International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Contents lists available at ScienceDirect International Journal of Heat and Mass Transfer journal homepage: www.elsevier.com/locate/ijhmt A numerical analysis of unsteady transport phenomena in a Direct Internal Reforming Solid Oxide Fuel Cell Maciej Chalusiak a, Michal Wrobel a, Marcin Mozdzierz a, Katarzyna Berent b, Janusz S. Szmyd a, Shinji Kimijima c, Grzegorz Brus a,⇑ a b c AGH University of Science and Technology, Department of Energy and Fuels, 30 Mickiewicza Av., 30-059 Krakow, Poland AGH University of Science and Technology, Academic Centre for Materials and Nanotechnology, 30 Mickiewicza Av., 30-059 Krakow, Poland Shibaura Institute of Technology, Department of Machinery and Control Systems, 307 Fukasaku, Minuma-ku, 337-8570 Saitama City, Japan a r t i c l e i n f o Article history: Received 22 August 2018 Received in revised form 14 November 2018 Accepted 22 November 2018 Available online 30 November 2018 Keywords: Solid oxide fuel cell Internal reforming kinetics Dynamic model FIB-SEM Microstructure a b s t r a c t In this paper, a transient microstructure-oriented numerical simulation of a planar Direct Internal Reforming Solid Oxide Fuel Cell (DIR-SOFC) is delivered. The performance criteria in a direct steam reforming for a fuel starvation scenario are analyzed in order to optimize the underlying process. The proposed two-dimensional multiscale model takes into account mass and heat transport, electrochemistry, as well as the intrinsic steam-reforming kinetics. In the paper, the methane/steam reforming process over the Ni/YSZ catalyst is experimentally investigated to verify the used chemical reaction model. A threedimensional digital microstructure representation of the commercial anode is analyzed using a Focused Ion Beam-Scanning Electron Microscope (FIB-SEM) and the nickel-pore contact surface is calculated to relate the reforming reaction rate to the catalyst’s active area. Based on the complete DIR-SOFC model, a parametric study is carried out, to simulate the dynamic response of a fuel cell for different design and operating conditions. The results prove the dominant impact of inlet fluid temperature and methane content on the calculated distribution of hydrogen across the channel, while the collected current density was found to be a less important factor. The simulations indicate, that in the case of the direct reforming, fuel starvation is likely to occur in the upstream of the anode channel, where the reforming reaction does not keep up with producing hydrogen. The obtained results provide a significant insight into safe and efficient control strategies for Solid Oxide Fuel Cells. Ó 2018 Elsevier Ltd. All rights reserved. 1. Introduction Growing popularity and rapid development of Solid Oxide Fuel Cells (SOFCs) stem for their potential to become a gamechanger in the field of clean power generation technologies. Due to a low level of pollutants’ emission, high operating temperature (600–1000 °C) and a wide range of utilized fuels, SOFC’s can play a key role in supporting conventional power systems [1,2]. A distinguishing feature of SOFCs with nickel-based anodes (e.g. Ni/YSZ) is the ability to internally reform a wide range of hydrocarbons, like methane. Together with increased tolerance to most of the impurities, it represents a significant advantage of this type of fuel cell. Direct internal reforming of methane on the DIR-SOFC anode prominently increases the system efficiency by recuperating waste heat from the electrochemical reactions to supply strongly endothermic reforming reaction, at the same time reducing the complexity, size ⇑ Corresponding author. E-mail address: brus@agh.edu.pl (G. Brus). https://doi.org/10.1016/j.ijheatmasstransfer.2018.11.113 0017-9310/Ó 2018 Elsevier Ltd. All rights reserved. and cost of the system by elimination of the external reformer [3,4]. This beneficial configuration of stack, however, engages various phenomena occurring simultaneously in one place and may be found highly challenging for thermal and substantial management. Endothermic steam reforming reaction may cause severe temperature gradients in the cell, leading to high thermal stresses, which, in the long run, may damage the cell’s ceramic components [5,6]. What is more, mismatching thermal effects in the DIR-SOFC may result in an anode cooling effect or insufficient hydrogen production. These issues were widely addressed in recent studies in terms of a steady state operation [7,8]. However, stages in which the cell’s operating conditions change in time, e.g. start-up or external load connection, are inevitable in its lifetime and it is extremely difficult to control the cell’s behaviour in such transient states. Therefore, it is crucial for optimization and management of the cell’s performance to understand the dynamic behaviour of DIR-SOFCs [9]. Yet, experimental studies of DIR-SOFCs under transient conditions is difficult, due to obstacles in non-invasive measurements of physical fields within the cell unit. Thus, numerical simulations M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 is a reasonable and cost-effective way to study DIR-SOFCs performance. With the introduction of Computational Fluid Dynamics codes, a much deeper understanding of the processes inside any fuel cell can be provided [8,10]. Dynamic behavior and the modeling of DIR-SOFC was of interest to several research groups in the past [11–14]. A dynamic model of DIR-SOFC stack was recently presented in the paper of Kupecki et al. [14]. The simulations of the stack operation took a complete zero-dimensional reforming model into account and the results confirmed the impact of reforming reaction kinetics on the thermal balance of the cell [14]. A highly endothermic reforming process may locally reduce the temperature, which leads to higher thermal stresses. The simulations indicate that temperature changes at the outlet are mostly affected by the high current values [14]. Aguiar et al. [11] investigated a timedependent response of an anode-supported DIR-SOFC to a step load change. The proposed one-dimensional model took into account mass and energy balances, as well as electrochemical and chemical reactions. In the model, a shift in power demands was represented by the current density step-change. The dynamic simulations indicate, that the overall SOFC temperature rises after positive load step-change, and drops in an opposite case. Additionally, the load increase was accompanied by an instant growth in hydrogen consumption. Huangfu et al. [15] developed a onedimensional SOFC model and performed its sensitivity analysis. In their study, they provided a transient distribution of temperature and anode gas in the channel and it shows, that the transition time due to the cells thermal capacity is in a range of seconds. An interesting study was also presented by Kang et al. [16], where the authors develop a one-dimensional dynamic model of a planar DIR-SOFC. Their results prove the dominance of the internal reforming process on the distributions of the anode gas composition and temperature, while various operating parameters change such as load current and fuel flow were introduced to the model. Nevertheless, as Bae et al. [17] noticed while analyzing the transient transport phenomena in SOFCs, more experimental and modeling studies should be performed to understand the effect of cells microstructure on transport processes and the overall dynamic response of SOFC thermodynamic variables. A limited amount of research, however, has been conducted to study the microstructure dependency on SOFCs performance, while recent studies by Kishimoto et al. [18] indicate that the Ni/YSZ anode microstructure affects the catalytic reaction rates and should be taken into account in CFD models. In the viewpoint of fuelling DIR-SOFC a substantial management appears as an important matter. Due to limitations originating from the kinetics of the methane reforming reaction, insufficient amounts of hydrogen may be produced, causing a fuel starvation phenomenon [19]. The authors observed, that during hydrogen starvation, acceleration of the degradation of the anode occurs mainly due to the mechanical stresses and disconnection of nickel particles. A large interest in developing the DIR-SOFC technology and a lack of current reports on three-dimensional dynamic models which include microstructure, inclines to a thorough research. Commercial SOFC cells’ manufacturers overcome the critical effects in the stagnation points, such as edges or corners, by using cassettes, in which the electrode does not cover the entire support. However, some problems in the longitudinal direction still need a solution. This work, therefore, aims to fill the void in the literature and to investigate the behaviour of DIR-SOFCs by time-dependent numerical simulations of a simplified twodimensional model that account for the anode microstructure, which reflects the qualitative characteristics in the core of the fluid flow. A complete model requires a detailed information about the methane steam reforming kinetics. Therefore, following the results 1033 from the literature, the utilized reforming reaction rate was verified using experimental data. Moreover, the anode microstructural parameters obtained with the FIB-SEM electron tomography were included in the rate equation to best reflect the real conditions. The empirical rate was then employed in the proposed numerical model. The computational scheme was built based on the Finite Volume Method. The scheme was thoroughly verified against the analytical benchmark solution and a good agreement was found. Having verified the accuracy of the computations we numerically analyzed the influence of the anode microstructure, load changes, fuel inlet flow, inlet temperature and fuel composition, on the DIR-SOFC performance. The parameters mentioned above, were presented in the numerical model by boundary and operating conditions. The calculated cross-sectional temperature and species distributions are shown as the results of the calculations and discussed with reference to the fuel starvation, in order to seek the properties of DIR-SOFC using which electrochemical and chemical reactions can be governed. 2. Mathematical model The computational domain refers to a 2D cross-section of the DIR-SOFC anode (fuel) channel. Its geometrical configuration is presented in Fig. 1. In the developed model, SOFC is fueled with methane (CH4 ) and a corresponding amount of steam (H2 O) to satisfy the set steam-tocarbon ratio (SC). Moreover, the hydrogen (H2), carbon monoxide (CO), carbon dioxide (CO2) and temperature (T) distributions are analyzed in the model. The following assumptions are made to describe the anode channel [1,20–22]: – flow in the fuel channel is unidirectional, laminar and time – independent; – atmospheric pressure is assumed in the calculation domain; – all fluids are incompressible, ideal gases; – heat transfer by radiation and fuel leakage is neglected; – all chemical and electrochemical reactions occur at the interface between the fuel channel and the anode; – the influence of current collectors, sealings, isolations, etc. are neglected in the model; – the anode is completely and fully penetrated by the gases; – hydrogen is assumed to be a primary fuel for the cell; – the thermal and substantial effects of the reforming reaction are considered as if they occurred in the entire volume of the anode. The fluid flow model is described in detail in Section 2.2. Fig. 1. Schematic configuration of the system. 1034 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 @u ¼ 0; @x 2.1. Heat and mass transfer model In the present model, the transport equations are derived by the volume-averaging method. Heat transfer inside the anode channel is governed by the following differential equation [21,23]: @ ðqC P T Þ þ $ ðquC P T Þ ¼ $ ðk$T Þ; @t ð1Þ 3 where q stands for density in [kg m ], C P is heat capacity, in [J kg1 K1], T stands for fluid temperature, in [K], k is the thermal conductivity of the mixture of the gases, in [W m1 K1], and u is the velocity vector, in [m s1]. The concentration of chemical species in the anode channel is governed by the mass conservation equation [23]: @ ðqY i Þ þ $ ðquY i Þ ¼ $ ðqDi $Y i Þ; @t ð2Þ ð3Þ Moreover, they are assumed to depend on the mixture temperature and composition. The thermophysical properties in Eqs. (3) are calculated locally according to the mixing laws, taken from [24]. The respective laws are described in detail in Appendix A. In the proposed model, the system of mass transport equations is solved with respect to the species mass fractions and the results are converted into partial pressures or molar fraction during post-processing e.g. for use in the steam reforming model, which is described in Section 2.3. 2.2. Fluid flow model In the general case, the time-independent fluid flow in the channel is governed by [23,25]: the continuity equation $ ðquÞ ¼ 0; where u ¼ uðyÞ is the only non-zero component of the velocity vector. The system of Navier-Stokes Eq. (5) yields: @ 2 u 1 @p ; ¼ @y2 l @y @p < 0: @x The above system of equations describes the Poiseuille-type flow. For the negative pressure gradient @p < 0, a positive value of velocity @x u is obtained. We assume that the fluid flows in the channel at a known rate N_ [m3 s1], defined as: N_ ¼ d Z H ð9Þ uðyÞdy; ð4Þ where d in [m] is the width of the unit cell in the MSTB test bench [26,22]. One can prove that the solution to the above system of equations is: uðyÞ ¼ 6N_ Hy y2 : 3 dH The methane/steam reforming process is widely known as a conventional process for producing hydrogen. The process consists of a set of numerous elementary reactions. However, two of them are proved to be largely dominant in the whole process [1,27–29]: the steam reforming reaction: 1 CH4 þ H2 O ! 3H2 þ CO DH0 ¼ 206:2 kJ mol ; CO þ H2 O CO2 þ H2 Rsh ¼ ksh pCO pH2 O þ ksh pH2 pCO2 ; ð6Þ p is the fluid pressure, in [Pa], while q, in [kg m3], and l, in [Pa s], are the fluid density and viscosity, respectively. In the presented model, a unidirectional, incompressible (q = const) and steady flow was assumed in the channel. As a result, the continuity Eq. (4) reduces to: ð12Þ The rates of the methane/steam reforming and water-gas shift reactions are taken from Sciazko et al. [31,32], where they were derived experimentally: where u is the fluid velocity vector, in [m s1], s defines the shearstress tensor, in [Pa] (here we assume a form for the Newtonian fluid): s¼l 1 DH0 ¼ 41:2 kJ mol : Since the reaction described by Eq. (11) is strongly endothermic, the temperature drops within the reaction area, if no heat source is present. Therefore a supply of thermal energy is needed. Moreover, the steam reforming reaction is slow, thus one needs to describe it by a pertinent rate equation. The water-gas shift reaction, described by Eq. (12), is fast and weakly exothermic and can be assumed to be in equilibrium at the reforming temperature [1,21,30]. ð5Þ 2 $u þ $uT $u ; 3 ð11Þ the Water-Gas-Shift reaction: _ cat 6:47 103 exp Rst ¼ w ð10Þ 2.3. Steam methane reforming model the system of Navier-Stokes equations $ ðquuÞ ¼ $p þ $s; ð8Þ 0 where Y i stands for the dimensionless mass fraction of species i and Di denotes the mass diffusion coefficient of species i in the mixture of gases, in [m2 s1]. The energy and mass conservation reflect the unsteady heat and mass transfer by both diffusion and convection in a single, gaseous, continuous gas mixture. In the studied system, all chemical and electrochemical reactions take place inside the anode, which, as a separate subdomain, is neglected in the model. Instead, the mass and heat effects of those reactions are introduced to the balances in the channel via the mass and heat fluxes. Equations, which describe the fluxes, are shown in Section 2.4. The physical parameters in Eqs. (1) and (2), the density q, the specific heat C P and the thermal conductivity k refer to the gas mixture, thus: q ¼ qmix ; C P ¼ C P;mix ; k ¼ kmix : ð7Þ þ 121 103 RT ! 0:083 p0:88 CH4 pH2 O ; ð13Þ ð14Þ where Eact kst ¼ Ast exp : RT ð15Þ The water-gas shift reaction rate, given by Eq. (14), reaches equilibrium rapidly, hence CO2 ; H2 ; CO oraz H2 O have to satisfy the equilibrium equation as follows [21,33]: K sh ! þ ksh pCO2 pH2 DG0sh : ¼ ¼ ¼ exp ksh pCO pH2 O RT ð16Þ M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 _ cat stands for the catalyst density, in [g m3], ast and bst In Eq. (13), w stand for the steam reforming reaction orders towards methane and steam, respectively. The species concentrations in the reaction site are expressed by their partial pressures, pCH4 and pH2 O , in [Pa]. The reforming rate constant kst , in [mol s1 g1 Paðast þbst Þ ], is described by Arrhenius constant Ast , in [mol s1 g1 Paðast þbst Þ ], activation energy Eact , in [J mol1], universal gas constant R, in [J kg1 K1], þ and temperature T, in [K]. In Eqs. (14) and (16), ksh and ksh are the rate constants of the forward/backward water-gas shift reaction, in [mol s1 m3 Pa2], DG0sh is the change in Gibbs free energy, in [J mol1]. Anode microstructure plays an important role in fuel cell operation, especially during the internal reforming process. It provides chemical and electrochemical reaction sites, as well as pathways for the migration of the reaction components. In this study the Ni/YSZ cermet anode is used, in which nickel serves as a reforming reaction catalyst. The effectiveness of this reaction depends strongly on the accessibility of the reaction sites to the methane penetrating the anode through the pores; in other words, nickel needs to be well exposed to the flowing methane and steam. Its exposure can be estimated by two parameters, which account for the local steam reforming reaction rate: the SNi-Pore – nickel-pore Ni contact area density, in [m2cat m3 ], and the A – Ni/YSZ specific surface area, in [m2cat g1 ]. The methodology for the evaluation of these parameters is described in Section 4. 2.4. Boundary and initial conditions The boundary conditions, employed in the model are shown schematically in Fig. 2. The anodic reactions are introduced through the dedicated non-local boundary conditions imposed at the channel/anode interface. The respective formulae for the boundary conditions are collected in Table 2. Their implementation assumes firstly calculating the overall effects for the whole volume of the anode V ano , in [m3], and the subsequent evaluation of the fluxes exchanged by the interface Aano , in [m2]. The heat generated by the methane/steam reforming reaction in Eq. (11) and Water-Gas Shift reaction in Eq. (12) is as follows [1]: Q st ¼ DH0st Rst V ano ; Q sh ¼ DH0sh Rsh V ano ; DH0st DH0sh ð17Þ ð18Þ where and stand for the enthalpy change in the methane/ steam reforming reaction and with the water-gas shift reaction, respectively, in [J mol1]. Hydrogen, the product of the reforming process, is utilized in the exothermic oxidation reaction: Fig. 2. Boundary conditions of the calculations’ domain. H2 þ O2 2 ! H2 O þ 2e ; 1035 ð19Þ which supplies the fuel cell with the electric charge. Heat from this reaction is generated as follows: Q form ¼ I DHfH2 O ; 2F ð20Þ where DHfH2 O is the enthalpy change in the water formation reaction, in [J mol1]. The electric charge flows through the anode generating Joule’s heat, according to: Q Joule ¼ I2 Rcell ; ð21Þ where I stands for the overall current collected from the unit cell, in [A], which is calculated as the current density iden , in [A cm2], multiplied by the cell area, Rcell stands for the unit cell resistance, in [X], and F denotes Faraday’s constant, in [C mol1]. The total heat flux Q total , in [W m2], through the anode surface is calculated as follows: 1 Q_ total ¼ Q þ Q st þ Q form þ Q Joule ; Aano st ð22Þ and applied in accordance with Table 2. The mass fluxes in the model are calculated in the same manner as the heat fluxes. Mass is produced/consumed in both chemical and electrochemical reactions only in the anode volume. The methane and steam entering the fuel channel take part in the reforming process, during which hydrogen, carbon monoxide and carbon dioxide are produced. Simulatenously to the reforming process, go the electrochemical reactions. The total sources/sinks of species are shown in Table 1. For example, the hydrogen generation flux due to the reforming process in the anode volume can be calculated as follows: J ref;H2 ¼ M H2 ð3Rst þ Rsh ÞV ano : ð23Þ Additionally, hydrogen is electrochemically oxidized, according to Eq. (19). As a result, high temperature steam is produced. The rate of the respective species production in the electrochemical reactions can be calculated according to Faraday’s law. For example, for H2 one obtains: J F;H2 ¼ I M H2 O : 2F ð24Þ In the proposed model, hydrogen and carbon monoxide are utilized as fuels. The total species flux density through the anode surface is as follows: i 1 X J_ total;i ¼ J ref;i þ J F;i ; Aano ð25Þ where i ¼ hH2 ; H2 O; CO; CO2 ; CH4 i: At the external boundaries of the domain, constant temperature and predefined mass fluxes are imposed, as presented in Fig. 2. On the inlet constant values of gas composition, fluid temperature and volumetric flow are applied. In this paper, the model sensitivity analysis is carried out inter alia, by introducing various Dirichlet boundary conditions to the domain, depending on the case (species or temperature). The upper boundary is assumed to be a gas-tight, adiabatic partition, which isolates the anode channel, so that no heat or mass is transferred through it. On the outlet a zero-gradient assumption is made both for the temperature and mass fraction gradients [25]. On the lower wall (the anode surface) steam reforming reaction, Q st , water-gas shift reaction, Q sh , Joule’s heat, Q Joule and electrochemical reactions, Q form , thermal effects are introduced, together with the mass flux of species i, J i . The inlet, outlet and upper boundary conditions are time independent in the model. The anode surface boundary condition is calculated 1036 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Table 1 Mass sources/sinks in the model. Species Mass source/sink due to methane steam reforming Mass source/sink due to the water-gas shift reaction Mass source/sink due to electrochemical reactions Total mass source/sink 2FI M H2 3Rst MH2 þ Rsh MH2 2FI M H2 2FI M CO Rst M CO Rsh M CO 2FI MCO MCO2 Rsh MCO2 +2FI M CO2 Rst M CH4 M H2 O Rst M H2 O Rsh M H2 O +2FI M H2 O H2 3Rst MH2 Rsh MH2 CO Rst M CO Rsh MCO CO2 0 Rsh MCO2 CH4 H2 O Rst M CH4 Rst M H2 O I 2F 0 Rsh MH2 O 0 I 2F Table 2 Boundary conditions. Position u Inlet @u @x @u @x Outlet Table 3 Catalyst properties. v T Yi Type Nickel content Particle size Specific surface area, Acat ¼0 v=0 T = Tinlet Yi = Yinlet i Ni/YSZ 60% 0.85 lm 5.2 m2 g1 ¼0 @v @x @T @x ¼0 @T @y @T @y ¼ Q total k @Y i @x @Y i @y @Y i @y ¼0 Bottom u=0 v=0 Top u=0 v=0 _ ¼0 ¼0 _ ¼ qJDi i ¼0 iteratively at every time step, based on the conditions from the previous iteration. Initially (t = 0 s), the mass fractions for hydrogen, methane, steam, carbon monoxide, carbon dioxide for all studied cases are 0. Moreover, it is assumed that nitrogen fills the anode channel completely at the initial time. The initial temperature of the gases in the fuel channel is set as 1023 K. 3. Experimental analysis Type Bed height Radius Length Stainless steel 1 mm 25.4 mm 450 mm were placed in the experimental set-up as shown in Fig. 3 (marked as T). All measurements presented in this paper were performed at atmospheric pressure. The geometrical configuration of the reactor is summarized in Table 4. 4. Relation between reforming reaction rate and microstructure morphology A reforming process experiment over the Ni/YSZ catalyst was carried out to verify the utilized chemical reaction model. A schematic view of the experimental setup is shown in Fig. 3. A stainless steel reformer was located in an electrical furnace, which can be heated up to 800 C. High purity methane was the fuel used in the experiment. It was supplied to the reformer via a flow controller and evaporator which was also used as a pre-heater. Water was fed to the system with a pump. The gas composition after the reforming process was analyzed by gas chromatography prior to which the steam had been separated by cooling down the gas mixture to 2 C. The reforming reaction tube was filled by a nickel supported on the yttria stabilized zirconia. The catalyst material was the industrial catalyst provided by AGC Seimi Chemical Co. Ltd. The properties of the catalyst material are shown in Table 3. Before feeding methane to the system, the catalyst material was treated for 2 h at an elevated temperature of 600 C by a mixture of 1 nitrogen (supplied at a rate of 150 ml min ) and hydrogen 1 Table 4 Reactor properties. (100 ml min ), to reduce NiO to metallic Ni. In order to avoid large temperature gradients in the reformer and a cooling effect of entering fluid, the reformer was partially filled with Al2 O3 balls. To control the thermal conditions of the experiment, four thermocouples In this study, the methane/steam reforming reaction rate on a conventional Ni/YSZ anode is investigated under varying temperature and gas conditions. To derive the reforming kinetics, the powder catalyst was used. The original rate (see Eq. (13)) was derived from a catalyst weight. However, recent studies by [19,18] using Focused Ion Beam-Scanning Electron Microscopy (FIB-SEM) indicate, that the reaction rate constant (see Eq. (15)) depends on the anode microstructure, as the underlying process takes place only at the nickel-pore interface. Thus, a respective rate equation is applicable only to a very specific anode. To overcome this obstacle, a generalized model was introduced [18]. It takes into account two parameters, the nickel-pore interface area density within the catalyst, SNi-Pore and Ni/YSZ specific surface area, ANi as a substitute for _ cat . By modifying the reaction rate equation the catalyst density w (see Eq. (13)), one obtains a universal formula suitable for any anode. Microstructural morphology and its influence on the reforming reaction rate was evaluated for a commercial state-of-the-art Ni/ YSZ anode provided by a leading SOFC stacks manufacturer, SOLD-POWER S.p.A [26,22]. In order to examine the anode Fig. 3. Schematic view on the experimental setup. 1037 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 microstructure, the 0.5 cm2 anode sample was extracted and impregnated with epoxy resin under vacuum conditions (Marumoto Struers KK). This procedure is essential for pore region recognition, as the resin fills the pores completely and yields black spots during SEM observation. A FIB-SEM tomography (Focused Ion Beam Scanning Electron Microscopy) was used to closely investigate the anode microstructure. The measurement procedure scheme is presented in Fig. 4a. A ”cut and see” cycle using FIB (cutting) and SEM (cross-section capturing) was repeated with nanoscale resolution to obtain a series of sample cross-sections. The images were then manually segmented in AVIZO software, where nickel, YSZ and pores were identified, and the 3D reconstructed microstructure was generated (see Fig. 4b) [26,22]. In-house codes were employed to calculate parameters, which describe the fabricated the anode microstructure (see Table 5). Based on the values of surface area of species, the nickel-pore interface area density SNi-Pore was calculated, according to the formula: SNi-Pore ¼ 1 Ni S þ SPore SYSZ : 2 ð26Þ Hence, the methane/steam reforming reaction rate for a unit Ni/YSZ reaction area can be formulated, as follows: SNi-Pore Eact st Rst ¼ cat Ast exp pbHst2 O : paCH 4 RT A ð27Þ 5. Numerical model The system of the governing differential equations given in Section 2 can be presented in the following general linearized form: wt @/ @/ @/ @ @/ @ @/ þ wx þ wy ¼ h þ h : @t @x @y @x @x @y @y ð28Þ The values of the respective coefficients from Eq. (28) are presented in Table 6. The numerical solution to the system will be sought in the framework of the Finite Volume Method [23]. The spatial discretization is based on the introduction of the control volumes over which the transport equations are integrated (details of the discretization scheme can be found in [23,25]). As for the approximation of the temporal derivative, we accept the formula: @/ðiþ1Þ /ðiþ1Þ /ðiÞ @/ðiÞ ¼2 ; @t Dt @t ð29Þ where /ðiÞ and /ðiþ1Þ represent the pertinent depending variables at time t and t þ Dt, respectively. It provides a parabolic distribution of / between two consecutive time instants [34]. The boundary condi- Table 5 Microstructure parameters. Connectivity [–] Grain size [lm] Tortuosity [–] Specific surface area, Sa [lm2 lm3] Nickel YSZ Pore 0.82 0.94 7.16 7.05 0.98 0.58 4.45 11.64 0.99 1.09 2.11 8.63 Table 6 Coefficients in Eq. (28) for the discretized governing equations. Eq. / wt wx wy h (1) (2) T Yi qCP q qCPu qu qCPv qv k qDi tions of the second type are introduced by assuming a second order approximation of the dependent variables in the proximity of the respective boundaries. Regular spatial and temporal meshes are used in the computations. The linear system of algebraic equations resulting from discretization of (28) is solved by means of the Gauss-Seidel method [25]. 6. Results 6.1. Verification of the proposed models 6.1.1. Verification using benchmarking function As the proposed algorithm of the solution relies on consecutive solving the linearized PDEs in the form (28), we need to identify the accuracy of the computations provided by our solver for each of such steps. To this end we shall introduce an auxiliary problem with a known analytical solution which preserves all the essential features of the original formulation. First, remark that the governing Eqs. (1) and (2) in their linearized versions can be represented by the following general formula: @ ða1 /Þ þ $ ða2 u/Þ ¼ $ ða3 $/Þ þ S; @t ð30Þ where / ¼ /ðt; x; yÞ stands for the respective dependent variable (temperature or mass fraction of species), coefficients ai refer to the material properties, while S is a source term. In the original problem S is identically zero, however we assume here that it can take an arbitrary value. Now, let the solution to the PDE (30) be of the form: ^ yÞ; /ðt; x; yÞ ¼ ect /ðx; provided that: Fig. 4. (On the left) The procedure scheme for the FIB-SEM tomography, (on the right) the 3D reconstructed sample using AVIZO. ð31Þ 1038 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Sðt; x; yÞ ¼ ect ^Sðx; yÞ; ð32Þ where c is some parameter. In the framework of this self-similar formulation, the original initial boundary value problem (described by PDE (30) together with the pertinent initial and boundary conditions) can be reduced to a boundary value problem, defined by the following PDE: ca1 /^ þ $ a2 u/^ ¼ $ a3 $/^ þ ^S; ð33Þ and the respective boundary conditions. Now, having an analytical solution to the Eq. (33) we can immediately extend it to its timedependent counterpart by relations (31) and (32). In this way, a full time-dependent analytical benchmark solution is obtained. It can be used to verify the accuracy and efficiency of the computations. In the following analysis a dedicated strategy to built the benchmark solution will be adopted. Namely, we shall assume ^ B ðx; yÞ, which presome form of potential solution to the Eq. (33), / serves all the essential features and boundary conditions of the ^ B ðx; yÞ to (33) and simoriginal problem. Then, on substitution of / ple algebra, we will arrive at the expression for the source term, ^ S, for which the equation is identically satisfied. In other words, instead of looking for some example of solution for (33), we accept a predefined benchmark function and then we find the form of the governing PDE for which the latter is fulfilled. Let the benchmark solution be: ^ B ðx; yÞ ¼ b þ exp b ð1 x=LÞ2 þ b ð1 y=HÞ2 ; / 1 2 3 ð34Þ where L and H are the constants defining the spatial extent of the domain (in x and y directions, correspondingly), while bj ðj ¼ 1; 2; 3Þ are some parameters to be taken as convenient. Note that, expression (34) preserves the respective natural boundary conditions (see Table 2) at x ¼ L and y ¼ H, while the normal derivative (and thus the flux) at y ¼ 0 is different from zero. Moreover, the ^ B , can be controlled spatial distribution of the benchmark function, / by the proper handling of the coefficients bj and powers of the x and y related terms (provided however that the respective powers are greater than 2). The highly non-linear form of the function (34) makes it a demanding benchmark for the computational algorithm, ^ B does not directly comply with most of as the spatial behaviour of / the standard discretization schemes based on the polynomial representation. The corresponding source term is defined as: ^ ^ ^ þ $ a2 u/ $ a3 $/ ; Sðx; yÞ ¼ SB ðx; yÞ ¼ ca1 / B B B ð35Þ where the velocity function u is taken from Section 2.2 [10]. The respective parameters of the benchmark solution used in our analysis are collected in Table 7. As a comprehensive investigation of the solver performance goes beyond the scope of this paper, we will identify and signalize only the most important trends. Let us start by presenting the spatial distributions of errors obtained for two different densities of a regular mesh (we assume that the number of points, N, is the same in both directions, x and y). In Fig. 5 we show the relative error of ^ for t ¼ 50 s. Two different mesh densities solution, d/, (N ¼ f20; 70g) were used in the computations for two different Table 7 Benchmark parameters. c a1 a2 a3 b1 b2 b3 L H 0.1 150 150 2 1 0.1 0.1 1 1 magnitudes of the time step, Dt (Dt ¼ f0:01; 0:1g s). As can be seen in the figure, the error level is very low regardless of the computational parameters in use (it does not exceed the value of 6 105 in ^ is observed. The location of any case). No sharp magnification of d/ the maximal error remains the same (x ¼ L; y ¼ H) for all considered variants. Furthermore, the error distributions are very similar ^ are different. It shows in the to each other, even if the levels of d/ figure that, when comparing the results obtained for the coarse spatial mesh, no significant improvement of the solution accuracy is observed while reducing the time step. However, the situation is different for the finer mesh. Here, by taking Dt ¼ 0:01 s instead Dt ¼ 0:1 s one decreases the error level by one order of magnitude. Thus, it is very important for the accuracy and efficiency of the computations to keep the proper balance between the density of spatial and temporal meshing. In other words, the overall error of solution depends on the interplay between the component errors introduced by the approximations of the temporal and the spatial derivatives. If, for some value of N, the latter is an accuracy limiting factor, then no improvement of the solution can be achieved by taking smaller time steps (and conversely, if the time step is sufficiently large, no error reduction can be obtained by finer spatial meshing). This trend can be also observed in Fig. 6, where the temporal evolutions of maximal and mean relative ^ B are shown. Notably, although the overall error is errors of / always greater for the coarse spatial meshing, the stabilization (saturation level) is achieved here faster. The rate of error convergence for different values of N is depicted in Fig. 7, were the mean errors are shown at time t ¼ 50s. The errors can be approximated by the exponential-type expression to give: for Dt ¼ 0:1 s ^ ðmeanÞ ¼ 6:32 104 þ 0:0112 expð0:08532NÞ; d/ ð36Þ for Dt ¼ 0:01 s ^ ðmeanÞ ¼ 1:34 104 þ 0:0102 expð0:08384NÞ: d/ ð37Þ The above expressions provide estimations for the maximal achievable accuracy which yields: 6:32 104 % for Dt ¼ 0:1 s and 1:34 104 % for Dt ¼ 0:01 s. We would like to emphasize, that a solution of a very good quality (sufficient for any engineering application) can be obtained even for larger time steps. The presented analysis proves that our solver is an efficient and very accurate computational tool. Thus, the credibility of numerical results described in the following subsections can be fully substantiated. 6.1.2. Chemical reaction model verification The chemical reaction model (see Eq. (13)) was verified in this section. It was used to predict the outlet gas composition of the reformer (reactor) from Fig. 3. The methane conversion rate xcr defined in Eq. (38) was chosen as a testing parameter, as it requires calculating the outlet content of all spiecies [33]: xcr ¼ xCO þ xCO2 ; xCO þ xCO2 þ xCH4 ð38Þ where xj denotes molar fractions of species j ¼ hCO; CO2 ; CH4 i at the reformer outlet. The calculated conversion rates were compared to the corresponding experimental measurements from [33]. The results are presented in Fig. 8. The dashed line represents the perfect correlation between the measured and the calculated conversion rates. A good agreement observed in Fig. 8 implies, that the chemical reaction model was elaborated correctly and accurately predicts the composition at the outlet of the reformer. M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 1039 Fig. 5. Solver benchmark verification results tor time t = 50 s. The influence of mesh density and time step: (a) N = 20, Dt = 0.01 s; (b) N = 20, Dt = 0.1 s; (c) N = 70, Dt = 0.01 s; (d) N = 70, Dt = 0.1 s. 6.2. Numerical simulation results 6.3. Reforming reaction rate In this section, a parametric study is carried out to simulate the performance of Direct Internal Reforming Solid Oxide Fuel Cell. We employed the mathematical and numerical models developed in Sections 2 and 5 to analyze the performance of DIR-SOFC. Furthermore, we selected the most important parameters for DIR-SOFC control to be investigated, namely: the nickel-pore inter- The reforming reaction rate Rst is a parameter, which introduces the microstructural properties to the model and prominently affects the DIR-SOFC operation. Its complete form in Eq. (27) depends on a number of parameters and influences the distributions of all species and temperature, which makes it an informative parameter to trace. Fig. 9 depicts a series of Rst plots after stabilization (20 s), showing how the investigated parameters (nickel-to- face area density SNi-Pore , the collected current density iden , the inlet and methane inlet mass fraction Y inlet CH4 , the inlet temperature T inlet the volumetric flow of inlet gas mixture N_ . The values of these parameters as well as some general parameters, which describe the model, are summarized in Table 8. Calculations are carried out for the time-independent boundary conditions (see Table 8) over the time interval sufficient to reach the steady state. The system response was analyzed on the basis of the temperature and mass fraction distributions of the respective species. Also, the averaged-by-height values of temperature and mass fractions are collected for the conducted studies. Numerical simulations are made for the 8 cm per 0.1 cm rectangular fuel channel for which a 100 30 uniform mesh was applied. In order to preserve the qualitative character of the elaborated model, the influences of the aforementioned parameters are studied in the core of the fluid flow and reflect the behaviour of the gases and the temperature in the DIR-SOFCs longitudinal direction. pore surface density SNi-Pore , steam-to-carbon ratio SC and methane mass fraction at inlet Y inlet CH4 ) influenced the reforming reaction rate, directly related to mass and heat source terms. A dominant impact of SNi-Pore can be observed in Fig. 9. An almost proportional increase in the reaction rate towards increasing nickel exposure in Fig. 9a is in agreement with the Eqs. (13) and (27), describing the reforming reaction rate, that account for the anode microstructure. In Eq. (13), the reaction order towards steam is low, thus negligible changes on Rst in Fig. 9b can be observed, when subjected to different SC ratios. For a similar reason, a high order of the reaction towards methane makes the Rst sensitive to the methane content changes, what was shown in Fig. 9c. Apart from the sensitivity analysis conducted for the stabilized operation of DIR-SOFC, a transient behaviour of the reforming reaction rate is presented in Fig. 10. One can notice that after the initial state, in which the reaction is completely suppressed due to the absence of methane and 1040 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Fig. 6. Maximal and mean relative errors of solution for: (a) and (c) – Dt = 0.1 s, (b) and (d) – Dt = 0.01 s. Table 8 Parameters used in simulation. Fig. 7. Mesh sensitivity analysis, mean error dependence on mesh density. Parameter Unit Values Fuel channel length Fuel channel width Initial H2 content Initial CO content Initial CO2 content Initial N2 content Initial temperature Steam-to-carbon ratio SC m m – – – – K – – 0.08 0.001 0 0 0 1 1023/1073 4.0 Inlet carbon monoxide Y inlet CO Inlet steam Y inlet H2 O – SC Y inlet CH4 0 Inlet carbon dioxide Y inlet CO2 – 0 hydrogen Y inlet H2 nitrogen Y inlet N2 – 0 Inlet Inlet Ni-Pore Ni-Pore interface S Current density iden Inlet CH4 content Y inlet CH4 Inlet temperature T inlet Inlet gas flux N_ – inlet 1 Y inlet H2 O Y CH4 2 3 m m 1 106/2.5 106/5.0 106 A cm2 – 0.15/0.35/0.55 0.05/0.1/0.15 K 1023/1073/1123 L min1 0.1/0.5/1/1.5 6.4. Transient analysis of the mass fractions and temperature distributions Fig. 8. Correlation plots of the measured against the calculated fractional conversions of methane during the experiment. steam, Rst grows rapidly as the gases fill the anode channel. After stabilization it gets a descending profile towards the outlet. These figures give a closer look at the intensity of the reforming reaction rate, revealing its scale and providing information for the interpretation of the further outcomes. This section presents the computation results of the transient evolution of the height-averaged mass fractions and temperature profiles from the initial state, up to the steady state. These profiles are presented in Fig. 11, in which temperature, hydrogen, methane and steam are included. Within the stabilization time, rapid changes occur in the anode channel of DIR-SOFC, as the reforming reaction is progressing. The temperature equalization at the beginning of the simulation (Fig. 11a) finishes quickly after about 4s as the endothermic reforming reaction starts to dominate the heat balance. The temperature drops drastically causing a large temperature gradient along the cell after stabilization. In Fig. 11b, after the initial absence of hydrogen, it gradually fills the channel, however the pace of the increase is bigger when the temperature reaches its M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 1041 highest value. It is worth noting that the transient trend in hydrogen appearing in the channel, according to Eq. (23), is directly related to the transient trend of Rst and in fact follows it tightly. However, due to the fact that the reforming reaction is considered to be slow [1,20,21,27,33,35], it does not keep up with producing hydrogen what results in a severe deficiency and the cell undergoes a fuel starvation in this region. Fig. 11c and d show the methane and steam average distributions, respectively. One can notice, that due to the dominant convective term, methane and steam spread the fastest in the channel and affect the reforming reaction rate according to Eq. (27). It is worth noting, that the profiles on all figures reach the steady state at similar time, this is around 20 s. 6.5. The effect of the anode microstructure Fig. 9. Methane steam reforming reaction rate profiles as a result of applying various: (a) SNi-Pore , (b) SC, (c) Y inlet CH4 values. The nickel-pore interface area SNi-Pore is one of the parameters which describe the anode mictrostructure. The area of the nickel exposed to the gases penetrating the anode is a measure of its ability to reform methane into hydrogen. In this section, the sensitivity analysis of the aforementioned parameter was carried out to address the problem of the anode microstructure impact on the fuel starvation and anode cooling effect. By extracting only one parameter to be changed, a deeper analysis of the numerical solution and finding the key factors is possible. Moreover, this section presents the transient distributions of species and temperature inside the DIR-SOFC with the extracted crucial microstructural parameter, Nickel-to-Pore contact area density SNi-Pore , which determines the rate of the reforming reaction in the cell. Figs. 12 and 13 present the results of the numerical simulations for the influence of SNi-Pore on the hydrogen and methane distributions in the anode channel, respectively. The first phenomenon to be observed is a severe hydrogen starvation shown in Fig. 12a. Only a negligible amount of hydrogen is produced on the anode exhibiting 106 m2 m3 of the nickel-pore interface area. This figure shows a state, when hydrogen consumption outbalances its production causing a prominent deficiency of fuel in the fuel channel. An anode with over twice as much reforming reaction sites, Fig. 10. Methane steam reforming reaction rate profiles in a function of time. inlet _ Conditions: SNi-Pore = 2.5e6 m2 m3, SC = 4.0, Y inlet = 1073 K. CH4 = 0.1, N = 0.5 L/min, T 2:5 106 m2 m3 , presented in Fig. 12b produces enough hydrogen to prevent the starvation state in almost the entire channel. However, due to the forced convection and the fact that the reforming reaction is slow, even the anode with improved microstructural properties, shown in Fig. 12b as well as in Fig. 12c, is not able to Fig. 11. Height-averaged profiles of the respective parameters’ distributions throughout time: (a) hydrogen, (b) temperature, (c) methane, (d) steam; simulation conditions: inlet SNi-Pore = 5.0e6, N_ = 0.1 L/min, SC = 4.0, iden=0.15 A cm2, Y inlet = 1073 K. CH4 = 0.1, T 1042 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Fig. 12. Distribution of hydrogen for the different nickel exposure coefficients in the anode microstructure SNi-Pore : (a) SNi-Pore = 1e6 m2 m3, (b) SNi-Pore = 2.5e6 m2 m3, (c) inlet inlet = 1073 K. SNi-Pore = 5.0e6 m2 m3. Conditions: N_ = 0.5 L min1, SC = 4.0, iden = 0.15 A cm2, Y inlet CH4 = 0.1, Y H2 = 0, T Fig. 13. Distribution of methane for the different nickel exposure coefficients in theanode microstructure SNi-Pore : (a) SNi-Pore = 1e6 m2 m3, (b) SNi-Pore = 2.5e6 m2 m3, (c) inlet inlet SNi-Pore = 5.0e6 m2 m3. Conditions: N_ = 0.5 L min1, SC = 4.0, iden = 0.15 A cm2, Y inlet = 1073 K. CH4 = 0.1, Y H2 = 0, T completely eliminate the hydrogen insufficiency from the fuel channel. One can notice, that despite producing sufficient amounts of hydrogen at the outlet section, the inlet section still suffers from hydrogen starvation. The results presented in this figure suggest utilizing certain pre-reforming methods to provide any hydrogen to the upstream section of the cell. The hydrogen production process in the DIR-SOFC is always related to methane consumption in the reforming process. Fig. 13 presents the methane distribution in the anode channel for the various nickel exposures in the anode’s microstructure. It can be observed, that methane consumption rises for larger SNi-Pore values and the region of methane depletion responds to the region of the intensified hydrogen production. Despite higher methane consumption it is worth noticing in Fig. 13c, that only a small amount of methane is converted, about 13% at the outlet, while the rest can still be utilized. The methane steam reforming is a strongly endothermic process and the amount of heat transferred in this process depends mostly on the reforming reaction rate Rst . Increasing the rate of the reforming by increasing the SNi-Pore thus results in large heat consumption, what is reported in Fig. 14. One can observe a drastic change in temperature distribution with the SNi-Pore . The temperature for the SNi-Pore ¼ 1 106 m2 m3 in Fig. 14a rises for about 10 K while for the SNi-Pore ¼ 5 106 m2 m3 in Fig. 14c it drops for Fig. 14. Distribution of temperature for the different nickel exposure coefficients in the anode microstructure SNi-Pore : (a) SNi-Pore = 1e6 m2 m3, (b) SNi-Pore = 2.5e6 m2 m3, (c) inlet inlet SNi-Pore = 5.0e6 m2 m3. Conditions: N_ = 0.5 L min1, SC = 4.0, iden = 0.15 A cm2, Y inlet = 1073 K. CH4 = 0.1, Y H2 = 0, T M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 about 5 K causing an anode cooling effect. In order to sustain as high temperature as possible, other performance control mechanisms need to be used. However, appropriate microstructural parameters of the used anode, as presented in Fig. 14b, can provide a desired uniform temperature distribution and the reduction of thermal stresses in the cell. The effect of an enhanced nickel exposure in the anode microstructure on DIR-SOFC performance can be also seen in Fig. 15, where the temperature and species distributions are averaged over the cell height. This figure presents their trends in terms of changing microstructural parameters of the anode. The numerical simulations for various SNi-Pore values confirm the existence of fuel starvation regions in the fuel channel. The most vulnerable turns out to be the inlet section, where no hydrogen is supplied (see Section 2.4) and not enough methane is converted in the reforming process. To obtain a complete study of the behaviour of DIR-SOFC towards the microstructural changes, transient plots of the studied variables are required. An investigation of the distributions especially near the anode surface is essential and therefore in Figs. 16 and 17, the distributions of mass fraction and temperature as a function of time was presented in the vicinity of the anode surface. Fig. 16 presents the transient distributions of temperature, hydrogen and methane for two simulated SNi-Pore values: 1 106 m2 m3 and 2:5 106 m2 m3 . A rapid temperature growth during the first 4 s is in agreement with the results from Fig. 11a, however the lower SNi-Pore applied here give two different trends for further temperature profile evolution. In Fig. 16a there are fewer reaction sites, so that the elctrochemical reaction balances the reforming reaction and the cell temperature stabilizes quickly. It also exhibits a rising profile towards the channel outlet, which is consistent with a descending profile of the reforming reaction rate from Fig. 9a. In Fig. 16b, in turn, there is enough reforming reaction sites for the chemical reaction to outbalance the exothermic electrochemical reactions, and after 4s the temperature drops, resulting in a more uniform temperature distribution when a steady state is reached. Fig. 16c and d present the results of hydrogen transient distribution when subjected to the analogical conditions as for Fig. 16a 1043 and b. One can observe, that both distributions exhibit similar trends in time with a distinguishing feature of the enhanced production of hydrogen after the initiatory temperature equalization. What is more, an increased temperature does not directly influence the hydrogen production but the heat can only be consumed in the reforming process, which is more intense when the higher SNi-Pore are imposed. As a result, there is an almost two times lower hydrogen production in Fig. 16c than in Fig. 16d and it also can be observed in Fig. 12b. A close relation between methane distribution and the reforming reaction rate, can be observed in Fig. 16e and f. As a result of the reforming reaction rate towards the methane close to 1 (see Eq. (13)), the methane transient distribution follows the Rst profile. What is more, due to proportionality between the Rst and the SNi-Pore , the increasing nickel exposure causes adequate methane consumption increase. Due to the intense changes in the temperature transient distribution observed in Fig. 16a and b a closer analysis of the phenomena inside an anode channel was carried out and shown in Fig. 17. In order to eliminate the impact of the temperature difference between the initial and inlet conditions, a similar simulation was made for identical temperatures of 1073 K and the initial time and inlet. A fine contribution of nickel exposure SNi-Pore on the anode cooling effect can be observed. An analysis of this Figure indicates the existence of the optimal microstructure, for which a uniform distribution in a steady state can be achieved and the initiatory temperature rise can be suppressed. 6.6. The effect of current density The reforming process in the DIR-SOFC aims to provide fuel for the cell, it is then responsible for hydrogen and carbon monoxide production. These species are then consumed in order to satisfy the external load applied to the cell, which in this model is represented by current density iden . In the developed model, the current density is assumed constant along the cell, which reflects the perfect conditions for the charge transport. Fig. 18 presents the results of the numerical calculations for the different current density values, ranging from 0.0 A cm2, which simulates the situation with Fig. 15. Height-averaged distributions of species and temperature for the different nickel exposure coefficient in anode microstructure SNi-Pore : (a) hydrogen, (b) temperature, inlet inlet (c) methane, (d) carbon monoxide, (e) carbon dioxide. Conditions: N_ = 0.5 L min1, SC = 4.0, iden = 0.15 A cm2, Y inlet = 1073 K. CH4 = 0.1, Y H2 = 0, T 1044 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Fig. 16. Transient distributions of temperature (first row), hydrogen (second row) and methane (third row) in the vicinity of the anode for different SNi-Pore : (a), (c), (e) inlet inlet SNi-Pore = 1e6 m2 m3, (b), (d), (f) SNi-Pore = 2.5e6 m2 m3. Conditions: N_ = 0.1 L min1, SC = 4.0, iden = 0.15 A cm2, Y inlet = 1073 K, T initial = 1023 K. CH4 = 0.2, Y H2 = 0, T no load, up to 0.55 A cm2. It shows a prominent impact of iden on the temperature distribution. One can observe an intensive cooling in Fig. 18a, which suggests a considerable mismatch between the heating and cooling processes in the cell, as a result of which a large temperature drop is reported. Nevertheless, increasing iden significantly intensifies Joule’s heat generation according to Eq. (21) as well as the heat released during water formation, according to Eq. (20). As a result, in Fig. 18 as well as in Fig. 19a, a considerable temperature rise is observed near the cell’s outlet, especially at the beginning of the simulation time. The heat generation increase following the enhanced iden increases the temperature in the vicinity of the anode so much that the hydrogen production due to the reforming process almost outbalances its enlarged consumption, what can be observed in Fig. 19b, in which the amount of hydrogen in the fuel channel decreases just slightly, for ca. 30%. None of the studied cases in this section, however, eliminates the starvation state at the inlet. Fig. 19b also implies, that supplying no hydrogen to the cell by the inlet may cause its starvation as the reforming process does not keep up with producing it. The result from Fig. 18a in particular shows a crucial insight into the transient behaviour of DIR-SOFC and reveals a large temperature gradient along the cell after 2s, which diminishes in time. M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 1045 Fig. 17. Transient distributions of gas temperature (identical initial and inlet boundary conditions), in the vicinity of the anode for different SNi-Pore : (a) SNi-Pore = 1e6 m2 m3, inlet inlet (b) SNi-Pore = 2.5e6 m2 m3, (c) SNi-Pore = 5.0e6 m2 m3. Conditions: N_ = 0.1 L min1, SC = 4.0, iden = 0.15 A cm2, Y inlet = 1073 K, T initial = 1073 K. CH4 = 0.1, Y H2 = 0, T Fig. 18. Distribution of temperature for the different collected current densities iden : (a) iden = 0.0 A cm2, (b) iden = 0.35 A cm2, (c) iden = 0.55 A cm2. Conditions: inlet inlet N_ = 0.5 L min1, SC = 4.0, SNi-Pore = 1e6 m2 m3, Y inlet = 1073 K. CH4 = 0.1, Y H2 = 0, T 6.7. The effect of methane inlet content Methane is directly converted into hydrogen and carbon monoxide in the reforming process and additionally is a major component of the reforming reaction rate, given by Eq. (27). Adjusting inlet composition proved to be an effective way to control SOFC systems performance. The numerical results for the different methane inlet content Y inlet CH4 are presented in Figs. 20 and 21. In our model methane is not electrochemically oxidized, but chemically converted to hydrogen in a strongly endothermic reaction. The changes in methane concentration directly influence the reforming reaction rate which grows together with methane content Y CH4 at the inlet. This explains the significant temperature drop accompanied by a boosted hydrogen production when more methane is supplied, which can be observed in Figs. 20 and 21. Moreover, a strong anode cooling effect is observed in Fig. 20c, 1046 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 methane inlet content Y inlet CH4 increase can be observed in Fig. 22, where height-averaged plots of temperature, hydrogen, as well as carbon monoxide and carbon dioxide are shown. An average temperature tends to drop when more methane is supplied, which is a conformation of the results from previous figures. The amount of all gases in the figure as a result of the reforming process tends to rise with Y inlet CH4 increase. One can see in Fig. 21c, that despite the hydrogen insufficiency at the inlet section, an appreciable amount of H2 is present at the outlet and which can be further utilized. This suggests recuperating the outlet hydrogen and redirecting it back to the inlet to cover the areas suffering from its absence. 6.8. The effect of inlet temperature Fig. 19. Height-averaged distributions of species and temperature for the different collected current densities iden : (a) temperature, (b) hydrogen. Conditions: inlet inlet N_ = 0.5 L min1, SC = 4.0, SNi-Pore = 1e6 m2 m3, Y inlet = 1073 K. CH4 = 0.1, Y H2 = 0, T while a contrary tendency to heat up the cell is observed when less methane is supplied (Fig. 20a) as a result of a domination of electrochemical reactions in the cell. Similarly to Figs. 14b and 18b, a very uniform distribution of temperature can be observed in Fig. 20b, which suggests the optimal conditions for the operation of the DIR-SOFC cell. In Fig. 21 it can be seen, that as a result of the Rst growth, the hydrogen starvation region is diminished when more methane is supplied to the cell. It is not, however, eliminated and the hydrogen starvation region still occupies some part of the channel volume. The overall trends as a response to various Figs. 23 and 24 show the results of the numerical computations for hydrogen and methane for the different inlet temperatures T inlet , which varied from 1023 K to 1073 K. In Fig. 23 an increase in hydrogen production can be observed while the inlet temperature grows. In fact, this is the result of the boosted methane conversion due to the higher temperature in the reaction site according to Eq. (27), which can be observed in Fig. 24. It is worth to notice, that for all cases, fuel starvation was not eliminated from the fuel channel and the deficiency of hydrogen can still be observed. However, hydrogen starvation is reduced prominently when a gas mixture of a higher temperature is provided to the system. Moreover, increasing the inlet temperature by 50 K can intensify hydrogen production two times and rise methane conversion by 5%, while an 100 K increase gives three times higher H2 production and a 13% higher CH4 conversion rate (see Fig. 25). inlet inlet inlet 1 _ Fig. 20. Distribution of temperature for the different methane content at inlet Y inlet , SC = 4.0, CH4 : (a) Y CH4 = 0.05, (b) Y CH4 = 0.1, (c) Y CH4 = 0.15. Conditions: N = 0.5 L min inlet SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet = 1073 K. H2 = 0, T inlet inlet inlet 1 _ Fig. 21. Distribution of hydrogen for the different methane content at inlet Y inlet , SC = 4.0, CH4 : (a) Y CH4 = 0.05, (b) Y CH4 = 0.1, (c) Y CH4 = 0.15. Conditions: N = 0.5 L min inlet SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet = 1073 K. H2 = 0, T M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 1047 Fig. 22. Height-averaged distributions of species and temperature for the different methane content at inlet Y inlet CH4 : (a) temperature, (b) hydrogen, (c) carbin monoxide, (d) inlet carbon dioxide. Conditions: N_ = 0.5 L min1, SC = 4.0, SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet = 1073 K. H2 = 0, T Fig. 23. Distribution of hydrogen for the different inlet temperature T inlet : (a) T inlet = 1023 K, (b) T inlet = 1073 K, (c) T inlet = 1123 K. Conditions: N_ = 0.5 L min1, SC = 4.0, inlet SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet H2 = 0, Y CH4 = 0.1. Fig. 24. Distribution of methane for the different inlet temperature T inlet : (a) T inlet = 1023 K, (b) T inlet = 1073 K, (c) T inlet = 1123 K. Conditions: N_ = 0.5 L min1, SC = 4.0, inlet SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet H2 = 0, Y CH4 = 0.1. Inlet temperature T inlet proves to be an effective means to control the DIR-SOFC work. Its increase has all of the positive results on the distribution of species, increases the hydrogen content, reducing hydrogen starvation at the inlet and increasing methane conversion. Higher temperature thus enables larger loads and when using hydrogen recuperation, it can completely eliminate its insufficiency from the cell. 6.9. The effect of gas inlet velocity Changing fluid flow velocity entering the cell is the basic control strategy for chemical reactors, including SOFC [12,36,37]. Not only can it influence the fluid velocity in the cell but also the temperature and the rate at which the reaction takes place inside the reactor. In this study DIR-SOFC system response was simulated for 1048 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Fig. 25. Height-averaged distributions of species and temperature for the different inlet temperature T inlet : (a) hydrogen, (b) methane, (c) steam. Conditions: N_ = 0.5 L min1, inlet SC = 4.0, SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet H2 = 0, Y CH4 = 0.1. three inlet volumetric flows: 0.5 L min1, 1 L min1 and 1.5 L min1. A time-independent flow was assumed, according to the profile from Section 6.9. The results of the conducted simulations for hydrogen and temperature were presented in Figs. 26 and 27, respectively. A remarkable growth of the hydrogen starvation region is noticed when the volumetric flow rises to 1.5 L min1, as it occupied more than 60% of the channel. Notably, for the smallest flow the fuel starvation region occupied about 15% of the channel and a significant amount of hydrogen could be produced, as it was not flushed (see Figs. 26a and 28a). From the viewpoint of fluid flow influence, similar trend was reported for the temperature distribution, presented in the twodimensional distribution in Figs. 27 and 28b as the heightaveraged plot. From the previous simulation results one can notice that a cooling effect is dominant for these conditions (see Figs. 14c or 15b). Hence, as shown in Fig. 27a, the slower the gas mixture goes through the channel, the more time the reforming process can take place and the stronger the cooling effect, as the gases reach the inlet temperature instantly (see Section 2.4). However, the larger the volumetric flow N_ inlet , the higher the temperature Fig. 26. Distribution of hydrogen for the different inlet volumetric fluid flow N_ inlet : (a) N_ inlet = 0.5 L min1, (b) N_ inlet = 1.0 L min1, (c) N_ inlet = 1.5 L min1. Conditions: SC = 4.0, inlet inlet SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet = 1073 K. H2 = 0, Y CH4 = 0.1, T Fig. 27. Distribution of temperature for different inlet volumetric fluid flow N_ inlet : (a) N_ inlet = 0.5 L min1, (b) N_ inlet = 1.0 L min1, (c) N_ inlet = 1.5 L min1. Conditions: SC = 4.0, inlet inlet SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet = 1073 K. H2 = 0, Y CH4 = 0.1, T 1049 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Fig. 28. Height-averaged distributions of species and temperature for the different inlet volumetric fluid flow N_ inlet : (a) hydrogen, (b) temperature, (c) methane. Conditions: inlet inlet SC = 4.0, SNi-Pore = 1e6 m2 m3, iden = 0.15 A cm2, Y inlet = 1073 K. H2 = 0, Y CH4 = 0.1, T at the outlet and the more uniform distribution is obtained, what inclines to a careful gas flow adjustment. Foundation for Polish Science, co-funded by the European Union under the European Regional Development Fund. 7. General conclusions Appendix A. Thermodynamic and transport properties of gas mixtures The dynamic behaviour of Direct Internal Reforming Solid Oxide Fuel Cell anode channel was analyzed using the microstructureoriented two-dimensional numerical model. The accuracy of the numerical solver was thoroughly verified using the benchmarking function and a conformal agreement at a level of 0.005% of relative error was reported. As the aim of this work was to reflect the impact of the Ni/YSZ anode microstructure, the reforming reaction rate was modified to include the actual accessibility of the material to reform methane to hydrogen. The FIB-SEM computer tomography was used to evaluate the essential quantitative parameters of the anode’s microstructure. A parametric study of a complete DIR-SOFC anode channel model was carried out to identify the system transient response to various boundary and operating conditions. The conducted parametric studies lead to the conclusion that the convective nature of the flow in the channel, as well as a non-equilibrium methane reforming, which occurs at a limited rate, are the key factors contributing to the existence of the hydrogen starvation in the vicinity of the channel inlet. Modifying specific process parameters, such as reducing the volumetric fluid flow or increasing the inlet gas temperature, matched to the anode’s methane reforming effectiveness, contributes to the diminishing of the hydrogen starvation state significantly. As it was indicated in the paper, the elaborated Nickel-to-Pore area has a dominant effect of the DIR-SOFC dynamic behaviour. Therefore, designing anode microstructure to balance the contribution of chemical and electrochemical reactions seems a promising strategy for future. Yet, in order to eliminate the hydrogen starvation completely, a pre-reforming process or recuperating of the waste hydrogen is required. Density The gas mixture density, applied in transport equations (Eq. (1), (2)), is calculated with the ideal gas equation of state: qmix ¼ ð39Þ where Mmix stands for the molar mass of mixture and is a weighed sum of molar mass species: Mmix ¼ X xi M i : ð40Þ i Specific heat The specific heat of gas mixture is estimated with empirical relationship by Todd and Young [38], utilizing the least squares method. Experimentally obtained, molar specific heat is temperature and composition dependent over the temperature interval from 273 K to 1473 K. Its value for species i is calculated as follows: C P;i ¼ 6 X ak rk ; where r ¼ T=1000 k ¼ 0 . . . 6: ð41Þ k¼0 The values of coefficients a0 ; . . . ; ak for all species are presented in Table 9. Thermal conductivity In order to calculate the thermal conductivity, the Wassiljewa equation was used [24]: 1 0 kmix Conflict of interest PMmix ; RT C n B X B xi k i C C; B ¼ n C BX A i¼1 @ xj Aij ð42Þ j¼1 The authors declare no conflict of interest. Acknowledgements The presented research is a part of ‘Easy-to-assemble Stack Type (EAST): Development of solid oxide fuel cell stack for innovation in polish energy sector’ project, carried out within the FIRST TEAM programme (project number First TEAM/2016-1/3) of the with the Mason and Saxena modification employed to compute Aij [24]: Aij ¼ 12 14 2 Mi 1 þ lli M j j 12 i 8 1þM Mj ; for i – j: ð43Þ 1050 M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 Table 9 Specific heat coefficients from Eq. (41) [38]. Species, i a0 a1 a2 a3 a4 a5 a6 CH4 H2 O CO CO2 H2 N2 47.964 37.373 30.429 4.3669 21.157 29.027 178.59 41.205 8.1781 204.6 56.036 4.8987 712.55 146.01 5.206 471.33 150.55 38.040 1068.7 217.08 41.974 657.88 199.29 105.17 856.93 118.54 66.346 519.9 136.15 113.56 358.75 79.409 37.756 214.58 46.903 55.554 61.321 14.015 7.6539 35.992 6.4725 10.35 Table 10 Parameters of gas mixture components [24]. Species Molar mass [g mol1] Van der Vaals radius CH4 H2 O H2 CO CO2 N2 16.04 18.015 2.016 28.01 44.01 28.155 3.758 2.641 2.827 3.69 3.941 3.798 The dynamic viscosity of species follows: r [Å] li in Eq. (43), is calculated as 1 li ¼ 16:64 M 2i T : ð=kB Þr2 ð44Þ The values of parameters in Table 10. r; and kB used in Eq. (44) are presented Diffusion coefficient The diffusion coefficient was calculated using Fuller’s method [24], beginning with the binary diffusion coefficient estimated with Eq. (45): 0:00143T 1:75 Dij ¼ pffiffiffiffiffiffiffih i2 ; P M ij ðRÞi1=3 þ ðRÞj1=3 ð45Þ where Ri denotes the atomic diffusion coefficient of species i and M ij are the harmonic means, which were obtained with Eq. (46): M ij ¼ 2 1 Mi þ 1 Mj 1 : ð46Þ Next, according to Blanc’s law, the diffusion coefficient of species (denoted as i) in the gas mixture was calculated [24]: Dmix;i ¼ n X Xi D ij i¼1;i–j !1 : ð47Þ Anode conductivity The electrical conductivity of the anode depends mainly on the electrical conductivity of nickel. In this paper the temperature dependence was applied, according to the following formula [39]: rel ¼ 4:5 105 1200 : exp T T ð48Þ References [1] G. Brus, Experimental and numerical studies on chemically reacting gas flow in the porous structure of a solid oxide fuel cells internal fuel reformer, Int. J. Hydrogen Energy 37 (22) (2012) 17225–17234. [2] D.P. Xenos, P. Hofmann, K.D. Panopoulos, E. Kakaras, Detailed transient thermal simulation of a planar SOFC (solid oxide fuel cell) using gPROMS, Energy 81 (2015) 84–102. [3] S.C. Singhal, K. Kendall, High Temperature Solid Oxide Fuel Cells: Fundamentals, Design, and Applications, Elsevier, 2015. [4] M.A. El-bousiffi, D.J. Gunn, A dynamic study of steam-methane reforming, Int. J. Heat Mass Transf. 50 (2007) 723–733. Lennard-Jones potential =kB [K] Atomic diffusion coefficient R [Å] 148.6 809.1 59.7 91.7 195.2 71.4 25.14 14.141 4.62 22.01 37.91 9.08 [5] H. Iwai, Y. Yamamoto, M. Saito, H. Yoshida, Numerical simulation of intermediate-temperature direct-internal-reforming planar solid oxide fuel cell, Energy 36 (4) (2011) 2225–2234. [6] S. Lee, H. Kim, K. Joong, J.-w. Son, J.-h. Lee, B.-k. Kim, W. Choi, J. Hong, The effect of fuel utilization on heat and mass transfer within solid oxide fuel cells examined by three-dimensional numerical simulations, Int. J. Heat Mass Transf. 97 (2016) 77–93. [7] S. Sohn, J.H. Nam, D.H. Jeon, C.-j. Kim, A micro/macroscale model for intermediate temperature solid oxide fuel cells with prescribed fullydeveloped axial velocity profiles in gas channels, Int. J. Hydrogen Energy 35 (21) (2010) 11890–11907. [8] Z. Zhang, D. Yue, G. Yang, J. Chen, Y. Zheng, H. Miao, W. Wang, J. Yuan, N. Huang, Three-dimensional CFD modeling of transport phenomena in multichannel anode-supported planar SOFCs, Int. J. Heat Mass Transf. 84 (2015) 942–954. [9] T.X. Ho, Dynamic characteristics of a solid oxide fuel cell with direct internal reforming of methane, Energy Convers. Manage. 113 (2016) 44–51. [10] F. Arpino, N. Massarotti, Numerical simulation of mass and energy transport phenomena in solid oxide fuel cells, Energy 34 (12) (2009) 2033–2041. [11] P. Aguiar, C.S. Adjiman, N.P. Brandon, Anode-supported intermediatetemperature direct internal reforming solid oxide fuel cell: II. Model-based dynamic performance and control, J. Power Sources 147 (1-2) (2005) 136–147. [12] Y. Komatsu, S. Kimijima, J.S. Szmyd, Numerical analysis on dynamic behavior of solid oxide fuel cell with power output control scheme, J. Power Sources 223 (2013) 232–245. [13] a. Chaisantikulwat, C. Diaz-Goano, E. Meadows, Dynamic modelling and control of planar anode-supported solid oxide fuel cell, Comput. Chem. Eng. 32 (10) (2008) 2365–2381. [14] J. Kupecki, K. Motylinski, J. Milewski, Dynamic modelling of the direct internal reforming (DIR) of methane in 60-cell stack with electrolyte supported cells, Energy Procedia 105 (2017) 1700–1705. [15] Y. Huangfu, F. Gao, A. Abbas-Turki, D. Bouquain, A. Miraoui, Transient dynamic and modeling parameter sensitivity analysis of 1D solid oxide fuel cell model, Energy Convers. Manage. 71 (2013) 172–185. [16] Y. Kang, J. Li, G. Cao, H. Tu, J. Li, J. Yang, One-dimensional dynamic modeling and simulation of a planar direct internal reforming solid oxide fuel cell, Chin. J. Chem. Eng. 17 (2) (2009) 304–317. [17] Y. Bae, S. Lee, K. Joong, J.-h. Lee, J. Hong, Three-dimensional dynamic modeling and transport analysis of solid oxide fuel cells under electrical load change, Energy Convers. Manage. 165 (December 2017) (2018) 405–418. [18] M. Kishimoto, N. Furukawa, T. Kume, H. Iwai, H. Yoshida, Formulation of ammonia decomposition rate in Ni-YSZ anode of solid oxide fuel cells, Int. J. Hydrogen Energy 42 (4) (2017) 2370–2380. [19] G. Brus, K. Miyoshi, H. Iwai, M. Saito, H. Yoshida, Change of an anode’s microstructure morphology during the fuel starvation of an anode-supported solid oxide fuel cell, Int. J. Hydrogen Energy 40 (21) (2015) 6927–6934. [20] G. Brus, J.S. Szmyd, Numerical modelling of radiative heat transfer in an internal indirect reforming-type SOFC, J. Power Sources 181 (1) (2008) 8–16. [21] M. Mozdzierz, G. Brus, A. Sciazko, Y. Komatsu, S. Kimijima, J.S. Szmyd, Towards a thermal optimization of a methane/steam reforming reactor, Turbul. Combust. 97 (1) (2016) 171–189. [22] G. Brus, H. Iwai, M. Mozdzierz, Y. Komatsu, M. Saito, H. Yoshida, J.S. Szmyd, Combining structural, electrochemical, and numerical studies to investigate the relation between microstructure and the stack performance, J. Appl. Electrochem. 47 (9) (2017) 979–989. [23] Suhas V. Patankar, Numerical Heat Transfer and Fluid Flow, Taylor and Francis, 1980. [24] B. Poling, J. Prausnitz, J. O’Connell, The Properties of Gases and Liquids, The McGraw-Hill Companies, 1997. M. Chalusiak et al. / International Journal of Heat and Mass Transfer 131 (2019) 1032–1051 [25] H. Versteeg, An Introduction to Computational Fluid Dynamics, Pearson Education Limited, 2004. [26] G. Brus, H. Iwai, A. Sciazko, M. Saito, H. Yoshida, J.S. Szmyd, Local evolution of anode microstructure morphology in a solid oxide fuel cell after long-term stack operation, J. Power Sources 288 (2015) 199–205. [27] M. Tomiczek, R. Kaczmarczyk, M. Mozdzierz, G. Brus, A numerical analysis of heat and mass transfer during the steam reforming process of ethane, Heat and Mass Transf. 54 (8) (2018) 2305–2314. [28] S. Hosseini, K. Ahmed, M.O. Tadé, CFD model of a methane fuelled single cell SOFC stack for analysing the combined effects of macro/micro structural parameters, J. Power Sources 234 (2013) 180–196. [29] G. Brus, An analysis of transport phenomena in an internal indirect reforming type solid oxide fuel cell, Energy Eng. (2011) 1–162. [30] Y.W. Kang, J. Li, G.Y. Cao, H.Y. Tu, J. Li, J. Yang, A reduced 1D dynamic model of a planar direct internal reforming solid oxide fuel cell for system research, J. Power Sources 188 (1) (2009) 170–176. [31] A. Sciazko, Y. Komatsu, G. Brus, S. Kimijima, J.S. Szmyd, A novel approach to the experimental study on methane/steam reforming kinetics using the Orthogonal Least Squares method, J. Power Sources 262 (2014) 245–254. [32] A. Sciazko, Y. Komatsu, G. Brus, S. Kimijima, J.S. Szmyd, An application of generalized least squares method to an analysis of methane/steam reforming process on a Ni/YSZ catalyst, ECS Trans. (2013) 2987–2996. 1051 [33] G. Brus, Y. Komatsu, S. Kimijima, J.S. Szmyd, An analysis of biogas reforming process on Ni/YSZ and Ni/SDC catalysts, Int. J. Thermodyn. 15 (1) (2012) 43– 51. [34] M. Wrobel, G. Mishuris, Efficient pseudo-spectral solvers for the pkn model of hydrofracturing, in: D. Bigoni, A. Carini, M. Gei, A. Salvadori (Eds.), Fracture Phenomena in Nature and Technology, Springer International Publishing, 2014, pp. 151–170 (Cham). [35] M. Mozdzierz, M. Chalusiak, S. Kimijima, J.S. Szmyd, G. Brus, An afterburnerpowered methane/steam reformer for a solid oxide fuel cells application, Heat and Mass Transf. (2018) 2331–2341. [36] Y. Komatsu, G. Brus, S. Kimijima, J.S. Szmyd, The effect of overpotentials on the transient response of the 300W SOFC cell stack voltage, Appl. Energy 115 (2014) 352–359. [37] Y. Komatsu, G. Brus, S. Kimijima, J. Szmyd, Experimental study on the 300w class planar type solid oxide fuel cell stack: investigation for appropriate fuel provision control and the transient capability of the cell performance, J. Phys: Conf. Ser. 395 (2012) 352–359. [38] B. Todd, J. Young, Thermodynamic and transport properties of gases for use in solid oxide fuel cell modelling, J. Power Sources 110 (1) (2002) 186–200. [39] U. Anselmi-Tamburini, G. Chiodelli, M. Arimondi, F. Maglia, G. Spinolo, Z. Munir, Electrical properties of Ni/YSZ cermets obtained through combustion synthesis, Solid State Ionics 110 (1) (1998) 35–43.