easier to distinguish. Thiss low w N2O flow raate is also goood from a saafety point off

28

Development and Test of an Experimental Apparatus to Study Thermal-Choking in IDEAL GASES and Self-decomposition in Superheated N2O

view because it limits the amount of N2O expelled from the Rayleigh apparatus during an experiment.

Compute density prior to filling: R

4. Apparatus Verification Procedure

Pressurize the tube to a new state point:

4.1 Pressure Test R

The final Rayleigh apparatus was hydrostatically tested to 1,300 psi, and showed no sign of water leakage. The maximum operating pressure (with N2O) is not expected to exceed 750 psig, and given its two phase nature, the water leak test is probably sufficiently valid. This is not necessarily true for gas tests, which are required for characterization of our apparatus. We tested the apparatus under air pressure of up to 800 psig (in a secure Test Cell laboratory environment), and observed slow leaks in the apparatus, which we set to quantify. We also developed a method for quantifying the volume of the apparatus as-built. This volume is not easy to determine accurately, given manufacturing tolerances and the manifold of small tee-tubes used for differential pressure measurements (visible in Figs. 1-4). 4.2 As-Built Volume The actual, as-built apparatus volume is an important parameter. For instance, it cross references the total fluid mass in the tube (measured indirectly, by integrating the rate output of a Coriolis Mass Flow meter) with an equation of state for average density. To be valuable, this estimate must be accurate, and may significantly differ from the nominal design value, 11.687 × 10-5 m3 for the straight-pipe volume in this case. Yet, measuring this volume directly is daunting, given the length scale of the apparatus and many tube manifolds connected to it for sensing purposes. An experimental volume calculation technique was developed to take advantage of “ideal gas” fluid property variations during tube pressurization. The procedure consists of isolating the tube, and filling it with air from a large reservoir up to a predetermined pressure, all the while recording its temperature and pressure. The method is summarized below:

(2)

(3)

The added mass in the tube may be found by integrating the flow rate through the Coriolis meter, and relates the mass at the two states: ∆ (4) where, p, T, m are the pressure, temperature and mass inside the pipe, respectively, R is the specific gas constant (air in this case), and V is the desired volume. Note that p and T are obtained through the average measurements of the eight ports, and ρ is then calculated through the equation of state. The only unknowns remaining are m1, m2 and V, so that the set of three equations has a solution, and we can compute the apparatus volume. To reduce the possibility of bias from the instruments, the procedure was repeated many times, at different pressures, until a normal distribution curve of volume was obtained. Two of these curves are shown in Figs. 6 and 7, for tube pressurization to 100 psig and 300 psig, respectively. The mean volume obtained from this method gave a most-likely, as-built volume of 14.5 × 10-5 m3. Subtracting the known volume for the straight pipes between the shutoff valve at the bottom and the

Volume (1e-5 m3)

Fig. 6 Experimental volume values distribution for 100 experiments to 100 psi.


Volume (1e-5 m3)

Fig. 7 Experimental volume values distribution for 100 experiments to 300 psi.

apparatus, and the same at the top, we obtain our best-estimate of the apparatus volume: Vexperimental = 12.0 × 10-5 m3, or approximately 2.7% larger than that of the nominal, calculated volume of the apparatus.

mass flow rate and pressure drop data with the curve prescribed by the perfect gas model, as shown in Figs. 8 and 9, respectively. We thus establish that in our case, the effective leak area is on the order of 1.9 × 10-10 m2. More importantly, the procedure described here allows for a relatively simple technique to monitor changes in effective leak area, which may be due to loosening of fittings, thermal expansion, wear, thread damage, etc., over time, and the effect of any change that we choose to impart on the apparatus. The same method could be applied with improved accuracy, if required, by using a gas with a smaller molecular structure, for example helium. For the purposes of our characterization exercise, air was considered sufficiently accurate.

5. Results With the apparatus properties suitably characterized,

γ

t (min)

1 Fig. 8

2

R

2 1

γ

.

γ

1

2γ 1 R

Pt and Tt versus time data are obtained through a series of experiments on the tube. The value for γ, a polytropic constant in this case, is found empirically. Finally, the leak area is inferred by matching the actual

Pressure (psig)

2

γ

curves for different

2γR γ 1

γR

R

Mdol (Kg/s)

4.3 Effective Leak Area Calculation Once pressurized to 800 psig with air and isolated, the apparatus was found to leak at a (initial) rate of about 2.5 psi/sec. We set to correlate this pressure drop with an estimated effective leak area size, which represents the sum area of all the sites where gas may escape. Given the high pressure drop, we consider choked flow across each actual leak orifice. The effective leaks area (Aeff) is then modeled as:

29

t (min)

Fig. 9 Pressure drop curves, for different

.

30


we ran two experiments with air, at different pressures and mass flow rate, with heat input of approximately 3.5 kW. The result of these runs is presented below.

6. Predictive Model for Perfect Gas Behavior The theory behind the behavior of friction-driven (“Fanno”) and heat-driven (“Rayleigh”) perfect gas flows is well-established [2]. Using an appropriate perfect gas surrogate, therefore, enables us to directly test our ability to capture the unique features of these flows, by verifying our results against accepted theoretical models, something that cannot be readily carried out with more exotic fluids, such as two-phase N2O. A great perfect gas surrogate for this purpose is helium: its specific heats remain constant over a large range of temperature, and its properties are accurately modeled by the Ideal Gas Law over a large range of conditions. Having established a basic experimental framework (basic design dimensions for the apparatus, instrument layout, flow rate capabilities, as-built volume, etc.), a perfect gas surrogate was used to characterize the apparatus friction factor. This is a function of surface roughness and Reynolds number, using a perfect gas surrogate to determine its value sets it for two-phase flows studies as well. 6.1 Fanno Flow Analysis In purely friction-driven, or Fanno flow, the static pressure and temperature along the flow pipe are not independent, but instead are a function of the pipe “effective” surface roughness and the stagnation temperature at inlet. By measuring both the gas temperature and pressure, at the same location, this gives us an indication of the quality of the resolution of the instrument, and whether the setup is properly calibrated. We used the following method, adapted from calorically perfect ideal-gas Fanno theory, and conservation of mass and energy. Define the mass flux G (mass flow rate per unit area):

· (5) Using the subscript “0” to denote the stagnation state, then, conservation of energy dictates that: · R

(6)

where, R represents the specific gas constant of the perfect gas, and P and T are the static pressure and temperature, respectively, and both are being measured in apparatus. So, for a given entry stagnation temperature in the apparatus, also measured, each measurement of temperature should be associated with a unique value of pressure, and vice-versa, regardless of the physical level of pipe friction. This provides a baseline method for verifying our data: we use the first data point (pressure and temperature, and mass flux) to determine the stagnation condition, and use the set of subsequent measured pressure points to determine how well our thermocouples are tracking the expected temperature field in the field. We repeat the exercise, using the temperature data to predict pressure. Ideally, the two sets of data match in both case, but the exercise illustrates if there are variations in the predicted data set. Figs. 10-13 show the results of this analysis. This is due to the variability in temperature measurements, where the instrument noise is of the same order as the change in temperature we are trying to detect here. We conclude that as far as Fanno modeling is concerned, the resolution of our thermocouple system is inappropriate for a proper characterization of the apparatus in that mode. For the sake of Fanno modeling, we also conclude that the pressure measurements must dictate the calculation of the temperature field. For low speed flow modeling with a two-phase fluid or even a perfect gas, significantly higher resolution thermocouple measurements are required, though those instruments may not be suitable for heated flow experiments (especially in perfect gas experiments), our main goal here. Based on the data shown above, using pressure to predict temperature, the roughness level of the apparatus falls somewhere between those of “drawn

31

Pressure (psi)

Flow rate (lbm/s)


Time (s)

Fig. 10 Pressures at Station 1 through Station 8, plus the mass flow rate (COFM) from the Coriolis mass flow meter, for Run #1. 5,000 4,500 4,000

Temperature (℉)

3,500

2,500 2,000

Power (W)

3,000

1,500 1,000 500 0 Time (s)

Fig. 11 Air temperatures at Station 1 through Station 8, plus the input power (nominally 3.5 kW for this test), for Run #1.

copper tubing” and “glass”. In other words, very smooth pipe walls, despite the instrument penetrations, and therefore suitable for pure Rayleigh flow experiments. 6.2 Rayleigh Flow Analysis There are two fundamental elements of analysis for heat-driven (Rayleigh) flows: first, there is no pressure

loss due to friction along the apparatus; and second, all heat added is properly accounted for. The first requirement was verified during our Fanno flow tests. The second requirement is investigated here. Certain assumptions must be made with respect to the apparatus before this analysis can be carried out. For instance, we assume that the temperature change in the fluid is consistent with a purely radial heat flux

Flow rate (lbm/s)


Pressure (psi)

32

Time (s)

Fig. 12 Pressures at Station 1 through Station 8, plus the mass flow rate (COFM) from the Coriolis mass flow meter, for Run #2. 5,000 4,500 4,000

Temperature (℉)

3,500

2,500

Power (W)

3,000

2,000 1,500 1,000 500 0 Time (s)

Fig. 13 Air temperatures at Station 1 through Station 8, plus the input power (approximately 3.5 kW), for Run #2.

through the pipe to the fluid, though this is not the case. Figs. 14-16 illustrate how this assumption may cause an error in measurements, because of the relatively low axial thermal resistance of the copper tubing. Fortunately, with the types of fluids being tested in the Rayleigh experiments of concern here (turbulent regime helium and two-phase flows), the convective

heat transfer coefficients are high, so that the effective Biot number of the apparatus remains high enough that convection may be expected to dominate axial conduction. Still, an axial heat loss through the ends of the apparatus may be expected. Heat flux to the apparatus is provided by a series of electric heaters that generate a total of approximately

Development and Test of an Experimental Apparatus to Study Thermal-Choking in IDEAL GASES and Self-decomposition in Superheated N2O "

33

(7)

Temperature (℉)

where, D is the inner diameter; " is the heat flux (52 is the mass flow rate; and Cp is the specific W/in2); heat of the gas. Moreover, since the mass flux is constant, the pressure may also be modeled implicitly by an energy balance to give: " · Longitudinal position (inches)

Pressure (psi)

Fig. 14 Comparison of calculated temp erature in the Rayleigh apparatus (in “Fannomode” inblue), using the measured pressure, as compared to the actual measured temperature (in black).

Longitudinal position (inches)

Fig. 15 Calculated pressure (in blue) based on the measured temperature, versus the actual measured pressure.

(8)

with: R·

(9)

This model (independent of fluid temperature downstream of the first thermocouple) allows us to determine the total level of heat loss of our apparatus, through three possible loss paths: (1) radially through the insulation; (2) axially through the end caps; and (3) axially along the copper tube itself. To illustrate the results of this analysis, we present two runs performed with heated helium, corresponding to two different flow rates: 0.0110 lbm/sec (Run #12, see Fig. 17), and 0.0145 lbm/sec (Run #18, see Fig. 19). The expected temperature plots, based on Eq. (7) above, are then modified to account for heat loss through the apparatus, until the curve matches the data points. Both runs showed good agreement when corrected assuming a 22% loss of nominal heat input, as shown in Figs. 18-20. Using that constant heat loss factor, both runs show good agreement with the theoretical prediction. The copper surface temperatures varied between 173 °F and

Fig. 16 Heat transfer modes in the apparatus. Some heat may be transferred axially upstream through the copper pipe (i.e., heat conduction, left to right).

3, 500 W along the insulated pipe surface, which works out to 52 W/in2 of pipe area exposed to the heated gas. The expected temperature of the gas as a function of axial distance along the pipe is therefore:

190 °F for the low flow rate run (Run #12) and 142 °F and 256 °F for the higher flow rate run (Run #18), giving us an estimate for the temperature tolerance of the heat loss characteristics of the apparatus, which may also be applied to test runs using a less-well characterized test gas, such as N2O.

7. Preliminary Predictive Model for N2O Behavior For two-phase flows, the CHF is defined as the

34

Development and Test of an Experimental Apparatus to Study Thermal-Choking in IDEAL GASES and Self-decomposition in Superheated N2O Heated helium (Rayleigh) Run #18 corrected heat flux

Temperature (℉)

Temperature (℉)

Heated helium (Rayleigh) Run #12

Distance along pipe (inches)

Fig. 17 Low mass flow rate heated helium test. Heated helium (Rayleigh) Run #12 corrected heat flux

Distance along pipe (inches)

Fig. 20 Same run, modeled using the same 78% actual heat input to the flow, as used in the low flow rate experiment model.

rather than flow boiling. In flow boiling, the critical Temperature (℉)

heat flux is a function of thermodynamic quality. As a fluid moves through a heated pipe and is evaporated, the quality increases. This results in a decrease in CHF. If the fluid remains in the pipe long enough, the applied heat flux will eventually exceed the critical heat flux. The location in the pipe at which this occurs is called the “dryout” point. As is the case in pool boiling, Distance along pipe (inches)

Fig. 18 Same run, modeled using 78% of the actual heat input to the flow. Heated helium (Rayleigh) Run #18

dryout in flow boiling results in an increase in surface temperature and is generally avoided. There are two types of dryout. If the fluid is low quality at the point of dryout, IAF occurs in which a liquid core is surrounded by a vapor ring. The second

Temperature (℉)

type of dryout is termed mist flow in which the fluid simply breaks into small droplets and is suspended in a vapor bulk. This type of dryout, which is the most likely for our N2O-cooling applications, typically occurs at high qualities when the fluid is in the annular flow regime prior to mist flow dryout. Note that heat transfer coefficients in the mist flow regime are Distance along pipe (inches)

Fig. 19 High mass flow rate, heated helium test. Once again, the modeled temperature distribution overestimates the actual, measured temperature as a result of heat loss through the apparatus.

heat load above which the liquid phase will no longer wet the walls of the pipe. If the applied heat flux exceeds this value, the flow is classified as film boiling

considerably higher than those in the IAF regime. Similarly, the observed wall temperatures are much lower in mist flow than IAF. This is due to the fact that mist flow liquid is able to intermittently wet the walls of the pipe, which is not the case in IAF. An analytical dryout model for N2O was developed to be used to predict the flow state in the current experimental apparatus. This model represents the


two-phase flow of N2O as comprised of a homogeneous fluid, that is, the properties of the two phases are averaged over each cross section of the pipe. In this way, a one-dimensional flow assumption is made that allows single-phase fluid dynamic principles to be applied. This model ran for two cases, each case having an N2O saturation pressure of 365 psia and a heat input rate of 3.32 BTU/s (3.5 kW), but with the two cases having different mass flow rates: 0.05 lbm/s and 0.25 lbm/s, respectively.

along the pipe, the copper wall temperature first drops slightly, and then begins to rise at an increasing rate as the N2O quality approaches one. The heat transfer coefficient as a function of distance along the pipe is shown in Fig. 25. As is expected, the heat transfer coefficient decreases as the quality increases. Notice that there appears to be a slight maximum at about 12 inches into the pipe, which corresponds to the location of the minimum wall temperature. The N2O quality at this location has a value of about 25%. In general, two-phase flows will dryout prior to reaching the saturated vapor line [3]. A significant limitation of the homogeneous flow model used by itself is that it includes no mechanism for determining dryout, other than simply assuming that it occurs right at the saturated vapor line. As heat transfer coefficients are much higher in the nucleate boiling regime than that in the mist flow or film boiling regime, this method will likely over predict heat transfer coefficients and under predict wall temperatures. It is therefore a very rudimentary approach, but it does provide an upper bound for when dryout occurs. And in the low flow rate case, dryout in the actual system would likely occur for quality in the range between 50% and 70%, corresponding to 40 inches and 65 inches into the pipe.

Quality

Results for the low flow rate case, 0.05 lb /s, are shown in Figs. 21-25. As is expected, the N2O quality increases as the N2O travels along the pipe, reaching almost 85% at the exit. Friction is ignored in the homogeneous model, so the pressure drop along the pipe is solely due to the heat added. The total pressure drop along the Rayleigh pipe more than 7 feet is approximately 1 psi for the low flow rate case, which is on par with the predicted Rayleigh pressure drop for the other models presented herein. Since the quality remains less than one, the N2O remains saturated, and accordingly, the temperature is seen to drop as the pressure drops. As the N2O evaporates as it flows along the pipe, it cools the copper walls of the pipe, which are simultaneously being heated. As a function of distance

35

Axial distance (in)

Fig. 21 Predicted N2O quality along the pipe for the low flow rate case.


Mixture pressure (psia)

36

Axial distance (in)

Copper wall temperature (℉)

Fig. 22 Predicted N2O pressure along the pipe for the low flow rate case.

Axial distance (in)


Fig. 23 Predicted N2O temperature along the pipe for the low flow rate case.

Axial distance (in)

Fig. 24 Predicted copper wall temperature along the pipe for the low flow rate case.

37

Coolant heat transfer coefficient (BTU/in2-sec-°R)


Axial distance (in)

Quality

Fig. 25 Predicted heat transfer coefficient along the pipe for the low flow rate case.

Axial distance (in)

Fig. 26 Predicted N2O quality along the pipe for the high flow rate case.

Results for the high flow rate case, m  0.25 lbm/s, are shown in Figs. 26-30. The overall results for the high flow rate case differ from those for the low flow rate case differ in the manner expected due to the greater cooling capacity available, for the constant input heating rate. The N2O quality reaches 35% at the exit, indicating that dryout likely would not be reached in the real apparatus under these flow conditions. The Rayleigh pressure drop remained small at approximately 3.5 psi, although it more than tripled from the low flow rate case. Since the quality remains less than one, the

N2O remains saturated, and accordingly, the temperature is seen to drop as the pressure drops. As in the low flow rate case, as the N2O evaporates as it flows along the pipe, it cools the copper walls of the pipe, which are simultaneously being heated. However, in this case, evaporation of the N2O provides enough cooling to keep the temperature of the copper wall dropping slightly as a function of distance along the pipe. The heat transfer coefficient as a function of distance along the pipe is shown in Fig. 30. In a practical sense, the heat transfer coefficient is essentially constant in this


Mixture pressure (psia)

38

Axial distance (in)

Predicted N2O pressure along the pipe for the high flow rate case.

Mixture temperature (℉)

Fig. 27

Axial distance (in)


Fig. 28 Predicted N2O temperature along the pipe for the high flow rate case.

Axial distance (in)

Fig. 29 Predicted copper wall temperature along the pipe for the high flow rate case.

39

Coolant heat transfer coefficient (BTU/in2-sec-°R)


Axial distance (in)

Fig. 30 Predicted heat transfer coefficient along the pipe for the high flow rate case.

case. However, a close examination of the details in Fig. 30 does show that again a slight maximum occurs at the location where the N2O quality has a value of about 25%. Whether the existence of this maximum is due to parameter changes with respect to temperature or a strange ramification of the homogeneous model is unknown at this point. The apparent discontinuities in wall temperature and heat transfer coefficient, which

contributed to the modeling of the experiments and this help is greatly appreciated. Portions of this work were carried out under NASA STTR Contract NNX11C107P [4]. The authors gratefully acknowledge this support.

References [1]

appear in both the low and high flow rate cases, are attributed to the discontinuous nature of the Shah heat transfer correlation used in the model. [2]

Acknowledgments The assistance of Mr. Terry Cooke and Mr. Jim Gerhardt, both of Cal Poly State University, in helping to design and conduct the experiments in this project is gratefully acknowledged. Michael Kerho, Chief Aerodynamicist with Rolling Hills Research, also

[3]

[4]

Karabeyoglu, A., Dyer, J., Stevens, J., and Cantwell, B. 2008. “Modeling of N2O Decomposition Events.” Presented at the Proceedings of the 44th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Hartford. Zucker, R. D., and Biblarz, O. 2002. Fundamentals of Gas Dynamics, 2nd ed.. Chichester: John Wiley and Sons. Chen, J. C. 1966. “Correlation for Boiling Heat Transfer to Saturated Fluids in Convective Flow.” Industrial Engineering Chemistry Process Design and Development 5 (3): 322-9. Murray, W. R. 2010. “A Refined Model for the Behavior of Nitrous Oxide (N2O) to Assess the Limits of N2O Cooling.” NASA STTR Phase I Grant.

D


DAVID

PUBLISHING

Green Product of Liquid Fuel from Plastic Waste by Pyrolysis at 900 °C Dianta Mustofa Kamal and Fuad Zainuri Department of Mechanical Engineering, Politeknik Negeri Jakarta, Depok 16425, Indonesia Received: September 21, 2014 / Accepted: November 05, 2014 / Published: January 31, 2015. Abstract: Alternative treatments to convert plastic waste into fuel currently receive great attention from researchers worldwide. The objective of the research is to obtain liquid fuel from pyrolysis of waste plastics that is safe for humans as well as environment, with a heating value and fuel quality meet the standardized compliant. The method used for the research is plastic waste pyrolysis heated at 900 °C, and the resulting vapor is condensed through a crossflow condenser. The method resulted in a liquid fuel with a calorific value of 46,848 J/g, which is greater than that of plastic waste processing at a temperature of 425 °C that is only 41,870 J/g. In addition, the nature of current method for treating plastic waste is considered more secure than that of plastic waste processing at the temperature of 425 °C. The reason for this is the fact that the percentage of compounds that could potentially be carcinogenic (boric acid and cyclopentanone) is reduced. Key words: Plastic waste, fuel, pyrolysis, green product.

1. Introduction As the highest consumption of fossil fuels country, Indonesia consumed petroleum for approximately 1.6 million barrels per day in 2005, while in 2006, it reached 1.84 barrels per day. Other countries such as Japan and Germany equally consume only less than 1 million barrels per day [1]. In 2013, the United States produced about 30 million tons of plastic each year total, but with only about 4% are recycled [2]. In addition to producing energy, the combustion of fossil energy sources also releases gases, including CO2 (carbon dioxide), NOx (nitrogen oxides), and SO2 (sulfur dioxide), which causes air pollution [3]. So it is highly necessary to find alternative fuels to be widely used that are environmentally friendly. Research developed at this time can be used as fuel instead of fossil fuels [4]. Previously, in 2009, Anggono [1] has conducted a research on the type of plastic waste from food packaging (Low Density Corresponding author: Dianta Mustofa Kamal, Dr., research field: renewable energy. E-mail: [email protected].

Polyethylene or LDPE) at a temperature of 425 °C heating, and the results show that compounds have properties such as flammable acetone and cyclopentanone (1.68% area) [5]. In the same year, Damanhuri [6] stated that the cyclopentanone compounds were cyclic ketone compounds that potentially exist carcinogenic gas (toxic). In addition, boric acid is also harmful if it is accidentally breathed in, since it may irritate mucous membranes that showed by sore throat, coughing, and short breathing. Basing on the background, Indonesia contributed to the decline in petroleum reserves and also the problem of energy crisis faced by the world today. Therefore, this study aimed to obtain liquid fuels resulted from pyrolysis of waste plastics that is safe for humans and environment, with a heating value and fuel quality that meet standardized-compliant.

2. Material and Method Shredded plastic waste and included in the converter and heated to a temperature of 900 °C, and the resulting vapor is condensed through a crossflow condenser.

Green Product of Liquid Fuel from Plastic Waste by Pyrolysis at 900 oC

Fuel oil is produced at the heating temperature of 900 °C, and the heating value is tested using Bomb Calorimeter and testing GC (gas chromatography). Tests conducted uses the calorific value-bomb calorimeter contained in Energy Conversion Engineering Laboratory, Polytechnic of Jakarta. GC-MS (GC mass spectrometry) analysis method is used to read both spectra contained in the combined method. GC test results if there are samples contain many compounds, which are evident from the many peaks (peak) in the GC spectra. Based on the data retention and time is already known from the literature, we know what compounds were present in the sample [7]. Next step is to incorporate the compound into the alleged mass spectroscopy instruments. This can be done because one of the uses of GC is to separate the compounds of a sample. After that, the results can be obtained from MS spectra at different charts.

3. Result and Discussion Pyrolysis process starts at temperatures around 230 °C [8] and happens in the absence of oxygen for thermal decomposition of organic material [9, 17]. Plastic is a synthetic organic material or semi-synthetic organic materials derived from petroleum and natural gas of plastic products, resulting in PET (polyethylene terephthalate), HDPE (high density polyethylene), PVC (polyvinyl chloride), LDPE (low density polyethylene), PP (polypropylene), PS (polystyrene), polyurethane and polyphenols, generating plastic waste that consists around 50%-60% of polyethylene, 20%-30% of polypropylene, 10%-20% of polystyrene, and 10% of polyvinyl chloride [10, 17]. For PE (polyethylene) medium and high density polyethylene, the melting point ranges from 120 °C to 135 °C. Low density polyethylene melting point ranges from 105 °C to 115 °C. HDPE is characterized by a density that exceeds or equals to 0.941 g/cm3. HDPE has a low degree of the ramifications and inter-molecular [16] strength and very high tensile strength. It functions

Fig. 1

41

Converter plastic waste into fuel.

as material for milk bottle, bottle/detergent packaging, packaging margarine, water pipes and bins. LDPE is characterized by a density from 0.910 g/cm3 to 0.940 g/cm3. LDPE has a high degree of the long and short chain branching, which means it will not turn into a crystalline structure. It also indicates that LDPE has a low tensile strength. LDPE can be found in the form of container since it is strong and in the form of plastic film applications, such as plastic bags and plastic wrap. LLDPE (linear LDPE) is characterized by a density between 0.915 g/cm3 and 0.925 g/cm3. LLDPE is a linear polymer with a short chain branching with a significant amount. LLDPE is used as material for cable wrap, toys, packaging caps, buckets, containers and pipe [15]. Sapriyanto [5] has tested a machine to convert plastic waste into fuel. The test material is 1 kg of plastic waste that is heated within 530 °C heating. All kinds of plastic are put into the machine. Then, within two hours, the machine produces liquid fuels as much as 300 mL. The test shows the calorific value of the fuel is plastic waste of 10,519 cal/g or 44,040.95 J/g, equivalent to the heating value of the premium is 10,285 cal/g or 43,061.24 J/g. In the same year, Ramadhan [11] also examined the oil obtained from the pyrolysis process of waste plastic. This study uses two types of plastic as a fixed variable, namely HDPE and LDPE, using the reactor with a diameter of 20 cm and height of 40 cm. Pyrolysis temperature is held at


42

250-420 °C and the reaction time lasts for 0-60 minutes. Oil produced in the pyrolysis process can be compared to kerosene and oil is a source of valuable chemicals, such as alcohols, organic acids, ethers, ketones, aliphatic and aromatic hydrocarbons. And gas is produced in the form of Cox, NOx, H2 and alkanes [6, 15, 19]. Suryo [12], in his study of the properties of a mixture of waste biomass pyrolysis oil and plastic waste PP (polypropylene), tries to investigate the density, viscosity and heating value. The pyrolysis oil resulted from the research is thus used to boil water on the stove. The efficiency of pyrolysis oil stoves is also tested using standard WBT (water boiling test). The research shows the obtained oil-fired stove efficiency is best at 30% biomass composition and 70% plasticat the temperature of 400 °C. Based on the decision of the Director General of Oil and Gas in 2008, the Ministry of Energy and Mineral Resources of the Republic of Indonesia, the standards and quality (specification) of fuel, in the form of oil marketed in the country, are as follows [13, 18]. At first, pyrolysis technology is considered as an environmentally friendly method [14] since the method Table 1

ultimately produces CO2 and H2O, the former is a non-toxic gas. But in its development, the cyclopentanone compounds as the result of the pyrolysis of cyclic ketones can potentially turn into carcinogenic gas (toxic). In addition, boric acid, also harmful if inhaled, can cause irritation of mucous membranes accompanied by sore throat, coughing, and breathing becomes short [3]. Cyclopentanone compounds can be identified through gas chromatography [7, 15, 17, 19]. The calorific value is compared to the value obtained with the standard and quality (specification) of fuel oil type of oil that is marketed in the country (Dept. of Energy and Mineral Resources of Indonesia, 2008) [19], to meet the calorific value of fuel, standards should be above 41,870 J/g. Based on test results, the value of the heat, produced of 11,189 cal/g or 46,848 J/g, thus can meet the standard calorific value of the fuel sold in the country. Based on the results of testing, the fuel produced at the heating temperature of 900 °C obtained levels of flammable compounds (2-propanone) increases, while potentially carcinogenic compounds (boric acid and cyclopentanone) reduced the percentage.

Specifications of fuel oil [18]. Limit

No. Characteristic

Units

1 2 3 4 5 6 7 8 9 10 11

MJ/kg kg/m3 mm2/dt % m/m °C °C % m/m % m/m % m/m % v/v mg/kg

NilaiKalori Densityat 15 °C Kinematics Viscosityat 50 °C Sulfur contain Melting point Flashing point Carbon Residual Ash contain Sedimen Water contain Aluminium + silikon

Intermediete fuel oil-1 Min. Marks. 41.87 991 180 3.5 30 60 16 -

Intermediate fuel oil-2 Min. Marks. 41.87 991 380 4.0 40 60 20 0.15 0.10 1.00 -

Source: Director general SK oil & gas, energy and mineral resources, 2008. Table 2

Test results calorific value (1 gram mass oil plastic).

Calorific value BBM plastic Premium Quality Standards of Ministry of Energy in Indonesia

Unit (cal/g) 11.189 11.245 10.000

Testing method ASTM D 240 D 1259 D 445 D 1552/2622 D 97 D 93 D 189 D 482 D 473 D 95 D 5184/AAS


43

Testing results of GC-MS

Fig. 2 Graphs the GC-MS testing fuel from plastic waste. Table 3 Data of important compounds of liquid fuels from waste plastics. No.

Peak

% Area

Expected compounds

Formula

1 2 3 4

1 2 3 4

64.69 27.08 0.09 6.93

2-propanon/acetone Boric acid Acetic acid

5

5

1.20

Cyclopentanone

C3H6O H3BO3 C2H4O2 C5H8O

4. Conclusion The processing of plastic waste at a temperature of 900 °C produces liquid fuel with calorific value of 46.848 J/g which means that this value is greater than that of the processing of plastic waste at a temperature of 425 °C which results in the calorific value of 41.870 J/g. Testing the fuel produced GC-MS showed that the levels of potentially carcinogenic compounds (boric acid and cyclopentanone) is reduced so that the percentage of plastic waste at a temperature of 900 °C has more secure properties than that of plastic waste processing at a temperature of 425 °C.

References [1] Anggono. 2009. “Pyrolysis of Waste Plastics to Getting Liquid Smoke and Determination of Chemical Components and Its compiler Ability Test for Liquid Fuels.” Sainsdan Terapan Kimia 3 (2): 174-82. [2] Sarker, M., and Rashid, M. M. 2013. “Food Container Waste Plastic Conversion into Fuel.” International Journal of Engineering and Applied Sciences 3 (1): 1-16. [3] Kays, W. M., and London, A. L. 1955. Compact Heat Exchangers. Palo Alto: National Press. [4] Abatneh, Y., and Sahu, O. 2013. “Preliminary Study on the Conversion of Different Waste Plastics into Fuel Oil.” International Journal of Scientific & Technology Research 2 (5): 226-9. [5] Sapriyanto, A., 2011. Plastic Waste Converters Being Oil. Student Creative Program of Politeknik Negeri Jakarta,

44


Indonesia. [6] Damanhuri, E., 2009. Management of Hazardous and Toxic Materials (B3). Bandung Institut of Technology. [7] Rodiansono, Triyono, W. T. 2007. “Preparation, Characterization and Activity Test of NiMo/Z and NiMo/Z-Nb2O5 Catalysts for Hydro Cracking of Waste Plastic Fraction to Gasoline Fraction.” Berkala MIPA 17 (2): 43-54. [8] Kreith, F., and William, Z. B. 1980. Basic Heat Transfer. New York: Harper & Row. [9] Kamal, D. M., 2013. “Polytech: Conversion Machine of Plastikinto Oil Fuel with Continuous System and Reservoir Wet-Steam Oil with 20 kg Capacities.” In Proceedings of AISC, 283-7. [10] Sari. 2011. The Calorific Value Combustion Optimization Bio-briquettes Coal Mixed with Coconut Shell Charcoal. Sebelas Maret University, Surakarta, Indonesia. [11] Ramadhan, A., and Munawar, A., 2011. “Plastic Waste Processing Using Pyrolysis Process into Oil.” Jurnal Ilmiah Teknik Lingkungan 4 (1): 44-53. [12] SuryoAji Wibowo, A. 2011. The Study of Characteristics of Mixed Waste Biomass Pyrolysis Oil and Waste Plastic Polypropylene (PP). Sebelas Maret University, Surakarta, Indonesia.

[13] Sarker, M., and Rashid, M. M. 2013.“Mixture of LDPE, PP and PS Waste Plastics into Fuel by Thermolysis Process.” International Journal of Engineering and Technology Research 1 (1): 1-16. [14] Napitupulu, F. H. 2006. Effect of Calorific Value (Heating Value) a Planning against Fuel Fuel Boiler Room Volume Method Based Determination of Calorific Value of Fuel Required. Mechanical Engineering, North Sumatera University. [15] Imam, M., 2005. The Nature and Characteristics of Plastic Materials and Materials Additives. Academy of Maritim, Semarang, Indonesia. [16] Munson, B. R., Young, D. F., and dan Okiishi, T. H. 2002. Fundamentals of Fluid Mechanics, 4th ed.. Hoboken: John Wiley & Sons. [17] Pavia, D. L., Gary, M. L., George, S. K., and Randall, G. E. 2006. Introduction to Organic Laboratory Techniques, 4th ed. Belmont: Thomson Brooks/Cole, 797-817. [18] Regulation of the Ministerof Environment No.13 Year 2009 regarding Standard Emissions Activity Oil and Gas Industry. [19] Zuhra, C. F. 2003. “Refining, Processing and Using of Crude Oil.” NorthSumatra University. Digitized by USU Digital Library.

D


DAVID

PUBLISHING

Numerical Simulation on Ice Shedding Phenomena in Turbomachinery Ryosuke Hayashi1 and Makoto Yamamoto2 1. Graduate School of Mechanical Engineering, Tokyo University of Science, Tokyo 125-8585, Japan 2. Department of Mechanical Engineering, Tokyo University of Science, Tokyo 125-8585, Japan Received: October 01, 2014 / Accepted: November 03, 2014 / Published: January 31, 2015. Abstract: In the jet engine, icing phenomena occur primarily on the fan blades, the FEGVs (fan exit guide vanes), the splitter, and the low-pressure compressor. Accreted ice disturbs the inlet flow and causes large energy losses. In addition, ice accreted on a fan rotor can be shed from the blade surface due to centrifugal force and can damage compressor components. This phenomenon, which is typical in turbomachinery, is referred to as ice shedding. Although existing icing models can simulate ice growth, these models do not have the capability to reproduce ice shedding. In the present study, we develop an icing model that takes into account both ice growth and ice shedding. Furthermore, we have validated the proposed ice shedding model through the comparison of numerical results and experimental data, which include the flow rate loss due to ice growth and the flow rate recovery due to ice shedding. The simulation results for the time at which ice shedding occurred and what were obtained using the proposed ice shedding model were in good agreement with the experimental results. Key words: Ice accretion, ice shedding, turbomachinery, multiphysics CFD (computational fluid dynamics).

Nomenclature Acell Ain CD Eac Eair Econ Ee,s Efri Eim Ein Eout Fa Fc f dd LWC mac me,s mim min

Area of surface cell Area of inlet droplet Drag coefficient Energy of accreted ice Energy of aerodynamics Energy of convection Energy of evaporation or sublimation Energy of friction Energy of impinging droplet Energy of runback-in Energy of runback-out Adhesion force between ice and the wall Centrifugal force acting on accreted ice Freezing fraction Droplet diameter Liquid water content Mass of accreted ice Mass of evaporation or sublimation Mass of impinging droplet Mass of runback-in

Corresponding author: Ryosuke Hayashi, Ph.D. student, research fields: CFD and Jet Engine. E-mail: [email protected].

mout MVD Ncom Nim Nin Qin Pim Red rd t Ud Uf Uin Ur Vd β ρd ρf τ Ω

Mass of runback-out Median volume diameter Total number of computational droplets Local number of impingement droplets Number of inflow droplets Mass flow rate of inflow droplets Local rate of impingement droplets Reynolds number of droplet Radial position of droplet Time Droplet velocity Flow velocity Inlet droplet velocity Relative velocity between air and droplet Droplet volume Local water distribution Droplet density Flow density Adhesion stress Rotational speed

1. Introduction At altitudes in which aircraft operates, clouds contain numerous super-cooled droplets, which impinge

46

Numerical Simulation on Ice Shedding Phenomena in Turbomachinery

and accrete on an aircraft body. This phenomenon is referred to as ice accretion. When an ice layer forms on an aircraft wing, the ice adversely affects the performance of the aircraft by increasing drag and reducing lift and may cause a crash. In a jet engine, ice accretion disturbs the inlet flow and can lead to severe performance degradation. Thus, ice accretion phenomena are a serious problem in aircraft operations. In order to overcome the problems associated with icing, a number of major research institutes around the world, including NASA (National Aeronautics and Space Administration) and ONERA (Office National d'Etudes et de Recherches Aerospatiales) have been investigating ice accretion phenomena both experimentally and computationally [1-4]. The icing phenomenon is caused by complicated interactions under various physical conditions, including the ambient temperature, the LWC (liquid water content), the temperature of the surface with which the droplets collide, the impingement position, the surface roughness, and the mass of the droplets. Generally, it is difficult to reproduce these complicated icing conditions experimentally. In addition, icing tests are costly. Therefore, CFD (computational fluid dynamics) is expected to provide a useful method by which to predict ice accretion phenomena, because of the rapid development of computers in recent years. Various icing simulations can be conducted through two-dimensional simulations using a NACA (National Advisory Committee for Aeronautics) airfoil for three-dimensional simulations of a jet engine [5-8]. Icing of the main wing and the tail wing have been investigated for more than 60 years, and de/anti-ice systems, such as de-icer boots and bleed air systems, have been proposed and are widely used in existing aircraft. However, practical de/anti-icing systems for jet engine icing have not yet been established. In a jet engine, the main icing components are the fan blade, the FEGV (fan exit guide vane), the nose cone, the splitter, and the low-pressure compressor. In addition, there have recently been instances of icing on

high-pressure compressors, in which the temperature is approximately 30 °C [9]. Based on the above considerations, icing phenomena in a jet engine have been actively investigated in recent days [10]. Engine icing leads to ice shedding, which is a phenomenon whereby accreted ice is shed from the wall surface. After ice shedding occurs, the jet engine injects the shed ice pieces, which damage the internal components, including the compressor blade and the casing. This phenomenon is very complicated because there are several unknown physical properties, such as the ice density, the adhesion force between accreted ice and the wall, and the contact force between ice pieces. Papadakis et al. [11] measured the aerodynamics forces and moments acting on a fragment similar to a piece of ice at a wind tunnel facility and simulated the trajectory of the shed fragment based on the data obtained from the wind tunnel test. Baruzzi et al. [12] considered a fluid-structure interaction on cube ice and solved the flying cube movement. However, they considered only shed ice pieces and ignored ice growth and factors affecting ice shedding. Existing icing models can reproduce only ice growth and do not include ice shedding. Therefore, existing icing models have a problem in that accreted ice is assumed to grow continuously with time. In the present study, we develop a new icing model that can consistently reproduce the flow field around the computational target, super-cooled droplets impinging on the walls, ice growth, and ice shedding. The proposed icing model, which includes ice shedding, is validated though comparison with experimental data for flow rate change due to ice growth and ice shedding, as measured by Murooka et al. [13].

2. Numerical Procedures The icing simulation code used herein includes iterative computations for the fluid motion, the droplet trajectory, the ice growth, the ice shedding and the grid modification. Each computation is described below in detail.


2.1 Flow Field The flow field is assumed to be three-dimensional, compressible, and turbulent. The governing equations are the Favre-averaged continuity, Navier-Stokes, and energy equations. Since a rotational frame of reference is used, the Coriolis force and centrifugal force are added as body forces. The Kato-Launder k-turbulence model (Kato and Launder [14]) is used to estimate turbulence. The governing equations are discretized using a second-order upwind TVD (total variation diminishing) scheme (Yee [15]) for the inviscid terms, a second-order central difference scheme for the viscous terms, and an LU-ADI (lower upper-alternating direction implicit) scheme (Fujii and Obayashi [16]) for the time integration. 2.2 Droplet Trajectory Droplet trajectory is computed based on a Lagrangian approach in order to obtain the local water distribution on a blade. The computation uses the following assumptions: (1) Droplets are spherical; (2) Droplets are sufficiently small and thus do not break up; (3) The forces acting on a droplet are drag, centrifugal force, and the Coriolis force; (4) Droplets do not interact with each other; (5) Droplets do not affect the flow field (one-way coupling); (6) The initial droplet velocity is equal to the gas velocity at the release point. The equation of the droplet motion is:

  1   dU d 3 U U  CD f dt 4 d dd r r      　　　  2  U d      rd





47

where, Red is the Reynolds number of a droplet based on the diameter and the relative velocity between the gas and the droplet. The droplet trajectory is computed in order to obtain the local water distribution, i.e., the impingement distribution of super-cooled droplets on a unit area at each second. The mass flow rate of the inflow droplet in the control volume Qin is as: Qin  AinU in LWC

(3)

The number of inflow droplets per second Nin is computed as: Q N in  in (4)  d Vd The local ratio of the number of impinging droplets to the total number of droplets Pim and the local water distribution  are expressed, respectively, as: N Pim  im (5) N com N (6)   Pim in Acell The local water distribution is a key parameter in icing simulation. 2.3 Thermodynamics The thermodynamic computation is carried out using the Messinger model proposed by Messinger [17]. Fig. 1 shows an image of the Messinger model, which is based on the mass and the energy balances in a control volume. The governing equations are: mim  min  mac  me,s  mout

(7)

Eim  Ein  Eair  E fri  Eac  Ee,s  Eout  Econ (8)



(1)

The second term on the right-hand-side represents the centrifugal force and the Coriolis force. The drag coefficient CD is expressed as: 24 (2) CD  1  0.15 R e d 0.687  R ed

The freezing rate f and the runback-out mass mout can be derived from Eqs. (7) and (8) as: f 

m ac m im  m in

m out  1  f mim  min   m e, s

(9) (10)

If f = 1, all of the mass in the control volume accretes,


48

Runback-in

Ice

Runback-out

Blade

Water

Ice Wall

Fig. 1

Schematic diagram of the Messinger model.

(a) Previous method

whereas, if f = 0, all of the mass in the control volume runs back to the next cell. 2.4 Grid System to Reproduce the Ice Layer In icing simulations, computational grid regeneration is required with the surface shape change due to accreted ice. In the previous simulation method [8], we regenerated the computational grid along the ice surface. This requires practical experience and know-how when the ice geometry is complicated. In the present study, the exposure time is long enough for the ice shedding phenomenon to occur. In addition, the ice geometry after ice shedding is very complicated. Thus, the grid regeneration used in the previous method is too difficult. In order to overcome this problem, we adopt the icing cell method (see Fig. 2), which is a new method for simulating the ice shedding phenomenon. In this method, the computational cells are classified as either a fluid cell or an ice cell. If accreted ice grows over a fluid cell, the cell is treated as an ice cell, and the ice cell is retreated as a fluid cell after ice shedding occurs. Therefore, unlike in the previous method, the icing cell method does not need to regenerate the computational grid along the accreted ice shape. Thus, we can reproduce both the ice growth and the ice shedding. 2.5 Ice Shedding Judgment In turbomachinery, such as a fan rotor, ice shedding occurs when the centrifugal force increases due to the growth of thick ice. Forces acting on ice accreted on the

Ice Blade

(b) Icing cell method Fig. 2

Grid systems for reproducing the ice layer.

rotor surface include the adhesion force between the accreted ice and the wall, the centrifugal force, the contact force between the ice pieces, the fluid drag, and the wind shear stress. Since the fluid drag and the wind shear stress are much smaller than the other terms (approximately three orders of magnitude), these terms can be ignored. Simulation of the ice contact force is very difficult, because there is poor understanding of this force. Therefore, we simplify the ice-shedding phenomenon and treat both the ice adhesion force and the centrifugal force as the force acting on the accreted ice. We attempt to reproduce the ice shedding phenomenon under this simplified assumption. The centrifugal force and the adhesion force acting on accreted ice are expressed, respectively, as:

Fc  i Acell Bi r2

(11)

Fa  Acell

(12)


49

Table 1 Computational conditions for validation. Airfoil type Chord length Angle of attack Median volume diameter Liquid water content Exposure time Free stream velocity Static pressure Static temperature

Unit (m) (deg.) (m) (g/m3) (s) (m/s) (kPa) (C)

NACA0012 0.53 4.0 20 1.30 480 58.1 95.61 -27.8

In Eq. (12), the adhesion stress is obtained from an ice adhesion test conducted by Murooka et al. [13] The wall material is aluminum. By use of Eqs. (11) and (12), we computed the forces acting on the accreted ice on each blade surface cell. If the centrifugal force exceeds the adhesion force, we judge ice shedding to have occurred and treat the cell as a shedding cell (i.e., fluid cell). Using the rotorcraft, Brouwers et al. [18] experimentally indicated that ice shedding tends to occur from the tip side. Therefore, if a cell is judged to be a shedding cell, we assume that all cells from the position of the cell to the tip of the rotor are shed simultaneously. Once ice shedding occurs, we repeatedly compute the flow field, the droplet trajectory, and the thermodynamics.

3. Validation of the Proposed Method 3.1 Validation Conditions We validated the proposed icing method (i.e., the icing cell method) using a NACA0012 airfoil because there are significant experimental data on this airfoil in the literature. In the present study, the overset grid method is used to clarify the icing area around the leading edge, where the icing phenomenon frequently occurs. The computational grid system is shown in Fig. 3. The icing cell method appeared to be sensitive to the sub-grid resolution. Therefore, we investigated three sub-grids having different grid resolutions: a coarse grid with a resolution of 181 × 41, a medium grid with a resolution of 241 × 71, and a fine grid with a resolution of 301 × 101. The total numbers of grid points included in the main grid for these resolutions are 23,112,

Fig. 3 Computational grids used in the validation (red: main grid; blue: sub-grid).

32,802, and 46,092, respectively. The computational domain is 20 chord × 20 chord. The validation conditions are listed in Table 1. These conditions are for the rime ice, which occurs at the very cold atmospheric temperature. 3.2 Validation Results The computational results predicted by the present method were compared to the results by our previous method, the computational results by NASA, and the experimental data measured by NASA [1]. The ice shape near the leading edge is shown in Figs. 4a-4c for the coarse, medium, and fine grids, respectively. The ice shape and the icing area obtained using the proposed method are approximately the same as the others. In addition, in the proposed method, the grid resolution appears to have little effect on the icing simulation. These results confirm that icing simulation using the icing cell method is reasonable.

4. Computational Conditions 4.1. Computational Target and Grids The computational target is a commercial axial blower (Showa Denki Co. Ltd., Kairyu series A2D6H-411) because there is no experimental data of the ice shedding in a jet engine. This axial fan was used in the experimental study (Murooka et al. [13], see Fig. 5). This axial fan has 12 rotor blades, although, for simplicity, only one rotor blade is simulated, under the assumption of geometrical periodicity. The computational grid, which is based on an overset grid method, is shown in Fig. 6a and is the main grid for the


NASA experiment NASA simulation Previous method Present method

0.1

y/chord

y/chord

0.1

0

−0.1


0.1

y/chord

50

0

−0.1 0

0.1


0

−0.1

0.2

0

x/chord (a) Coarse grid

0.1

0.2

x/chord (b) Medium grid

Fig. 4

Comparison of icing shapes used in the validation.

Fig. 5

Computational target.

0

0.1

0.2

x/chord (c) Fine grid

(a) Overview of the computational grid

Table 2 Computational conditions. Rotating speed Mass slow rate Static temperature LWC MVD Exposure time

(rpm) (kg/s) (K) (g/m3) (m) (s

1,800 5.21 268.15 1.54 30 480.0

passage. Fig. 6b shows the sub-grid around the blade. Fine cells are set around the leading edge and the pressure side of this sub-grid, because the ice layer is easily formed there. The total number of grid points is 6,272,284. 4.2 Computational Condition The computational conditions used in the present study are listed in Table 2. The droplet trajectory simulation is conducted for 500,000 droplets that were

(b) Enlarged view of the sub-grid Fig. 6

Computational grids.

randomly distributed at the upstream boundary. The initial droplet velocity is equal to the local gas velocity, which is obtained from the flow field computation. The total pressure, total temperature, and flow angle are imposed, and the Mach number is extrapolated at the


upstream inflow boundary. The inflow turbulent kinetic energy is assumed based on 0.1% turbulence of the free stream. Adiabatic, no-slip conditions and the wall function are prescribed for the stationary walls and the rotating surfaces. The exit static pressure is specified.

droplets impinge on the pressure side, particularly around the leading edge. Ice layers are expected to form in such areas because the surface temperature is below the freezing point. 5.2 Ice Growth and Ice Shedding

.

L.E

(a) 10 s

(b) 20 s

(c) 30 s

(d) 40 s

(e) 50 s

(f) 60 s

(g) 70 s

(h) 80 s

(i) 90 s

(j) 100 s

(k) 110 s

(l) 120 s

(m) 130 s

(n) 140 s

(o) 150 s

(p) 160 s

T.E.

L.E

L.E.

The static temperature and the local water distribution on the blade surface are shown in Figs. 6 and 7, respectively. “L.E.” means the leading edge, and “T.E.” means the trailing edge. These two parameters are of great importance in icing simulation. The surface temperature is below the freezing point except at the blade tip on the pressure side, as shown in Fig. 6. Most

.

5.1 Static Temperature and Local Water Distribution

T.E

Fig. 8 shows the ice growth and ice shedding on the

5. Results and Discussion

T.E.

51

273.15

267.45 (K) (a) Pressure side

1.5 × 10

T.E.

L.E.

L.E.

Static temperature.

T.E.

Fig. 7

(b) Suction side

9

10.0

0.0 (m-2·s-1) (a) Pressure side Fig. 8

Local water distribution.

(b) Suction side

0.0 (mm) (q) 170 s (r) 171 s Fig. 9 Temporal evolution of the ice layer on the pressure side.


52

Normalized volume flow rate

1

0.9

0.8

0.7

Experiment Simulation

0

200

400

Time (s) Fig. 10 Change of volume flow rate.

pressure side. The ice accretion is concentrated on the leading edge, at which the temperature falls to the freezing point and the local water distribution is sufficiently high, as mentioned above. Ice shedding occurs 171 seconds after ice growth begins. The position at which ice shedding occurs is near the leading edge at 71% of span, where a relatively strong centrifugal force acts on the ice layer and thus thick ice accretes. This tendency of ice shedding from the blade tip is similar to the experimental results reported by Brouwers et al. [18]. Finally, the flow rate change due to the ice growth and the ice shedding is shown in Fig. 9. The volume flow rate in Fig. 10 is normalized by the initial volume flow rate of the experiment. In the experiment, ice shedding occurred 179 seconds after the start of the test. The predicted time of ice shedding occurrence was 171 seconds. Thus, the estimation of the ice shedding obtained using the proposed method is reasonable. However, the recovery rate is considerably lower than that of the experiment. In the experimental study, the decrease in the flow rate due to ice growth was 27.6% before the first ice shedding, and the recovery was 21.8% after ice shedding. On the other hand, the flow rate decrease in the present study was 34.4%, and the flow rate recovery was 10.6%. This is thought to be due

to the difference in the ice shedding volume. In the experiment, the flow rate after ice shedding was 94.3% of the initial flow rate, which means that larger ice pieces were shed from the rotor blade. Moreover, we simulated only the initial ice shedding. In the experiment, the volume flow rate recovers gradually after the initial ice shedding occurs, because ice shedding occurs more than once. Therefore, we need to run the simulation for a longer time until ice shedding occurs several times. The present study confirmed that the proposed ice shedding model has high predictive performance for the ice shedding time. However, the model still includes a large error for the ice shedding volume.

6. Conclusions We have developed a new icing simulation model that can reproduce both the ice growth and the ice shedding phenomena and validated the proposed model by comparing the obtained flow rate change due to ice growth and ice shedding to experimental data for an axial fan. The main results of the present study are as follows: (1) The proposed ice shedding model has high predictive performance for the occurrence time of ice shedding. (2) The proposed ice shedding model has an error for the flow rate recovery due to ice shedding because the volume of the piece of ice in the model was lower than that shed in the experiment. In the future, we intend to improve the present ice shedding model for accurately predicting the ice shedding volume. In addition, we intend to add an ice-piece tracking model. Finally, we intend to apply the proposed icing simulation model to a rotor-stator interaction field in a jet engine.

Acknowledgments The ice adhesion force and the flow rate change due to icing were obtained experimentally by T. Murooka (IHI Corporation), S. Shishido (IHI Corporation), R.


Hiramoto (Hokkaido Institute of Technology), and T. Minoya (Hokkaido Institute of Technology). The present study was supported in part by JSPS KAKENHI Grant Number 25286101.

References [1]

[2]

[3]

[4]

[5]

[6] [7]

[8]

Wright, W. B., Gent, P. W., and Gufford, D. 1997. DRA/NASA/ONERA Collaboration on Icing Research. NASA Contractor report. Veres, J. P., Jorgenson, P. C. E., and Wright, W. B. 2011. Modeling the Effects of Ice Accretion on the Low-Pressure Compressor and the Overall Turbofan Engine System Performance. Technical report. Nilamdeen, S., and Habashi, W. G. 2009. “FENSAP-ICE: Modeling of Water Droplets and Ice Crystals.” In Proceeding of the 1st AIAA Atmospheric and Space Environments Conference, 600-10. Presteau, X., Montreuil, E., Chazottes, A., and Vancassel, X. 2009. “Experimental and Numerical Study of Scallop Ice on Swept Cylinder.” In Proceedings of the 1st AIAA Atmospheric and Space Environments Conference, 563-78. Ozgen, S., and Canıbek, M. 2009. “Ice Accretion Simulation on Multi-element Airfoils Using Extended Messinger Model.” Heat and Mass Transfer 45 (3): 305-22. Hospers, J., and Hoeijmakers, H. 2011. Numerical Simulation of SLD Ice Accretions. SAE Technical Paper. Aliaga, C. N., Aubé, M. S., Baruzzi, G. S., and Habashi, W. G. 2011. “FENSAP-ICE-Unsteady: Unified In-flight Icing Simulation Methodology for Aircraft, Rotorcraft, and Jet Engines.” Journal of Aircraft 48 (1): 119-26. Hayashi, R., Kawakami, K., Suzuki, M., Yamamoto, M., Shishido, S., Murooka, T., and Miyagaw, H. 2011. “Numerical Simulation of Icing Phenomena in Fan Rotor-Stator Interaction Field.” In Proceedings of the 11th

[9]

[10]

[11]

[12]

[13]

[14]

[15] [16]

[17]

[18]

53

International Gas Turbine Congress, 1-5. Jeanne, G., Mason, J., Strap, W., and Chow, P. 2006. “The Ice Particle Threat to Engines in Flight.” In Proceedings of the 44th AIAA Aerospace Sciences Meeting and Exhibit, 2445-65. Veillard, X., and Habashi, W. G. 2011. “Icing Simulation in Multistage Jet Engine.” Journal of Propulsion and Power 27 (6): 1231-7. Papadakis, M., Yeong, H., and Suares, I. G. 2007. “Simulation of Ice Shedding from a Business Jet Aircraft.” In Proceedings of the 45th AIAA Aerospace Sciences Meeting and Exhibit, 1-25. Baruzzi, G., Lagace, P., Aube, M., and Habashi, W. G. 2007. “Development of a Shed-Ice Trajectory Simulation in FENSAP-ICE.” In Proceedings of SAE International Conference on Aircraft and Engine Icing, 3360-1. Murooka, T., Shishido, S., Hiramoto, R., and Minoya, T. 2011. “Surface Coating Effect on Protection of Icing for Axial Fan Blade.” In Proceedings of SAE International Conference on Aircraft and Engine Icing, 1-6. Kato, M., and Launder, B. E. 1993. “The Modeling of Turbulent Flow around Stationary and Vibrating Square Cylinder.” In Proceedings of the 8th Symposium on Turbulent Shear Flows, 1-6. Yee, H. C. 1987. Upwind and Symmetric Shock-Capturing Schemes. Technical report. Fujii, K., and Obayashi, S. 1987. “Practical Application of Improved LU-ADI Scheme for the Three-Dimensional Navier-Stokes Computations of Transonic Viscous Flows.” AIAA Journal 25: 369-70. Messinger, B. L. 1953. “Equilibrium Temperature of an Unheated Icing Surface as a Function of Airspeed.” Journal of the Aeronautical Sciences 20 (1): 29-42. Brouwers, E. W., Palacios, J. L., Smith, E. C., and Peterson, A. A. 2010. “The Experimental Investigation of a Rotor Hover Icing Model with Shedding.” In Proceedings of American Helicopter Society the 66th Annual Forum, 1-17.

D


DAVID

PUBLISHING

The Shift from “Grid-Tie” to Partly “Off-Grid” Balint David Olaszi and Jozsef Ladanyi Department of Electric Power Engineering, Budapest University of Technology and Economics, Budapest H-1111, Hungary Received: May 05, 2014 / Accepted: June 18, 2014 / Published: January 31, 2015. Abstract: This paper deals with that basic economy grounds which support to shift from “grid-tie” to partly “off-grid” PV (photovoltaic) systems. By using our own written VBA (visual basic) Simulation, we can predict the size of energy storage capacity and support our basic assumption. Key words: Partly off-grid PV system, energy storage simulation.

It is well-known that the underestimated household PV (photovoltaic) systems changed the way of the European electricity generation. The question is that how we can regulate and plan the effects of the upcoming boom in the case of the partly off-grid systems. The fundamental reason behind the ongoing shift from “grid-tie” to “off-grid” PV systems is that basic economics, which is mostly powered by the continuously widening gap between the PV feed-in prices and the household electricity prices [1, 2]. In this paper, the term “off-grid” means that the PV systems are only virtual in off-gird work, because these are connected to the low voltage network (so-called “the grid”), therefore, the only sense of the partly connected off-grid system is the minimization of the energy consumption from the grid, which leads to saving.

2. Economic Prediction Germany is one of the leading countries in PV systems installation and it has a really high price of household electricity and a fast decreasing PV feed-in tariff which is a good combination for such a system. Corresponding author: Balint David Olaszi, Ph.D. student, research fields: distributed energy storage and electric engineering. E-mail: [email protected].

Therefore, it is chosen to show the following prediction. Today, the gap between the household electricity price and feed-in tariff price is approximately 15 Euro cents. Using regression analysis for the German household electricity price and a yearly 7.5% decreasing in feed-in tariff, the gap is likely to be widened to 24% by 2020 in a moderate scenario (Fig. 1). This gap between the PV feed-in tariff and the household electricity price is significantly narrower in other European states; however, the trend is the same as that in Germany. Although, they have not reached the tipping point yet or if they did, then the gap is too narrow.

Cent/kWh

1. Introduction

60

60%

50

55%

40

50%

30

45%

20

40%

10

35%

0 2002

30% 2007

2012 2017 Year Fig. 1 German (dots) household and (triangle) PV feed in tariff prices, and part of taxes and subsidizations in the German electricity price (square).

The Shift from “Grid-Tie” to Partly “Off-Grid”

3. The Structure of the Energy Storage Simulation

55

where, Tc: PV cell temperature (oC); Ta: Ambient temperature (oC); TNOCT: PV cell nominal operating temperature (oC); I(t): Irradiation in the function of time (W/m2). The cell temperature is also important which has a strong effect on the efficiency of the PV modules as well as on the inverter [4]. To create a real temperature distribution, we have chosen the double cosine model which is calculated by using the mean temperature of a given day and the exact hour of the hottest and coldest hour [5]. In this simulation, the weather modification parameter is constant, because in such a short term, this index cannot be predicted. Additionally, there is an option to select the suitable type of the PV module, inverter and set the right number of the modules. The program chooses the selected PV module matrix which is the function of the solar irradiation and the cell temperature, the inverters matrix is the function of the DC (direct current) power and the outside temperature. The DC power of the PV modules is corrected averagely by 5% wiring and dust losses [6].

As is well-known, the determination of the adequate size of the electricity storage is going to be more important in the future to support the shift from “grid-tie” to the partly “off-grid” PV electricity generation. It leads us to the conclusion that there will be a need for such an adequate PV electricity storage optimization method program. In our work, we created a macro program based on Visual Basic, which helps us to set a wide range of weather conditions and custom parameters to get as suitable prediction for irradiation as possible. One of them is the albedo which can significantly influences the irradiation level on the PV array. The albedo represents the reflection of solar radiation on the nearby objects and it varies between urban and rural areas a lot. The typical urban albedo value is approximately 0.16. Furthermore, the temperature distribution plays another important role in the simulation by affecting the PV cell temperature which changes the efficiency of the PV modules. The temperature of the PV cell can be calculated with the Eq. (1) [3].

4. Simulation Results .

As a matter of fact, the first thing that should be

(1)

/

Irradiation calculation PV module matrix

DC/AC converter matrix

PV cell temperature calculation PV system AC power Weather conditions PV system parameters

Storable energy Dem and profil

Fig. 2

Process diagram of the simulation.


56

defined is that in what season the optimization should

1.00

be run. Obviously, if the simulation took place in summer, then the energy storage would be oversized; and in winter, it would be undersized. It leads us to the conclusion that the optimal size of the energy storage capacity should be defined in spring or in autumn, because it covers the most part of the year in case of

Power (kW)

0.50 0.00 -0.5000:00

04:48

09:36

00:00

-1.00 -1.50 -2.50

Time

Fig. 3 Daily energy consumption/production of 3.34 kWp PV system in Budapest on March 21.

PV cell temperature (oC)

50

We have done a simulation on a 3.34 kWp PV system in Budapest on March 21. The daily production (Fig. 3) and the temperature of the PV cell (Fig. 4) were registered during the simulation. The user interface of the simulation program is shown in Fig. 5. The adjustable parameters are the

40 30 20 10 0 04:48

09:36

Time

Fig. 4 PV cell temperature.

followings:

1.14

0.16

47.36

0.34

17.31

1,009

0.35 19.19 0.5 11.93

-13.86

13.12 Simulation parameters and results.

19:12

-2.00

storable PV energy. To create the summary of the daily energy consumption, we have to assume an average daily consumption. In this scenario, we used an average daily energy demand chart of a yearly 4,000 kWh consuming household [7].

Fig. 5

14:24

6.59

14:24

19:12


 Nday: on what day the simulation will run;  Temp. PV module_ NOCT: it is the nominal operating in PV cell temperature in °C;  Cloud index: value one represents no power decrease and value five represents 80% power decrease caused by clouding;  Alpha: Alpha is a parameter in the Angstrom’s turbidity formula which is the function of the aerosol size. For most natural atmosphere, Alpha is between 0.8 and 1.8;  Albedo: it shows that how much of the incoming irradiation the nearby objects reflect;  O3: ozone absorbs ultra viola radiation, therefore, it has a significant effect on the irradiation, especially in cities where there are high smog levels;  AOD: Aerosol Optical Depth stands for the air practical content;  The inverter and the PV module can be chosen, the program will load the chosen matrix;  Demand profile can be adjusted according to how much energy the consumer uses. According to the simulation results, the needed capacity of the energy storage would be 13.12 kWh and it stores (13.12/(0.95 × 0.85 × 0.8)) 8.47 kWh of “AC (alternating current)” energy, but it is still higher than the needed storable energy which covers the consumption (6.59 kWh) of the chosen day (March 21) [8]. (It is important to know that the average lead-acid battery energy efficiency is 85% for charging/discharging and the average DC/AC conversion efficiency is 95%, also taking into consideration the 80% DoD (depth of discharge) level.) Taking into consideration all of the mentioned simulation results, then a good guess would be 14 kWh to the size of the energy storage capacity.

5. Payback Time Calculation As we mentioned, the partly “off-grid” shift is only powered by the price difference between the PV feed-in tariff and the household electricity price. To

57

support our basic assumption, there will be a simple payback time calculation. The average retail price of one kWh deep cycle lead-acid battery cost is around 125 Euro and its life cycle life is around 1,000 cycles at 80% DoD [8]. It means that in the whole lifetime of the partly “off-grid” system there will be need for two more battery swaps. (taking into consideration the estimated battery capacity which has 30% over capacity.) The average retail price of a 3 kW battery charger is 2,000 Euro. The yearly income is calculated by the number of days and the saved “AC” energy multiplied by the gap between the feed-in tariff and the household electricity price. As the matter of fact, the payback time calculation takes into account a 2% of inflation rate which is used to calculate the NPV (net present value) price of batteries. Furthermore, the author made a prediction to widening price gap (5%) based on their former economic analysis. The calculation shows that the payback time would be around 15-16 years (Table 1)), but if everything goes as predicted, then, the payback time might decrease to around ten years until 2020. It is important to know that actual PV feed-in tariff price was used in the payback time calculation. However, if there is a slightly major bigger PV system (10-40 kWp) [2], then only 90% of the PV energy is paid in the feed-in system (Table 2) and 10% Table 1 Payback time calculation in case of PV feed-in tariff is available. 14 kWh battery NPV 1st battery swap

Cost (Euro) 1,750 1,650

NPV 2nd battery swap

1,550

Battery charger 3 kW Installation Total cost Yearly

2,000 250 7,200 Income (Euro) 360

Inflation

2%

Yearly price gap widening

5%

Payback time (years)

15.75

58


Table 2 Payback time calculation in case of PV feed-in tariff is not available. Total cost Yearly Inflation Yearly price gap widening Payback time (years)

Cost (Euro) 7,200 Income (Euro) 695 2% 5% 8.24

of the PV energy generation could be stored optionally. It leads to the fact that in such a case, the payback time is shorter. When there is no 100% feed-in opportunity for a PV system, then a partly “off-grid” system is good investment.

the battery capacity reaches a certain level of SoC (state of charge), but to be the precious, it should be the function of the battery capacity. Therefore, the morning peak load could decrease without having any effects on the consumer habits, or the rent ability of the investment.

References [1]

[2]

[3]

6. Conclusion As a conclusion, the European regulators must act fast to create a proper regulations to the partly “off-grid” battery chargers before they spread all over Europe. If the payback time is fewer than ten years, then, it could be a huge jump in the investment volume. The perfect regulation in the case of the “off-grid” charger controller would be that the charger completely satisfies the night peak and the morning peak load as well, but if it is not set properly, then, it will release all of the charge at night when the grid load is low. Furthermore, the battery charger must qualify for all of the IEC (International Electrotechnical Commission) standards. Indeed, we think that there should be further studies about the optimal discharging profile. We would suggest that the discharge at night should stop, when

[4]

[5]

[6]

[7]

[8]

Photovoltaik guide Journal “PV-Feed in Tariff Prices.” Accessed August 11, 2014. http://www. photovoltaikguide.de/. Eurostat. “Eurostat Harmonisierter Verbraucherpreisindex.” Accessed August 11, 2014. http://ec.europa.eu/eurostat/de/web/products-data-in-focu s/-/KS-QA-14-004.htm. Fesharaki, V. J., Dehghani, M., and Fesharaki, J. J. 2011. “The Effect of Temperature on Photovoltaic Cell Efficiency.” Presented at 2011 International Conference on Emerging Trends in Energy Conservation, Tehran, Iran. Taha, H. 1997. “Urban Climates and Heat Islands: Albedo, Evapotranspiration, and Anthropogenic Heat.” Energy and Buildings 25 (2): 99-103. Bilbao, J., Miguel, A. H., and Kambezdis, H. D. 2001. “Air Temperature Model Evaluation in the North Mediterranean Belt Area.” Journal of Applied Meteorology 41 (8): 871-84. Sulaiman, S. A., Hussain, H. H., Leh, N. S. H. N., and Razali, M. S. I. 2011. “Effects of Dust on the Performance of PV Panels.” World Academy of Science, Engineering and Technology 58: 558-93. Veldman, E., Gibescu, M., Slootweg, H. J., and Kling, W. L. 2013. “Scenario-Based Modeling of Future Residential Electricity Demands and Assessing Their Impact on Distribution Grid.” Energy Policy 56: 233-47. Christiana Honsberg, Stuart Bowden, PV Education “Battery Technologies.” Accessed March 10, 2013. http://pveducation.org/pvcdrom/batteries/remaining.

D

Journal of Energy and Power Engineering 9 (2015) 59-67 doi:10.17265/1934-8975/2015.01.007

DAVID

PUBLISHING

Stability Monitoring and Control of Generation Based on the Synchronized Measurements in Nodes of Its Connection Alexander Fishov1, Maria Shiller2, Anton Dekhterev2 and Vladimir Fishov2 1. Automated Electrical Power Systems Department, NSTU (Novosibirsk State Technical University), Novosibirsk 630073, Russia 2. Novosibirsk Regional Dispatching Office of System Operator of the United Power System, JSC (Novosibirsk RDO SO UPS, JSC), Novosibirsk 630007, Russia Received: October 04, 2014 / Accepted: November 13, 2014 / Published: January 31, 2015. Abstract: Synchronized distributed measurements of mode parameters create a technical feasibility for development and implementing new technologies of control the mode stability and the admissibility of EPS (electric power system) mode. Discussion will focus on different models obtained from data synchronized measurements for operational and automatic emergency control without EPS being totally controlled. According to the proposed technology, the generator’s output power restrictions are determined in real-time by the terms a static stability using the generators’ mode model as a multipole with connection nodes of generators’ electromotive forces (the matrix of SMA (self and mutual admittances) of electromotive forces of generators). Potential applications of the technology are distribution network with the main substation and generators of commensurable capacity, and transmission network with large power plants (generators) distributed into the network. The one-level control system for all of generators with defining the generator’s power limits relative to the main substation is implemented in the first case. In the second case, the two-level control system is brought in, based on the separation of large and small generation motion. The results of the method and technology efficiency verification are shown in the paper, by both computer simulations of the power system modes and its physical model. Key words: Power system, static stability, synchronized measurements, normal mode, transient mode, post emergency mode, matrix of self and mutual admittances.

1. Introduction 

The control of the mode stability during the operation of EPS (electric power system) is essential in the systems of technological and emergency control schemes, operating control for providing stability of the parallel operation of generators either in normal, post emergency mode or in relatively short quasi steady mode, which happens after the immediate attenuation of electromechanical transients. The currently applied methods of control are based on the usage of mathematical models of the electric Corresponding author: Alexander Fishov, professor, research fields: power system stability, power system emergency control, and automated control system. E-mail: [email protected].

transmission system, which are based on the grid topology and parameters of all of its elements [1]. The integration of the distributed generation into electric transmission systems with large power plants in the area of distribution grid increases the complexity of the problem and considerably complicates a required control system. Furthermore, the efficiency of its usage decreases. This presents the technological barrier to the development of distributed generation and offers an opportunity for new technology to control the stability, which would enable more preferable conditions for integration of distributed generation in the current EPS and grid. Synchronized distributed measurements of modes’ parameters create a technical feasibility for

Stability Monitoring and Control of Generation Based on the Synchronized Measurements in Nodes of Its Connection

60

development and implementing new technologies of control the mode stability and the acceptability of EPS mode which are able to solve the problem. These measurements enable to obtain models for taking into consideration limits towards stability of EPS mode to the operating and automatic emergency control which suit current mode of the grid but without being totally controlled. Such feasibility is explored in previous works [2-4] as well as in the current works [5-7].

where, i and j are indexes which show number of generators in the scheme. is inside power of i and are elements of matrix generator. ,

2. Theoretical Basis

Synchronized measurements of active and reactive power, vector measurements of voltage in the nodes of power connection allow for each of the time periods (power modes) calculate matrix and . Rewrite Eq. (3) for n-1 generator for one mode and we have more generally undefined system of linear independent equations relative to SMA vector of

2.1 Model of Generators Mode and Control of Some Network Parameters The proposed technology of controlling the limits for generators to distribute the capacity by the condition of a static stability in real time is based on the model of mode of generators in terms of multi-port circuit nodes of connection of their electromotive forces (the matrix of SMA (self and mutual admittances) of generator electromotive forces of generators). In initial condition, model does not contain another node of commonly used equivalent circuit of EPS besides nodes of generator EMF. This means that loads of the systems are shown as linear bridges. Counting or miscounting voltage regulation depends on equivalent circuit being used and the way of describing changes of the EMFs. Let us show electrical mode of EPS generators in the current time period as passive linear multi-port with EMF connected to the nodes. Mode of n generators is described by system of equations:

S  diag  E  Y*  E*

and

.

Eq. (2) can be written as:

Si  Ei  ( E* )т  y(i)* where,

(3)

is column matrix (vector) SMA of i

generator which conforms to i line of matrix

.

generators, unit vector y :

S  A  y* where,

(4)

is right-angled matrix, where coefficients are

producing EMF with index coincide with index of conductivity, though in every line, element with distinct first index of the line are cleared. For making system of Eq. (4) defined (with square matrix ) or redefined, it should be complimented using measurements for different time periods of transitional mode together with redistribution power between generators. In the result of solving defined or redefined system of Eq. (4), there are parameters of multi-port (SMA of generator EMFs.) With necessity of control inside parameters of the

(1)

grid (voltage in some nodes, power flow in the dedicated

where, is matrix (vector ) of inside active and reactive powers of generators on the nodes of EMF. is square matrix of SMA of generator EMFs. is

synchronized measurements in the appropriate nodes

matrix of vectors of generators EMFs. Mode of each of generators is described as: n

Si  Ei  Y E j 1

* ij

* j

(i  1, 2,..., n)

section), it should be installed additional units of of the grid and power lines. By the results of measurement, identification

of complex coupling

coefficients with EMF of generators and then defining needed controlled parameters in quasi steady and post (2)

emergency modes are possible.


61

2.2 The Model Identification

2.3 Using the Model for Control

The necessary condition for actual SMA matrix identification (having system of Eq. (4) without degeneracy) is occurrence of steady state changes (redistribution of power between generators). It is possible by two ways: The first is by using sporadic disturbance of normal mode of EPS in the result of emergency disturbance. In this case, SMA matrix identifies in the time period of transient mode which is connected with attenuation of electromechanical transients and the application domain is control of acceptability quasi steady and steady PEM (post emergency mode). The second is by using artificial changes of mode, e.g., by using short time unloading of generator by active power. In that case, there can be provided the control of limits of acceptability of normal mode (limits of producing power by generators). The sufficient condition of actual SMA matrix identification is a representation of generators that characterized by in-phase rotors motion as the one equivalent generator, so far as in general case in complicated multi-machine EPS, a power redistribution between generators (or power plants) during disturbances or operations occurs between some of generators only. An absence of mode changes of residual generators does not allow identifying their mutual admittances due to a degeneracy of the system of equations. The obtained model in this case shows the structure of the mutual generators motion during the power redistribution between them. The overdetermination of Eq. (4) is required for stable results under a presence of the measuring error and method error conditioned by the substitution of a real object mode by its specified model. In Fig. 1, the mutual admittances identification results for the scheme containing three generating nodes and infinite buses, are presented. It can be seen from Fig. 1 that the measurements redundancy increase improves the stability of identification results of SMA matrix values.

The model obtained can be used in calculating the limits on the power output of each network generators for the steady state, quasi steady, and post emergency mode taking into account the work of excitation regulators. During the operating control of the output power of the generators, while operator is controlling the stability with the suggested technology, several actions should be performed. Firstly, the operator reduces the

Fig. 1 The equations set overdetermination impact on the stability of identification results of the SMA matrix during electromechanical transients after disturbance (identification “window” is 3-5 s): (а) rotor angle oscillations of generators; (b) mutual reactive admittance from SMA matrix subject to the fourfold overdetermination; and (с) mutual reactive admittance from SMA matrix subject to the tenfold overdetermination.


2.4 Advantages and Peculiarities of the Method The proposed method assumes the stability control in node coordinates (generators active power coordinates), that provides following advantages:

200 180 160 140 120 100 80 60 40 20 0

Active power (MW)

2,000 1,800 1,600 active power limit 1,400 1,200 1,000 800 600 current active power 400 rotor angle 200 0 (a) 6 reactive admittance (10-3 Ω) 5 4 3 2 1 active admittance (10-4 Ω) 0 -1 -2 0 6 7 1 2 3 4 5 (b) Time (s) Fig. 2 Generator’s power control with monitoring of restrictions on the output power and finishing with an intentional violation of the stability limit: (a) operator observed parameters; and (b) identified admittance (window 1-5.5 s). Mutual admittance

generator’s output power by some small quantity, which is sufficient for identification of SMA matrix during the redistribution of power in the system. After the procedure of identification and definition of the generator’s output power, restrictions were done, then, the desired change of the generation mode can be realized. Furthermore, during the process of changing, the control of the restrictions in real time continues (Fig. 2). If unexpected changes occur in normal network conditions, also the power output limits are defined to be used, if necessary, to prevent violations of stability automatically or manually. Specifically, in respect to control of turbine power to provide the required stability margin in post-emergency mode, it is can be changed the level of long-time turbine off-loading. This will improve the controllability of EPS and reduce redundancy control actions. The structure of restrictions control system for different purposes networks has the features: (1) Distribution network with supply center (node connection to the main network) and generators in different nodes. In this case, a single-level control system is implemented for all generators of the distribution network. Limit power of each generator are determined to direction of weighting generator-supply center. (2) Transmission network EPS containing large power stations (generators) and on small nodes distributed generation. In this case, two-level control system is implemented on the basis of separation of motions large and small generation. Parallel work stability of large generators is provided by traditional technology. Stability areas with distributed generation, according to the proposed technology under controlled conditions movement of large generators.

Rotor angle (degrees)

62

(1) The absence of a necessity of total equipping of EPS by measuring means. It is enough to install recorders at the power plants buses only. (2) Network structure independence. As far as an association with specified network sections is not required. (3) Informativity. Because the structure of the identified SMA matrix shows the structure of the generators rotor motion including information about an existence of in-phase moving generator groups which stability is under the threat in current network conditions. It allows choosing the vector of regime loading appropriated for current conditions when post emergency stability limits are calculated. (4) Universality. Because stability margin control is possible as for a transmission grid so for electrical networks with distributed generation. (5) Availability in the turbine power control loop.

3. The Result of Method and Technology Efficiency Verification 3.1 Verification via Simulation Modeling Comparative calculations using the SCS (software


63

and computing system) “Mustang” held for Surgut EPS of United Power System of Russia, a complete circuit is shown in Fig. 3, and the equivalent (based on an actual matrix SMA EMF equivalent generators) is shown in Figs. 4a and 4b. Parameters of the initial mode of the relevant matrix SMA EMF generators used for subsequent weighting are shown in Table 1. Matrix SMA identified on the calculation of the transient process after a disturbance EPS mode. In the calculations with weighting as the infinite buses was selected the large condensing plant (named Reftinskaya GRES) and consistently each of the five generating nodes of the circuit was weighted. Results of calculations limiting on the stability condition active power generating units from the matrix SMA EMF equivalent generators and complete calculations, the equivalent circuit model of

Fig. 3 Scheme 500 kV of the Surgut EPS.

ReftGRES

Tyumen

NVGRES

SGRES-1-220

SGRES-2

SGRES1-500

the Surgut power unit controlled EPS are shown in Table 2. It should be noted that in the general case, the calculation error limit active power generating units from the matrix SMA EMF defined: measurement errors regime parameters, structure and amplitudes mutual motion generator rotor on the interval (window) identification matrix SMA, linearity load model, replacement of group generators with in phase moving rotors in the transition process to one equivalent generator, composition accounted regime restrictions, as well as the reliability of the convergence process solutions of the equations of the steady state in weighting procedures. 3.2 Verification via Physical Models The proposed method was estimated in conditions approximated to actual operating conditions using the digital-analog-physical simulator (electro-dynamic simulator—EDS) of High Voltage Direct Current Power Transmission Research Institute, JSC and the EDS of Novosibirsk State Technical University. Verification of physical and mathematical models

Fig. 4 Equivalent Surgut EPS: (a) equivalent circuit diagram; and (b) equivalent computational scheme.

(created in the Mustang Software) was performed by comparison of process oscillograms. 3.2.1 Schemes for the Testing Schemes of test power systems are presented in the Fig. 5. 3.2.2 Testing at the Test Center of NSTU The electrodynamic model of the NSTU test center is equipped by system of the synchronized


64

measurements SMART-WAMS manufactured by CJSC RTSoft for registration the transient mode parameters. Software and hardware complex SMART-WAMS based on multifunctional transducers (MIP-02) allows to organize up to 6 registration points of parameters, the electromechanical transient calculated on 20 ms interval with a unified timestamp. MIP-02 is connected to the secondary circuits of Table 1

Parameters of the initial regime for weighting based on the relevant matrix SMA EMF generators.

Number of node 1 2 3 4 5 6 Table 2

current and voltage circuits breakers that ensure the measuring point in different parts of the circuit depending on the location of the circuit breaker. Points synchronized measurements including current and voltage vectors of the positive sequence are denoted by in Figs. 5a and 5b. Fig. 6 shows the oscillation of the generator mode parameters by disturbance after autoreclosing of circuit

Active power of generator (W) 703.4 1,203.0 4,800.0 1,165.0 865.0 2,008

Name ReftGRES Tyumen SGRES-2 SGRES-1-220 NVGRES SGRES-1-500

EMF (kV) 522.0 556.1 546.7 583.3 583.5 545.4

Reactive power of generator (MVAr) 527.6 262.2 1,428.8 791.9 523.5 593.4

Rotor angle (rad) 0.0000 0.2067 0.4947 0.1114 0.3980 0.4185

Estimated limiting active power generating units.

Power system model (including AER)

Name nodes weighting

Tyumen SGRES-2 E = var, SGRES-1-220 Ug = const NVGRES Model SGRES-1-500 generator voltage Tyumen regulation E = var, SGRES-2 Ug = const, SGRES-1-220 while NVGRES Qg < Qmax SGRES-1-500

G2 В21

Т1

B31

L1 L7

Limit active power generating units (MW) Complete digital model of EPS Matrix SMA EMF generators 8,856.3 8,937.4 7,207.5 7,208.8 3,633.9 3,598.5 3,269.0 3,266.9 4,411.0 4,413.5 5,326.6 5,357.0 6,378.7 6,327.5 2,753.1 2,780.3 2,437.3 2,409.0 3,588.0 3,588.8

B12 B13

Disagreement regarding the complete digital model EPS (%) 0.92 0.02 0.97 0.06 0.06 0.57 0.80 0.99 1.16 0.02

Infinite buses

(a)

(b) (d) (c) Fig. 5 Schemes of physical models of test power systems: (a), (b) schemes at the test center of NSTU that differ by a type of a loading connected to generator buses; and (с), (d) schemes at the test center of NIIPT with radial and ring topologies.


Fig. 6 Change of operation parameters of the generator and determination limit of the output power with subsequent verification. Table 3 Generator output active power limits (the scheme is presented in Fig. 5а). Generator’s mode of reactive power

From the weighting Lines L1, L7 are switched on Overexcitaion 4,150 Underexcitation 2,307 Line L1 is switched on Overexcitaion 3,486 Underexcitation 1,980

The power limit (W) From the From SMA matrix measurement 4,220 2,317

4,100 2,150

3,553 2,000

3,560 1,990

breaker B21 (Fig. 5a) with defining the restriction on the generator’s output active power and its further test by loading the generator to its hold-off. Results of assessment of the output active power limit obtained by means of a weighting method and by means of the SMA matrix are presented in Table 3 in compare with measured value. 3.2.3 Testing at the Test Center of NIIPT Measurements of operation parameters were performed by means of the fault recording system developed at the test center of NIIPT. The recording of instantaneous voltages and currents values were performed with the rate of 2,400 samples per second. By using measured values of the current and the voltage of a generator, the active power, the reactive power, effective values of the voltage and the mutual phase were estimated with the period of 30 ms. In Fig. 7 for the scheme presented in Fig. 5b, in the condition of a tripped two-phase fault at the sending end of the line L2, changes of the active power of the

65

generator during the transient process (P) and the real time estimation of the output power limit of a generator by the terms of static stability by using the SMA matrix are presented. In Fig. 8a, changes of operation parameters of a generator during the electromechanical process obtained by using the physical model are shown. In Fig. 8b, a real time estimation of the output power limit of a generator by means of the SMA matrix is presented. The output active power limits that were obtained by means of a weighting method and by means of a matrix of the SMA using of a physical model are 8.0 and 8.2 kW respectively. Results of assessment of the output active power limit that were obtained by means of a weighting method and by means of the SMA matrix in conditions of changes of operation parameters using physical models (for schemes with two generators) are presented in Fig. 9 and Table 4. Results obtained by using physical models in conditions approximated to actual operating conditions confirm a validity of abstract theorems that are presented in Section 2.

Fig. 7 The current active power (P) and the output active power limit (Pmax) of the generator (the simulation modeling, the window of the identification of the SMA is 2-12 s).

66

Stability Monitoring and Control of Generation Based on the Synchronized Measurements in Nodes of Its Connection Table 4 Output power limits of the generator G2 (Fig. 5c) and the equivalent of the in-phase group of generators that consists of generators G1 and G2 (Fig. 5d). Scheme Fig. 5c Fig. 5d

The power limit (W) From the weighting From the SMA matrix 4,700 4,570 7,450 7,400

6. Conclusions

Fig. 8 The oscillogram and an estimation of the output power limit obtained by using a physical model: (a) changes of operation parameters of generator in the condition of a tripped two-phase fault at the sending end of the line L2; and (b) the current active power (P) and a real time estimation of the output power limit of the generator (Pmax), a window of the SMA identification is 2-8 s.

It is possible to identify an actual matrix of the SMA (self and mutual admittances) of generator EMFs by using the synchronized (phasor) measurements at generator buses and to get a real time assessment of limit conditions of a power system by the terms of a static stability via matrix of the SMA. It is advisable to perform a real time monitoring of static stability limits in a post-emergence mode of a power system in two stages: the first stage is a short-term quasi steady mode that comes immediately after the electromechanical transient attenuation; the second stage is long-term post-emergency state. It is useful to get static stability limits at each stage and then use those assessments in the emergency automation. The proposed technology is focused on the use of real-time and does not require data on the topology, parameters and loads of the electric network.

Acknowledgments The authors thank Victor Denisov, Ph.D., associate professor of Automated Electrical Power Systems Department NSTU, Leonid Grashchenkov, Michael Plyuschev, Alexander Nazarenko, specialists Novosibirsk RDO SO UPS, JSC for their invaluable assistance in the preparation of equipment for carrying out experiments on the EPS physical model at the test center of NSTU. Fig. 9 Assessments of output power limits using a physical model: (a) the output power limit of the generator G2 (Fig. 5c), a window of the SMA identification is 2.5-5.5 s; and (b) the output power limit of an equivalent of the in-phase group of generators that consists of generators G1 and G2 (Fig. 5d), a window of the SMA identification is 2.5-4.5 s.

References [1]

[2]

Ayuyev, B. I., Davydov, V. V., and Erokhin, P. I. 2010. “Optimization Model of Limiting Operating Modes of Electric Systems.” Electricity 11: 2-12. Mohammadi-Ivatloo, B., Shiroei, M., and Parniani, M. 2011. “Online Small Signal Stability Analysis of Multi-machine Systems Based on Synchronized Phasor


[3]

[4]

[5]

Measurements.” Electric Power Systems Research 81 (10) 1887-96. Chakrabortty, A., Chow, J. H., and Salazar, A. 2010. “A Measurement-Based Framework for Dynamic Equivalencing of Large Power Systems using WAMS.” In Proceedings of 2010 ISGT Innovative Smart Grid Technologies, 1-8. Chusovitin, P. V., and Pazderin, A. V. 2013. “Monitoring of Power System Stability Based on a Dynamic Equivalent Determined from Vector Measurements.” Electricity 2: 2-10. Toutoundaeva, D. V., and Fishov, A. G. 2011. “Improvement of the Power System Stability Margin

[6]

[7]

67

Control and Standardization Technique.” Scientific Herald of Novosibirsk State Technical University 2 (43): 147-60. Soboleva, M. A., and Fishov, A. G. 2013. “Determining the Limiting Electric Power System Operating Conditions on the Basis Admittances Matrices with Respect to the EMFs of Equivalent Generators.” Electricity 8: 9-14. Fishov, A. G. 2013. The Way of Stock Control the Stability of the Mode of Synchronous Electric Machines Included in the Electricity Network. RU Patent 2,500,061 C2, filed December 02, 2011 and issued November 27, 2013.

D


DAVID

PUBLISHING

Identifying the Factors for Ensuring Customers Actively Engaged in Smart Grids Linda Hull1, Even Bjørnstad2, Yvonne Boerakker3, Magnus Brolin4, Yeoungjin Chae5 and Duncan Yellen6 1. Strategic Asset Management, EA Technology Ltd., Chester CH1 6ES, UK 2. Enova SF, Oil and Energy, Trondheim N-7030, Norway 3. Policy Advisory and Research, DNV GL, Arnhem 6800, Netherlands 4. Energy Technology Department, SP Technical Research Institute of Sweden, Göteborg 400 22, Sweden 5. Planning Division, KPX, Bitgaramro 625, Korea 6. Smart Grid Delivery, EA Technology Ltd., Chester CH1 6ES, UK Received: July 30, 2014 / Accepted: October 10, 2014 / Published: January 31, 2015. Abstract: Smart grids are expected to become an essential component of the future energy system. The technical potential of smart grids is far reaching and increasingly well understood, and smart grids are now in the early phases of market deployment in several regions, particularly, in Europe and the US. Less understood than the technical aspects is how and to what degree end users (i.e. the customers) are willing and able to embrace smart grid technologies and the changes in mindset associated with this transition. This article reports the main findings from an IEA (International Energy Agency)-DSM (demand side management) project addressing the role of customers in a smart grid deployment scheme, specifically how customer behavior may restrict the technical potential of smart grids from being realized. With a model of household energy behavior as the theoretical point of departure, the research builds on experiences from various smart grid pilot studies, together with consumer research within similar domains, to identify behavioral challenges that are likely to hamper adoption of “smart grid behaviors”. Based on this insight, a set of recommendations to minimize customer resistance to smart grid deployment is suggested. Key words: Smart grid, energy behavior, consumer, barriers.

1. Introduction Security of supply and sustainability are two strategic issues related to the future energy system. The concept of “smart grid” represents an important part of the solution to these issues, particularly in relation to the functioning of the electricity grid. This article reports the key findings from an international project carried out within the DSM (demand side management) implementing agreement of the IEA (International Energy Agency), explores how end users of electric energy interact with smart grid related initiatives, how behavioral barriers may limit the efficiency of smart

Corresponding author: Linda Hull, senior consultant, research fields: demand side management and energy networks. E-mail: [email protected].

grid, and how these barriers could be addressed. The key hypotheses of the project is that the behavior of end users is a key factor of the success of smart grid initiatives, and that the factor is not well understood by implementers of smart grid initiatives. Further, smart grid research has had a main focus on system technologies and has, to a much lesser degree, addressed the human dimension—the customer behavior of the smart grid. With this article, we propose a platform for analyzing the customer issues, and aim at revealing the most important “traps” to be avoided by smart grid implementers. We combine both theoretical knowledge and empirical work in this process, the findings of which may function as useful guidelines for smart grid implementers and policy makers. The article is structured in the following way: the

Identifying the Factors for Ensuring Customers Actively Engaged in Smart Grids

authors start with a brief overview of the main issues characterizing the smart grid and energy behavior research (Section 2). Next, in Section 3, the authors outline the theoretical platform upon which their empirical interpretations rest. Section 4 states the typical goals of smart grid initiatives, while Section 5 discusses the typical pitfalls related to these, by giving examples of challenging behavioral issues and how they could be handled. Conclusions are presented in Section 6.

2. Smart Grids: A Brief Review One of the major future global challenges concerns the need for a sustainable and reliable energy supply. This is being addressed in several strategic documents and roadmaps on an international level, such as the IEA Energy Technology Perspectives 2014 [1] and the Europe 2020 Strategy [2]. Concerning electrical power, it is foreseen that renewable power sources, such as wind and solar, will increase significantly and will play an important role in the future power system. It is also projected that the importance of electricity will increase in the future, on the one hand, due to a large-scale introduction of electric vehicles, increased use of heat pumps, etc., and on the other hand, due to the need to achieve carbon emission reduction targets. This development implies a new diversity in the power system, and a transformation of the system from being centralized towards a more decentralized one [3]. This constitutes a paradigm shift, resulting in new challenges requiring new solutions to ensure the reliability and energy security of the system. This leads to the idea of a smart grid, addressing and providing the solutions to these issues. End users will play a key role in delivering effective smart grids. Due to this central role of consumers (grid customers), research has been carried out concerning the system and market impact of measures on the demand side. Examples of such research are presented in Refs. [4-8]. Also simulations on a detailed level have been made to assess the technical potential of

69

flexibility in households [9-12]. All of these are based on a theoretical technological and/or statistical approach to describe the consumption of electricity. A related topic is the control of residential electrical loads, including heating and cooling, electric vehicles, appliances etc.. Technical solutions aiming at optimizing residential load patterns are presented in Refs. [13-15]. The future role of electric vehicles, such as smart charging and applied as local storage, has been addressed in Refs. [16-18]. In these papers, the technological functions and performance of control devices are described and analyzed. However, still lacking in the above described analyses is a discussion on how to motivate and entice consumers to adopt and/or to use the described technology. This is a complex issue, not in the least due to its inter-disciplinary nature, involving several research areas, such as technology, economics, psychology and sociology. The main aim of this paper is to give insight into this issue, namely, the factors ensuring the engagement of consumers in smart grids. For this purpose, the focus in terms of the “consumer” is on households and small business end users of electric energy and their “energy behavior” (see definition in Section 3). Understanding the energy behavior of the target group is fundamental for all implementation of energy related policies, programs and technologies. The theoretical underpinning of this concept has evolved, but we may take the neoclassical economics theory as a point of departure. The traditional economics position is that energy behavior is largely understood as an outcome of an investment type decision, where the expected present value of the investment is the key decision parameter. Empirical evidence has challenged this position by demonstrating that consumers often make decisions that are not “rational” from a purely economic perspective. This so called “energy paradox” was debated during the 1990s, and a range of explanations was suggested to account for these seemingly irrational behaviors [19-22]. In addressing the purely economic perspective of

70


smart grids, financial incentives can be used to motivate the demand side. Examples of pricing schemes, tariffs and the impact on consumption can be found in Refs. [5, 23, 24]. However, financial incentives are often not sufficient, and may not be effective for all consumer segments. Economic incentives may be enhanced by visualization of electricity consumption, environmental impact, etc. to induce changes in energy consumption patterns. Research has been performed to analyze the impact of different visualization techniques on consumers, and examples are presented in Refs. [25, 26]. In addition, a large number of demonstrations and pilots worldwide have been carried out. Many of these have addressed the issue of flexibility in residential areas, where different technologies and offers have been deployed. A few examples are the SDG & E (San Diego Gas and Electric) Reduce Your Use Day (US) [27], PG & E (Pacific Gas and Electric) Smart Rate (US) [28], Jeju Island Test Bed (KR) [29], Customer Led Network Revolution (UK) [30], and the introduction of time-of-use tariffs in Italy (IT) [31]. There are also a few reports giving an international overview of various demonstrations and pilots, including Refs. [32, 33]. An overall conclusion from the review of these studies is that there is a need to look beyond the purely economic and technical mechanisms to understand the engagement of consumers in smart grids. In parallel with, and partly as a consequence of these research findings, theoretical models of energy behavior have evolved that challenge the neoclassical model. Space prohibits a comprehensive discussion of these; however, some interesting characteristics may be mentioned. A social psychological approach, such as the Theory of Planned Behavior, focuses more on fundamental personal traits, such as norms, attitudes and beliefs, in explaining decisions and behaviors [34]. This and similar approaches widen the basis of explanatory variables from the narrower focus on economic variables. A much used approach to explain

technology diffusion, suggested by Rogers [35], is also useful to explain how certain energy behaviors spread in society. This theory rests on an assumption of “psychological heterogeneity” among consumers, whereby the change-oriented risk taking innovators and early adopters implement certain energy behaviors (technologies) at a much earlier stage than the mainstream consumers. An important practical application of this theory is for change agents to exploit the dynamics of the differences between these consumer segments. Economic and psychological models typically focus on the individual decision maker in understanding energy behavior. There are theoretical positions that question this approach on ontological grounds, and claim that the energy behavior of an individual is not to be understood as discrete “energy decisions”, but rather is the result of different everyday practices, trivial habits at home and in the workplace. “Practice theory” exemplifies this position, where focus is more on the general context of living, the practice of everyday life with the routines and technologies involved, and the energy use implied by this [36]. It might be that the wide mix of energy behaviors, ranging from daily routine behaviors to single “once in a lifetime” investment decisions, is not explained by a single theory. It is our view, however, that an implementer of smart grid technologies should have some theoretical reference points for guidance in order to understand the target group (customer base) better, and thereby be better able to justify and tailor the implementation to customer needs, address the relevant barriers, and thus, achieve a more optimal smart grid implementation process. From this overview of smart grid issues, we therefore turn in the next section to a more specific discussion of theories of energy behavior.

3. Energy Behaviors and Practices As demonstrated in the literature review above, there is a wide ranging and complex set of factors that influence the way that consumers interact with smart


grid related activities, i.e., their energy behaviors. In general terms, an energy behavior refers to any action or decision taken by the consumer (decision maker) that affects the energy use of the consumer. More specifically, any given “energy behavior” should be defined in terms of a few key elements. Firstly, it involves a decision maker; the individual or other entity that makes the decision and performs the behavior. The second element is a well-defined outcome or action. This could be switching off lights, buying low energy light bulbs, reducing indoor temperature or deciding when to use a washing machine. Further, the context is relevant, i.e., does the action take place at home, at work or in the car? Finally, the time is also an important element, both in terms of defining the point in time of the initial decision and the duration/frequency of the action if it is repeated: Is it a one-off action that takes place today, or is it repeated over several weeks and months? Once the behavior is well defined, a theoretical model of behavior can be used to help understand the factors that influence the decision maker’s choice over whether to perform the behavior or not. 3.1 Energy Behavior Model A number of models or frameworks of understanding exist. No single model or framework of

Fig. 1 Theoretical model of energy behavior.

71

understanding is considered to be ideal, but they are a useful tool to help understand and achieve an outcome that depends on behavior change. Some models focus on individuals actors; others focus on the individual in his/her social environment. Some focus only on the behavior per se; other models are more concerned with the wider social, material and economic context within which the behavior is performed. Some theoretical positions focus on one-off behaviors; others on habitual day-to-day activities. Some focus on discrete actions; others consider a complex set of inter-related actions. It is believed that valuable insights can be gained from considering both the characteristics of the individual, and also the physical, social and political context within which the decision is made. We therefore chose to base our work on the following model of energy behavior (Fig. 1), which represents a synthesis of several theoretical positions [34, 37]: This model gives several messages. Firstly, it indicates, by the “core” represented by the shaded boxes, that the characteristics of the decision maker represent an important determinant of behavioral decisions; Secondly, there is a distinction between discrete one-off decisions, which are often the concrete outcome of a deliberate (planned) decision process, and

72


habitual behaviors, which become routine over time once established and where the actual “decision” may be more blurred. The process of establishing a “pristine” behavior may thus be different from the process of changing an established habit. Thirdly, there are factors outside the control of the decision maker, such as external barriers and facilitators, and the general societal context that also need to be understood when analyzing an energy behavior. Regulations, market conditions, prices, climate, infrastructure, etc. are examples of such variables. In empirical analyses across countries/regions, it is important to be aware that differences in the societal context may become an important variable that explains different “rationalities” at the individual level. Our model of energy behavior thus accounts for central elements of both the individualistic and the more system-oriented models discussed in the literature review. 3.2 Diffusion of Behaviors As mentioned above, the core of the energy behavior model is the individual, and the “psychological characteristics” of the individual decision maker represents important explanatory variables of behavior. There is much evidence that different decision makers make different decisions in a given situation. Rogers [35] suggests that different psychological characteristics of the consumers, decision makers, explain these differences. Applied to new technologies, this manifests itself in the typical observation that the diffusion over time of a new technology is often described by an S-curve. An interpretation of this process is that “innovators” and “early adopters” first adopt the new technology, later to be followed by the “mainstream” and, possibly, the “laggards”. This process is explained by different psychological profiles of the different customer segments (see Fig. 2). We suggest that this theory is relevant also for the introduction of smart grid technology. Some consumer groups are curious, open and proactive towards the new

technology, and are eager to use it. Others are skeptical, even suspicious to new technologies, including smart grid technology. These may even refuse to have the technology installed in their home. A particular problem should be mentioned in this respect. Diffusion processes are often illustrated by a smooth and continuous curve, indicating that the process proceeds by its own force once it has gained some momentum. However, as pointed out by Moore [38], this diffusion process may very well stop after the innovators and early market consumers have adopted the technology. This means that there might be a “chasm” separating these early market segments and the mainstream market, as illustrated in Fig. 2. A smart grid deployment process should therefore pay specific attention to the problem of crossing the chasm.

4. Outcomes Required A range of smart grid related initiatives have been trialed and, in some cases, rolled out to large numbers of consumers. Those involving an element of consumer energy behavior change have focused on delivering one or more of the following outcomes: (1) An overall reduction in energy consumption. This may seem the logical place to start, but if the primary goal is to reduce peak demands, it may be more effective to focus on changing the pattern of consumption. (2) A different, but enduring, underlying pattern of consumption. This can be used to address situations where peak demands occur over extended periods of time (i.e., over several months of the year).

Innovators

Early adopters

Early majority

Late majority

Fig. 2 Diffusion of innovations model.

Laggards


(3) A pattern of consumption that responds dynamically to the varying operational requirements faced by the electricity system. Achieving this outcome is beneficial for dealing with short term constraints that require changes to the pattern of demand on an infrequent and unpredictable basis. (4) Enabling industry stakeholders to access and utilize energy consumption information. Improved information about electricity usage and voltage levels on the distribution networks offers the potential to improve the efficiency of the electricity system. Smart meter data can help network operators optimize the way they manage and operate their networks and can reduce the need for network investment. Which of these outcomes are required will depend on the specific context of the smart grid initiative, and this context is likely to vary between countries and regions. This context may also be influential on the behavioral responses of the initiative, as pointed out by our theoretical model.

5. Designing Customer Offerings There are a number of different factors that need to be taken into consideration when designing a smart grid initiative to ensure that consumers are willing to engage. This implies that consumers “sign up” to the initiative, undertake any actions that enable them to participate (i.e., install technologies) and deliver the required outcomes. 5.1 Provide Choice When an individual is unhappy with a situation (s), he can react by exiting (choosing something else) or voicing their concerns (protesting). Therefore, it is important to provide consumers with an element of choice. These protests can, in some cases, be extreme. For example, one woman in Houston, Texas, brandished a gun at a utility worker to prevent the installation of a smart meter in her home. However, it is also important to ensure that consumers are not faced with too many choices;

73

otherwise, they can be paralyzed by an inability to choose from the myriad of options available to them [39]. This is attributed to a number of factors including concerns that they may make the wrong choice (the anticipation of regret) and the difficulty of assessing the trade-offs between the different options. Therefore, it is important to provide consumers with choices, but not too many. 5.2 Provide Tangible Benefits A significant amount of analysis has been undertaken, and is currently on-going, to demonstrate the benefits of smart grids. Many of these studies focus on the benefits to industry stakeholders or on the benefits to society as a whole. However, it is important to ensure that they also provide tangible benefits directly to consumers themselves. A review of a number of surveys of consumer attitudes and views towards smart grid related activities shows that consumers generally say that they prefer/expect to receive a financial reward [40]. Tangible benefits are not just limited to direct financial benefits, but include other aspects, such as improved comfort, improved health or reduced environmental impact. 5.3 Take Care When Framing the Initiative The way that smart grid initiatives are framed has an important impact on the way they are assessed and evaluated by consumers. As indicated in the energy behavior model in Fig. 1, an individual’s decision to perform an action depends on a number of factors, including their own views, opinions and beliefs. This implies that different segments of consumers react (behave) in different ways in response to a given initiative. So it is not possible to make generalizations that apply to all individuals. However, some general observations can be made about how the framing options can have an important impact on the decision making process. This includes focusing on the avoidance of losses

74


rather than achieving benefits, taking care over the reference point used to compare benefits and the timing of payments and benefits. An example of “avoidance of losses” is provided below. Many consumers are risk averse, i.e., they are reluctant to take a course of action that has an uncertain outcome. They are more likely to select an option with a certain outcome, even if the expected outcome from an alternative, but uncertain, option is higher. However, individuals are much more willing to “take a gamble” where losses are involved. The following is a well cited example of the powerful effect of framing on decision making that was originally developed by Tversky and Kahneman [41]. Consider a scenario where an unusual disease is expected to kill 600 people. Two alternative programs have been proposed to combat the disease, with different outcomes expected. The way that these programs are framed has a significant impact on which of the proposed programs is preferred. If the options are framed in terms of how many people are saved (option 1: 200 people saved, option 2: 1/3 probability that 600 people will be saved, but 2/3 probability that no-one will be saved), the majority (72%) select option 1. If, however, the options are framed in terms of how many people will die (option 3: 400 people will die, option 4: 1/3 probability that no-one will die but 2/3 probability that everyone will die), the majority (78%) select option 4. Therefore, there may be advantages in framing the initiative in terms of avoiding waste or losses rather than in terms of the benefits that could be achieved. 5.4 Ensure Consumer Concerns Are Addressed Consumer concerns must be identified and addressed. The results of previous field trials and case studies provide a useful starting point for identifying concerns, but it is important to realize that the results of one study will not necessarily apply to a different group of consumers or within a different context.

For example, smart meters have been rolled out to all of Enel’s consumers in Italy. The roll out was driven by Enel’s desire to improve the cost effectiveness of their metering activities, and consumers were not provided with any choice over whether or not they wished to have a smart meter. Despite the lack of choice, no strong opposition was raised by consumers. The same is not true for consumers in the Netherlands, for example. A mandatory roll-out was proposed, with consumers facing a possible €17,000 fine or six months imprisonment for refusal. Consumers and consumer groups raised concerns over the violation of right to privacy and the possibility that the data could be misused. This led to the Dutch First Chamber refusing to approve the Smart Metering Bill. A review of a number of smart grid related case studies and surveys of consumer opinions conducted as part of the project shows that there is a wide range of issues that present barriers to consumer engagement. Consumer concerns relating to health, safety and data privacy are reasonably well documented, but it is important that other concerns are not overlooked. 5.4.1 Disruption and Inconvenience Consumers cited a number of concerns relating to disruption to their property or routines. In particular, the impact on time is important. Many relate to the installation process itself, i.e., the inconvenience associated with the time spent waiting for installers to arrive and during the installation process itself. In addition, householders cite that they can be put off by the installation process itself, i.e., due to the need to move possessions in order to allow the installation to take place or the need to redecorate after the installation has been completed. Although seemingly innocuous, these concerns can have a significant impact on consumer engagement. For example, large numbers of consumers are put off having free or subsidized loft insulation fitted within their home due to upheaval and disruption. This is particularly true when large volumes of possessions are stored in the loft which would need to be relocated.


Consumers also cite a general dislike of people coming into their home, for example, to install new technologies. This relates to not only the inconvenience caused, but also concern over the potential for damage. For example, an Australian demand response trial involved installing technology to allow the air-conditioning equipment of householders to be controlled remotely. The trail reported that householders were more likely to participate if their air-conditioning units were located outdoors [42]. 5.4.2 Financial Commitments and Uncertainty over Benefits The anticipation of regret, i.e., of being worse off, is an important factor influencing consumer willingness to participate. The possibility that they will end up paying more than they do at the moment that outweighs the possibility that they will end up paying less. Some schemes remove the risk by providing assurances that consumers will not pay more under the new initiative than they would have done on their existing tariff. Whilst, this may help consumers enroll, it may not necessarily provide a cost effective approach for stakeholders. For example, the principle of not exposing the consumers to any risks can constitute a barrier to a larger rollout by a distribution operator. It may lower revenue and thereby decrease the possibilities for the network operator to invest and operate the system. It is essential to provide consumers with information that is as accurate and reliable as possible, rather than “best case” scenarios. Claims that smart meters can help householders save up to 10% on their bills typically represent the upper limit on the savings obtained from trials. In practice, many consumers achieve savings much lower than this, leading to dissatisfaction and disappointment. Lack of confidence over the level of benefits could be even more important where an individual has to make an upfront investment in order to participate, i.e., they know how much they need to invest but may not have certainty over when (or if) they will get a return

75

on their investment. However, it is important to note that ensuring a financial incentive in return for participation is not sufficient to ensure that consumers will take action. This does not imply they are behaving irrationally, but rather that there are other factors that are considered to be more important.

6. Conclusions Over the coming years, smart grids offer the potential to support the move to a low carbon future. They could help to deliver a fundamental change in the way that the balance of electricity supply and demand is managed. In particular, they enable a coordinated approach whereby the actions of all users connected to the energy system can be integrated. This includes the actions of the consumers themselves. However, there is a real risk that if consumers do not adopt new approaches to the way that they consume electricity, smart grids may not be able to achieve their full potential. A neo-classical economic analysis of the potential benefits and losses does not accurately predict whether a consumer will adopt a particular initiative or technology. Rather, there are a number of different factors and elements that influence the decision maker; only some of which have been identified in this paper. Understanding these factors, addressing consumer concerns and ensuring that smart grids provide tangible benefits to consumers are important aspects to be taken into consideration when designing smart grid initiatives. A number of general guidelines can be identified in order to help ensure a smart grid initiative that is more likely to be adopted by consumers: a selection of which have been identified here. However, the decision of whether or not to engage is always made by an individual. The factors that influence the decision making process are wide-ranging and complex. As a result, an initiative that is successful for one group of consumers


76

may not necessarily be effective with another group in a similar context due to the differing views and beliefs of the individuals involved. Likewise, what works for one group of consumers may not succeed with another group of like-minded consumers due to the existence of differing opportunities and barriers.

[9]

[10]

[11]

Acknowledgments The authors wish to thank the following for supporting this research: Enova SF, Norway; EA Technology, UK; DNV GL, Netherlands; SP Technical Research Institute of Sweden, Sweden; KPX, Korea and all of the members of the stakeholder teams that supported the project.

[12]

[13]

References [1]

[2]

[3] [4]

[5]

[6]

[7]

[8]

International Energy Agency. 2014. Energy Technology Perspectives 2014—Harnessing Electricity’s Potential. Paris: IEA Publications. European Commission. 2010. Communication from the Commission: EUROPE 2020—A Strategy for Smart, Sustainable and Inclusive Growth. A report by COM, Brussels. International Energy Agency. 2011. Technology Roadmap: Smart Grids. Paris: IEA Publications. Moghaddam, M. P., Abdollahi, A., and Rashidinejad, M. 2011. “Flexible Demand Response Programs Modeling in Competitive Electricity Markets.” Applied Energy 88 (9): 3257-699. Albadi, M. H., and El-Saadany, E. F. 2008. “A Summary of Demand Response in Electricity Markets.” Electric Power Systems Research 78 (11): 1989-96. Ma, O., Alkadi, N., Cappers, P., Denholm, P., Dudley, J., Goli, S., Hummon, M., Kiliccote, S., MacDonald, J., Matson, N., Olsen, D., Rose, C., Sohn, M. D., Starke, M., Kirby, B., and O’Malley, M. 2013. “Demand Response for Ancillary Services.” IEEE Transactions on Smart Grid 4 (4): 1988-95. Olsen, D. J., Matson, N., Sohn, M., Rose, C., Dudley, J., Goli, S., and Kilicote, S. 2013. Grid Integration of Aggregated Demand Response, Part 1: Load Availability Profiles and Constraints for the Western Interconnection. Technical report. Hummon, M., Palchak, D., Denholm, P., Jorgenson, J., Olsen, D. J., Matson, N., and Kiliccote, S. 2013. Grid Integration of Aggregated Demand Response, Part 2: Modeling Demand Response in a Production Cost Model. Technical report.

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21] [22]

Widén, J., and Wäckelgård, E. 2010. “A High-Resolution Stochastic Model of Domestic Activity Patterns and Electricity Demand.” Applied Energy 87 (6): 1880-92. Richardson, I., Thomson, M., Infield, D., and Clifford, C. 2010. “Domestic Electricity Use: A High-Resolution Energy Demand Model.” Energy and Buildings 42 (10): 1878-87. Capasso, A., Grattieri, W., Lamedica, R., and Prudenzi, A. 1994. “A Bottom-up Approach to Residential Load Modeling.” IEEE Transactions on Power Systems 9 (2): 957-64. Gottwalt, S., Ketter, W., Block, C., Collins, J., and Weinhardt, C. 2011. “Demand Side Management—A Simulation of Household Behavior under Variable Prices.” Energy Policy 39 (12): 8163-74. Pedrasa, M. A. A., Spooner, T. D., and MacGill, I. F. 2010. “Coordinated Scheduling of Residential Distributed Energy Resources to Optimize Smart Home Energy Services.” IEEE Transactions on Smart Grid 1 (2): 134-43. Mohsenian-Rad, A. H., and Leon-Garcia, A. 2010. “Optimal Residential Load Control with Price Prediction in Real-Time Electricity Pricing Environments.” IEEE Transactions on Smart Grid 1 (2): 120-33. Fernandesa, F., Moraisa, H., Valea, Z., and Ramosa, C. 2014. “Dynamic Load Management in a Smart Home to Participate in Demand Response Events.” Energy and Buildings 82 (Oct.): 592-606. Papadaskalopoulos, D., Strbac, G., Mancarella, P., Aunedi, M., and Stanojevic, V. 2013. “Decentralized Participation of Flexible Demand in Electricity Markets—Part II: Application with Electric Vehicles and Heat Pump Systems.” IEEE Transactions on Power Systems 28 (4): 3667-74. Peterson, S. B., Whitacre, J., and Apt, J. 2010. “The Economics of Using Plug-in Hybrid Electric Vehicle Battery Packs for Grid Storage.” Journal of Power Sources 195 (8): 2377-84. Sortomme, E., and El-Sharkawi, M. A. 2011. “Optimal Charging Strategies for Unidirectional Vehicle-to-Grid.” IEEE Transactions on Smart Grid 2 (1): 131-8. Hassett, K. A., and Metcalf, G. E. 1993. “Energy Conservation Investment: Do Consumers Discount the Future Correctly?.” Energy Policy 21 (6): 710-6. Jaffe, A. B., and Stavins, R. N. 1994. “The Energy Paradox and the Diffusion of Conservation Technology.” Resource and Energy Economics 16 (2): 91-122. Metcalf, G. E. 1994. “Economics and Rational Conservation Policy.” Energy Policy 22 (10): 819-25. GGolove, W. H., and Eto, J. H. 1996. Market Barriers to Energy Efficiency: A Critical Reappraisal of the Rationale for Public Policies to Promote Energy Efficiency. A report

Identifying the Factors for Ensuring Customers Actively Engaged in Smart Grids by Lawrence Berkeley National Laboratory. [23] Bartusch, C., Wallin, F., Odlare, M., Vassileva, I., and Wester, L. 2011. “Introducing a Demand-Based Electricity Distribution Tariff in the Residential Sector: Demand Response and Customer Perception.” Energy Policy 39 (29): 5008-25. [24] Herter, K. 2007. “Residential Implementation of Critical-Peak Pricing of Electricity.” Energy Policy 35 (4): 2121-30. [25] Ellegård, K., and Palm, J. 2011. “Visualizing Energy Consumption Activities as a Tool for Making Everyday Life More Sustainable.” Applied Energy 88 (5): 1920-6. [26] Bonino, D., Corno, F., and Russis, L. D. 2012. “Home Energy Consumption Feedback: A User Survey.” Energy and Buildings 47 (April): 383-93. [27] SDG & E. 2014. “Reduce Your Use Day.” http://www.sdge.com/save-money/reduce-your-use/reduc e-your-use-rewards. [28] PG & E. 2014. Accessed September 29, 2014. http://www.pge.com/. [29] Korea Smart Grid Institute 2012, “Jeju Island Test Bed.” Accessed July 11, 2013. http://www.smartgrid.or.kr. [30] Northern PowerGrid 2013, “Customer Led Network Revolution (CLNR)” Accessed April 24, 2013, http://www.networkrevolution.co.uk. [31] Maggiore, S., Gallanti, M., Grattieri, W., and Benini, M. 2013. “Impact of the Enforcement of a Time-of-Use Tariff to Residential Customers in Italy.” In Proceedings of CIRED the 22nd International Conference on Electricity Distribution, 1-4. [32] Uyterlinde, M. 2014. “Report on Case Analyses, Success Factors and Best Practices.” Accessed October 28, 2014. http://www.s3c-project.eu/.

77

[33] Mulder, W., Kumpavat, K., Faasen, C., Verheij, F., and Vaessen, P. 2012. Global Inventory and Analysis of Smart Grid Demonstration Projects. DNV KEMA report. [34] Fishbein, M., and Ajzen, I. 2010. Predicting and Changing Behavior: The Reasoned Action Approach. New York: Psychology Press. [35] Rogers, E. M. 2003. Diffusion of Innovations. New York: Free Press. [36] Nicolini, D. 2012. Practice Theory, Work, and Organization: An Introduction. Oxford, UK: Oxford University Press. [37] Egmond, C., and Bruel, R. 2007. Nothing Is as Practical as a Good Theory: Analysis of Theories and a Tool for Developing Interventions to Influence Energy-Related Behavior. A report by SenterNovem. [38] Moore, G. A. 2002. Crossing the Chasm: Marketing and Selling Disruptive Products to Mainstream Customers. New York: HarperCollins. [39] Abbott, B., Mannella, A., Mcnamara, K., and Tripathy, A. 2007. “Optimising Choices for How People Really Buy, Not How We Think They Buy, Diamond Management and Energy Consultants.” Accessed September 29, 2014. http://www.slideshare.net/easiegmann/predictably-irratio nal-customers. [40] Hull, L. 2013. Interaction between Consumers and Smart Grid Related Initiatives. IEA DSM task 23 report. [41] Tversky, A., and Kahneman, D. 1986. “The Framing of Decisions and the Evaluation of Prospects.” Studies in Logic and the Foundations of Mathematics 114: 503-20. [42] Essential Services Commission of South Australia. 2007. Demand Management Program Interim Report No.1. June 2007.

D


DAVID

PUBLISHING

Development of a Genetic Algorithm for Evaluating the Performance of Overhead Power Distribution Lines and Proposing Solutions to Improve Distribution Line Safety Fabio Romero1, Alden U. Antunes1, Dario Takahata1, André Meffe1, Carlos C. B. Oliveira1, Fernando L. Lange1 and Hamilton Souza2 1. Department of Engineering Research and Development, Daimon Engineering & Systems, São Paulo 01310200, Brazil 2. Department of Engineering, AES Eletropaulo Distribution Utility, São Paulo 06460040, Brazil Received: July 08, 2014 / Accepted: September 04, 2014 / Published: January 31, 2015. Abstract: This paper aims to present and discuss the use of a power flow methodology based on Gauss elimination method to evaluate the performance of distribution network taking into account the neutral conductor absence at specific sections, and a development of a methodology based on GA (genetic algorithm) capable of evaluating alternative solutions in different bars of the feeder, in order to propose appropriate solutions to improve the distribution network safety. Besides the technical aspects, the proposed GA methodology takes into account the economic feasibility analysis. The results of power flow simulations have shown that the presence of single-phase transformers along with the absence of the neutral conductor at specific sections of the MV (medium voltage) network may increase the Vng (neutral-to-ground voltage) levels of the feeders involved, jeopardizing the system’s safety. On the other hand, the solutions proposed by the GA methodology may reduce the network Vng levels and improve the safety conditions, providing values close to the ones found before the neutral conductor theft. Key words: Genetic algorithm, power flow, distribution line safety, power flow simulation.

1. Introduction The performance evaluation of MV (medium voltage) systems has a great importance for utilities for planning and operation of distribution network purposes. The occurrence of neutral conductor theft has been increasing due to the easiness associated with the reselling of the copper and aluminium conductors at the black market. The unpredictability of this action may cause undesirable consequences to the electric utilities, such as impacts at the energy quality, increment of expenses and time for the maintenance crews to repair and/or reinstall the conductors, equipment damages, increase of the step and touch potentials etc.. Corresponding author: Fabio Romero, engineer, research fields: electrical power distribution planning, distribution losses, smart grid, protection of distribution lines, grounding, lightning electromagnetic fields, and power energy quality. E-mail: [email protected].

AES (Applied Energy Service) Eletropaulo (Brazil) has been suffering the aforementioned setbacks due to the absence of neutral conductor wires in its electric distribution network. The Utility is responsible for supplying about 6.3 million customers in the State of Sao Paulo, Brazil. The aim of this work is to present a genetic algorithm methodology to evaluate some alternatives along the MV network, in order to recommend appropriate solutions to reduce the Vng (neutral-to-ground voltage) levels and, consequently, improve the network safety. The power flow analysis, based on Gauss elimination method [1, 2], evaluates the impact of neutral conductor absence at specific sections of the network regarding the performance and safety of distribution network, whereas, the GA (genetic algorithm) methodology proposes technical-economic solutions to mitigate the damages

Development of a Genetic Algorithm for Evaluating the Performance of Overhead Power Distribution Lines and Proposing Solutions to Improve Distribution Line Safety

due to the neutral conductors theft. The analyses presented in this paper have been part of the activities of an R & D (Research and Development) Project carried out in partnership with AES Eletropaulo and the results have been applied in two different feeders of the utility.

2. System Summarized Description AES Eletropaulo’s distribution system is characterized by the presence of a common neutral conductor for the MV (medium voltage) and LV (low-voltage) circuits, multi-grounded at every 300 m and at every pole where there is an equipment installed (distribution transformers, reclosers, capacitor banks, voltage regulators etc.). The common multi-grounded neutral is employed in order to provide a low impedance path back to the substation [3]. The Gauss Elimination Method [1, 2] and GA methodology have been implemented in the Interplan software [4] and the simulations were performed for the feeder RGR-104, due to the high incidence of its neutral conductor theft. Feeder RGR-104 is a 13.2 kV circuit with 51 km long and, according to AES Eletropaulo, the neutral conductor is absent in about 20% of the total length. Neutral conductors installed are 3/0 AWG (American Wire Gauge) aluminium wire at the main feeder and 1/0 AWG aluminium wire at the laterals. Feeder RGR-104 is characterized by possessing Y-∆ (wye-delta) transformers, multigrounded neutral at every 300 m and at every pole where there is an electric distribution equipment installed, and also at poles adjacent to equipment. To evaluate the feeder’s performance, simulations through Interplan power flow software have been employed, using as input, feeder RGR-104’s geo-referenced data and phase and neutral current measurements at the substation. The methodology adopted in order to calculate the power flow allows, as well as other parameters, the phase and neutral current, neutral-ground voltage and electrical losses to be simulated. Table 1 shows the measured

79

Table 1 Measured and simulated values of phase and neutral currents at the substation of feeder RGR-104. ID (A)

IE (A)

IF (A)

IN (A)

445

452

417

32

values of phase and neutral currents at the substation of the feeder RGR-104. The contributions for the ground resistances of the consumers’ pole grounding and adjacent feeders are not considered in the simulations. Therefore, unless otherwise indicated and based on measurements along the feeder, the average ground resistivity (ρg) has been assumed to be 300 Ω.m and the average Rg (ground resistance) to be 50 Ω for grounded poles. The analysis with the GA proposed methodology takes into account the alternative solutions presented in the Table 2.

3. GA (Genetic Algorithm) Methodology A methodology that uses the technique of GA [1] has been proposed, based on research of optimization methods for technical and economical assessments, and on research of alternative solutions to mitigate the problems due to the theft of neutral conductors in distribution networks. GA methodology is a particular class of evolutionary algorithms that uses techniques inspired by evolutionary biology of Darwin’s natural selection. The elements, which define a state of the problem from a population, are represented by chromosomes. These, in turn, are represented by genes. Thus, using some concepts of inheritance as crossover, mutation and natural selection of the fittest ones after a few generations, one expects to find the best solutions. GA methodology is based on a combinatorial analysis, in which one seeks an optimal solution. The basic principle of GA technique is the adaptation of the population due to the inheritance of good genes of previous generations, and the good characteristics of individuals will be passed to future generations, according to the evaluation process. Fig. 1 shows the GA flowchart.

80


Table 2 Alternative solutions evaluated by GA methodology to mitigate the damages due to the neutral conductor absence and improve the feeder safety condition. Grounding configuration options 1,500 mm length four parallel rods 1,500 mm length six parallel rods 2,400 mm length four parallel rods 2,400 mm length six parallel rods Neutral conductor options

Total installation cost price (USD) 130.00 193.00 145.00 220.00 Total installation cost price (USD/km)

Reinstallation of 1/0 AWG aluminium wire in the MV line 1,450.00 sections without neutral conductors Reinstallation of 3/0 AWG aluminium wire in the MV line 1,790.00 sections without neutral conductors Installation of bimetallic steel/ 2,520.00 aluminium wire 7N9 53% IACS Installation of bimetallic steel/ 3,480.00 aluminium wire 7N7 53% IACS IACS: International Annealed Cooper Standard.

One defines as an individual, the set of grounding solutions and conductors substitution of a specific configuration of the network, which will be represented by a finite size vector. The principle is analogous to the adaptation of living organisms in different environments. Initially, one draws the initial set of individuals, in which the population will be chosen, selecting the best suited beings, to apply later on the possibility of mutation and crossover. Finally, one assesses all individuals of this generation and, if a parameterized maximum number of generations are reached, one meets the most adapted individual. Otherwise, the cycle restarts from a selection of the best beings. 3.1 Steps of GA Methodology for Determining Adequate Solutions to Improve Network Safety 3.1.1 Selection of Individuals The selection of individuals is carried out by employing the tournament method, i.e., one picks up randomly three individuals in the population and one selects only the most adapted individual. 3.1.2 Mutation The aim of mutation is to generate new individuals with diverse characteristics, i.e., create greater variety of genes.

Fig. 1 Genetic algorithm flowchart.

3.1.3 Recombination A recombination or crossover is required to increase the variety of individuals, or to increase the population of the most suited ones. 3.1.4 Assessment The evaluation is performed through a merit index, which is composed by the grades assigned to the variables of “technical loss at the grounding” and cost of the proposed alternatives (Table 2). The technical loss variable is related to the value of Rg and the leakage current through the grounding rod, i.e., the smaller the value of Rg for a given leakage current, the smaller the technical loss in the ground; and consequently, the lesser the Vng. The merit index is assessed by the difference between the technical loss of the solution candidate individual and the initial technical loss of the individual (neutral stolen feeder configuration), represented by a certain cost to obtain this candidate, from the initial individual. 3.2 Codification of the Individual The encoding of GA individuals is accomplished


81

through a vector divided in two segments: grounding genes and conductor substitution genes. Fig. 2 presents a vector (chromosome) in its simplified version for a case of n grounding configurations and m neutral conductor substitution options. One points out from Fig. 2 that each candidate (individual) is formed by a set of poles (with a specific Rg) and branches (with a specific neutral conductor). The first part of the encoding vector is composed by the grounding configuration options (Table 2), in which each gene (grounded bar) will be set to the values “0” to “x”, which are the grounding arrangement alternatives. Thus, contrary to conventional binary encoding, where there are only two possibilities in which the gene may worth, i.e., either “0” or “1”, each gene may alternatively be worth “0” to “x”. In this paper, the genes (grounding bars, whether where there

Fig. 2 Codification of a finite size vector (representing a chromosome).

 MV branch with bimetallic steel/aluminum wire 7N7 53% IACS. 3.3 Evaluation Function The GA should find the best solution to maximize the evaluation function feval (losses, cost) as follows: f eval  losses, cos t  

is equipment or where the bars are grounded every 300 m) may have five grounding configuration options (Table 2):  poles without grounding;  1,500 mm long four parallel rods;  1,500 mm long six parallel rods;  2,400 mm long four parallel rods;  2,400 mm long six parallel rods. The second part of the encoding vector is composed by the neutral conductor options (Table 2), in which the options for each gene of conductor substitution (absent neutral conductor branch) will be set to the values “0” to “y”, which are the conductor substitution alternatives. Thus, contrary to conventional binary encoding, where there are only two possibilities in which the gene may worth, i.e., either “0” or “1”, each gene may alternatively be worth “0” to “y”. In this paper, the genes (absent neutral conductor branches) may only have five neutral conductor options:  MV branch without neutral conductor;  MV branch with 1/0 AWG aluminium wire;  MV branch with 3/0 AWG aluminium wire;  MV branch with bimetallic steel/aluminum wire 7N9 53% IACS;

 Initial _ Configuration _ Losses  Analysed _ Individual _ Losses    Instalation _ Costs  

(1) where,  Analysed_Individual_Losses: economic value of the grounding losses of the analyzed alternative, based on the study period (years), on the initial year energy cost (US$/MWh), on the market growth rate (percentage per annum) and internal rate of return (percentage per annum);  Initial_Configuration_Losses: economic value of grounding losses of the system original configuration, i.e., network with some branches without neutral conductors, adopting the same period and economic indices;  Installation_Costs: economic value of the chosen alternative, considering the same parameters aforementioned. Fig. 3 shows the results of a simulation, with the proposed alternatives in Table 2. Fig. 3 shows that the best alternative is the one which presents a grounding energy loss value around US$3,770.00, and an installation cost of approximately US$50,000.00 in a six-year period.

82

Development of a Genetic Algorithm for Evaluating the Performance of Overhead Power Distribution Lines and Proposing Solutions to Improve Distribution Line Safety Installation cost vs. grounding energy loss

Fig. 3 Alternative solutions presented by GA proposed methodology.

4. Conclusions This paper has evaluated the performance of MV distribution networks due to the lack of neutral conductors in some of its branches and proposed technical solutions to mitigate the damage caused by the operation of the network under these operative conditions. Feeder RGR-104 has been used to simulate due to its high incidence of neutral conductor theft. To evaluate the performance of the feeder taking into account the network neutral conductor thefts, a new methodology for power flow method based on Gauss Elimination has been developed and implemented into Interplan software [4]. The simulation results show that the absence of neutral conductors, in conjunction with the imbalance caused by the wye transformer connection along the feeder, greatly increases the levels of neutral-ground voltage, decreasing the security of the network. A new methodology based on GA has been

developed to improve the network security, proposing appropriate solutions (including changes in grounding systems and replacement and/or installation of new conductors) to keep the network Vng levels within adequate technical and safety conditions. As an alternative to eliminate the frequent thefts of aluminum neutral conductors in certain regions of the feeder, bimetallic conductors have been used in branches with high theft incidence. The results presented in this paper may be used as reference for planning improvements in the performance of distribution networks. However, it should be stressed that the results tend to be conservative because one has neglected the influence of consumer’s pole grounding and the groundings of adjacent feeders near the studied feeder, and also the variations in ground resistance and soil resistivity should be taken into account.

References [1]

[2]

[3]

[4]

Antunes, A. U., Méffe, A., Takahata, D., Kamikoga, D. H., Romero, F., and Lange. F. L. 2010. Specification of Power Flow Methodology and Technical Economic Analysis of Alternative Solutions to Mitigate Damages Due to Neutral Conductor’s Theft. Technical report for AES Eletropaulo Distribution Utility, Sao Paulo, Brazil. Antunes, A. U., Antonelli, D., Nanni, M., and Romero, F. 2010. Specification Models for Power Flow Calculations. Technical report for AES Eletropaulo Distribution Utility, Sao Paulo, Brazil. Zipse, D. W. 2004. “The Hazardous Multigrounded Neutral Distribution System and Dangerous Stray Currents.” In Proceedings of Petroleum and Chemical Industry Conference IEEE, 23-45. Oliveira, C. C. B., Kagan, N., Guaraldo, J. C., El Hage, F. S., Meffe, A., and Filho, M. M. 2004. “Interplan—A Tool for Planning High, Medium and Low Voltage Networks.” Presented at IEEE the Transmission and Distribution Conference and Exposition, Denver, USA.

D


DAVID

PUBLISHING

MPPT Control System for PV Generation System with Mismatched Modules Chengyang Huang1, Kazutaka Itako1, Takeaki Mori1 and Qiang Ge2 1. Department of Electrical and Electronic Engineering, Kanagawa Institute of Technology, Kanagawa 243-0292, Japan 2. Department of Energy and Power Engineering, Yangzhou University, Yangzhou 225127, China Received: September 26, 2014 / Accepted: November 05, 2014 / Published: January 31, 2015. Abstract: Nowadays, in a household PV (photovoltaic) generation system, it is generally connecting PV modules in series and then output to the power-conditioner. However, when PV modules are mismatched, it will lead to a wrong MPPT (maximum power point tracking) to all modules and a power decreasing of the whole system. Aiming at this problem, this paper presents the idea which improves the MPPT without changing the conventional power-conditioner, by adding a Buck type DC-DC (direct current) converter behind each module. Simulations of PSIM (power simulation) and experiments are taken to prove this theory. The result shows that, by this idea, the generated power of the conventional PV generation system can be greatly increased under the condition of mismatch. Key words: PV generation system, MPPT, Buck converter, mismatched modules.

1. Introduction Nowadays, in a household PV (photovoltaic) generation system, it is generally connecting PV modules in series and then output a power-conditioner. The power-conditioners which have been widely used in some developed countries generally consist of DC-DC (direct current) converter and inverter. In order to execute PV modules in MPP (maximum power point), the DC-DC converter in power-conditioner has the function of MPPT (maximum power point tracking). DC is converted into AC (alternating current) in inverter and then connecting the output to the grid or an AC load. But, under the condition of mismatch (modules of different output power, shadows, dirtiness, etc.), this structure of system can led to the problem of mismatch losses [1-3]. The problem has been discussed frequently and some new ideas of solar power system have been proposed. One of the most popular theories is that executing MPPT control to each PV module Corresponding author: Kazutaka Itako, professor, Dr. Eng., research field: power electronics. E-mail: [email protected].

respectively, and then outputting to the inverter. But it is difficult to achieve this idea in some countries (such as Europe, Japan) in which PV generation system has already been widely used. In order to realize this idea, the conventional power-conditioner which has both DC-DC converter (for MPPT control) and inverter in it must be replaced by new product which is expected only to consist of inverter. However, the price of conventional power-conditioner is about 5,000 dollars, and it will bring great financial losses if power-conditioner is changed. Therefore, in order to solve the mismatch problem and reduce financial losses to the minimum, the most ideal way is to add a DC-DC converter behind each module without changing the conventional power-conditioner [4]. The Buck converter is adopted as DC-DC converter. In this paper, simulations of PSIM (power simulation) and experiments are taken to prove this theory. The result shows that, by adding Buck converter for MPPT control to each module, the generated power of the conventional PV generation system can be greatly increased under the condition of mismatch.

MPPT Control System for PV Generation System with Mismatched Modules

84

2. The Problem of the Conventional System The conventional system’s structure is shown as Fig. 1. It consists of PV array and power-conditioner. The P & O method (perturbation and observation method) is the most widely used as MPPT control. It is executed by periodically perturbing (incrementing or decrementing) the array terminal voltage and comparing the PV output power with that of the previous perturbation cycle. If the power is increasing, the perturbation will continue in the same direction in the next cycle, otherwise, the perturbation direction will be reversed [5-8]. The flowchart of this method is represented in Fig. 2. However, when the PV array is mismatched (for example, one module has a partial shadow), the PV characteristics of the entire array will change from one

peak into two peaks (A and B) as shown in Fig. 3. Because of the series connection, when one module is partially shaded, the output current of the entire array will be forced to be decreased. It will lead to a low power output which is peak B in Fig. 3. However, because the bypass diode in modules operates, the module with partial shadow can be bypassed and other modules are able to generate the MPP power (Peak A in Fig. 3). However, the P & O method will drive the operating point towards the peak B which has low power output instead of the MPP (A).

3. The Improved System This paper proposed the method which executes the MPPT control to each PV module respectively without changing the conventional power-conditioner.

Boost converter AC load Power conversion Fig. 1

The conventional PV generation system. Start

Fig. 2

The flowchart of P & O method.


85

MPP as shown in Fig. 5a, while the module with a partial shadow will has a internal mismatch problem as shown in Fig. 5b. In this condition, the total generated power of the improved system will be not only higher than the B in Fig. 3, but also higher than the MPP (A in Fig. 3) of the conventional system when mismatch happened.

4. Simulation Fig. 3

The operating point when mismatch happened.

As shown in Fig. 4, the improved system consists of PV modules, power-conditioner and the added Buck converters executing MPPT control to each PV module respectively. Also taking “one module has a partial shadow” as the mismatch condition, with executing MPPT control respectively, the normal modules will not receive the bad effect of mismatch and operate at

As shown in Figs. 6 and 7, in order to prove the theory, the conventional system and the improved system are simulated by PSIM software. The situation of “partial shadow” is taken as the condition of mismatch. In order to verify the theory through experiments, the simulation is made based on the actual parameters. 4.1 The Simulation of PV Modules In order to simplify the system, two pieces of PV modules are set in the simulation, one PV module is

Buck converter Buck converter

Boost converter AC load

Buck converter

Power conversion

Fig. 4 The improved PV generation system.

(a) Normal modules Fig. 5

The operating point of each module.

(b) Module with shadow


86

7.5 mH

0.0001 Ω

PV1

500 uF

C-BLOCK

PV2

Fig. 6

470 uF

80 V

PI

20 KHZ

The simulation of conventional system.

Fig. 7 The simulation of improved system.

supposed to be covered by shadow. The type of PV modules used in the experiment is GT133S manufactured by KIS (Kindness Intelligence Solar Service). This PV module has two clusters of solar cells which are connected in series. Each cluster has a bypass diode. In order to make the simulation close to the real one, PV module is simulated based on the actual parameters by the model of ”Solar Module (functional model)” in PSIM. The simulation of PV module is shown in Fig. 8. Two clusters of solar cells are simulated by two models of “Solar Module” which parallel connected with a bypass diode respectively. The two models are

connected in series. As shown in the Table 1, in order to make the models match with the electrical characteristics of the module, voltage parameters of each model is set to half while the current parameters keep the same because of the series connecting. When one of the solar cells in PV module is partly covered by shadow, the output current of this cluster will be reduced proportionally, and the output voltage will stay the same. So, according to this characteristic, for simulating the PV module which is covered by shadow, one of the “Solar Module” models, the current parameter should be decreased. It is set to one quarter of the original in the simulation: Ipm = 0.7625 A.


87

the improved system, the average output powers of two PV modules are 49.9 W and 14.1 W, so the total power generation is 64.0 W. It is raised to 110.5% as compared with conventional system. Therefore, the simulation proved that, the generated power of the system can be improved under the condition of mismatch by adding Buck converters. Fig. 8

5. Experiments

The simulation of the PV module.

Table 1 Parameters of each “solar module” model. PV modules (GT133S) Model (normal) Model (with shadow)

Pm 50 W 25 W 6.25 W

Vpm 16.4 V 8.2 V 8.2 V

Ipm 3.05 A 3.05 A 0.7625 A

4.2 The Simulation of Power-Conditioner As shown in Fig. 9, the model of power-conditioner primarily consists of a Boost converter and a battery. The battery can be regarded as an inverter because the input of the inverter is usually controlled at a constant voltage. The MPP is tracked by P & O method in “C-BLOCK”. 4.3 Results of Simulation As shown in Figs. 10 and 11, without considering the efficiency of added Buck converters, the average power generation of conventional system is 30.4 W. In

Fig. 9

The simulation of power-conditioner.

Based on the simulation results, the experiment is taken. 5.1 Methods of the Experiment The picture of PV modules which are taken in the experiments is shown in Fig. 12. In order to make a comparison, four pieces of modules, which have the same electrical characteristics, are divided into two groups. The electrical characteristics of modules are shown in Table 1. One module from each group is covered by a board of the same size to realize the situation of mismatch. The covered area is three quarters of one solar cell. The experiments are divided into two groups. Group 1 is made based on the structure of conventional household PV generation system. It consists of two modules, a Boost converter and an electronic load (constant-voltage mode). The Boost converter and the


Power (W)

88

Time (s) The result of conventional system.

Power (W)

Fig. 10

Time (s) Fig. 11

The result of improved system.

electronic load which are connected in series, function as the power-conditioner. Group 2 implements the scheme of the improved system. It consists of two modules which are connected with a Buck converter (for MPPT control), a Boost converter and an electronic load (power-conditioner), respectively. All the parameters, system structures and control methods are the same with the simulation. The generated power of two groups will be measured and a comparison will be taken. 5.2 Results of the Experiment The date of the experiment is February 21, 2014. The results of Group 1 are shown in Fig. 13. The results of Group 2 are shown in Fig. 14. The system condition that, the power loss of the

Fig. 12

Two groups of PV modules.

added Buck converters is close to or even greater than the increased generated power, must be avoided. In other words, the efficiency of Buck converters must be confirmed. In Fig. 15, the average of the input power (Buck 1 + Buck 2) is 59.5 W, while the average of the output power (Buck 1+ Buck 2) is 52.1 W, so, the total efficiency of two Buck converters is 87.6%. Fig. 16 shows the total power generation of two


The output voltage of modules

30.0

2.0

The output current of modules

15.0

1.0

0.0

Current (A)

Voltage (V)

3.0

Current(A)

voltage(V)

45.0

0.0 0

1

2

3

4

5

6

7

8

9

10

Time(s) Time (s)

Fig. 13 The output of modules in Group 1. 20.0

8.0

10.0

4.0 The output current of module 1 and 2

5.0

Current (A)

voltage(V)

6.0 The output voltage of module 1 and 2

Current(A)

Voltage (V)

15.0

2.0

0.0

0.0 0

1

2

3

4

5

6

7

8

9

10

Time(s) Time (s)

Fig. 14 The output of modules in Group 2. 80.0 The input power of buck 1 and buck 2

Power (W) Power(W)

60.0

The output power of buck 1 and buck 2

40.0

20.0

0.0 0

1

2

3

4

5

6

7

8

9

10

Time(s) Time (s) Fig. 15 The input and output power of added Buck converters in Group 2. 100 Solar radiation

60

600

40

200

Group 1

0 10:11

10:44 Time

Time (s)

Fig. 16 Total power generation of a day.

400

Group 2

20 0 9:38

800

11:18

11:51

Solar radiation( W/m2)

Power(W) Power (W)

80

Solar radiation (W/m2)

1000 1,000

89


90

groups in 2.5 hours. The total power generation of Group 1 (the conventional system) is 68.71 Wh while Group 2 (the improved system) is 123.73 Wh (including the power loss of added Buck converters). The power generation of Group 2 is raised to 80.1% as compared with Group 1 (take 87.6% as Buck converters’ efficiency, 107.01% raised if the power loss is not included). So it is proved that the idea of adding Buck converters can actually improve generated power of the system under the condition of mismatch.

[2]

[3]

[4]

6. Conclusion This paper proposes the idea that solving the mismatch problem of conventional household PV generation system without changing the conventional power-conditioner. In PSIM simulation and experiments, it is proved that, by adding a Buck converter behind each module respectively, the generated power of the conventional system can be greatly increased under the condition of mismatch.

[5]

[6]

[7]

References [1]

Sanz, A., Vidaurrazaga, I., Pereda, A., Alonso, R., Roman, E., and Martinez, V. 2011. “Centralized VS Distributed (Power Optimizer) PV System Architecture Field Test Results under Mismatched Operating Conditions.” In Proceedings of 2011 the 37th IEEE Photovoltaic

[8]

Specialists Conference, 2435-40. Qiu, Y. N., Betts, T. R., and Gottschalg, R. 2009. “Electrical Mismatch within Single Junction Amorphous Silicon and Micromorph Tandem Thin Film PV Modules.” In Proceedings of 2009 the 34th IEEE Photovoltaic Specialists Conference, 911-6. Paul, P., Ghosh, S. K., Ghosh, K., and Mukherjee, D. 2011. “Impact of Mismatch Losses Arising in Crystalline and Amorphous Silicon PV Modules an Indian Experience.” In Proceedings of 2011 World Congress on Sustainable Technologies, 153-5. Huang, C. Y., Itako, K., Mori, T., and Ge, Q. 2014. “MPPT Control Method Using Boost Type DC-DC Converter for PV Generation System with Mismatched Modules.” Presented at 2014 International Conference on Power Engineering, Energy and Electrical Drives, Singapore. Subudhi, B., and Pradhan, R. 2012. “A Comparative Study on Maximum Power Point Tracking Techniques for Photovoltaic Power Systems.” IEEE Transactions on Sustainable Energy 4 (1): 89-98. Hua, C. C., Lin, J. R., and Shen, C. M. 1998. “Implementation of a DSP-Controlled Photovoltaic System with Peak Power Tracking.” IEEE Transactions on Industrial Electronics 45 (1): 99-107. Koutroulis, E., Kalaitzakis, K., and Voulgaris, N. C. 2001. “Development of a Microcontroller-Based Photovoltaic Maximum Power Point Tracking Control System.” IEEE Transactions on Power Electronics 16 (1): 46-54. Femia, N., Petrone, G., Spagnuolo, G., and Vitelli, M. 2005. “Optimization of Perturb and Observe Maximum Power Point Tracking Method.” IEEE Transactions on Power Electronics 20 (4): 963-73.

D


DAVID

PUBLISHING

Application of an Analytical Method to Locate a Mixing Plane in a Supersonic Compressor Emmanuel Benichou1 and Isabelle Trébinjac2 1. Turbomeca, Groupe SAFRAN, Bordes 64511, France 2. Laboratoire de Mécanique des Fluides et d’Acoustique, UMR CNRS 5509, Ecole Centrale de Lyon, Ecully Cedex 69134, France Received: September 12, 2014 / Accepted: November 03, 2014 / Published: January 31, 2015. Abstract: In order to achieve greater pressure ratios, compressor designers have the opportunity to use transonic configurations. In the supersonic part of the incoming flow, shock waves appear in the front part of the blades and propagate in the upstream direction. In case of multiple blade rows, steady simulations have to impose an azimuthal averaging (mixing plane) which prevents these shock waves to extend upstream. In the present paper, several mixing plane locations are numerically tested and compared in a supersonic configuration. An analytical method is used to describe the shock pattern. It enables to take a critical look at the CFD (computational fluid dynamics) steady results. Based on this method, the shock losses are also evaluated. The good agreement between analytical and numerical values shows that this method can be useful to wisely forecast the mixing plane location and to evaluate the shift in performances due to the presence of the mixing plane. Key words: Supersonic compressor, shock wave, pressure loss, RANS, mixing plane.

Nomenclature Symbol

P Pt Pt

t



Static pressure Stagnation pressure in the impeller frame Circumferentially averaged stagnation pressure Circumferential pitch Density

V Vn

Velocity in the impeller frame

Vt 1, Vt 2

Tangential velocity components

r a

Radius Speed of sound Perfect gas constant Rotation speed Position of the bow shock on the profile symmetry axis Coordinates of a point in the profile frame Mach number



 x0 x, y M

Normal velocity component

Corresponding author: Emmanuel Benichou, Ph.D. student, research fields: aerodynamic instabilities in centrifugal compressors, including rotor-stator interactions and flow control issues using boundary layer aspiration. E-mail: [email protected].

µ

  eB d

 θ K

Mach angle Angle of the flow Angle of the bow shock Axial distance between points x0 and B Detachment distance of the shock wave Pitchwise distance on which the flow is considered isentropic Circumferential direction Total pressure loss coefficient

Subscript B, C 1, 2 ∞

Relative to points B and C Calculated in Section 1 or Section 2 Value of the quantity at infinite upstream

Exponent *

Value of the quantity at M = 1

1. Introduction The need for compact, efficient high pressure ratio compressors often results in high rotation speeds. In some cases, the entry flow may therefore be supersonic over the entire- or upper-span. The resulting physics of the flow field in the entry zone can be complex because

92

Application of an Analytical Method to Locate a Mixing Plane in a Supersonic Compressor

of the interaction between compression and expansion waves [1-3]. Besides, shock waves propagating upstream the blades are dissipative and must necessarily be taken into account in the prediction of the stage performances [4]. Current CFD (computational fluid dynamics) offers three main categories for simulations: RANS (Reynolds-averaged Navier-Stokes), LES (large eddy simulation) and DNS (direct numerical simulation). RANS simulations approximate the mean effect of turbulence, while DNS enables a full resolution of the Navier-Stokes equations, from the smallest turbulence scale (Kolmogorov scale) up to the integral scale. LES corresponds to a filtered DNS: only the largest turbulence scales are resolved, the smallest ones being modeled. The more turbulence scales are resolved, the finer the mesh must be, and thus the more expensive the simulation becomes. In a current engine design process, only RANS simulations can be carried out. This sort of simulation is based on the Reynolds decomposition and turbulence models are added to close the set of equations. U-RANS (unsteady RANS) simulations are generally not affordable in a conception approach, because of their high CPU cost. That is why the only tool typically available for designers today consists in steady RANS simulations. In the case of multi-row turbomachinery, these simulations rely most of the time on the use of mixing planes, which average the data in the circumferential direction, and thus do not let the non-uniformities in the flow field transmit in the upstream or downstream direction. This article focuses on the entry zone of a supersonic compressor. The filtration of the shock pattern upstream the blades by the mixing plane raises the issue of the mixing plane location. The present paper compares numerically different locations to point out the influence of the mixing plane on the flow field, notably in terms of stagnation pressure change. In a first part, the numerical results are qualitatively analyzed and the role of the mixing plane is highlighted.

An analytical model is then used to reproduce the shape of the shock waves at the leading edge of the blades. Finally, a method is given that enables to judge the reliability of steady simulations on supersonic configurations.

2. Test Case and Numerical Procedure The test case is a centrifugal unshrouded impeller designed and built by Turbomeca. In the present study, only the front part of it is concerned. There is therefore no need giving the compressor geometry and performance in this paper. Only one operating point is examined, and it corresponds to the sonic blockage region. The flow, supersonic over 60% of the span, is examined at several section heights between h1 and h2 in Fig. 1. A simulation performed without any mixing plane is used as reference (Fig. 2a). In this one, the inlet block is rotating with the impeller, and there is no particular interface. In order to evaluate the change in performance induced by the mixing plane approach, three different mixing plane locations are numerically tested (Fig. 2b). They are labeled “a”, “b” and “c” in Fig. 1. In those three simulations, the inlet block is fixed, like a stator row and the periodicity enables to use a smaller domain since the flow upstream of the mixing plane is circumferentially uniform. Computations were performed with the elsA software developed at ONERA [5]. The code is based on a cell-centered finite volume method and solves the

h1 h2 Air inlet Fig. 1 Meridional sketch of the compressor inlet part.


93

for convective fluxes and a 2nd-order centered scheme for viscous fluxes. An LU implicit phase (lower upper decomposition) is associated to the backward-Euler scheme for time integration. The near-wall region is described with y+ < 13. The inlet condition imposes the velocity angles and the standard stagnation pressure and temperature. The turbulent values are determined from a free-stream turbulence rate of 5% (resulting from previous measurements). The outlet condition imposes a uniform value of static pressure. The walls are described with non-slip and adiabatic conditions. The steady state enables to simulate only one blade passage, the azimuthal boundaries being periodic. Blade-to-blade surfaces are then extracted from all simulations at the same section heights between h1 and h2. At the mixing plane interface, a circumferential average using Riemann invariants is computed at both upstream and downstream faces. The resulting values (m1-m5) are then applied to the adjacent face with a non-reflective boundary condition:

Outlet

Splitter blade

Main blade Periodic boundaries

Inlet



(a)

Mixing plane

 =0

1  m1  S  ( P   aVn )dS  m  1 ( P   aV )dS n  2 S  1  2 m3   ( P   a )dS S  1  m4  S  Vt dS  m  1 V dS  5 S  t 1

2

with S   d S the total surface of the interface. (b) Fig. 2 (a) 3D view of the domain without mixing plane; and (b) 3D view of the domain with mixing plane.

compressible RANS equations on multi-block structured meshes. The k-l model of Smith [6] (chosen according previous work [7]) is used for turbulence modeling. The set of equations are resolved in the relative frame of each row, using the Roe space scheme

3. Numerical Results Fig. 3 shows the relative Mach number in the reference case (without any mixing plane), in a blade-to-blade surface. The white lines indicate the three mixing plane locations and the purple line shows the iso-contour M = 1. With the mixing plane located in (c), the subsonic zone between the leading edge and the shock wave is clearly cut.

94


(a) (b) (c)

Section 1

Section 2

Without mixing plane

Case (a)

Fig. 3 Relative Mach number without mixing plane.

Fig. 4 shows that in the four configurations, the detachment distance of the shock remains constant, which tends to prove that the mixing plane gives a correct value of the averaged Mach number. Indeed, as explained in the following, the detachment distance only depends on the blade geometry and inlet Mach number. The white lines represent Mach iso-contours from 0.7 to 1.4. However, the shape of the subsonic zone is seriously affected. The more the mixing plane is located downstream, the less the shock waves can extend upstream. Thereby, according to the position of the mixing plane, the total pressure loss is under-estimated. The change in stagnation pressure can be quantified with the value of the loss coefficient K calculated as:

Pt 1

with Pt , the momentum-averaged relative stagnation pressure integrated on the whole surface at Section 2 (located at the blade leading edge) and on the whole surface at Section 1 (located upstream) as:

 P  V t

n

 dS

S

 V

n

S

Case (c)

Pt 2

K  1

Pt 

Case (b)

 dS

Fig. 4 Relative Mach number in the four test cases.

As is expected, the more the mixing plane is located downstream, the lower the losses are, and consequently, the more the massflow is over-estimated. Table 1 gives the difference between overall inlet massflow in cases

Application of an Analytical Method to Locate a Mixing Plane in a Supersonic Compressor Table 1

Performance shift to the reference configuration.

Mixing plane located in (a) Mixing plane located in (b) Mixing plane located in (c)

Massflow +0.04% +0.35% +0.80%

95

LE

K/Kref -1.2% -8.3% -21.2% (c)

(a), (b), (c) compared to the reference configuration without mixing plane.

(b)

The effect of the mixing plane can also be shown by

(a)

the evolution of the entropy along a constant span height in the supersonic region (black arrow in the meridional view in Fig. 5). By preventing the shock wave to extend upstream, the mixing plane introduces a

Without mixing plane

shift in the global level of entropy. These data show that the result of the steady state simulation significantly depends on the mixing plane location both in terms of performance (massflow and loss) and flow topology. The objective of the following part is to propose a method which can be used to: (1) forecast the location of the mixing plane minimizing the shift in performance; (2) forecast the change in performance for a given mixing plane location.

4. Analytical Description Let us consider a supersonic incoming flow compressor with subsonic axial velocity component. Depending on the inlet Mach number and on the back pressure level, two different regimes can exist:

Fig. 5 Evolution of entropy along the rotation axis z, at a constant span height.

Moreover, the presence of splitter blades here is likely to influence the shock system at the leading edge of the main blades through potential effects. Thus, the back pressure has to be imposed very low. The two inputs of this model are the upstream Mach number and the geometry of the blade leading edge. The detachment distance is calculated with Moeckel’s method [8], which assumes that the detached shock has a hyperbolic shape (Fig. 6). The equation of the hyperbola is specified by its asymptote (of angle ) and the position of point C located on the sonic line [BC], which is supposed to be straight.

 the unstarted regime, characterized by a detached, quasi-normal shock across the passage;  the started regime, characterized by an attached oblique shock.

yC

In case of a blunt leading edge, the shock cannot be

C

C

C

C’

strictly attached and a small subsonic area exists upstream the blade. The detachment distance of the shock is obviously smaller in the case of started regime than that of unstarted regime. The model presented hereafter is only valid for a

M 1 O

x0

A

started regime. This is the first reason why the study

d

takes place near the sonic blockage: the shock wave has

eB

to remain attached to the leading edge of the blades.

Fig. 6 Sketch of a detached shock.

B

B x

96


In order to apply Moeckel’s model, the upstream flow is supposed to be two-dimensional, uniform and the profile is approximated by a symmetric shape. The thermodynamic shock relations for ideal gas are used. The maximum entropy rise, corresponding to a normal shock, is located on the streamline passing by the leading edge. This line and the sonic line are assumed to be straight. The available equations are:  The equation of the hyperbola:

yc2  ( xc2  x02 ) tg 2  

(1)

 The equation of the tangent at point C:

tg C 

(2)

calculated. In the present case, a normal shock is considered at point C: 1



 1  * a*  2 2  1  1 2      1 M 1 1 1       2  C* aC*   1    1 M 

(9)

It would also be thinkable to consider an oblique

uncertainty in this method and a degree of freedom for measurements but it is rather difficult to give a formula which works for all types of blades).

(3)

tg 2 

The steps of the detachment distance calculation are thus:

 The equation of the distance eB: tg eB  yC 2 C  ( yC  y B )tg C tg  

tg 2 C  tg 2  

(4)

tg 2  

In order to calculate the ordinate of point C, yC, the continuity equation is written between the segment [OC’] upstream the shock wave and the sonic line [BC]:

yC  yB cos  C

 From the value of the inlet Mach number M,  is calculated with: 1    arcsin (10) M

and the values of the deviation C and shock angle C are deduced from the shock relations.  Point B which belongs to the profile is identified from its tangent B which has to be equal to C.  Eqs. (3), (4) and (6) are then used to calculate the detachment distance d:

(5)

which leads to:

yC  yB

that Eq. (8) in the way the stagnation pressure ratio is

the user (playing on this ratio enables to fit CFD or

tg 2 C  tg 2 

V yC  C* aC* BC  C* aC*

It is important to notice that a choice is possible at

sonic line. This is at the same time a source of

xC 2 tg   yC

 yC

(8)

shock to evaluate the thermodynamic state along the

which, combined with Eq. (1), gives:

x0  yC

 * a * Pt    C* aC* Pt C

d  eB   xB  xA 

(11)

This procedure, initially thought for symmetric isolated profiles, leads to a geometric representation

1 1

 V  a cos  C  * a*  a *  * C

*  * C

(6)



2



the hyperbola and the sonic line. The part of the shock wave confining this subsonic region is

Since a shock wave is isenthalpic, we can write:

 1 2    V  M  (1  2 M  )     1  * a*  

of the subsonic zone (in grey in Fig. 6), between

responsible for most of the total pressure losses.

 1 2(1   )

Therefore, it is crucial to let it fully extend upstream (7)

the blades. The consequences of a mixing plane cutting the subsonic pocket are examined in the following.


97

5. Comparison between Analytical and Numerical Results The analytical method previously described for started regimes enables to draw the shape of the detached bow shock and the subsonic zone. Fig. 7 superimposes the analytical shape (in black) on the CFD results without any mixing plane (the yellow line is the iso-contour M = 1). Qualitatively, the subsonic area is well approximated. Assuming that the major part of the losses are due to the part of the shock confining the subsonic part, it is seen that the analytical calculation gives a pretty good evaluation of the minimal distance to be put between the mixing plane and the leading edge (dmin in Fig. 7). However, two weaknesses should be pointed out. Firstly, the model is very sensitive to the upstream Mach number, in particular in the low supersonic Mach

dmin

Fig. 7 Application of Moeckel’s method with M ≈ 1.3.

numbers region (between M = 1.0 and M = 1.1). Fig. 8 shows the typical evolution of the detachment distance as a function of the upstream Mach number. Furthermore, when the upstream Mach number decreases, the shape of the subsonic pocket tends to become more elliptic. Fig. 9 shows that with M = 1.1, the detachment distance and the orientation of the shock wave are still well estimated but Moeckel’s method cannot reproduce the numerical result near the leading edge. Secondly, the choice in the calculation of the loss across the shock leads to the thermodynamic state of the sonic line, because in the end, it directly drives the value of the detachment distance. Since Eq. (9) is non linear, the higher the upstream Mach number, the more important the way the loss is calculated becomes. The main benefits of Moeckel’s model are that only the geometry of the profile and the upstream Mach number are needed. As a consequence, this is a tool which can easily be employed in a pre-design phase. It should be underlined that the sonic blockage is the most “favorable” case because when the mass flow decreases, the compressor will pass from started

Fig. 8 Evolution of the detachment distance for a given profile.

Fig. 9 Application of Moeckel’s method with M ≈ 1.1.

98


regime to unstarted regime. Thus, the shock pattern moves upstream, so that the impact of the mixing plane can only become stronger.

Section 1

Section 2

6. Shock Loss Prediction Although current blade design relies on numerical optimization, including transonic bladings [9], many analytical shock loss models exist, as described in Refs. [10, 11], for example. Bloch, Copenhaver and O’Brien also give an interesting approach based on Moeckel’s method [12], adapted to unstarted regime. Since the other objective of this paper is to analytically forecast the change in pressure loss for a given mixing plane location compared with the reference case (without any mixing plane), the shock loss associated to a started regime has to be evaluated. Fig. 10 shows the relative total pressure calculated in the reference case (without mixing plane): the low relative stagnation pressure zone at the leading edge plane corresponds to the projection of the strongest part of the shock on the leading edge plane in the mean flow direction. In order to be consistent with the numerical fields, only the part of the shock wave which is downstream the mixing plane location is kept and this portion of the upper branch of the hyperbola is projected in the plane of the leading edge (Fig. 11). The losses are then calculated in the plane of the leading edge as if the fluid going through the bow shock was seeing a normal shock and the rest of the incoming flow remained unperturbed. The losses are evaluated with the coefficient K defined as:

  Pt 2 t     K  1    t   t Pt 1

max

min

Incoming flow

Fig. 10 Relative stagnation pressure in a blade-to-blade surface. Mixing plane

t 

LE

Isentropic part : no pressure loss

(12)

where,  is the length of the projected hyperbola, t is the pitch and Pt 2 / Pt 1 refers to the total pressure ratio across a normal shock (see Eq. (9)). The value of K is of course overestimated, since a part of the incoming fluid actually goes across an oblique shock and the bow shock extending upstream becomes rapidly evanescent. But it enables to evaluate the shock loss in the most unfavorable situation. Note



Normal shock loss

Bow shock cut by the mixing plane Incoming flow Fig. 11 Loss calculation for a given Mach number with a given mixing plane location.


that in case of no mixing plane, the model is equivalent to considering a normal shock over the whole pitch. This choice keeps the present tool simple, but it would be possible to make a more sophisticated model:  First, by improving the shock wave approximation. A good example is given by Ottavy [13], by coupling Levine’s method [14] to predict the unique incidence seen by the blade profile with Moeckel’s detachment distance calculation.  By discretizing the shock wave, so that the flow angle and the pressure loss change from a normal shock on the blade axis to an oblique shock and integrating the result on a pitch to the neighboring blade. For a given profile, it is possible to plot the loss as a function of the upstream Mach number in the different configurations (Fig. 12). The case without mixing plane corresponds to the shock loss across a normal shock. Depending on its location, the mixing plane has no more impact beyond a certain Mach number. For example, with a mixing plane located in Section (b) (green curve), it can be seen that beyond M = 1.46, the mixing plane has no more influence. It means that if the upstream Mach number is higher than this value, the steady results can be considered reliable. For a given mixing plane location, the discrepancy compared to the reference case firstly increases with the Mach number. Indeed, the length  continuously increases but the pressure loss is increasing faster due to the shock wave. Then the upper branch of the hyperbola is more and more straightened up together with a decrease in the detachment distance until its projection covers the whole pitch. At that step, there is no more difference with or without a mixing plane and a steady RANS simulation can be considered as reliable (Fig. 13). This analytical tool has been applied to the three mixing plane positions and to the reference case with no mixing plane. The input data are the circumferentially-averaged Mach numbers in Section (b) coming from the CFD with no mixing plane and the geometry of the front part of the blades.

99

(a)

(b) Fig. 12 (a) Evolution of K for a given profile; and (b) zoom from Fig. 12a.

M = 1.20

M = 1.50

M = 1.80

Fig. 13 Effect of the mixing plane on the bow shock for different upstream Mach numbers.

100


Fig. 14 Analytical and numerical loss as function of the upstream Mach number for different mixing plane locations.

Fig. 14 compares the analytical and numerical values obtained for the losses. Different section heights between h1 and h2 have been tested, so that both the profile and the Mach number were changing. The values of K are plotted as a function of the upstream Mach number and are calculated as: Pt 2

K 1

Pt 1

where, 2π

 P  V t

Pt 

n

 rd

0

2π

 V

n

 rd

0

In Fig. 14, the analytical evolution corresponding to case (a) is the same as the reference one for Mach numbers greater than 1.30. This means that the major part of the hyperbola is contained downstream the mixing plane for the corresponding section heights. According to this criterion, any mixing plane should be located upstream Section (a) (Fig. 3) for the present impeller. Nevertheless, despite the good qualitative agreement between analytical and numerical results in Fig. 7, we can observe quantitative discrepancies in the shock loss.

First of all, the simulation without mixing plane describes a different loss evolution around M = 1.20 than those with a mixing plane. This is a well-known problem with mixing planes in general: the loss is radially redistributed. The influence of the location of the mixing plane is clearly visible in the numerical curves. The slopes are not so far from the analytical ones. But there is a sort of offset between the numerical and the analytical results. It is probable that for the low Mach numbers, the shock loss is low compared to the viscous loss. To compare properly the two curves families, it would be necessary to take from the Navier-Stokes simulations only the shock loss, as done in Ref. [12] by subtracting the friction loss from measurements. The discrepancies are also due to the strong hypotheses made in the analytical method, which consists in a two-dimensional approach and due to the choice made for the loss calculation. Real blade profiles are also often cambered and not symmetric in order to produce lift, which is not taken into account in the present model. And finally, the numerical loss at the highest Mach numbers (close to the section height h2) is suddenly increasing, near the shroud boundary layer. It is likely that friction loss dominates shock loss in this area. Thanks to this simple model, the order of magnitude of the under prediction of the losses due to the introduction of a mixing plane is easily evaluated. It has been tested that this approach gives acceptable results from inlet Mach number larger than 1.1. Once again, the major drawbacks of this method are that it gives wrong predictions for lower Mach numbers and that the shock formulas which are used in it are very sensitive. This is maybe one of the reasons why Bloch et al. [12] had to take into account an “effective” leading edge radius in their model dedicated to predict the shock loss through the lower branch of Moeckel’s hyperbola, in supersonic compressor cascades operating in unstarted regime. Indeed, they increased the leading edge thickness until the analytical results


matched the experimental ones. This reminds us of the difficulty of implementing a shock loss model that fits all types of profiles, under various operating conditions.

[2] [3] [4]

7. Conclusion Steady state numerical simulations performed with a mixing plane approach show that the results, in terms of mass flow and losses, significantly depend on the mixing plane position. The operating point chosen here corresponds to the sonic blockage of the compressor but for near-surge points, it would be even more important. The fact that this study takes place near the blockage enables to propose an analytical method in order to forecast this change in performance. The validity of this analytical method is checked by comparing its results with the numerical ones in the entry zone of a transonic compressor. Analytical and numerical results show good agreement. This tool may be useful for transonic compressor design: first, to have an idea of the minimal distance that should be put between the mixing plane and the leading edge of the blades, and then to know how representatively the steady simulations can be expected.

Acknowledgements We would like to thank Turbomeca which supported this study, together with ONERA which collaborated on the numerical simulation. This work was granted access to the HPC resources of CINES under the allocation 2013-2a6356.

References [1]

Lichtfuss, J. J., and Starken, H. 1974. “Supersonic Cascade Flow.” Progress in Aerospace Sciences 15: 37-149.

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

101

Kantrowitz, A. 1946. The Supersonic Axial-Flow Compressor. NACA technical report. Chauvin, J. 1970. Supersonic Turbo-Jet Propulsion Systems and Components. AGARD report No. 120. Trébinjac, I., Ottavy, X., Rochuon, N., and Bulot, N. 2009. “On the Validity of Steady Calculations with Shock-Blade Row Interaction in Compressors.” In Proceedings of the 9th International Symposium on Experimental and Computational Aerothermodynamics of Internal Flows, ISAIF9-062. Cambier, L., and Gazaix, M. 2002. “elsA: An Efficient Object-Oriented Solution to CFD Complexity.” Presented at 2002 the 40th AIAA Aerospace Science Meeting and Exhibit, Reno, USA. Smith, B. R. 1995. “Prediction of Hypersonic Shock Wave Turbulent Boundary Layer Interactions with k-l Two-Equations Turbulence Model.” Presented at 1995 the 33rd AIAA Aerospace Sciences Meeting and Exhibition, Reno, USA. Rochuon, N. 2007. “Analysis of the Three-Dimensional Unsteady Flow in a High Pressure Ratio Centrifugal Compressor.” Ph.D. thesis, École centrale de Lyon. Moeckel, W. E. 1942. Approximate Method for Predicting Form and Location of Detached Shock Waves. NACA technical report. Burguburu, S., Toussaint, C., Bonhomme, C., and Leroy, G. 2004. “Numerical Optimization of Turbomachinery Bladings.” Journal of Turbomachinery 126 (1): 91-100. König, W. M., Hennecke, D. K., and Fottner, L. 1996. “Improved Blade Profile Loss and Deviation Angle Models for Advanced Transonic Compressor Bladings: Part II—A Model for Supersonic Flow.” Journal of Turbomachinery 118 (1): 81-7. Schobeiri, M. T. 1997. “Advanced Compressor Loss Correlations, Part I: Theroretical Aspects.” International Journal of Rotating Marchinery 3 (3): 163-77. Bloch, G. S., Copenhaver, W. W., and O’Brien, W. F. 1999. “A Shock Loss Model for Supersonic Compressor Cascades.” Journal of Turbomachinery 121 (1): 28-35. Ottavy, X. 1999. “Laser Anemometry Measurements in an Axial Transonic Compressor. Analysis of the Unsteady Periodic Structures.” Ph.D. thesis, École centrale de Lyon. Levine, P. 1956. The Two-Dimensional Inflow Conditions for a Supersonic Compressor with Curved Blades. WADC technical report 55-387.

D


DAVID

PUBLISHING

Thermal Design of Power Transformers via CFD Ralf Wittmaack Siemens AG, EM TR LPT GTC RES PN, Katzwanger Str. 150, Nürnberg 90461, Germany Received: August 18, 2014 / Accepted: September 30, 2014 / Published: January 31, 2015. Abstract: At Siemens, an in-house CFD (computational fluid dynamics) code UniFlow is used to investigate fluid flow and heat transfer in oil-immersed and dry-type transformers, as well as transformer components like windings, cores, tank walls, and radiators. This paper outlines its physical models and numerical solution methods. Furthermore, for oil-immersed transformers, it presents an application to a HV (high voltage) winding in a traction transformer of locomotives, cooled by synthetic ester. Key words: Thermal design, CFD, physical models, numerical methods.

1. Introduction The life time of power transformers is substantially influenced by chemical degradation processes occurring in the electrical insulation. Since the speed of these processes depends significantly on temperature, the proper prediction of component temperatures during transformer operation is a crucial part of the design process. There are several sources of heat in the transformer. If a time varying voltage is applied, magnetic hysteresis effects and eddy currents lead to no-load losses in the steel sheets of the core. In addition, during normal operation, the electrical currents cause ohmic and stray load losses. To keep the temperatures of the transformer components within acceptable limits, appropriate cooling is essential. Depending on the type of transformer, this is normally accomplished via natural or forced convection of the cooling fluids air or oil. In addition to mineral and silicone oil, natural and synthetic ester fluids are also used. Thanks to its flexibility and accuracy, CFD (computational fluid dynamics) is increasingly being used to analyse transformer thermal design. This follows the trend established in other branches of Corresponding author: Ralf Wittmaack, Dr.-Ing., research fields: CFD code development and application on nuclear safety, gas turbine development, aerospace research, and electrical transformers. E-mail: [email protected].

advanced technology development like aerospace, automotive, and power generation, where CFD simulations are indispensable parts of the product development cycles. Employing commercial CFD codes, several detailed studies of disc-type transformer windings were performed by Torriano et al. [1] and Jiao [2]. Moreover, extended full geometry CFD analyses coupled to electromagnetic simulation of the load and no-load losses in core and windings were presented by Smolka et al. [3, 4]. Furthermore, combined oil and air flows in distribution transformers were investigated with commercial CFD codes by Fonte et al. [5] and Gastelluritia et al. [6]. Our intention is to provide a simulation method that may be used for detailed CFD analyses on fine grids as well as for simplified coarse grid studies. The in-house code UniFlow is designed to be applicable also by users with limited experience in CFD. For this reason, e.g., material attributes are employed for a convenient coupling of fluid and solid regions in conjugate heat transfer simulations.

2. Physical Models and Numerical Methods 2.1 Physical Models Our physical model is aimed at investigating flows

Thermal Design of Power Transformers via CFD

with several kinds of heat transfer in a complex geometry. It simulates the flow of single-component, incompressible Newtonian fluids in a three-dimensional geometry. In addition to the fluids, in gaseous or liquid state, several structural materials are considered as hydrodynamic obstacles and thermodynamic heat structures. The hydrodynamics is described by the continuity equation and the Navier-Stokes equation. For the simulation of turbulence, the algebraic eddy viscosity model of Baldwin and Lomax [7] is available. To simulate the transition between laminar and turbulent flows, algebraic transition models of Drela [8] and Mayle [9] are on hand. For temperature dependent density or material properties of the viscous stress tensor, the hydrodynamics of the fluid is coupled to the thermodynamics. For this reason, internal heat transfer (by convection and conduction) and heat generation by internal sources as well as heat transfer to the surroundings are modelled via a heat transport equation. To allow for the simulation of phase transitions, it is provided in enthalpy formulation. At the rigid boundaries, heat conduction is considered. For coarse grids, convective heat transfer coefficients may be employed at solid-liquid interfaces. Radiant heat transfer is simulated at structural material surfaces. The material properties (density, dynamic viscosity, specific heat, heat conductivity, and convective heat transfer coefficient) depend on the temperature. Solids may have orthotropic heat conductivity. 2.1.1 Dynamic Equations Our dynamic equations are written in Cartesian coordinates. The continuity equation for incompressible flow is:

  ρv m  = 0 x m

(1)

According to Landau and Lifshitz [10], where ρ is density and v velocity, x is the space coordinates and we use Einstein’s summation convention for the space direction index m. Introduction of the continuity

103

equation into the Navier-Stokes equation [10] leads to a momentum equation in strong conservation form:

ρ

 v vm  vi   p + m  ρvi vm  μ  i + i  =  i + ρgi (2) t x  x  xm x 

where, t is time, p pressure, and g gravitational acceleration. After inclusion of the continuity equation, our heat transport equation in strong conservation form reads:

ρ

h  + m t x

 T   = Pd  ρ h v m  λ x  m  

(3)

Here, h is specific enthalpy, T temperature, λ heat conductivity, and Pd density of the heat sources or sinks. 2.1.2 Radiant Heat Transfer Model Radiant heat transfer may be simulated between structural material surfaces adjacent to the fluid. The employed radiation model assumes that the radiating surfaces are boundaries of a hollow space with linear dimension much greater than their distance. It is applicable for, e.g., parallel plates and concentric cylinders. With this simplifying assumption, the power received by surface “a” via the heat transfer from surface “b” is:



4

Pab = cab Aa Tb  Ta

4



(4)

According to Baehr and Stephan [11], where,

cab =

σ  1 Aa  1 +   1 εa Ab  εb 

(5)

Here, A is area of a radiating structural material surface, T surface temperature, σ = 5.67051 × 10-8 W/(m2·K4) Stefan-Boltzmann constant, and ε emissivity of a structural material surface. Computation domain nodes undergoing radiant heat transfer may have their radiation partner nodes inside the computation domain or at the boundary. 2.2 Numerical Methods For the numerical representation of our model, we

104


developed a finite volume method and employ boundary fitted, curvilinear, non-orthogonal, block-structured grids. The blocks may be connected via one-to-one or patched couplings. The arrangement of the dynamic variables in the control volumes of the grid is collocated at the node centre. The dynamic equations are solved sequentially. For the solution of the momentum, pressure-correction, and heat transport equations, we use implicit schemes. The system of continuity and momentum equations is solved by a SIMPLE [12], SIMPLEC [13], or PISO algorithm [14]. To speed up the code execution and to simplify the estimation of discretisation errors, a FAS (full approximation scheme) multi-grid algorithm is employed [15]. It is a geometric approach with standard coarsening applied to the outer iterations, visiting the grid levels in V-cycles. For steady-state problems, it operates as a FMG (full multi-grid) algorithm, whereas for transient problems, the algorithm starts at the finest grid. For the efficient solution of sparse linear equations, several algorithms are available. The parabolic momentum and heat transport equations may be solved with SIP (strongly implicit procedure) solvers that are modified to handle block couplings via the residual vector, as outlined by Ferziger and Peric [12]. Additionally, for the elliptic pressure-correction equation, an aggregation-based algebraic multi-grid algorithm of Notay [16] is available.

Fig. 1 Outside view of transformer without top cover.

Fig. 2 Active part of transformer without tap changer.

3.1 Geometry Model The single phase transformer has two limbs to make it more compact. We analyze one of the limbs and consider a periodic three-dimensional 7.5° segment of the circumference. The periodicity comes from 24 spacers along the circumference. Fig. 3 shows the initial specific enthalpy of an axial part of the angular segment of the winding. It explains the location of the materials: grey indicates ester, red pressboard, blue conductor, and light blue Nomex. This axial part consists of half a coil from the 92 coils in total.

3. Application In this section, we describe an investigation of a HV (high voltage) winding of a locomotive transformer of high speed trains. The goal is to find the maximum temperatures in the insulation materials to allow for the selection of appropriate materials that withstand the thermal load. To cool the winding, a pump driven flow of synthetic ester is used. Figs. 1 and 2 show an outside view with removed top cover and the active part of the transformer without tap changer.

Fig. 3 Location of materials in axial part of circumferential winding segment.


3.2 Grid A single block hexahedral grid with 706,560 nodes is used and 222,880 nodes represent the ester while the rest correspond to structural materials. The node lengths are in the range of 0.62 mm to 5.1 mm. 3.3 Boundary Conditions At the inlet, the pump driven oil flow occurs with a velocity of 0.9 m/s and a temperature of 364.55 K. All computation domain boundaries are adiabatic, except the inlet and outlet. 3.4 Material Properties Structural materials considered in this simulation are copper and its adjacent Nomex 410 isolation as well as pressboard and Nomex 994 isolation. In our grid, the winding is split up into an inner and an outer part, made up of the copper conductor and Nomex 410. This is aimed at separating the inner part that consists mainly of Cu, from the outer region, where the dominant part of the Nomex 410 insulation is located. Their mixture material properties are calculated via: n mat

n mat

i =1 1

i =1

ρmix =

 αi ρi , c p mix =

x

i

cp i ,

 n mat α  V ρ λmix =   i  , αi := i , xi := αi i (6)  i=1 λi  V ρmix   Here nmat is no. if considered structural materials, V volume, cp specific heat at constant pressure,  volume fraction, and x mass fraction. Several modes of operation exist in the transformer, e.g., rated current and overload. In our simulation, we consider rated current that corresponds to an average power density of the load losses Pd = 1.389 MW/m3. The load losses are calculated by a Maxwell solver and subsequently mapped to the CFD grid. 3.5 Non-dimensional Numbers and Boundary Layer Thickness With the width of a winding segment of l = 74.5 mm as characteristic length, the ester inlet velocity of 0.9

105

m/s, and the average ester temperature of 368 K, the Reynolds number R is 6,875 and the Prandtl number is 140. The ester flow along a winding segment resembles flow along a flat plate, where the transition from laminar to turbulent flow occurs between R = 3.5 × 105 and 3.5 × 106, according to Ref. [17]. This indicates that the ester flow is laminar, i.e., the hydrodynamic and thermal boundary layer thickness may be estimated according to Ref. [10]:

δh 

l R

; δt  δ h Pr

-

1 3

(7)

This leads to δh = 0.9 mm and δt = 0.17 mm at the end of a winding segment, i.e., the grid of our simulation of the traction transformer windings is too coarse to resolve the thermal boundary layer. To compensate for that, convective heat transfer coefficients are used at all interfaces of ester and structural materials. These were calculated via a detail model and a fine grid. 3.6 Simulation Results As a result of the axial flow barriers, there is a meandering oil flow in the radial channels. The maximum radial velocity component is similar to the inlet velocity, while the maximum vertical velocity is considerably higher than the inlet velocity. The maximum value occurs at the inner side, where the flow area is smaller. Figs. 4 and 5 show the radial and axial velocity components. Fig. 6 shows the calculated pressure in the ester. There is some stagnation pressure at the axial flow barriers. Since the limbs are oriented horizontally, we run our simulations without gravitational acceleration. For this reason, there is no hydrostatic contribution to the pressure. The temperature of ester and structural materials is shown in Figs. 7 and 8, at the ester and the spacer side of the winding. It indicates that the maximum temperatures are encountered at the spacer side. The maximum Cu temperature of 384.2 K is calculated at the spacer side of the windings. At that side of the HV windings, the temperature differs only

106


Fig. 4 Radial component of oil velocity.

slightly between Cu and the adjacent insulation materials Nomex 994 and pressboard. The maximum temperature of the structural materials is also very close. The maximum temperature in the Nomex 944 of 384.14 K and in the pressboard of 384.09 K are particularly interesting. According to Ref. [18], enhanced thermal ageing and cracking of the ester may occur at temperatures above 403 K. The maximum temperature of ester in the simulation is more than 27 K below this threshold.

6. Conclusions We analysed the thermal design of the HV windings of a traction transformer in the steady state with the CFD code UniFlow. This winding is cooled by Fig. 5 Axial component of oil velocity.

Fig. 6 Pressure of oil.

Fig. 7 Temperature at ester side of winding.

Fig. 8 Temperature at spacer side of winding.

synthetic ester in the OD (oil direct) mode. In addition to the result shown here, the thermal design of cast resin transformers can be studied, including radiant heat transfer between core, windings, and radiation cylinders. This is demonstrated in Ref. [19]. Other applications are related to detailed analyses on segments of disc windings with respect to, e.g., modelling of material compositions, width of oil channels. Another field of application is oil flows in transformer cores. Moreover, combined oil and air flows are analysed in the context of fin type distribution transformers. This is aimed at optimisation of the thermal efficiency of the fins and other tasks. Furthermore, combined oil and air flows in radiators can be investigated. In addition to steady state analyses, transient processes are investigated. One interesting type of transient occurs at the cold start of a transformer. This matters in particular for oil transformers where the dynamic viscosity is very high at low temperatures, especially for ester fluids. The future work will include validation of the calculated results via experimental data. The reason why this has not yet been accomplished is that detailed measurements inside power transformers are quite demanding, e.g., as a result of the high electric and magnetic fields. However, a comprehensive verification


of UniFlow was performed via analytical solutions. This covers steady viscous flow between parallel plates and concentric cylinders, potential flow around a cylinder, transient heat conduction in a parallelepiped, as well as heat convection and heat-up by a source. Furthermore, the simulation of transitional air flow onto a flat plate was validated via experiments outlined by Schlichting [17].

[7]

[8]

[9]

References [1]

[2]

[3]

[4]

[5]

[6]

Torriano, F., Pichler, P., and Chaaban, M. 2012. “Numerical Investigation of 3D Flow and Thermal Effects in a Disc-type Transformer Winding.” Applied Thermal Engineering 40: 121-31. Jiao, Y. 2012. “CFD Study on the Thermal Performance of Transformer Disc Windings without Oil Guides.” M.Sc thesis, KTH School of Industrial Engineering and Management. Smolka, J., and Nowak, A. J. 2008. “Experimental Validation of the Coupled Fluid Flow, Heat Transfer and Electromagnetic Numerical Model of the Medium Power Dry-Type Electrical Transformer.” International Journal of Thermal Sciences 47 (10): 1393-410. Smolka, J., Biro, O., and Nowak, A. J. 2009. “Numerical Simulation and Experimental Validation of Coupled Flow, Heat Transfer and Electromagnetic Problems in Electrical Transformers.” Archives of Computational Methods in Engineering 16 (3): 319-55. Fonte, C. M., Campelo, H., Sousa, R. G., Dias, M. M., Lopes, J. C. B., and Lopes, R. 2011. “CFD Analysis of Core-Type Power Transformers.” Presented at the 21st International Conference on Electrical Distribution, Frankfurt, Germany. Gastelluritia, J., Ramos, J. C., Larraona, G. S., Rivas, A., Izagirre, J., and del Río, L. 2011. “Numerical Modelling of

[10]

[11] [12] [13]

[14]

[15] [16]

[17] [18] [19]

107

Natural Convection of Oil inside Distribution Transformers.” Applied Thermal Engineering 31 (4): 493-505. Baldwin, B. S., and Lomax, H. 1978. “Thin Layer Approximation of and Algebraic Model for Separated Turbulent Flows.” AIAA-paper, NO. 78-0257. Drela, M., 1998. MISES Implementation of Modified Abu-Ghannam/Shaw Transition Criterion, MIT Aero-Astro, Boston, MA, USA. Mayle, R. E. 1991. “The Role of Laminar-Turbulent Transition in Gas Turbine Engines.” ASME Journal of Turbomachinery 113 (4): 509-36. Landau, L. D., and Lifshitz, E. M. 1989. “Fluid Mechanics.” In Course of Theoretical Physics. Oxford: Pergamon Press. Baehr, H. D., and Stephan, K. 2006. Heat and Mass Transfer. Berlin: Springer-Verlag. Ferziger, J. H., and Peric, M. 1999. Computational Methods for Fluid Dynamics. Berlin: Springer-Verlag. van Doormal, J. P., and Raithby, G. D. 1984. “Enhancements of the SIMPLE Method for Predicting Incompressible Flows.” Numerical Heat Transfer 7 (2): 147-63. Issa, R. I. 1986. “Solution of Implicitly Discretized Fluid Flow Equations by Operator Splitting.” Journal of Computational Physics 62 (1): 40-65. Trottenberg, U., Oosterlee, C. W., and Schüller, A. 2001. Multigrid. New York: Academic Press. Notay, Y. 2010. “An Aggregation-Based Algebraic Multigrid Method.” Electronic Transactions on Numerical Analysis 37: 123-46. Schlichting, H. 1979. Boundary-Layer Theory. New York: McGraw-Hill. Vosen, H. 1997. Kühlung und Belastbarkeit von Transformatoren. Berlin: VDE-Verlag GmbH. Wittmaack, R. 2014. “Thermal Design of Power Transformers via CFD.” Presented at the 11th World Congress on Computational Mechanics, Barcelona, Spain.

D


DAVID

PUBLISHING

Static Analysis of Eolic Blade through Finite Element Method and OOP C++ Tiago Freires1, Silvia Caroline1, Raimundo Menezes Jr2 and Moises Meneses Salvino3 1. Alternative and Renewable Energy, Department of Renewable Energy Engineering, Center Federal University of Paraiba, João Pessoa 58051-900, Brazil

2. Laboratory of Numerical Methods Distributed Processing, Department of Renewable Energy Engineering, Federal University of Paraiba, João Pessoa 58051-900, Brazil 3. Energy Efficiency Lab Hydraulics and Sanitation, Federal University of Paraiba, João Pessoa 58051-900, Brazil Received: July 27, 2014 / Accepted: September 30, 2014 / Published: January 31, 2015. Abstract: This work deals with a description of an elastic analysis of eolic blade (preprocessing, processing and post-processing stages). The eolic blade geometry is approximated by flat finite elements in which the membrane effects are evaluated using the FF (free formulation) finite element and the flexure effects are calculated using DKT (discrete shear triangle) finite element. The pre-processing stage is implemented using OpenGL library, to provide the graphical construction for geometry, mesh orientation, and other requirements of the finite element model. For the processing stage is built a specific dll (dynamic link library) library implemented in C++ language for the FF and DKT elements analysis. The post-processing stage has been built using specific dialogs to present all results in the graphic interface, where the static displacements of the eolic blade model are shown. Key words: Flat shell, FEM (finite element method), DKT, FF.

1. Introduction One of the most important problems involving generation of energy through wind source is the structural study of the blades. These structures have a high degree of complexity and its construction in general is based on international norms (expensive experimental tests) for certification of their safety in use. One of the potential tools to study and design of eolic blades is the numerical methods [1, 2] (FDM (finite difference method), FEM (finite element method), BEM (boundary element method), etc.). In this paper, it is described an attempt to use the object-oriented programming C++ and finite element method numerical capabilities to build a tool related to analysis of eolic blades with geometry approximated

Corresponding author: Raimundo Menezes Jr, Ph.D., research field: mechanical engineering. E-mail: [email protected].

by triangular flat elements. In addition, standard drawing functions from OpenGL library are used to provide a more friendly and efficient pre-processing to input element geometry, mechanical properties, element connectivity, and boundary constraints. In the processing stage, the stiffness matrices related to membrane and bending effect and equivalent nodal forces (wind pressure) are evaluated by functions written in oriented object language C++ and compiled in one dll (dynamic link library) called only in the process of analysis. Finally, in the post-processing stage, all results obtained in the analysis process are shown in specific dialogs box.

2. Pre-processing The pre-processing is the stage where input data (such as geometry, node and element numbering, mechanical properties, boundary conditions, loading) are set to perform the calculation of the discretized

Static Analysis of Eolic Blade through Finite Element Method and OOP C++

Fig. 1

109

Full discretized model.

problem. A full discretized model of a wind generator in the tool of analysis is shown in Fig. 1. The tool provides all drawing capabilities necessary to incorporate remaining FEM input data information (elasticity modulus, poisson ratio, thickness, etc.), and the OpenGL library is used in order to implement the drawing routines and to assign structural analysis data. OpenGL is a cross-language API (application programming interface) for writing applications that produce 2D and 3D computer graphics [3]. The interface consists of over 250 different function calls which can be used to draw complex three-dimensional scenes from simple primitives [4, 5]. In Fig. 2, a piece of source code to draw the triangle finite element is visualized. The parametric tool use OpenGL when the user selects the menu option “Pre-Processing -> Parametric Modeling”, so that this corresponds to execute the instructions in the source code as shown in Fig. 2. In Fig. 3, the source code of the function to draw the nodes of the finite element is shown. When the file data with coordinate nodes, elements connections are opened, the mesh of the eolic blade in environment is automatically generated, producing a graphical representation depicted in Fig. 4.

The next step in pre-processing is related to the definition of structural element properties. This procedure can be done in specifics dialogs for the elements. All material properties are set by user and applied to the model. For the nodes, the wind load and boundary conditions that represent the fixed position for the eolic blade must be assigned. This can be done in environment dialog box, where the all six degrees in all nodes of the model can be accessed and one by one conveniently prescribed by user, as shown in Fig. 5. The final representation of the model, with the wind load and nodal constraints applied, can be visualized in Fig. 6. In this stage, the model is done for analysis.

3. Processing The processing is stage of the analysis in which main calculations (such as elemental stiffness matrix evaluation, structural stiffness matrix assembling, nodal equivalent force vector evaluation, algebraic system solution) are done. In present paper, the structural eolic blade problem is analyzed by superposition of bending and membrane effects using FEM. For membrane effects, FF (free formulation) finite element originally developed by Ref. [6] is used.

110

Fig. 2

Static Analysis of Eolic Blade through Finite e Element Me ethod and OO OP C++

Functtion to draw th he triangle finiite element.

Fig. 3 Functtion to draw noodes of the finiite element.

Fig. 4

Discrretized blade model. m


Fig. 5

Element and node properties.

Fig. 6

Wind load and fixing conditions for the eolic blade.

The main characteristics of this element are triangular flat geometry, three DOF (degrees of freedom) by node (two displacements on plane and one drilling rotation perpendicular to the plane) located at each triangle vertex. For bending effect analysis is used the flat triangular element DKT (Discrete Kirchhoff Theory), that have three DOF (one transverse displacement and two slopes) by node located at each triangle vertex. This element was developed to deal with thin bending plate problems and its formulation has been thoroughly discussed in Refs. [7-9]. The eighteen DOF of the

Membrane effects

Bending plate effects

Fig. 7

DKT and FF finite elements association.

111


112

Fig. 8

Classes and Functions.

wind turbine tower analysis element (bending + membrane problems) are shown in Fig. 7. The analysis is done using the Saproms.dll library, in which functions and classes are implemented in oriented object language C++. For sake of conciseness, only the principal classes and functions of the Saproms.dll are described in Fig. 8. The main function that uses methods and objects from Saproms.dll is “solver”. This function works in two steps: in the first step, the objects and functions in Saproms.exe collect the input data assigned in pre-processing stage, and store them in vectors. In the second step, these vectors are processed by objects and methods from Saproms. dll to perform the structural calculations. From the point of view of mathematics, the “solver” function for the static analysis uses the principle of total strain energy. The governing equation of the problem is given by: where,

∑

matrix of the structure, and , displacement and nodal forces.

(1) is the stiffness are the vectors of

The equivalent nodal force vector is obtained from the external work done by the wind loads are expressed by: ,

,

d

(2)

where, , , , are the displacement and the wind loading in the element, A is the area of the element, Te is the external work done by the wind. The equivalent nodal force vector is equal to the vector derived from the work of external loads compared to degrees of freedom, so that the establishment of interpolations to , and , along the area of the element is necessary, assuming a linear variation in Fig. 9: 1 (3) 1 (4) where, and are coordinates of the finite element surface. Substitute Eqs. (3) and (4) in Eq. (2), 1

1 d

(5)

By minimizing the potential energy due to external loads:

Static Analysis of Eolic Blade through Finite Element Method and OOP C++   Te   wi  Fi    Te    Fj     w j F    k   T  e  w  k

    (1     ) 2       (1     )  A  (1     )    

(1     )

2  .

113

 60 60 60  12 30 12    12 12 30    6 6 3 A  6 18 6    T  ~ 6 18  180  6 1 3 2   2 3 1    3 12 3   3 3 12 

(1     )   g i      . dA  g j  2   gk     

(6) After calculating the integral, we have the vector of nodal loads that can be given by (see Fig. 10):  gi   Fi       F j   T~  g j  g  F   k  k

(8)

where, T ~

2 1 1 A 1 2 1  12  1 1 2

(6)

Another formulation to the equivalent nodal force vector was proposed by Refs. [9-14], called pseudo-consistent equivalent nodal force vector, where rotational load degrees are calculated considering the external work done by the wind loads expressed by:

 1      1     1     1    1 1  1    2 T Te  u ST 2A   ~ ~ 1   2 0 0  1     2  1   2  1   3    1   3

 2   2 3 2  3  22 4 3

     2   2   2  dd g 3  ~ 2 2   3   3  4 

Fig. 9

Fig. 10

Interpolating for

,

and

,

.

Wind loads convertedto nodal equivalent forces.

(8) Again, after calculating the integral, we have the vector of nodal loads that can be given by (see Fig. 11):  gi   Fi     T   Fj   S T  g j  ~ ~ g  F   k  k

where, S  G 1 Q ~

The T is given by: ~

~

~

(7) Fig. 11 Wind loads convertedto pseudo-consistent nodal equivalent forces.


114

1 The G and Q in Eq. (9) are matrices given in Eqs. (11) and (12). The explicit values of Eq. (11) can be ~

~

obtained from Ref. [9].  Q11 Q  21  Q31   Q41  Q51 Q  Q61 ~ Q  71  Q81   Q91 Q10,1 

Q12 Q22 Q32 Q42 Q52 Q62 Q72 Q82 Q92 Q10, 2

Q13 Q23 Q33 Q43 Q53 Q63 Q73 Q83 Q93 Q10,3

Q14 Q24 Q34 Q44 Q54 Q64 Q74 Q84 Q94 Q10, 4

Q15 Q25 Q35 Q45 Q55 Q65 Q75 Q85 Q95 Q10,5

Q16 Q26 Q36 Q46 Q56 Q66 Q76 Q86 Q96 Q10,6

Q17 Q27 Q37 Q47 Q57 Q67 Q77 Q87 Q97 Q10,7

Q18 Q28 Q38 Q48 Q58 Q68 Q78 Q88 Q98 Q10,8

Q19  Q29  Q39   Q49  Q59   Q69  Q79   Q89   Q99  Q10,9 

0 0 0 0 0 0 0 0 0   1  1 0 0 0 0 0 0 9 9 / 2   11/ 2  11/ 2 0 1 0 0 0 9 0 0  9 /12   0 0 27 9 / 2 9 / 2 9 / 2 45 / 2 45 / 2 9 / 2   18  9 0 0 0 0 0 0 18  9 / 2 45 / 2 G 1    ~ 0 0 0 18 0 0  9 / 2 0 45 / 2  9  27 / 2 0 0 0 0 27 / 2 27 27 27 / 2 27 / 2   0 0 0 27 / 2 27 / 2 27 27 / 2 0  27  27 / 2  9 / 2 9 / 2 0 0 0 0 0 0 27 / 2 27 / 2  27 / 2 27 / 2 0 9/2 0 0 0 0 0   9 / 2

Fig. 12

Displacement results.

(9)

(10)


115

YY 0.8

XX

0.6 0.4 0.2 0

Fig. 13

Displacement results.

4. Post-Processing After discussing the mathematical aspects of the wind load model, this section finally shows the post-processing results. In user-friendly environment, the displacements results can be accessed by graphical outputs as viewed in Fig. 12. The graphical results for the displacement variation on the length of the blade are shown in Fig. 13. The mechanical properties and other input data considered for the model analysis are presented in Fig. 5. The load for the wind pressure in each node in X direction on surface of the eolic blade is 1.000e-02 N.

5. Conclusion In this paper, a tool for analysis of eolic blades was presented. The main attractive feature of this structural analysis tool is the dll program for the processing stage implemented in C++ language for the FF and DKT finite elements used to discretize the eolic blade. In addition, a user-friendly environment was implemented using OpenGL library to provide the graphical construction for geometry, mesh orientation, and other requirements of the finite element model.

[2] [3]

[4]

[5]

[6]

[7]

[8]

[9]

References [1] Menezes Junior, R. A., Mendonca, A. V., Paiva, J. B., and

[10]

Mendonça, A. V. 2010. “A User-Friendly Environment for Planar and Space Frames Using the Boundary Element Method.” In Proceedings of Ect 2010 the Seventh International Conference on Engineering Computational Technology, 94-115. Zienkiewics, O. C. 1977. The Finite Element Method, 2nd ed.. New York: Mcgraw-Hill. Foley, J. D., van Dam, A., Feiner, S. K., and Hughes, J. F. 1995. Computer Graphics: Principles and Practice in C. Upper Saddle River: Addison-Wesley. Shreiner, D., Woo, M., Neder, J., and David, T. 2005. OpenGL Programming Guide: The Official Guide to Learning OpenGL, Version 2, 5th ed.. Upper Saddle River: Addison-Wesley. Hearn, D. D., Baker, M. P., and Carithers, W. 2010. Computer Graphics with Open GL. Upper Saddle River: Prentice Hall. Bergman, P. G., and Felippa, C. A. 1985. “A Triangular Membrane Element with Rotational Degrees of Freedom.” Computer Methods in Applied Mechanics and Engineering 50 (1): 25-69. Batoz, J. L., Bathe, K. J., and Ho, L. W. “A Study of Three-Node Triangular Plate Bending Elements.” International Journal for Numerical Methods in Engineering 15 (12): 1771-812. Batoz, J. L., and Lardeur, P. 1989. “A Discrete Shear Triangular Nine d.o.f. Element for the Analysis of Thick to Very Thin Plates.” International Journal for Numerical Methods in Engineering 28 (3): 533-60. Batoz, J. L., and Dhatt, G. S. 1992. Modélisation des Structures par éléments Finis. Vol 3: Coques. Paris: Hermes, 448-55. Vlasov, V. Z. 1961. Thin-Walled Elastic Beams.

116


Washington: National Science Foundation. [11] Taranah, B. S. 1968. “Torsional Behaviour of Open Section Shear Wall Structures.” Ph.D. thesis, University of Southampton. [12] Onu, G. 1990. “Inclusion of Warping Shear Effect in the Thin-Walled Core Element for Multistory Building.” Computers & Structures 35 (2): 175-82. [13] Cantin, G., and Clough, R. W. 1968. “A Curved

Cylindrical Shell Finite Element.” AIAA Journal 6 (6): 1057-62. [14] Hrennikof, A., and Tezcan, S. S. 1966. “Analysis of Cylindrical Shells by the Finite Element Method.” Presented at the Symposium of Problems of Interdependence of Design and Construction of Large Span Shells for Industrial and Civic Buildings, Leningrad, Russia.

Development of a Genetic Algorithm for Evaluating ...

Development of a Genetic Algorithm for Evaluating ...

Suggest Documents

Algorithm for evaluating the effectiveness of a high-rise development ...

Development and Application of a Genetic Algorithm for Variable ...

development of a hybrid fuzzy genetic algorithm model for ... - SciELO

The Development of a Genetic Algorithm Code for ... - CiteSeerX

Strategies for multiobjective genetic algorithm development - Core

Implementation of a Genetic Algorithm for Jiles

A KRYLOV SUBSPACE ALGORITHM FOR EVALUATING ... - CiteSeerX

A Genetic Algorithm Tutorial

development of a genetic algorithm based ... - World Scientific

Development of a Hybrid Agent-Genetic Algorithm ... - CiteSeerX

Development of a Hybrid Agent-Genetic Algorithm Approach to ...

Development of a Multi-Objective Genetic Algorithm ...

Development of a Benchmarking Methodology for Evaluating ...

a genetic algorithm approach to a 3d highway alignment development

A Hierarchical Genetic Algorithm for Topology ...

A genetic algorithm for multiprocessor scheduling

A Genetic Column Generation Algorithm for ...

A Hierarchical Genetic Algorithm for Topology

A Hybridized Lagrangian Genetic Algorithm for

A parallel Genetic Programming algorithm for

A Continuous Genetic Algorithm Designed for the

A Segregated Genetic Algorithm for Constrained ... - CiteSeerX

TimGA - A GENETIC ALGORITHM FOR DRAWING UNDIRECTED ...

Towards a Genetic Programming Algorithm for ... - CiteSeerX