Experimental Assessment, Model Validation, and ... - ACS Publications

Article pubs.acs.org/IECR

Experimental Assessment, Model Validation, and Uncertainty Quantification of a Pilot-Scale Gasifier M. Hossein Sahraei,† Marc A. Duchesne,‡ Robin W. Hughes,‡ and Luis A. Ricardez-Sandoval*,† †

Department of Chemical Engineering, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 Natural Resources Canada, CanmetENERGY, Ottawa, Ontario, Canada K1A 1M1

‡

ABSTRACT: This contribution presents a new set of petroleum coke dry gasification tests performed on pilot-scale gasifier. Dry gas composition and flow rate, temperature distribution, conversion, and pollutant formation taken from the experimental tests and respective calculations were used to validate the prediction capabilities of a reduced order model (ROM) developed for the same gasifier. The ROM predicted the experimental observations for conversion in the range of 48−90%. This study confirms that a systematically developed ROM (with a fixed framework) can predict the behavior of a gasifier under different operating conditions with reasonable accuracy. Moreover, this study investigates the variability in the ROM’s key outputs in the presence of uncertainty in the feed and model parameters, i.e., the volatile percentage of the fuel, solid particle diameters, angle of multiphase flow jet, and recirculation ratio. These parameters affect the feedstock’s properties and the mixing/laminar flows within different zones of the gasifier. Insights gained from the uncertainty quantification study revealed significant variability in the conversion, peak temperature, and steam percentage in the syngas; while the dry syngas composition does not seem to be significantly affected by the uncertainty of the parameters considered. materials for construction.5 Some major commercial-scale entrained-flow gasifier technologies available in the market have been developed by GE Energy and Shell. In their designs, the fuel is injected as dry-feed (Shell) or slurry-feed (GE) in single-stage units and produce syngas at temperatures of 1,473−1,753 K (GE) and 2,073−2,273 K (Shell).6 Apart from commercial-scale units, the performance of pilot-scale and labscale entrained-flow gasifiers for different types of fuel has been experimentally investigated. Table 1 presents some available experimental data in the literature for bench- and pilot-scale entrained-flow gasifiers. As shown in this table, most of the experimental studies have been performed on the gasification of biomass and coal. On the other hand, studies on petroleum coke gasification are rather limited. Such tests are required to investigate the expected operation of large-scale entrained-flow gasifiers and develop the corresponding instrumentation and control systems. In general, the gasification experiments have been performed for limited test conditions due to associated experimental expenditures. Therefore, mathematical models are suitable and are cost-effective tools that can be developed to predict the performance of the gasifier for a wide range of operating conditions and under different scenarios. Models based on computational fluid dynamics (CFD) can predict comprehensive information regarding the design, feasible operating tests, and multiphase flow patterns inside a

1. INTRODUCTION Recent efforts in electricity production have been directed toward reduced-carbon energy production from fossil fuels since renewable energy sources may not be sufficient to fulfill the world’s energy demands. Consequently, solid fuels such as coal and petroleum coke are expected to continue playing a major role in supplying the world’s energy demands for the upcoming decades.1 Gasification is an attractive solid fuel-fired technology since it provides substantial environmental benefits over combustion-based processes, such as less expensive gascleaning equipment, higher efficiency when integrated to a carbon capture unit, and lower SOx and NOx emissions during power production.2 Development of efficient gasification units that can handle different feedstocks has gained recent attention as the fuel flexibility allows the industry to adapt to changes in fuel costs and sustain a cost-effective energy production scheme.3 One of the features that determine the suitability of a gasifier to handle different types of fuels is the mixing zone of the particles and gases inside the unit. Although fluidized-bed and entrained-flow gasifiers can deliver satisfactory mixing zones, the latter can also provide higher throughput for attractive commercial-scale designs. Entrained-flow gasifiers can process most of the commercially available solid fuels as long as the fuel can be pulverized into fine particles and does not include high ash contents.4 The high operating temperature of these gasifiers, typically above the ash melting temperature, enables high carbon conversion to syngas. However, high operating temperatures within the unit can have a negative impact on the burner and refractory life, hence requiring the use of expensive © 2016 American Chemical Society

Received: Revised: Accepted: Published: 6961

February 19, 2016 May 12, 2016 June 2, 2016 June 2, 2016 DOI: 10.1021/acs.iecr.6b00692 Ind. Eng. Chem. Res. 2016, 55, 6961−6970

Article

Industrial & Engineering Chemistry Research Table 1. Experimental Studies on Bench- and Pilot-Scale Entrained-Flow Gasifiers work

gasifier

feedstock

Niu et al.7

2 kg/h lab-scale, two-stage

diesel oil

Xu et al.8 Lee et al.9

16 kg/h pilot-scale, two-stage 1 ton/day pilot-scale, cogasification 1 ton/day pilot-scale pressurized pilot-scale atmospheric 150−200 kg/h pilot-scale 2 ton/day pilot-scale bench-scale 24 kg/h lab-scale

coal petroleum coke/ coal coal biomass biomass biomass biomass coal

40 kg/h air-blown 40−55 kg/h pressurized pilotscale 2−5 kg/h pressurized pilot-scale Texaco pilot-plant single-stage and two-stage pilot-scale, cogasification pressurized gasification facility 2 ton/day pilot-scale 50 kg/h pilot-scale

biomass biomass

Wagner and Whitty10 Hernandez et al.11 Leijenhorst et al.12 Ogi et al.13 Zhou et al.14 Soelberg et al.15 and Highsmith et al.16 Senapati and Behera17 Weiland et al.18 Tremel et al.19 Kasule et al.20 Ren et al.21 Hernandez et al.22 Lewis et al.23 Watanabe and Hara24 Duchesne et al.25

coal coal coal biomass/coal petroleum coke coal coal

gasification system. However, these models are computationally expensive even for stationary conditions. On the other side of the spectrum, process simulators provide modules that predict the behavior of gasification systems using ideal reactors, which generally cannot accurately capture the main features of the gasification process such as conversion, reaction systems, compositions, and temperature profiles. As a result, there is a need to develop practical models that can predict the fundamental aspects of gasification processes with reasonable accuracy at low computational costs. In recent years, reduced order models (ROMs) have been recognized as a suitable alternative to overcome the limitations imposed by computationally intensive CFD and simplified (ideal) reactors in process simulators.26−29 These low-order models are comprised of a reactor network that represents distinct flow regions inside the gasification system. The low-order model describes the distribution of momentum, heat, and mass transfer in specific zones of the gasification unit. Each reactor is designed based on the physical structure and flow pattern of the corresponding zones in the unit. A review of the ROMs developed for various entrained-flow gasifiers is available in the literature.28 One of the advantages of ROMs for complicated systems such as gasifiers is the ability to perform modeling analyses which are computationally expensive for CFD models. One such analysis is uncertainty quantification with a large number of samples. Application of mathematical techniques to evaluate the behavior of models in the presence of uncertainty has been widely recognized in academia and industry.30−33 To the authors’ knowledge, only two studies have addressed parameter uncertainty in gasification models; however, none of these studies have used a ROM for this purpose. Gel et al. applied a nonintrusive uncertainty quantification method to study the variations of temperature, pressure, and heating rate on the performance of a fluidized-bed gasifier.34 In that study, a C3Mbased computational coal kinetics model was employed to investigate the system’s sensitivity and perform uncertainty quantification. Shastri and Diwekar used CFD simulations to study the effect of coal composition on a gasifier’s operation.35

provided data outlet composition and conversion, wall temperature at certain points outlet composition and temperature outlet composition, conversion and cold gas efficiency outlet syngas composition outlet gas composition transient data for syngas composition and temperature outlet gas composition outlet gas composition, cold gas efficiency, carbon conversion outlet gas composition including nitrogen and sulfur pollutants outlet gas composition, temperature transient data for gas composition and temperatures at certain points transient outlet gas composition outlet gas composition outlet gas volume fraction outlet gas composition temperature profile temperature profile and volume fraction temperature profile, outlet gas composition and slag properties

Due to the computational costs of the CFD model, only 15 samples were employed in that study; accordingly, the distributions provided are limited since they may not capture the actual distribution of the process outputs due to parameter uncertainty. As the computational costs of ROM are much lower than those needed for CFD simulations, a large number of samples can be employed for the uncertainty analysis. Accordingly, accurate distributions of ROM’s outputs in the presence of model uncertainty can be used to quantify the potential financial risk associated with commercial performance of the technology. Decisions can then be made as to whether more research is needed or the level of certainty can be accepted and proceed with scale-up. The present contribution investigates pressurized and dry gasification of petroleum coke in CanmetENERGY’s pilot-scale gasifier by performing experimental and modeling assessments. Our group has developed a ROM for CanmetENERGY’s gasifier. Simulation results generated by the ROM were previously compared with CFD simulation results.28,36 However, the ROM has not been validated using actual experimental data. It is therefore the primary aim of this study to also investigate and validate the capabilities and limitations of the ROM in predicting the results of experimental tests, which have been performed at a wide range of operating conditions. Note that the experimental tests presented here were performed recently and have not been published elsewhere. To the authors’ best knowledge, this is the first time that a ROM is validated with experimental tests of petroleum coke gasification, while the previous ROM studies were focused on coal.8,27,37 Moreover, the ROM employed in this work was developed under the assumption of perfectly known model parameters. However, certain model parameters are uncertain and represent sources of mismatch between the actual experimental observations and the ROM predictions, i.e., model-plant mismatch. Accordingly, there are uncertainties within the key model parameters and feed that need to be considered to verify their effect on the ROM’s outputs. Accordingly, an uncertainty analysis on the expected behavior 6962

DOI: 10.1021/acs.iecr.6b00692 Ind. Eng. Chem. Res. 2016, 55, 6961−6970

Article

Industrial & Engineering Chemistry Research

Figure 1. Schematic diagram of CanmetENERGY’s pressurized entrained-flow gasification system.

thermocouples are indicated on Figure 1 as TC1 through TC4. A nitrogen-cooled gas sampling probe (indicated in Figure 2a) was inserted into the gasifier at the same elevation as

of an entrained-flow gasifier has been performed in this study using our ROM. The structure of this article is as follows: Section 2 describes the experimental tests performed on CanmetENERGY’s gasifier. Model description of the gasifier and the uncertain parameters are presented in Section 3. Model validation and the uncertainty analysis performed in this study are presented in Section 4. Concluding remarks are stated in Section 5.

2. EXPERIMENTAL DESIGN The CanmetENERGY pressurized entrained-flow slagging gasifier (Figure 1) is a gasification facility that converts 1 tonne/day of different slurry/dry feedstocks to syngas.25 Its inner diameter and length are 0.20 and 2.1 m, respectively; the sampling probe of the gasifier is located at the length of 1.8 m from the feed injectors. The pilot gasifier is refractory-lined with a thickness of refractory and insulation materials, including mullite, Ziralcast ceramics, and Kaowool board/paper, of 0.2 m. For the present study, the gasifier was configured for singlestage dry-fed gasification. The feeding system used nitrogen for conveying and is described in a previous study.38 Oxygen was injected through the burner by eight jets that impinged on the fuel stream at high velocity. Steam, preheated up to 500 K, was passed through the outer burner annulus with a low velocity. A quench vessel beneath the gasifier allowed for discharge of slag on a batch basis after test completion. Before testing, the gasifier was preheated overnight using natural gas and oxygen-enriched air. Once the desired temperature was attained, natural gas and oxygen injection were terminated. The reactor was then pressurized by injecting nitrogen. At the desired pressure, nitrogen injection was ceased. Fuel flow was then established, followed by a progressive ramping up of oxygen injection to obtain the desired temperature. Oxygen flow was adjusted to maintain a constant temperature at thermocouple TC4 (Figure 1). Automation and data collection for the pilot gasifier facility was accomplished using an ABB Freelance 2000 distributed control system.39 The temperature profile within the gasifier was monitored using type B thermocouples protruding past the hot face and into the reaction chamber by ∼5 mm. The locations of the

Figure 2. a) Burner region and sampling probe of the gasification system and b) the reactor network for CanmetENERGY’s gasifier.28

TC4. Syngas was extracted through the probe to provide the syngas composition inside the reactor. Dried gas analysis was performed via gas chromatography capable of measuring CO, CO2, CH4, H2, O2, COS, H2S, and N2 once every 2 min. Online infrared carbon monoxide and carbon dioxide analyzers were used to validate syngas compositions generated by the gas chromatograph. Dry syngas flow exiting the reactor was estimated by performing a nitrogen mass balance based on the nitrogen injection flow rate and the dry syngas composition. Note that the nitrogen injection flow rate includes conveying gas and nitrogen in the petroleum coke. The dry syngas 6963


Article


The momentum transfer equations take into account terms for convection momentum transfer, pressure drop, and viscous forces between gas, solid particle, and wall. In the energy balance equations, heat of homogeneous and heterogeneous reactions, conduction, convection, and energy transfer between gas, sold particle, and refractory layers are considered. Moreover, the mass balance equations account for reactions, diffusion, and convective mass transfer. The corresponding differential equations and additional algebraic expressions that form the ROM are presented in ref 28. As the fuel enters the gasifier through the injection nozzles, it decomposes into vapor, volatiles, ash, and carbon through a process known as coal pyrolysis. The ROM considers this feature through the application of submodels for drying and devolatilization. Among the volatile products, the combustible gases react with oxygen and energy is released in the system, which is then used by gasification reactions to produce syngas. The relevant homogeneous reactions were incorporated in the ROM by gathering kinetic data from the literature, whereas the kinetic parameters of heterogeneous reactions were obtained from PC Coal-lab software for the petroleum coke used in the operating conditions of the experimental tests.40,41 The reactions have been reported in a previous study28 and have been shown to predict the initial reactivity of particles with reasonable accuracy. However, as internal particle structure was not considered in the kinetic equations, they do not account for the influences of pore evolution and intrinsic chemistry. Moreover, the ROM considers the heat transfer through the gasifier’s refractory layers by using effective temperaturedependent heat transfer coefficients for insulating layers. Overall, the model parameters of the ROM were calculated using CFD simulations (such as recirculation ratio, jet angle, and reactor network size), from PC Coal-lab software (heterogeneous reaction and devolatilization rates), or taken from the literature (such as homogeneous reaction rate and physical properties of gases). Detailed information on the model parameters has been provided in the authors’ previous work.28 The development of the ROM required the specification of parameters that were not known with certainty and were fixed to their expected (more likely) values. However, these parameters can take multiple values during operation depending on the operating conditions of the system and structure of the multiphase flow patterns. Furthermore, the gasifier’s feed streams may also experience uncertainties (variability) in their operating conditions that cannot be easily measured online. In the present study, fuel particle diameter, volatile matter percentages in fuel, angle of jet multiphase flow, and recirculation ratio of the reactor network are considered to be key uncertain parameters that affect the ROM’s prediction capabilities in terms of conversion and syngas production. As there is variation during operation in the size of particles, a distribution can be used to account for the fuel’s particle size, rather than employing a single (averaged) particle size in the analysis. The volatile percentage of the fuel can also vary during gasifier operation by two mechanisms: 1) changes in the petroleum coke’s composition and 2) changes in the volatility of petroleum coke components under different operating conditions, e.g. higher temperature increases the volatility of components. To study the effect of variability of volatiles in the gasifier, the volatile percentage of the fuel is considered as an uncertain parameter. Note that the volatile percentage of the fuel affects the proximate analysis as a whole. In this work, the

composition was obtained by gas chromatography with the dried syngas. Carbon conversion was estimated by performing a carbon mass balance with the injected fuel and dry syngas. The operating conditions of the experimental tests are presented in Table 2. As shown in this table, these tests were conducted at Table 2. Operating Conditions of the Experimental Tests test 1 fuel (kg/h) 47.1 steam (kg/h) 9.9 oxygen (kg/h) 28.4 nitrogen (kg/h) 11.9 pressure (bar) 16 petroleum coke composition proximate analysis (as received)

ultimate analysis (dry, ash-f ree)

test 2

test 3

test 4

50.7 21.8 34.2 11.0 16

41.2 10.7 37.2 12.1 16

52.3 0 30.4 11.4 16 mass fraction 0.046 0.127 0.005 0.822 0.042 0.061 0.018 0.015 0.864

ash volatiles moisture carbon hydrogen sulfur nitrogen oxygen carbon

different operating conditions, i.e., injected fuel flow rates vary from 41.2 to 52.3 kg/h, injected steam flow rates vary from 0 to 21.8 kg/h, and injected oxygen flow rates vary between 28.4 and 37.2 kg/h. A high operating pressure is often desired for these systems as it decreases the volume of the gasifier and reduces capital costs (larger reactor volume is not required) for a given throughput. Hence, the set of experimental tests was performed at 16 bar.

3. MODELING METHODOLOGY The modeling section of this study considers a reduced order model (ROM) of an entrained-flow gasifier. This popular reactive modeling approach accounts for compartmentalized flow regions of the gasifier by using a continuous reactor network. Each individual reactor describes the distribution of heat and mass transfer in a specific flow field of the gasifier and interacts with the adjacent reactors. The goal in a ROM is to capture the key performance characteristics of the actual gasifier by using an appropriate arrangement of reactors and reduce the computational time of solving the model. In order to develop a reactor network for a gasifier, CFD models or experimental images are required to indicate the streamlines of the multiphase flow inside the system. In the present work, a one-dimensional reduced order model (ROM) for CanmetENERGY’s gasifier based on CFD simulation was employed.28 The details of the CFD simulation and streamlines of the multiphase flow are fully described elsewhere28,36 and have not been discussed here for brevity. A brief description of the ROM is presented next. According to the flow structure of each zone obtained by CFD simulations, mixing flow fields are modeled using continuously stirred tank reactors (CSTRs), whereas laminar regions are modeled using plug flow reactors (PFRs). The reactor network (shown in Figure 2b) considers two jet expansion zones (JEZ), two external recirculation zones (ERZ), and a downstream zone (DSZ). Velocity, temperature, and mass distributions are obtained for each zone of the reactor network by simultaneously solving the respective equations. 6964


Article

Industrial & Engineering Chemistry Research Table 3. Experimental and Modeling Results for Conversion and Dry Gas Molar Flow Rate and Percentages dry molar percentages percentage of carbon conversion test test test test

1 2 3 4

dry gas flow rate (kmol/h)

CO2

H2S

N2

exp.

ROM

exp.

ROM

exp.

ROM

exp.

ROM

exp.

ROM

60 69 90 48

59.8 68.0 91.0 52.8

3.53 4.31 4.12 3.01

3.27 4.30 4.18 3.38

6.4 10.7 6.4 1.7

5.4 9.2 5.3 2.0

1.83 1.78 1.34 1.58

1.69 1.53 1.57 1.77

12.8 10.2 11.0 14.6

12.4 9.9 10.8 14.0

uncertainty in the volatile percentage is also reflected in the carbon percentage of the proximate analysis, because there is low moisture and ash in petroleum coke. The inlet flow rates affect the geometry of the expansion and recirculation zones within the gasifier. Therefore, some model parameters such as jet angle and recirculation ratio are also considered as uncertain parameters since flow rates are prone to change during operation. The geometry of fuel and oxygen injectors forms a multiphase flow which is dictated by an angle, typically referred to as the jet angle. Due to severe operating conditions inside the gasifier, it is often difficult to measure the jet angle. For the present ROM, this parameter was estimated based on the flow patterns of streamlines presented in our previous work.28 As mentioned above, the recirculation ratio is one of the sensitive parameters in the present ROM.36 This parameter represents the molar flow from the outlet of JEZ2 to ERZ2, over the total inlet molar feed flow to JEZ1 (see Figure 2b). In addition, the recirculation ratio is also sensitive to variations in the feed’s flow rates; hence, it was considered as an uncertain parameter in the present analysis. Accounting for these uncertain parameters allows us to study the performance of CanmetENERGY’s gasifier more thoroughly by using a ROM which considers parameters that are better described with probability density functions. The distributions considered for uncertain parameters and the method used to propagate uncertainty are described in the next section.

Figure 3. Dry syngas compositions of the experimental tests compared to ROM simulation results.

the experimental measurements. As shown in Figure 3, the highest CO mole fraction (0.546) was achieved for test 3, whereas the highest H2 was reached in test 2, where the steam flow rate was high. For the latter test, higher CO2 is produced when compared to the other tests, which can be due to the water−gas shift reaction and steam gasification. The results for other components, i.e. CO2, H2S, and N2, are presented in Table 3. During devolatilization and char gasification, sulfur is released in the form of hydrogen sulfide (H2S). The H2S formation mechanism is described elsewhere.37 As no global kinetic expressions were available for sulfur reactions, rate expressions with extremely high constants are employed as suggested in our previous work.28 Note that in the reducing environment of the gasifier, sulfur is mostly present in the form of H2S rather than COS. As shown in Table 3, the experimental dry syngas H2S molar fractions deviate from ROM predictions by 0.14 to 0.25 percentage points. Also, according to this table, the predicted dry gas flow rates by the ROM were in reasonable agreement with experimental tests. In the case of CH4, since the heterogeneous reaction of CH4 formation implemented in the ROM has a slow rate, the ROM predicted lower CH4 content (0.0006−0.001%) compared to the experimental tests (0.05− 0.2%). Another reason for lower predictions of CH 4 concentration may be the lack of a devolatilization rate in the ROM. Due to high temperatures within the gasifier, devolatilization reactions are assumed to be instantaneous reactions. Accordingly, all the CH4 released from the fuel is predicted to instantly react with oxygen near the injection point in the gasifier. In order to further demonstrate the distribution of the species along the gasifier’s length, molar fraction profiles of the ROM for test 3 (which has the highest conversion) are presented in Figure 4. According to this figure, oxygen is consumed quickly at the inlet of the gasifier due to volatile and char combustion. During the conversion of fuel, molar fractions of CO, H2, and H2S gradually increase following the length of the gasifier. Furthermore, the molar fraction of CO2 peaks at x

4. RESULTS AND DISCUSSION 4.1. Model Validation. In order to validate the ROM of the gasifier unit considered in this study and verify its capabilities in predicting experimental results, a set of petroleum coke gasification tests has been performed (Table 2). Note that the modeling results indicated that the original base-case condition under which the ROM was developed would have caused operational difficulties. These difficulties were due to high temperature distribution (up to 2900 K) along the gasifier, which can seriously damage the gasifier’s wall and measurement devices.28 Hence, this test was not considered as part of the experimental design. The results from the experimental tests and the ROM are presented in Table 3. According to this table, test 3 has the highest oxygen to fuel mass ratio (0.9); therefore, the highest conversion was achieved for this test, i.e., 90%. The lowest conversion (48%) was achieved in test 4 where steam was not used. As shown in Table 3, the ROM predicts well the conversion measured in the different experimental tests. Figure 3 presents the dry CO and H2 compositions obtained from the ROM and experimental tests at the sampling probe. The bars in this figure represent the relative measurement errors for gas composition (±5%) based on measurements with calibration gases. Although the ROM predicts slightly different CO (lower) and H2 (higher) compositions, the results are in reasonable agreement with 6965


Article


centerline and near-wall temperature predictions as the temperature is assumed to vary axially but not radially. Based on thermocouple measurements, the temperature typically decreases along the gasifier’s length. The temperature at the sampling probe is between 1,498 and 1,580 K for the test conditions. As shown in Figure 5, the centerline temperature increases sharply at the beginning of the reactor (JEZ1) reaching a peak value in this section, which is mostly due to the combustion of volatiles and recirculation of gases. Among the test conditions, test 3 has the highest peak temperature (2,504 K) due to the highest oxygen to fuel ratio. By depletion of oxygen in the system, the heat required for endothermic reactions overcomes the heat released by exothermic reactions, and the temperature decreases as the gases move toward the sampling probe. Based on these observations, the temperatures predicted by the ROM are in reasonable agreement with the experimental data. The results of conversion, temperature profile, dry gas flow rates, and compositions from the model validation have demonstrated the capabilities of the ROM to describe the performance and operation of CanmetENERGY’s gasification unit at different operating conditions. Although the ROM was developed based on a CFD simulation with operating conditions different than the experimental conditions presented in this work, it was still capable of predicting the outputs with reasonable agreement with experimental results. Nevertheless, among the tests, the highest deviation between the ROM’s output and experimental test result was observed for test 4 where the flow rate of steam was zero. Such deviation was expected since zero flow rate of the steam alters the streamlines

Figure 4. Trend of dry gas molar fractions along the gasifier’s length for test 3.

= 0.13 m (where oxygen is depleted) and decreases afterward due to gasification reactions. The temperature profiles obtained from the experiments and the ROM (centerline and wall) are presented in Figure 5. Experimental values were taken from thermocouples near the wall at locations shown in Figure 1. Near the burner, JEZ1 and JEZ2 in the ROM provide centerline temperature predictions, while ERZ2 provides near-wall temperature predictions. Further from the burner, the ROM’s DSZ provides both

Figure 5. Temperature profile along the gasifier’s length for the experimental tests. 6966


Article

Industrial & Engineering Chemistry Research Table 4. Description of Uncertain Parameters parameter

mean

SD

volatile percentage particle diameter (μm) recirculation ratio jet angle (degrees)

0.127 79

0.015 10

1.1 17

0.2 1

observations The mean value is taken from proximate analysis, while the standard deviation is assumed. Experimental observation for particle size distribution of petroleum coke. CFD calculations and sensitivity analysis.36 Mean value is calculated from the flow patterns of a CFD simulation.28 Generally, a jet angle of 16−24° exists in experimental and industrial investigations.43

of the multiphase flow compared to the streamlines of the case under which the ROM was developed. The average CPU time required by the ROM to compute the results for a single operating condition was approximately 4−5 min (Core i7, 3.4 GHz with 8 GB of RAM); performing the same simulation using CFD would require approximately 10 days (32 Cores, 2.9 GHz). Although highly variable, the cost to perform an experimental test is estimated at $30,000 USD. Therefore, the present ROM is a computationally efficient tool that can be used to perform further analyses where it would be costly to perform either experimental tests or CFD simulations. 4.2. Uncertainty Analysis. In this work, Monte Carlo (MC) sampling has been used as the uncertainty propagation method. MC sampling takes advantage of a large number of sample points chosen randomly from the distribution of uncertain parameters and a primary model to propagate uncertainty and therefore quantify output variability, typically represented as a probability distribution function (PDF). In the present study, normal (Gaussian) distributions in the uncertain parameters were considered as it is often regarded as an appropriate approximation in most engineering applications. Also, this type of distribution can be approximated to other types of probability distributions such as binomial or Poisson.42 The mean, standard deviation, and basis of selection for the uncertain variables are presented in Table 4. The highest conversion was achieved for test 3; therefore, this test was selected to perform the uncertainty analysis since it is expected that the gasifier will more likely operate at high conversion levels. In the present study, 2,500 MC samples that comply with the uncertainty descriptions provided in Table 4 were randomly generated for each uncertain parameter. Simulations of the ROM’s gasifier were carried out for each combination in the realizations of the uncertain parameters obtained from the MC sampling technique. Probability distribution functions (PDF) were generated from the MC simulation results to show the effect on output variability due to uncertainties in model parameters and the fuel’s characteristics outlined in the previous section. Increasing the number of MC samples used in the present uncertainty analysis will not significantly improve the accuracy in the outputs’ PDFs, i.e., the relative errors in the means and standard deviations are expected to be less than 1%; however, they will significantly increase the computational costs. Based on the predicted distributions of outputs, upper and lower bounds evaluated at 10% confidence intervals, i.e., 95% and 5% respectively, are presented in Table 5. According to this table and the distribution obtained for conversion (Figure 6), the average conversion is 90.7%, while there is a wide range of variability, i.e. from 86.9% to 94.1% (for the confidence interval of 10%), with a standard deviation of 2.3%. Higher volatiles in fuel, larger particle diameters, and large recirculation ratios have direct effects on coal conversion. Similarly, high variability was also observed for the temperature inside the gasifier at the sampling probe (the outlet) and at the position at which the

Table 5. Probabilistic Bounds Evaluated at 10% Confidence Intervals carbon conversion (%) Tsampling probe (K) Tpeak (K) CO molar fraction H2 molar fraction CO2 molar fraction H2O molar fraction

lower bound (5%)

upper bound (95%)

86.9 1,517 2,374 0.525 0.265 0.047 0.069

94.1 1,700 2,686 0.537 0.280 0.063 0.115

Figure 6. Carbon conversion distribution (the dot represents the experimental result).

highest (maximum) temperature was recorded (peak temperature) (Figure 7). The average outlet (sampling probe) and peak temperatures obtained from the present uncertainty analysis are 1,604 K and 2,529 K, respectively. The main source of variability for the temperature distribution is expected to be directly correlated to the percentage of volatiles. As the volatile percentage increases, more hydrogen and hydrocarbons react with oxygen, causing higher peak temperatures inside the gasifier. According to the results shown in Figure 7, the standard deviation of sampling probe and peak temperatures are 56 and 77 K, respectively. Note that the standard deviations in temperature for the other locations inside the gasifier (not shown here for brevity) are lower than that observed for the peak temperature. Temperature variability may cause safety concerns as it increases the risk of damaging the refractory. Note that the cooling water section installed at the top of the gasifier can potentially reduce the risk of such safety concern. On the other hand, a higher peak temperature leads to a higher temperature distribution inside the gasifier, which will improve conversion at the expense of increased risk. The molar distributions of gas species are presented in Figure 8. Note that the presented values for the molar fraction of CO, CO2, and H2 are based on a dry gas composition. The mean values for CO and H2 molar fractions are 0.532 and 0.273, respectively. The lowest variability was observed for the CO 6967


Article


expected since the standard deviation of volatile material has less impact on the total amount of carbon (above 80% of the petroleum coke is carbon). On the other hand, the considered volatile variability influences the major volatile components produced from the fuel, i.e. H2. As there are several reactions that convert H2 to H2O and vice versa, the variability for the combined H2 and H2O molar fractions was expected to be greater than the variability of the combined CO and CO2 molar fractions. The results shown in Figures 5−7 indicate that the resulting distributions for the outputs analyzed in this study are non-normal, where the observed non-normality is higher for gas composition compared to temperature and conversion. In order to perform the present uncertainty analysis, a total of 175.4 h of CPU time was required (Core i7, 3.4 GHz with 8 GB of RAM). Although the cost can still be considered significant, performing the same analysis using CFD simulations will result in prohibitive computational costs. Note that performing one realization of the uncertainty analysis using the CFD model will roughly take 168 h (32 Cores, 2.9 GHz). Accordingly, the present uncertainty analysis cannot be performed using CFD or experimental tests due to prohibitive computational and economic costs. Therefore, the results of uncertainty analysis cannot be accessed using any other sources and therefore cannot be compared with CFD simulations. On the other hand, the proposed ROM is able to provide access to this information and can be used for example to evaluate the availability of the system under different operating conditions and in the presence of model parameter uncertainty.

Figure 7. a) Distribution of temperature at the sampling probe (the dot represents the experimental result) and b) distribution of peak temperature inside the gasifier.

5. CONCLUSIONS This study addresses experimental assessment, model validation, and uncertainty quantification of model and feed parameters for CanmetENERGY’s gasifier. Experimental tests were conducted with petroleum coke at high pressure and with different fuel, oxygen, and steam flow rates. The performance of

molar fraction with a standard deviation of 0.003, whereas the highest variability was obtained for the molar fraction of H2O (standard deviation of 0.011). A narrow variability for the combined CO and CO2 molar fractions (Figure 8) was

Figure 8. Molar fraction distributions of the major species inside the gasification unit. 6968


Article


(9) Lee, S. H.; Yoon, S. J.; Ra, H. W.; Son, Y.; Hong, J. C.; Lee, J. G. Gasification characteristics of coke and mixture with coal in an entrained-flow gasifier. Energy 2010, 35, 3239−3244. (10) Wagner, D. R.; Whitty, K. J. Extractive gas-phase sampling of the reaction zone of a pressurized entrained-flow coal gasifier. Fuel Process. Technol. 2015, 137, 157−163. (11) Hernández, J. J.; Aranda-Almansa, G.; Bula, A. Gasification of biomass wastes in an entrained flow gasifier: Effect of the particle size and the residence time. Fuel Process. Technol. 2010, 91, 681−692. (12) Leijenhorst, E. J.; Assink, D.; Van de Beld, L.; Weiland, F.; Wiinikka, H.; Carlsson, P.; Ohrman, O. G. W. Entrained flow gasification of straw- and woodderived pyrolysis oil in a pressurized oxygen blown gasifier. Biomass Bioenergy 2015, 79, 166−176. (13) Ogi, T.; Nakanishi, M.; Fukuda, Y.; Matsumoto, K. Gasification of oil palm residues (empty fruit bunch) in an entrained-flow gasifier. Fuel 2013, 104, 28−35. (14) Zhou, J.; Chen, Q.; Cao, X.; Mei, Q.; Luo, Z.; Cen, K.; Zhao, H. Biomass−oxygen gasification in a high-temperature entrained-flow gasifier. Biotechnol. Adv. 2009, 27, 606−611. (15) Soelberg, N. R.; Smoot, L. D.; Hedman, P. O. Entrained flow gasification of coal: 1. Evaluation of mixing and reaction processes from local measurements. Fuel 1985, 64, 776−781. (16) Highsmith, J. R.; Soelberg, N. R.; Hedman, P. O.; Smoot, L. D.; Blackham, A. U. Entrained flow gasification of coal: 2. Fate of nitrogen and sulphur pollutants as assessed from local measurements. Fuel 1985, 64, 782−788. (17) Senapati, P. K.; Behera, S. Experimental investigation on an entrained flow type biomass gasification system using coconut coir dust as powdery biomass feedstock. Bioresour. Technol. 2012, 117, 99− 106. (18) Weiland, F.; Nordwaeger, M.; Olofsson, I.; Wiinikka, H.; Nordin, A. Entrained flow gasification of torrefied wood residues. Fuel Process. Technol. 2014, 125, 51−58. (19) Tremel, A.; Haselsteiner, T.; Kunze, C.; Spliethoff, H. Experimental investigation of high temperature and high pressure coal gasification. Appl. Energy 2012, 92, 279−285. (20) Kasule, J. S.; Turton, R.; Bhattacharyya, D.; Zitney, S. E. Mathematical Modeling of a Single-Stage, Downward-Firing, Entrained-Flow Gasifier. Ind. Eng. Chem. Res. 2012, 51, 6429−6440. (21) Ren, Y.; Xu, S.; Li, G. Experimental study on the operational performance of an advanced two-stage entrained-flow coal gasifier. Energy Fuels 2014, 28, 4911−4917. (22) Hernandez, J. J.; Aranda-Almansa, G.; Serrano, C. Cogasification of biomass wastes and coal-coke blends in an entrained flow gasifier: an experimental study. Energy Fuels 2010, 24, 2479− 2488. (23) Lewis, A. D.; Fletcher, E. G.; Fletcher, T. H. CO2 gasification rates of petroleum coke in a pressurized flat-flame burner entrainedflow reactor. Energy Fuels 2014, 28, 4447−4457. (24) Watanabe, H.; Otaka, M. Numerical simulation of coal gasification in entrained flow coal gasifier. Fuel 2006, 85, 1935−1943. (25) Duchesne, M. A.; Hughes, R. W.; Lu, D. Y.; McCalden, D. J.; Anthony, E. J.; Macchi, A. Fate of inorganic matter in entrained-flow slagging gasifiers: Pilot plant testing. Fuel Process. Technol. 2014, 125, 18−33. (26) Monaghan, R. F. D.; Ghoniem, A. F. A dynamic reduced order model for simulating entrained flow gasifiers. Part II: Model validation and sensitivity analysis. Fuel 2012, 94, 280−97. (27) Gazzani, M.; Manzolini, G.; Macchi, E.; Ghoniem, A. F. Reduced order modeling of the Shell−Prenflo entrained flow gasifier. Fuel 2013, 104, 822−37. (28) Sahraei, M. H.; Duchesne, M. A.; Yandon, R.; Majeski, A.; Hughes, R. W.; Ricardez-Sandoval, L. A. Reduced order modeling of a short-residence time gasifier. Fuel 2015, 161, 222−232. (29) Lee, H. H.; Lee, J. C.; Joo, Y. J.; Oh, M.; Lee, C. H. Dynamic modeling of Shell entrained flow gasifier in an integrated gasification combined cycle process. Appl. Energy 2014, 131, 425−440.

the system was initially assessed under different ratio of inlet flow rates by conducting the experimental tests. Furthermore, the prediction capabilities of a reduced order model (ROM) was compared with the gasification tests. The ROM was able to predict the conversion, dry gas compositions, and temperature distribution with satisfactory agreement when compared to experimental observations. The highest variation of ROM results and experimental data was observed for the test where the steam flow rate was set to zero. According to the results, using a fixed framework in the reactor network to capture the streamlines of the multiphase flow is sufficient to predict (with reasonable agreement) the outcome of gasification tests. As the proposed ROM requires a relatively low computational time to perform a single simulation, the performance of the ROM was evaluated under uncertainty in the feed and the model parameters. The calculated bounds, at 10% confidence level, indicate that conversion can fluctuate between 86.9% and 94.1% for the test condition considered in this analysis (which has conversion of 90%). The highest temperature variability observed corresponds to the peak temperature (2,374 to 2,686 K). The variability of peak temperature may result in unexpectedly high temperatures and stresses due to thermal cycling that can damage equipment with associated safety hazards in combustion zone of the gasifier. The insight gained from this analysis, which cannot be accurately performed using CFD simulations due to computational limitations, can be used as a guideline to specify suitable and safe operation policies for the gasification systems.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: 1-519-888-4567 ext. 38667. Fax: 1-519-888-4347. Email: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This study was supported by the Government of Canada’s Program of Energy Research and Development and the ecoENERGY Innovation Initiative.

■

REFERENCES

(1) Shafiee, S.; Topal, E. When will fossil fuel reserves be diminished? Energy Policy 2009, 37, 181−189. (2) Buchan, B.; Cao, C. Coalfired generation: proven and developing technologies; Office of market monitoring and strategic analysis florida public service commission:2004. (3) Nunes, L. J. R.; Matias, J. C. O.; Catalão, J. P. S. Biomass waste co-firing with coal applied to the Sines Thermal Power in Portugal. Fuel 2014, 132, 153−157. (4) Luque, R.; Speight, J. Gasification for synthetic fuel production; Woodhead Publishing: 2014; pp 36−37. (5) Steynberg, A.; Dry, M. Fischer−Tropsch Technology, 1st ed.; Elsevier Science: 2004. (6) Sahraei, M. H.; McCalden, D.; Hughes, R.; Ricardez-Sandoval, L. A. A survey on current advanced IGCC power plant technologies, sensors and control systems. Fuel 2014, 137, 245−59. (7) Niu, M.; Yan, Z.h.; Guo, Q.; Liang, Q.; Yu, G.; Wang, F.; Yu, Z. Experimental measurement of gas concentration distribution in an impinging entrained-flow gasifier. Fuel Process. Technol. 2008, 89, 1060−1068. (8) Xu, S.; Ren, Y.; Wang, B.; Xu, Y.; Chen, L.; Wang, X.; Xiao, T. Development of a novel 2-stage entrained flow coal dry powder gasifier. Appl. Energy 2014, 113, 318−323. 6969


Article

Industrial & Engineering Chemistry Research (30) Li, Z.; Ierapetritou, M. G. Robust optimization for process scheduling under uncertainty. Ind. Eng. Chem. Res. 2008, 47, 4148− 4157. (31) Nease, J.; Adams, T. A., II Application of rolling horizon optimization to an integrated solid-oxide fuel cell and compressed air energy storage plant for zero-emissions peaking power under uncertainty. Comput. Chem. Eng. 2014, 68, 203−219. (32) Bahakim, S. S.; Rasoulian, S.; Ricardez-Sandoval, L. A. Optimal design of large-scale chemical processes under uncertainty: A rankingbased approach. AIChE J. 2014, 60 (9), 3243−3257. (33) Rasoulian, S.; Ricardez-Sandoval, L. A. Uncertainty analysis and robust optimization of multiscale process systems with application to epitaxial thin film growth. Chem. Eng. Sci. 2014, 116, 590−600. (34) Gel, A.; Chaudhari, K.; Turton, R.; Nicoletti, P. Application of uncertainty quantification methods for coal devolatilization kinetics in gasifier modeling. Powder Technol. 2014, 265, 66−75. (35) Shastri, Y.; Diwekar, U. Stochastic Modeling for Uncertainty Analysis and Multiobjective Optimization of IGCC System with Single-Stage Coal Gasification. Ind. Eng. Chem. Res. 2011, 50, 4879. (36) Sahraei, M. H.; Yandon, R.; Duchesne, M. A.; Hughes, R. W.; Ricardez-Sandoval, L. A. Parametric analysis using a reactor network model for petroleum coke gasification. Energy Fuels 2015, 29, 7681− 7688. (37) Monaghan, R. F. D.; Ghoniem, A. F. A dynamic reduced order model for simulating entrained flow gasifiers: Part I: Model development and description. Fuel 2012, 91, 61−80. (38) Kus, F.; Duchesne, M.; Champagne, S.; Hughes, R.; Lu, D.; Macchi, A.; Mehrani, P. Pressurized pneumatic conveying of pulverized fuels for entrained flow gasification. Powder Technol. 2016, 287, 403−411. (39) Freelance product catalog. 2013. https://library.e.abb.com/ public/d2b2567827d584ac482576d5002cca6f/T568%20-EN.pdf (accessed 10.06.2016). (40) Niksa Energy Associates LLC. http://www.niksaenergy.com/ products/pc-coal-lab/ (accessed 04.02.2016). (41) Niksa, S.; Liu, G.-s.; Hurt, R. H. Coal conversion submodels for design applications at elevated pressures. Part I. devolatilization and char oxidation. Prog. Energy Combust. Sci. 2003, 29 (5), 425−477. (42) Ricardez-Sandoval, L. A. Optimal design and control of dynamic systems under uncertainty. Comput. Chem. Eng. 2012, 43, 91−107. (43) Kong, X.; Zhong, W.; Du, W.; Qian, F. Compartment modeling of coal gasification in an entrained flow gasifier: A study on the influence of operating conditions. Energy Convers. Manage. 2014, 82, 202−211.

6970
