Modeling of a Coal Syngas Based SOFC

Transcription

1 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 Modeling of a Coal Syngas Based SOFC S. R. Pakalapati, F. Elizalde-Blancas, I. B. Celik Mechanical and Aerospace Engineering Department, West Virginia University, Morgantown, WV, USA Abstract One of the distinctive features of Solid Oxide Fuel Cells is the fuel flexibility they offer which enables the utilization of coal derived gas (coal syngas) as fuel. However, coal syngas contains a wide range of impurities in addition to methane, hydrogen, carbon monoxide, carbon dioxide and nitrogen. Research in this area is focused on determining the effect of various impurities on the performance of the anode and understanding the underlying mechanisms of degradation. Numerical modeling is especially helpful in such research endeavors due to detailed information that could be obtained from the simulations. In the light of this background a three-dimensional, in-house SOFC simulation code (DREAM-SOFC) has been modified to include species commonly found in coal syngas namely CH 4, H, CO, CO, H O and N. Due to the fact that anodes containing Ni can act as catalysts for the methane reforming reaction and the water gas shift reaction the kinetics of these two reactions are also considered. Furthermore the present model takes into account the electrochemical oxidation of both H and CO. As an application, a SOFC button cell which is being tested at West Virginia University is simulated. Since there is no experimental data yet, the results are interpreted using some consistency checks. 1.0 Introduction One of the most promising fuels for Fuel Cell based power generation technology is coal syngas. However, coal syngas is known to be a dirty gas containing a wide range of impurities in addition to methane, hydrogen, carbon monoxide, carbon dioxide and nitrogen. Research in this area [1,, 3, 4] is being focused on determining the effect of various coal impurities on the performance of the anode and understanding the underlying mechanisms of degradation. Such knowledge would be necessary to develop new anode materials that can withstand coal syngas impurities. To expedite this new research, detailed numerical simulation tools are needed for modeling fuel cells operating on coal syngas. In the light of this background, a three-dimensional CFD code (DREAM-SOFC) developed in house [5, 6] has been modified to include species commonly found in coal syngas such as CH 4, H, CO, CO, H O and N [7]. Due to the presence of these species and the fact that anodes containing nickel can act as catalysts for the methane reforming reaction and the water gas shift reaction, the code accounts for the kinetics of these two reactions explicitly. Although many researchers consider the water gas shift reaction to be at equilibrium [8, 9, 10, 11] since it is faster than the methane reforming reaction, in this study no equilibrium assumption is made for any of the reactions. Unlike many models where only H oxidation is considered [1, 13, 14] in this study an electrochemistry model has been implemented based on the model used by Nishino et al. [15] for simultaneous H and CO electrochemical oxidation. The calculation of the activation overpotentials is carried out by using the Butler-Volmer equation instead of the empirical relations like those proposed by Achenbach [16]. The computational domain consists NMS_07_P09 53

2 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 of the anode and cathode electrodes as well as the electrolyte. Flow channels are not included in the model but their influence on the performance of the fuel cell is introduced as boundary conditions at the electrode surfaces. These boundary conditions were estimated from axisymmetric flow simulations performed using the commercial software FLUENT..0 Modeling Three dimensional numerical model used in this study, DREAM SOFC, was originally developed for simulating SOFCs working on H -H O mixtures. The details of this model are reported elsewhere in the literature [5, 6] and are not repeated here. Here, we present the new sub-models that were implemented in the code to extend its capabilities to: (i) simulating operation with multi-component fuel composition, (ii) modeling bulk chemical reactions inside the electrodes, (iii) accounting for simultaneous electrochemical oxidation of both CO and H. Diffusion Inside the porous electrodes the only transport mechanism considered is diffusion, i.e. permeation effects are neglected. The flux of each species in the multi-component mixture is modeled using Fick s law along with an effective multi-component diffusivity. The effective diffusivity of i th specie is given by 1 eff ε 1 αim, Yi 1 Di = + K τ Dim, D i (1) where ε and τ are the porosity and tortuosity respectively. The Knudsen diffusivity D K i, is taken into account in Eq. (1) and defined as 1/ K 8RT Di = r 3 π M i () where M i represents the molecular weight of the species i and r is the mean pore radius. The multi-component molecular diffusivity D i,m is given by D im, 1 Yi = Y k k i Dik, (3) where Y i is the mole fraction of the species i and D i,k is the binary diffusion coefficient for species i and k which is calculated using the Chapman-Enskog theory and given by 3 1/ T ( Mi + Mk) / MiM k Dik, = pσ ik, ΩD (4) For more detailed information on the diffusion model see [11, 19] Reforming Coal syngas usually contains some amounts of methane. Since SOFCs operate at high temperatures and Ni is present on the anode electrode, the conditions are favorable for the well known methane reforming reaction CH 4 + H O CO + 3H (5) Further the carbon monoxide can react with steam to produce more hydrogen through the water gas shift reaction (WGS) NMS_07_P09 54

3 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 CO + + H O CO H (6) In the present methane reformation model only the global reactions described by Eqs. (5) and (6) are considered unlike the detailed mechanisms presented in [1, 13, 17]. The reaction rates for the methane reforming reaction and the water gas shift reaction are given by Ns d[ C R rea i ] prod rea ν rs = kr( νri νri ) [ Cs] dt r= 1 s= 1 (7) where R represents the number of reactions and N s the number of species present in the mechanism; k r is the rate constant (forward and backward), ν is the stoichiometric coefficient and [C s ] denotes the concentration of the reactants present in reaction r. The equilibrium constants and forward rate constants for reactions (5) and (6) are given in terms of pressures by Gemmen et al. [7]. All these relations were converted to concentration basis using the following relations p c Kmr ( ), c p Kmr = K wgs = K (8a, b) wgs RT where the superscripts c and p mean concentration and pressure basis respectively. The subscripts mr and wgs stand for methane reforming and water gas shift reaction respectively. The relation between rate constants expressed in concentration and pressure basis are Forward f ( ) f f f kmr, c = RT kmr, p, kwgs, c = kwgs, p (9a, b) Backward f f b kmr, c k b wgs, c k = mr, c, k c wgs, c c K = mr K (9c, d) wgs Because the methane reforming and water gas shift reactions occur inside the anode, mass and energy sources have to be considered. The net mass source/sink due to the bulk chemical reactions for each of the species present is given by d[ Ci ] Smi, = Mi dt (10) Equation (10) is introduced as a source term in the species conservation equation for species i. The endothermic and exothermic nature of the methane reforming reaction and water gas shift reaction respectively is incorporated into the model through the energy source term in the energy conservation equation given by d[ CH4 ] d[ CO] Se = Δ Hmr + ΔHwgs dt dt (11) where ΔH mr and ΔH wgs are the enthalpy change of the methane reforming reaction and water gas shift reaction respectively. Electrochemistry The electrochemistry model implemented in this study considers the electrochemical oxidation of both hydrogen and carbon monoxide. The model uses the idea proposed by Nishino et al. [15] for partitioning the current between the two oxidation reactions and for calculating the resulting electric potential difference across the anode and cathode. The electrode reactions considered here are given by NMS_07_P09 55

4 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 Anode: H O H O e = + + = + + CO O CO e (13) At the cathode oxygen is reduced by the reaction = O + 4e O (14) Two overall reactions can be established as a consequence of the two oxidation reactions (1) and (13) given by 1 H + O HO (15) 1 CO + O CO (16) The Nernst potential for reactions (15) and (16) is given respectively by 0 RT P HO EH = E ln H 0.5 F PH P O (17) 0 RT P CO ECO = ECO ln 0.5 F PCOPO (18) The potential across the electrolyte due to the two electrochemical reactions occurring in parallel is given by EH η H η O = E c Ea (19) ECO ηco ηo = E c Ea (0) where η H, η CO and η O are the activation overpotentials for hydrogen, carbon monoxide and oxygen respectively. The calculation of these overpotentials is performed by using the Butler- Volmer equation given for each participating species s, i.e. H, CO, O separately αnf i = i exp η exp RT ( 1 α ) nf η RT s s s 0, s s s (1) where i 0,s is the exchange current given in terms of molar fractions Y H E act, H Y CO i0, H = c 1 exp, i0, CO = c Y H, ref RT Y CO, ref NMS_07_P09 56 (1) (a, b) Equations (a, b) were obtained by curve fitting to the experimental data reported in [18] in the temperature range of ºC. The activation energy barrier in Eq. (a) is 110.3kJ/mol. It is important to note that in Eqs. (19) and (0) the concentration overpotentials are not explicitly expressed since the partial pressures at the anode-electrolyte and cathode-electrolyte interface are used in Eqs. (17) and (18) for the calculation of the Nernst potential. Similarly the ohmic overpotential is accounted for in the charge conservation equation. The parameter α in Eqs. (1) is the charge transfer coefficient which in this study is assumed as 0.5; n s is the number of electrons transferred in the electrochemical reactions (1), (13) and (14). The current densities i H and i CO are the current densities driven by hydrogen and carbon monoxide respectively. i O =i tot is the total current density given by i = i + i tot H CO (3) Equations (19) and (0) along with Eq. (1) form a set of equations to solve for hydrogen and carbon monoxide activation overpotentials and current densities.

5 Status, Needs and Future Applications. Morgantown WV, October 11-1, Application In this work an anode supported button solid oxide fuel cell operating with coal syngas is simulated. The fuel fed to the button cell studied in this work is similar to the coal syngas obtained from an entrained flow gasification type as reported in [7]. The only difference is that this study does not consider the presence of N in the fuel stream, so the compositions by volume used are: 9.1% CO, 8.5%H, 11.8%CO, 9.7% H O and 0.9% CH 4. Air is supplied to the cathode side. The button cell is operated at atmospheric pressure and the fuel and air stream temperature is set at 103K. The fully three-dimensional computational domain of the button cell consists of the anode and cathode electrodes as well as the electrolyte. Flow channels are not included in the model but their influence on the performance of the fuel cell is introduced as boundary conditions at the electrode surfaces facing the flow channels. These boundary conditions were obtained from axisymmetric flow simulations using the commercial software FLUENT. At the boundaries surrounding the domain (lateral surfaces) the zero flux boundary condition was prescribed for all the scalars (except the species concentrations) being solved (i.e. temperature, mass fractions, electric potential). At anode/fuel interface (top) and cathode/air interface (bottom) boundaries different boundary conditions were applied depending on the variable being solved. For the electric field at the bottom boundary zero potential was prescribed while at the top boundary the potential was part of the solution since the total current is specified. The temperature at these boundaries was set as 103K. For the species conservation equations Dirichlet boundary conditions were prescribed at the electrode/gas channel interfaces. The physical domain is spatially discretized using cartesian grids and blocking the cells outside the active domain. The anode, cathode and electrolyte thicknesses are 780μm, 00μm and 0μm respectively. The cell active area is around cm. The number of grids used in length and width directions of the cell is 50. In the other direction (along the thickness of the cell) 0 grid points were used in the anode and 10 nodes each in the cathode and electrolyte. The discretization of the governing equations was performed using the control volume technique. The discretized equations were solved using DREAM-SOFC. 4.0 Results and discussion Figure 1(a) shows the simulated VI curve for the button cell operating on typical coal syngas mixture. The current densities used here are well below the cell limiting current and are in the range where the cell losses are dominated by ohmic over potential. The calculated current density distribution was found to be uniform over the active surface of the cell. The ratio of current produced by H oxidation to that produced by CO oxidation, predicted by the electrochemistry model, is shown in Fig 1(b). Contrary to the usual assumption of a constant value, Fig 1(b) shows that the ratio of current partition between H and CO could vary with the cell current density. However, it is interesting to note that for most part, the ratio is around.5 as proposed by Suwanwarangkul et al. [4] and as the current increases it seems to approach a limiting value of about 3.5 close to that used by Gemmen and Trembly [7]. Temperature distribution in each case was more or less uniform over the thickness of the cell varying slightly along the radial direction with a maximum variation of about 1 o C from the center of the cell to the edge observed for the highest current density case simulated (0.8 A/cm ). NMS_07_P09 57

6 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 Cell Voltage (V) ih/ico Average Current Density (amp/cm) Average Current Density (amp/cm) (a) cell voltage (b) Ratio between current produced by H and CO Figure 1: Variations of cell parameters with the average current density Figure (a) shows the predicted cell temperature at different current densities. As expected the cell temperature increases with the current density and the dependence is non-linear. Three different heat sources in the cell are plotted against the current density in Fig (b). Entropy heating is due to the irreversibility during the electrochemical oxidation of H and CO, and bulk reaction heat source is due to endothermic steam methane reaction and slightly exothermic water gas shift reaction. These values were calculated for the whole cell by summing the local amounts in each control volume. As would be expected, the ohmic heat source exhibits quadratic variation with current density. The entropy and bulk reaction sources seem to vary linearly with the current density though the former is increasing and the latter is decreasing with increasing temperature. It is also interesting to note that at low current densities the cell heating is dominated by the heat release from the bulk reactions whereas at higher current densities entropy heating is dominant. The decrease in the bulk reaction heat generation with increasing current alludes to the possibility of water gas shift reaction being less favored at high current densities. This indeed is the case as seen from the profiles of the water gas shift reaction rates (Fig 3a). These rates are expressed in terms of moles of H O produced and thus negative values mean that the net reaction is forward. At low currents the WGS reaction rates are higher with the highest values being at the anode fuel channel interface. But as the current increases the reaction rates slow down especially near the active anode/electrolyte interface, even becoming positive (i. e. net backward reaction) for higher currents. A possible reason for this reversal is that at higher current densities as both CO and H are oxidized at a higher rate; being a smaller molecule H diffuses faster to the reaction site than CO thus shifting the concentrations to levels where reverse WGS reaction is favored. Thus less heat is produced from the WGS reactions at high current densities. In addition, methane reforming increases slightly with increasing current density (Fig. 3a). Thus at high current densities, more heat is consumed by the endothermic reforming reaction. NMS_07_P09 58

7 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 Cell Temperature Heat Generation (watts) Entropy Ohmic bulk reaction Average curren density (amp/cm) Average Current Density (amp/cm) (a) Maximum Cell Temperatures (b) Various heat sources in the cell Figure : Variations of cell parameters with the average current density (a) Distributions predicted by the model (b) Profiles with and with out fixed ratio assumption Figure 3: Reaction rate profiles along the anode thickness at different current densities Figures 4(a&b) show mass fraction distributions within the anode. It is observed that at low current densities, the hydrogen concentration inside the anode is higher than that at the fuel boundary. This is due to the high rates of forward WGS reaction which supplies more hydrogen than required for the current production. However, as current density increases, higher electrochemical consumption of H along with lower WGS reaction rates lowers the hydrogen mass fractions inside the anode. Since CO is consumed during the forward WGS reaction, the concentrations of CO inside the anode are always less than those at the fuel channel boundary (see Fig 4a). Carbon dioxide mass fraction profiles shown in Fig. 4(b) are consistent with the expected trends (i.e. increases inside the anode) since it is being produced from the water gas shift reaction as well as from the electrochemical oxidation of carbon monoxide. To assess the implications of the constant ratio assumption between the H and CO currents additional simulations were performed for two average cell current densities where this ratio is set to be 4:1 at an operating current density of 0.7 A/cm. Figs 4(c&d) show that fixing the ratio did not NMS_07_P09 59

8 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 greatly affect the concentration distribution inside the anode. However, the profiles shown in Fig 3(b) reveal that the rates of WGS and reforming reactions could be widely different for these two cases. This could mean that though the concentration profiles look similar for these two cases, the amount of heat generated due to the bulk chemical reactions may be different. Indeed it was found that the bulk reaction heat at 0.7 amp/cm current density increased 8 % when the current ratio was fixed. Also, the entropy generation during CO oxidation is much higher than that for H oxidation. Thus when the proportion between these two reactions changes the entropy heat generation will also be affected. The present simulations show that, entropy heat decreases by about 15 % when the current ratio is fixed at 4:1. However, the cell voltage for fixed current ratio was practically same as that when the current ratio was not fixed. (a) H and CO (b) H O and CO (c) H and CO (d) H O and CO Figure 4: Concentration profiles along the anode thickness at different current densities NMS_07_P09 60

9 Status, Needs and Future Applications. Morgantown WV, October 11-1, Conclusions The in house SOFC 3D simulation code was extended to include complex fuel mixtures such as coal syngas. Fick s law type relations with carefully estimated effective multi-component diffusivities were used to model diffusion fluxes. Chemical kinetics was used to account for the water gas shift reaction and steam reforming reaction. The new electrochemistry sub model used in this study allows for simultaneous electrochemical oxidation of CO and H. A button cell operating on typical coal syngas was simulated using the model and results were obtained for different operating current densities. It was found that the ratio between the current produced from the H oxidation and that from CO oxidation increase with the cell current and it ranges ca. 1 to 3.5. The cell heat generation is dominated by heat produced from bulk chemical reactions at the low operating current, whereas at high currents it is dominated by the entropy heat generation and ohmic heat. The methane reaction rates increased with the operating current whereas the water gas shift reaction rates decreased. It was further found that though the assumption of constant ratio between CO and H currents does not greatly alter the concentration distributions, it may result in significantly different heat generation and thus different temperature distribution. Acknowledgements This worked was supported by DOE EPSCoR, WV state EPSCoR and WVU Research Corporation under grant No: DE-FG0-06ER4699 References [1] J. P. Trembly, R. S. Gemmen, D. J. Bayless (007) The effect of coal syngas containing AsH 3 on the performance of SOFCs: Investigations into the effect of operational temperature, current density and AsH 3 concentration, Journal of Power Sources, 171, pp [] J. P. Trembly, R. S. Gemmen, D. J. Bayless (007) The effect of coal syngas containing HCl on the performance of solid oxide fuel cells: Investigations into the effect of operational temperature and HCl concentration, Journal of Power Sources, 169, pp [3] J. P. Trembly, A. I. Marquez, T. R. Ohrn, D. J. Bayless (006) Effects of coal syngas and H S on the performance of solid oxide fuel cells: Single-cell tests, Journal of Power Sources, 158, pp [4] R. Suwanwarangkul, E. Croiset, E. Entchev, S. Charojrochkul, M. D. Pritzker, M. W. Fowler, P. L. Douglas, S. Chewathanakup, H. Mahaudom (006) Experimental and modeling study of solid oxide fuel cell operating with syngas fuel, Journal of Power Sources, 161, pp [5] S. Pakalapati, I. Yavuz, F. Elizalde-Blancas, I. Celik, M. Shahnam, Comparison of a multidimensional model with a reduced order pseudo three-dimensional model for simulation of solid oxide fuel cells, Proceedings of the 4th International Conference on Fuel Cell Science, Engineering and Technology, Irvine, CA, June 19-1, 006 [6] S. R Pakalapati, A new reduced order model for solid oxide fuel cells, Ph.D. Thesis, West Virginia University, Morgantown, WV, 006 [7] R. S. Gemmen, J. Trembly (006) On the mechanism and behavior of coal syngas transport and reaction within the anode of a solid oxide fuel cell, Journal of Power Sources, 161, pp [8] S. Nagata, A. Momma, T. Kato, Y. Kasuga (001) Numerical analysis of output characteristics of tubular SOFC with internal reformer, Journal of Power Sources, 101, pp NMS_07_P09 61

10 Status, Needs and Future Applications. Morgantown WV, October 11-1, 007 [9] P. Aguiar, D. Chadwick, L. Kershenbaum (00) Modelling of an indirect reforming solid oxide fuel cell, Chemical Engineering Science, 57, pp [10] D. Sánchez, R. Chacartegui, A. Muñoz, T. Sánchez (006) Thermal and electrochemical model of internal reforming solid oxide fuel cells with tubular geometry, Journal of Power Sources, 160, pp [11] H. Yakabe, M. Hishinuma, M. Uratani, Y. Matsuzaki, I. Yasuda (000) Evaluation and modeling of performance on anode-supported solid oxide fuel cell, Journal of Power Sources, 86, pp [1] V. M. Janardhanan, O. Deutschmann (006) CFD analysis of a solid oxide fuel cell with internal reforming: Coupled interactions of transport, heterogeneous catalysis and electrochemical processes, Journal of Power Sources, 16, pp [13] H. Zhu, R. J. Kee, V. M. Janardhanan, O. Deutschmann, D. G. Goodwin (005) Modeling Elementary Heterogeneous Chemistry and Electrochemistry in Solid-Oxide Fuel Cells, Journal of the Electrochemical Society, 15, pp A47-A440 [14] J. Li, G. Cao, X. Zhu, H. Tu (007) Two-dimensional dynamic simulation of a direct internal reforming solid oxide fuel cell, Journal of Power Sources, 171, pp [15] T. Nishino, H. Iwai, K. Suzuki (006) Comprehensive Numerical Modeling and Analysis of a Cell-Based Indirect Internal Reforming Tubular SOFC, Journal of Fuel Cell Science and Technology, 3, pp [16] E. Achenbach (1994) Three-dimensional and time-dependent simulation of a planar solid oxide fuel cell stack, Journal of Power Sources, 49, pp [17] E. S. Hecht, G. K. Gupta, H. Zhu, A. M. Dean, R. J. Kee, L. Maier, O. Deutschmann (005) Methane reforming kinetics within a Ni-YSZ SOFC anode support, Applied Catalysis A: General, 95, pp [18] A. M. Sukeshini, B. Habibzadeh, B. P. Becker, C. A. Stoltz, B. W. Eichhorn, G. S. Jackson (006) Electrochemical Oxidation of H, CO and CO/H Mixtures on Patterned Ni Anodes on YSZ Electrolytes, Journal of the Electrochemical Society, 153, pp A705-A715 [19] R. B. Bird, W. E. Stewart, E. N. Lightfoot, Transport Phenomena, John Wiley & Sons, NY, 00 NMS_07_P09 6