Energy 141 (2017) 673e691

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Numerical investigations to determine the optimal operating

conditions for 1 kW-class flat-tubular solid oxide fuel cell stack
Kashif Rashid a, b, Sang Keun Dong b, a, *, Muhammad Taqi Mehran c
a
Department of Energy System Engineering, Korea University of Science and Technology (UST), 217, Gajeong-ro, Yuseong-gu, Daejeon, 34113, Republic of
Korea
b
Thermal Energy System Laboratory, Korea Institute of Energy Research (KIER), 152, Gajeong-ro, Yuseong-gu, Daejeon, 34129, Republic of Korea
c
Department of Advanced Energy and Technology, Korea University of Science and Technology (UST), 217, Gajeong-ro, Yuseong-gu, Daejeon, 34113,
Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: In this study, a detailed three-dimensional numerical model is developed which simultaneously as-
Received 20 April 2017 similates the transport processes, the electrochemical and chemical reactions to optimize the perfor-
Received in revised form mance of 1 kW-class flat-tubular solid oxide fuel cell stack while operating on H2 and pre-reformed
20 July 2017
methane fuels. The unique feature of this CFD (computational fluid dynamic) model is that it encom-
Accepted 18 September 2017
Available online 19 September 2017
passes the electrochemical oxidation of H2 and CO as well as internal steam reforming reactions
including radiation heat transfer analysis in the full stack. A CFD model validated with the experiments
performed in-house is utilized to explore the optimal operating conditions by investigating the effect of
Keywords:
Hydrogen and carbon monoxide parallel
pre-reforming rate, air/fuel inlet temperatures, oxidant utilizations and radiative heat transfer effect on
electrochemical reactions the temperature distributions. The numerical results elucidated that temperature and the current density
Internal steam reforming distributions can be regulated by adjusting the methane conversion in the pre-reformer. It is also
Methane conversion observed that neglecting the CO electro-oxidation in the modeling can underestimate the stack per-
Thermal management formance; whereas increasing the inlet temperature increases the stack performance. The oxidant uti-
Radiative heat transfer lization analysis established that higher oxidant utilization adversely affects the stack performance due
to higher cathodic activation polarization losses. Radiation heat transfer analysis demonstrates that it
curtails peak temperature and minimizes temperature gradients of the stack components.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction conventional power generation to SOFC-based electricity genera-

The global warming due to the greenhouse gas emissions is
becoming an alarming issue, and many efforts are being made to
explore an environmentally friendly systems to replace the con-
ventional and less efficient combustion-based power generation
technologies [1e4]. Solid oxide fuel cell (SOFC) has shown great
promise to serve an alternative electricity generation technology.
SOFC offers many attractive features such as high energy conver-
sion efficiency along with the low emission of greenhouse gasses,
and fewer mechanical parts than the conventional systems [5,6].
The ability of SOFC to reform hydrocarbon internally, disseminate
them as a prime candidate which will lead the transformation from
* Corresponding author. Thermal Energy System Laboratory, Korea Institute of
Energy Research (KIER), 152, Gajeong-ro, Yuseong-gu, Daejeon, 34129, Republic of
Korea.
E-mail address: skdong@kier.re.kr (S.K. Dong).
tion, as they can be operated on different types of hydrocarbon
fuels such as methane, syngas, biogas, coal, gasoline, etc. [7e9].
As SOFC's are manufactured from ceramic materials, therefore,
the high temperature is required to attain the better ionic con-
ductivity and this high operating temperature of SOFC offers
additional potential for co-generation applications [10,11]. The
natural gas (NG) is being extensively studied and utilized as a most
attractive fuel for medium sized SOFC-based stationary applica-
tions. The methane is a major constituent of NG, which should be
reformed either internally or externally because electrochemical
oxidation of methane in SOFC is not favorable [12]. The strong
endothermic internal steam reforming reaction (MSR) utilizes the
waste heat of irreversible polarizations and joule heat [13,14]. The
internal reforming in SOFC's suppresses the need to cool the stack
through a vast amount of excess air supplied across the cells. This
attractive feature significantly reduces the cost of operation and
hence improves the system efficiency [15]. However, Yoshida et al.

https://doi.org/10.1016/j.energy.2017.09.082
0360-5442/© 2017 Elsevier Ltd. All rights reserved.
674 K. Rashid et al. / Energy 141 (2017) 673e691

[16], Stimming et al. [17] and Klein et al. [18] stated that intense
endothermic reforming reaction creates a problem in SOFC stable
operation because of thermal management which is severely
affected by the temperature gradients. It is, therefore, essential to
have an appropriate understanding and perhaps the control over
the reforming reaction in the SOFC's to operate the stack at opti-
mum efficiency with internal steam reforming [19e21].
When H2 and CO present in the fuel stream, the total current
produced is the summation of both the reactant species at the
anode/electrolyte interface due to the electrochemical reactions
(ECR). CO electro-oxidation directly influences the performance of
stack, whereas, overpotentials losses affect the output voltage
directly [22]. In SOFC modeling studies, it is pervasive practice to
neglect the ECR of CO or in some cases, polarizations losses are not
fully implemented [23e28]. Li et al. [29], Haberman et al. [30] and
Hofman et al. [31] have made assumptions in their work that when
H2 and CO both are present in the fuel stream, CO only participate in
the water gas shift reaction (WGSR) instead of ECR because of very
fast kinetics of WGSR. These simplified assumptions for electro-
chemical modeling are debatable from the accuracy point of view.
However, some authors have considered ECR of CO as well in their
research work. The summary of some of the research work related
to ECR of H2 and CO is presented in Table 1.
During the literature review, it was established that all the
developed models have incorporated temperature dependent
transport properties, chemical kinetics and material properties of
cell and stack. In order to predict the fuel cell performance
correctly, it is, therefore, indispensable to develop a model that can
accommodate various modes of heat transfer in solid as well as in
flow channel components. Owing to the high operating tempera-
ture of SOFC, radiative heat transfer (RHT) becomes an important
mode of heat transfer that cannot be neglected (especially for stack
level study) because temperature gradients are crucial in deter-
mining the thermal stresses and to envisage the cell failure. How-
ever, in most of the existing models, RHT is either neglected
because of isothermal assumption [39e43] or simplified form of
the heat transfer model is used to estimate radiative heat transfer
[44e47]. There are numbers of model exist in the literature that
includes RHT, however, they do not assimilate the effects of CO2 and
water vapors in the anode electrode despite their high concentra-
tion. It is considered that these participating gases absorb, emit and
scatter radiation of certain wavelength [48].
All of the studies discussed above are limited either to a button
cell or a single planar/tubular SOFC cell. The stack level study, for
other design of SOFCs such as FT-SOFC, which models/implements
ECR's of both H2 and CO in detailed along with chemical reactions
MSR/WGSR, is either missing or very limited. And secondly, RHT
modeling is either neglected completely or selected features are
incorporated to avoid complexity and to reduce computational
time. Bessette et al. numerically exhibited that single cell have
different attributes than those of the stack, and it is an inaccurate
methodology to predict the output characteristics of SOFC stack
based on the unit cell results [49]. In the present work, a detailed
three-dimensional (3D) numerical analysis is performed on a 1 kW-
class anode-supported FT-SOFC 80 cells stack fueled by humidified
hydrogen and pre-reformed methane. The current CFD model
exclusively accounts for intricate transport phenomena (mass,
momentum, and heat transfer), electrochemical reactions
(including both H2 and CO) with overpotential losses such as acti-
vation, ohmic, and concentration, as well as MSR/WGSR on the real-
scale stack. The experiments are performed with both the fuels to
validate the modeling. The validated model is employed to examine
the effect of pre-reforming rate, neglecting the electro-oxidation of
CO, different inlet temperatures, and oxidant utilization on the

Table 1
Selected studies related to ECR of H2 and CO.

Authors Approach Limitation Model assumption

Petruzzi et al. [32]
Suwanwarangkul et al. [33]
Iwai et al. [34]
Ni et al. [35]
Andersson et al. [36]
Ni et al. [37]
B. Lin et al. [38]
- Two methodologies adopted to handle the CO
presence in the fuel stream
- WGSR was considered in the first strategy
- Whereas, in the second approach both ECR and
WGSR modeling of CO were implemented
- ECR's of both H2 and CO were considered
- Developed model further applied to predict
carbon formation
- MSR and WGSR modeling incorporated along
with ECR of H2 and CO
- Both ECRs are treated parallel reactions
- Activation, ohmic and concentration losses
included
- Pre-reforming rate was used to control local
temperature and current density
- ECR's of both H2 and CO were considered
- MSR and WGSR modeling also included
- Different reforming models included for
comparisons
- An alternative method proposed for OCV
calculation
- Finite element approach
- ECR's of both H2 and CO were considered
- Both ECRs are treated parallel reactions
- MSR and WGSR modeling also included
- Activation, ohmic and concentration losses
included
- ECR's of both H2 and CO were considered
- Only WGSR included
- Model developed to predict the non-uniform flow
velocity distribution among the cell channels
- Model developed to predict the cell to cell
performance variations
- Concentration polarization neglected 3D, transient
- Pressure drops neglected
- Heat transfer by convection only
- Isothermal modeling for button SOFC 2D, steady state
- Ohmic loss for electrolyte only
- WGSR occurred at anode/electrolyte interface
- Single planar SOFC quasi-3D, steady state
- No radiation heat transfer (RHT)
- Adiabatic condition on side walls
- Isothermal conditions in flow channel
- Selected portion of single SOFC selected for study 2D, steady state
- Fixed exchanged current densities used for anode
and cathode
- No thermal BC specified on the outer walls
- Selected portion of single SOFC selected for study 2D, steady state
- Local thermal equilibrium (LTE) applied
- No radiation heat transfer
- Symmetry BC at the top and bottom side walls
- No RHT 3D, steady state
- LTE in electrodes
- Pressure gradient neglected in porous flow
- Isothermal BC at the outer walls of stack
 K. Rashid et al. / Energy 141 (2017) 673e691 675

stack performance to optimize the stack operating conditions. RHT Table 2

is also included in the modeling to improve the temperature dis- Geometric dimensions of single FT-SOFC.

tributions inside the cell components and also near the side walls. (mm)
There are two major objectives of this study; first to improve the Cell length 200
modeling presented in our previous work [50] and secondly, to Anode channel dimensions (L x W x H) 200  1.2 x 1.2
develop an improved SOFC system design for medium-scaled res- Number of channels 21
idential power unit by exploring optimum operating conditions for Rib width 1.15
Cathode electrode dimensions 45  168
safe and long-term operation.
Anode thickness 0.725
Cathode thickness 0.020
2. Experimental setup Electrolyte thickness 0.010

Fig. S1 (a, b) shows the experimental setup of 1 kW-class FT-

SOFC stack installed in a residential power generating unit (RPG)
with the essential balance of plant (BOP). The BOP system com-
prises of heat exchangers (HX's) positioned on both sides of the
stack; a steam reformer mounted over the stack and startup burner
embedded between the stack and the steam reformer.
The stack consists of 80 anode-supported flat tubular SOFC (FT-
SOFC) cells, vertically mounted on the manifold. The stack was first
operated with 3% humidified hydrogen then pre-reformed
methane was supplied as a second fuel to validate the numerical
modeling. In the first case, the system was heated up to 760
C
through startup burner by supplying 1.2 lpm methane. When the
system and equipment's temperature stabilizes to 760
C; 18 lpm
hydrogen gas after passing through the humidifier was supplied at
the manifold inlet, whereas 150 lpm pre-heated air was supplied
through the exit holes of both the heat exchangers (75 lpm through
each HX) positioned across the stack. After attaining steady tem-
perature, startup burner was switched off; an external load was
applied to draw electrical current ranging from 0 to 18.48 A.
The stack output voltages were recorded for each of the corre-
sponding current drawn once the system got stabilized. In the
second case, 4.55 lpm methane was supplied to the external steam
reformer and 10 ccm of water was supplied to the steam generating
tube when the reformer inlet temperature reaches about 650
C.
Similarly, an external load is applied to draw electrical current
ranging from 0 to 18.7 A when the temperature of the stack is
steadied at 760
C. The stack output voltages were recorded for
each of the corresponding current drawn once the system got
stabilized. To maintain stable reforming operation and air/fuel pre-
heating; spent fuel and the excess air were burned in the com-
bustion chamber to provide necessary heat energy when startup
burner was cut off.
3. Model development
A computational domain of anode-supported flat tubular SOFC
(FT-SOFC) stack is presented in Fig. 1 along with the necessary di-
mensions whereas geometric parameters for single FT-SOFC are
presented in Table 2.
The stack consists of 80 cells divided into four assemblies each
containing 20 cells positioned vertically in an air enclosure. The fuel
is supplied to these cells with the help of manifold attached at the
bottom. The manifold is modeled in such a way to distribute the
incoming fuel gas uniformly to the anode channels. The pre-heated
air is supplied across the stack through the exit holes of heat
exchangers.
3.1. Reactions considered for modeling
In the present analysis, pre-reformed methane mixture (H2,
H2O, CH4, CO, CO2) is considered as fuel and when a fuel is
comprised of methane and carbon mono oxide, methane steam
reforming (MSR) and water gas shift reactions (WGSR) are taking
place in the anode.
CH4 þ H2 O / CO þ 3H2 ðMSRÞ (1)
CO þ H2 O / CO2 þ H2 ðWGSRÞ (2)
The volumetric reaction rates of MSR and WGSR are adopted
from the work of Lehnert et al. [51] and the rate expressions are
summarized in Table 3. Hydrogen and carbon monoxide reacts with
oxygen ions (O2) at the anode-electrolyte interface (TPB) to pro-
duce steam, carbon dioxide respectively and releases electrons.

H2 þ O2 /H2 O þ 2e ðReaction at anode

 electrolyte interfaceÞ (3)

CO þ O2 /CO2 þ 2e ðReaction at anode

 electrolyte interfaceÞ (4)

1
O þ 2e /O2 ðReaction at cathode  electrolyte interfaceÞ
2 2
(5)

3.2. Electrochemical models

The electrochemical modeling is employed to determine the

local current density distribution at the operating voltage. The
equilibrium/open circuit potentials for the electrochemical re-
actions of H2 and CO are locally calculated by the following Nernst
equations [35,36] and their expressions are provided in Table 4.
Fig. 1. Fluid domain with necessary dimensions. The activation polarization is an additional potential required to
676 K. Rashid et al. / Energy 141 (2017) 673e691

Table 3
Model equations for the calculation of internal steam reforming.

Reforming reaction rate (mol m3 sec1) Rsr ¼ kþ  3

MSR PCH4 PH2 O  kMSR PCO ðPH2 Þ
Shift reaction rate (mol m3 sec1) Rsh ¼ kþ 
WGS PCO PH2 O  kWGS PCO2 PH2
kþ
MSR (mol m
3
Pa2 s1) 2.3  108
k
MSR (mol m
3
Pa4 s1) 1.4  1020
kþ
WGS (mol m
3
Pa2 s1) 1.5  107
k
WGS (mol m
3
Pa2 s1) 1.4  107

Table 4
Nernst potentials and exchange current density calculations.
" #
Nernst potential (H2) PH2 ðPO2 Þ0:5
o þ RT ln
EH2 ; Nernst ¼ EH o ¼ 1:253  2:4516:104 *T
EH
2 2F PH2 O 2

" #
Nernst potential (CO) PCO ðPO2 Þ0:5
o þ RT ln
ECO;Nernst ¼ ECO o ¼ 1:46713  4:527: 104 *T
ECO
2F PCO2

   
Current densities 2F hact;an; H F hact;an; H
IH2 ¼ RT
3F dan; expH2 RT
2
 exp RT
2

   
RT 2F hact;an;CO F hact;an;CO
ICO ¼ 2:53F dan;CO exp RT  exp RT

   
2F hcat;O 2F hcat;O
IO2 ¼ RT
4F dcat;O2 exp RT
2
 exp RT
2

Interface conduct ivies   0:266

_ PH2 O
dan; ¼A 12000
H2 H2 exp R:T Keq; H2 PH2

  !0:266
_ PCO2
dan;CO ¼ A CO exp 12000
RT Keq; CO PCO
 
_
dcat;O2 ¼ A O2 exp 13000 PO0:50
RT 2
!  
Equilibrium constant 2FEHo o
2FECO
Keq; H2 ¼ exp RT
2
, Keq; CO ¼ exp RT

Adjustable parameter (validation) _ _ _

A H2 ¼ 2:15  1012 ; A CO ¼ 2:0  1012 ; A O2 ¼ 1:95  1011

overcome the energy barrier (activation energy) to precede the presented by Suwanwarangkul et al. [33] and Negata et al. [52] are
electrochemical reactions at the interfaces at the expected rate. The considered and detail explanation can also be approached in
Butler-Volmer equation is used to determine the relationship of Ref. [53]. The summary of the selected expressions for exchange
electrodes activation overpotentials and current densities because current density is described in Table 4.
of ECRs. The ohmic polarization is the resistance encounter by the

    movements of ions and electrons through the electrolyte, and
anF h ð1  aÞnF h electrodes respectively. The effective electronic conductivities of
I ¼ Io  exp  (6)
RT RT the anode and cathode electrodes and the ionic conductivity of the
electrolyte is adopted from the work of Andersson et al. [36] and
The exchange current density (Io) at the electrode/electrolyte
they are displayed in Table 5. The overall ohmic loss can be evalu-
interface depends upon many factors like material properties, ECR
ated by combining the conductivities of electrodes and electrolyte.
kinetics, operating conditions as well as reactant/product concen-
trations. A general Arrhenius type expression is employed for its " #
calculation. tan tcat te
hohm ¼ I þ þ (8)
san;eff scat;eff se;eff
 
Ea
Io ¼ k exp  (7) When the current is being drawn from the system, reactants are
RT
consumed in the vicinity of the electrode-electrolyte interface. The
In the present work, the exchange current densities expressions concentration difference of reacting species between the bulk and

Table 5
Physical properties of cell components*.

Anode electrode Cathode electrode Electrolyte

Component material Ni-YSZ LSM YSZ

Density (Kg m3) 6380 6864 5950
Specific heat (J Kg1 k1) 440 607 400
Thermal conductivity (W m1k1) 3.6 10 2.7
Porosity (ε), % 40 35 e
Average particle diameter, dp (mm) 2.5 2.5 e
Mean pore radius, rm (mm) 0.5 0.5
Tortuosity (x) 3.5 2.9 e

 
 
 
Effective electronic/ionic εan εcat εe
san;eff ¼ 9:510 1150 scat;eff ¼ 4:210 1200 se;eff ¼ 3:34  104 exp 10300
7 7
exp exp
conductivities (S.m1) T T xan T T xcat T xe
 K. Rashid et al. / Energy 141 (2017) 673e691 677

Table 6
Diffusion coefficient calculations.

Binary diffusion coefficient h i 1

=
2

0:001858 T 3=2 1
MA
þ M1
DAB ¼
B

Pv2AB UD
Multicomponent diffusion coefficient DAmix ¼ yB yC
1yA
yD
DAB þDAC þDAD þ…/
qffiffiffiffiffi
Knudsen diffusion coefficient DK;A ¼ 4850 dpore T
MA
Effective gas diffusion coefficients (m2 s1) H2/H2O/CH4/CO/CO2/O2/N2 4.73e5/1.20e5/1.97e5/1.74e5/1.41e5/1.27e5/1.30e5

at the TPB becomes significant because of the mass transport pro- 3.3. Electrical circuit model
cess. These mass transport phenomena take place in two steps (i)
external diffusion i.e. gasses diffuse from channels to the electrode The electrochemical reactions are schematically demonstrated
surface (ii) in/out transport of reactants/products within the porous as an equivalent circuit in Fig. 2. In the present work, two different
electrodes. The expressions used for the calculated of diffusion fuels (H2 and pre-reformed methane) have been considered for the
coefficients are summarized in Table 6 [54]. On the other hand, stack analysis; therefore, two separate approaches are adopted to
concentration polarization depends on many factors such as; type handle the ECR of H2 fuel and reformed gas mixture (considering
of the electrochemically active species, the thickness of diffusion ECR of both H2 and CO). In Fig. 2 (a), only the ECR of H2 is consid-
path as well as porosity/tortuosity of the electrodes. The expres- ered, and in this method, all the resistances are following series
sions for concentration polarization are presented as [55,56]: circuit model and ohm's law is used to calculate circuit voltage. In
the second circuit, Fig. 2 (b), the model comprises of two electro-
! motive forces (EMF) generated by electrochemical oxidations of H2


 P
PH and CO and their relevant activation and concentration resistances.
RT O H
hconc
an;H2 ¼ ln  2
2 (9) Whereas ohmic loss (due to ionic and electronic resistance), acti-
2F PH2 PH2 O
vation and concentration resistances of oxidant at the cathode are
connected in series in the model. The average current density at the
! TBP is maintained invariable whereas local current density varies


 P
PCO
RT CO over the reaction surface and is calculated based on heat/mass
hconc
an;CO ¼ ln  2
(10) transfer together with both electrochemical reactions. The two
2F PCO PCO2
ECR's given in equations (3) and (4) are represented in the circuit
diagram as two parallel reactions. Mass and charge transports
! phenomenon will regulate these reactions to acquire distinct cell


RT PO2 output voltage.
hconc ¼ (11)
cat;O2
4F PO 2
V cell ¼ EH2 ; Nernst  hact;H2  hact;O2  hconc conc
an;H2  hcat;O2  hohm
(12)

V cell ¼ ECO;Nerst  hact;CO  hact;O2  hconc conc

an;CO  hcat;O2  hohm (13)

I ¼ IO2 ¼ IH2 þ ICO (14)

The exchange current densities of H2 and CO at anode and O2 at
the cathode are calculated from the interface conductives
mentioned in Table 4 by adopting the method provided by
Refs. [22,32,57]. The Kirchhoff's junction and loop rules are applied
to solve the equivalent circuit model, and the total local current
density obtained at the anode current collector is the summation of
ECR's of H2 and CO. Whereas, the contribution of H2 and CO oxi-
dations towards the total current is determined locally by iteration
method in such a way that similar potential difference across the
parallel circuit part is obtained [34].

3.4. Radiative heat transfer modeling

To execute the heat transport analysis of the system, it is

Fig. 2. Schematic diagram of equivalent electrical circuits (a) ECR of H2 only (b) ECR of
both H2 and CO.
imperative to identify locations where heat is being produced and
consumed. The majority of the heat energy is generated at elec-
trolyte interfaces with the electrodes through reversible and irre-
versible processes. Most of the produced heat is recovered from the
system through anode fuel channels and air by convection but it did
not cover all the energy balance. To design a real system, it is,
therefore, important that radiative heat transfer (RHT) process
should be included in the modeling. The RHT mechanism not only
involves radiation transfer in electrodes, electrolyte and
678 K. Rashid et al. / Energy 141 (2017) 673e691
participating gases but it also involves surface-to-surface radiation
energy transfer in the channels [58]. In the present work, DO
(discrete ordinates) method is selected to approximate the RHT as it
covers the entire range of optical thicknesses and allows to solve
problems for both surface-to-surface and radiation in participating
media. This method also allows non-gray properties and specular
surfaces. Scattering is modeled using a scattering coefficient and a
scattering phase function is defined isotropic. As in the present
study, gases are also considered as participating media in RHT and
their properties are assumed composition dependent by adopting
weighted sum of gray gas model (WSGGM). The porous electrodes
are treated as opaque [59] and electrolyte considered as optically
thin, non-diffusing gray medium (isotropic) [44]. For angular dis-
cretizations in DO model, default settings are opted to avoid
computational delay.
3.5. CFD models and solution procedure
The governing equations for the continuity, momentum, energy
and species conservations are summarized in Table 7 along with
their related source terms. In the continuity equation, Sr is the
source term which characterizes the mass consumption and gen-
eration because of ECR's at the electrode-electrolyte interfaces and
MSR in anode electrode. In the momentum equation, u is the ve-
locity component in x, y, z directions; density (r) of the gaseous
mixture is computed by ideal gas law and viscosity (m) by Suther-
land's Law using the kinetic theory of gasses [60]. Sm is the mo-
mentum source term which contains viscous loss term (described
by Darcy's Law) and inertial loss term (defined by Power Law) [61].
Where a is the permeability (m2) and C is the inertial resistance
factor. In the energy equation, Cp is the specific heat of the gas
mixture and keff is effective thermal conductivity, whereas Sh is the
source term (Wm3) comprise of the electrochemical heat gener-
ation source term and methane steam reforming heat sink source
term.
ECR heat source includes ECR's entropy change (△S) due to both
H2 and CO and irreversible polarization losses [36,62e64]. Internal
reforming reaction consists of two reactions, MSR (eq. (1)) which is
endothermic in nature whereas WGSR (eq. (2)) is a weak
exothermic. In the specie equation, Yi is the mass fraction and Di,m is
the effective mass diffusivity of component i in a gaseous mixture of
fuel and oxidant streams (Table 6). Ssp is species source term, which
determines the rate of consumption and generation of each gas
component in the multicomponent gas mixture due to the
Table 7
Governing equations for mass, momentum, and heat transport calculations.
Continuity V$ðruÞ ¼ Sr
Momentum V$fðruÞug ¼ V$ðP þ m$VuÞ þ Sm
Energy V$ðrCp uTÞ ¼ V$ðkeff V TÞ þ Sh keff ¼ εkf þ ð1  εÞks
Species V:ðruYi Þ ¼ V:ðrDi;m V:Yi Þ þ Ssp
electrochemical and chemical reactions.
The mesh was constructed to attain a finer mesh at the reaction
sites and in the regions of possible high gradients. A finer mesh was
made at the anode-electrolyte interfaces, and at the location where
the sharp deviations in flow distribution were expected; whereas
relatively coarser mesh was made away from interfaces to accom-
plish required goal. The non-conformal mesh strategy was adopted
[61] and Ansys® ICEM CFD was utilized to construct hexahedral
mesh for all the stack components. The accuracy of the numerical
calculations was carried out by grid independence to ascertain the
optimum number of nodes. Fig. 3 (a, b, c) show grid quality of
manifold, 80 cells, and air enclosure at the selected region.
First, the single cell was tested for several numbers of nodes to
obtain an optimized mesh and then optimized mesh was applied to
the full-scale 80 FT-SOFC stack. The optimal mesh resolution of a
single cell on the active surface is presented in Fig. 3 (b). The final
mesh size of the stack is about 52 million elements.
To solve the intricate transport, ECR's and chemical processes
simultaneously, some simplifications are required to be made to get
the numerical solution accurately and efficiently. The following
assumptions are made during the CFD calculations:
1) Steady state analysis.
2) Laminar and incompressible flow.
3) CH4 only involved in MSR; all other possible reactions are
neglected.
4) Dry reforming is neglected because of high water contents in the
pre-reformed gas mixture.
5) MSR/WGSR is only taking place in the anode.
6) The mass flow inlet boundary condition is applied at the inlet of
the manifold and for incoming preheated air stream.
7) The pressure outlet boundary condition is applied at anode and
cathode outlets, and gauge (reference) pressure of 0 Pa is set.
8) Side walls of HX with air inlet holes are assumed isothermal as it
is expected that temperature at the walls will be close to the air
inlet temperature.
9) External emissivity (0.8) and external radiation temperature
(1050 K) are fixed on other walls and their boundary type is
opaque [65].
The main operating parameters opted from the experiment are
presented in Table 8 whereas physical properties of the cell com-
ponents are provided in Table 5. Different pre-reformed methane
fuels (based on CH4 conversion) are supplied at the manifold inlet
Sr;anðECRþMSRÞ ¼ SH2 þ SH2 O þ SCH4 þ SCO þ SCO2 Sr;cat ¼ SO2


Sm ¼  ami ui þ 12Ci jujui
Sh ¼ ShECR þ ShMSR
In anode (Heat sink):
ShMSR ¼ Rsr Hsr þ Rsh Hsh Hsr ¼ ð206205:5 þ 19:5175 TÞHsh ¼ 45063  10:28 T
In electrolyte (Heat source):
   
I
ShECR ¼ ðT DSECR Þ$2Fte
þ htotal I 1
te DSECR ¼ SH2 O  2SO2  SH2 þ SCO2  12SO2  SCO htotal
¼ Nerst potential  Net voltage
[35]
In anode:
Ssp;H2 ¼ 3Rsr þ Rsh ; Ssp;H2 O ¼  Rsr  Rsh Ssp;CH4 ¼  Rsr ; Ssp; CO ¼ Rsr  Rsh Ssp;CO2 ¼ Rsh
At anode-electrolyte interface:
IH2 IH ICO ICO
Ssp;H2 ¼  2Fte MH2 ; Ssp;H2 O ¼ 2Ft2e MH2 O Ssp; CO ¼ 2Ft e
MCO ; Ssp;CO2 ¼ 2Ft e
MCO2
At cathode-electrolyte interface:
IO
SSPcat ¼ 4Ft2e MO2
 K. Rashid et al. / Energy 141 (2017) 673e691 679

Fig. 3. Mesh quality of stack components (a) Manifold (b) 80 FT-SOFC (c) Air Enclosure.

Table 8 Table 10
Operating parameters for 80 cells stacka. Air flow rates at different oxidant utilization (for 80 cells stack).

Operating temperature (K) 1033.15 Oxidant Utilization (%) Air Flow rate
Operating pressure (Pa) 101325
(LPM) (Kg s1)
Operating average current density (Am2) 2473
Fuel inlet (kg s1) 3% humidified H2 3.463e5 20 122.6 2.631E-03
Fuel inlet (kg s1) pre-reformed methane 2.1850e4 30 81.704 1.754E-03
Air inlet (kg s1) 2.962e3 40 61.28 1.315E-03
a 50 49.02 1.052E-03
Calculated parameters from STX operated stack.

to determine the influence of internal reforming on the stack per-
formance. Steam to carbon ratio (S/C) of 2.5 is fixed in the pre-
reformer, Whereas various input fuel compositions are summa-
rized in Table 9. During the simulation, the effect of oxidant utili-
zation on stack performance and thermal management are also
tested. The necessary airflow rates based on different oxidant uti-
lizations against the reference fuel are provided in Table 10.
Being the complex nature of the SOFC structure, commercial
CFD software Ansys® Fluent is selected as a solver because of it is
broadly established accuracy [66]. The user-defined function (UDF)
in-house code written in C language is implemented in the solver to
compute ECR and MSR models and to generate an electrical field.
The finite volume approach is employed to discretize the governing
equations. A brief description of the solution procedure is
Table 9
Pre-reformed methane composition.
CH4 Conversion (%) Fuel Type Mole fraction
presented in Fig. S2. At the beginning, the solution is initialized
with UDF macro, and then after each iteration USER DEFINED
ADJUST macro provided by Ansys® Fluent is used to compute the
ECR and MSR models to determine the current densities distribu-
tion and the rate of MSR/WGSR. The source terms generated from
the solution of ECR/MSR models are implemented in the solver
(CFD models) to update the mass, flow field, species and temper-
ature distributions in the cells at the relevant zones. This updated
information is subsequently utilized to compute the electro-
chemical and chemical models again. The solution procedure is
repeated until the desired degree of accuracy is achieved. The
pressure-based solver is selected to execute CFD calculations with
second-order upwind interpolation scheme is employed to solve
the model equations whereas the convergence criteria on the re-
sidual scale are set to 103 for continuity, momentum, and species
whereas 106 for energy equation. However, for DO-model, a first-
order upwind scheme with 0.6 under-relaxation factor (URF) is
used to avoid divergence.

CH4 H2O H2 CO CO2 4. Results and discussion

99 [24] F1 0.01 0.15 0.65 0.15 0.04
90.1 [STX] F2 0.023 0.13 0.655 0.11 0.082 Before exploring the optimized operating conditions of SOFC
79.2 [STX] F3 0.038 0.301 0.509 0.076 0.076 stack, presented modeling has been validated with the experiments
75.4* [STX] F4 0.051 0.269 0.523 0.0723 0.085 performed at STX (Daegu, South Korea) facility. The numerical
30 [24] F5 0.171 0.493 0.263 0.0294 0.0436
calculations are carried out for different average current densities
680 K. Rashid et al. / Energy 141 (2017) 673e691
Fig. 4. Model validation with H2 and pre-reformed methane fuels.
(I_ave) i.e. 0-2500 Am2 and the stack voltage is iteratively adjusted
during the simulation in such a way that proposed I_ave be ob-
tained [34]. The stack is simulated with both 3% humidified H2 and
75.4% pre-reformed methane mixture at 1033.15 K, and the
Fig. 5. Comparison of H2 and pre-reformed methane performance (a) temperature distributions (b) current density distributions (c) polarization losses (d) nernst potentials.
performance curves are compared with the experimental data.
During simulations, for all average current densities, utilization of
fuel and oxidant are kept invariable, i.e., ~60% and ~20% respec-
tively and the same material properties of SOFCs are employed
those were used in experimental stack cells (Table 5). Meanwhile,
adjustable parameters of interface conductivities are adjusted to fit
the I-V-P curves with experimental data and their final values are
displayed in Table 4. Fig. 4 shows a comparison of simulated results
with the experimental data for both the fuels.
The numerical results exhibited good agreement with the
experimental results except at the lower current densities, where
simulated stack voltages are slightly higher because of some as-
sumptions like no contact resistance, the current collector with
infinite electrical conductance, etc. However, the maximum dif-
ference between the simulated and experimental data is less than
5% [67].
In this section, a brief comparison of the stack performance,
running with 3% humidified H2 and pre-reformed methane
(reference case, F4) is made. From Fig. 4, it is clearly evident that the
stack running with H2 fuel exhibited much higher performance, i.e.,
1216 W as compared to 1123 W. One of the reasons for higher
performance is that thermodynamically higher mole fraction
(partial pressure) of H2 steers to higher stack output power.
Second reason is that higher EMF and current density are ob-
tained through H2 fuel. All the pre-reformed fuel cases stated in
Table 9 have less than 66% mole fraction of H2 in their gaseous
mixture in comparison to 97% for H2 fuel. In the subsequent part,
 K. Rashid et al. / Energy 141 (2017) 673e691 681

reduction in the stack performance with reformed gas is more
elaborated by temperature Fig. 5 (a) and current density distribu-
tions by Fig. 5 (b) along the fuel stream. Figs. 5 (a) and 6 (b) show
that temperature first increases near the inlet (Y < 0.04 m) due to
ECR reactions then decreases steadily towards the end due to the
depletion of H2.
However, in the case of reference fuel, temperature first de-
creases due to MSR then increase because of ECRs. The current
density distribution for H2 fuel in Figs. 5(b) and 6(f) shows a steep
decline, whereas, for reference fuel, current density first rises due
to the addition of ECR species (H2 and CO) until Y ¼ 0.1 m then it
decreases due to the consumption of H2 and CO and fast depletion
of internal reforming reaction. In Figs. 5(c) and 6 (c, d, e), a com-
parison of polarization losses due to activation, ohmic, and accu-
mulative, have been presented to explore the reason why H2 fuel
exhibited higher stack performance. Only the ohmic polarization
has shown similar results, rests of losses for reference case fuel are
much greater at all the regions on the active surfaces. Similarly,
nernst potential distribution of H2 fuel along the flow stream (Fig. 5
(d)) is significantly higher than the individual reference fuel
because of the much higher H2 partial pressure in the fuel mixture.
From the comparison of EMF's and system polarization losses, it can
be deduced that H2 is a preferable fuel to get higher stack perfor-
mance, however, its ease of availability and storage are still chal-
lenging issues. In the subsequent sections, different methodologies
have been presented to obtain the optimal stack performance from
the same stack.
4.1. Effect of pre-reforming on the stack performance
In this section, the numerical results of 5 different cases of fuel
(based on pre-reformed methane conversion) are evaluated to

Fig. 6. H2 fuel distributions (a) H2 mole fraction (b) temperature distributions (K) (c) H2 activation resistance (Um2) (d) O2 activation resistance (Um2) (e) total polarization
resistance (Um2) (f) current density distributions (Am2).
682 K. Rashid et al. / Energy 141 (2017) 673e691

determine their effect on the performance of the stack. Fig. 7 (a, b)
demonstrate the variations of localized temperature and current
density distributions along the flow stream for all the 5 cases
(Table 9). In the case of fuel F1 and F2, the strong endothermic MSR
has limited effect on their temperature distributions, as most of the
CH4 is already consumed in the pre-reformer. Whereas in the case
of fuel F3 and F4, MSR has a visible effect of on the temperature
distribution that can be seen in Fig. 7 (a). However, in the case of F5,
a strong effect on temperature distribution is evident because CH4
is present at the inlet of the manifold with highest mole fraction
among all the fuels examined. The current density distributions for
the fuel (F1 to F4) exhibited similar trends over the whole active
surfaces. Near the inlet of Y-axis, F1 and F2 have shown higher
current densities in comparison to F3 and F4 due to the higher

Fig. 7. Effect of pre-reforming (along Y-axis) on (a) temperature distributions (b) current density distributions (c) mole fraction distributions of CH4 (d) total polarization losses (V)
(e) nernst potential distributions (V) (f) stack performance curves for different fuel cases.
 K. Rashid et al. / Energy 141 (2017) 673e691
molar concentration of H2, CO, and higher EMFs. However, at the
middle of Y-axis, fuel with higher CH4 content at the inlet exhibited
higher current densities due to the production of H2 and CO (MSR
and WGSR). F5 is the only fuel which has shown largest variations
of temperature and current density distributions due to their lower
content of H2 and CO and relatively higher CH4 mole fraction at the
inlet.
However, as CH4 is reformed internally, an appreciable amount
of H2 and CO is available in the middle of Y-axis causing tempera-
ture and current density to reach maximum or even goes beyond
other fuels. It is clearly evident from Fig. 7 (a, b) that local tem-
perature and current density distributions are strongly affected by
the pre-reformed methane conversion rate. The reason for the
difference in distributions of different fuels can also be envisioned
by probing Fig. 7 (c), which demonstrates CH4 mole fraction dis-
tributions along the main flow stream. The sharp decline in the
mole fraction means fast MSR reaction rate and it is also ascertained
from Fig. 7 (c) that higher the pre-reforming conversion rate, more
gradual is the internal reforming rate.
The distribution of total polarization over the whole active
surfaces and generation of electromotive force (EMF) are also
compared to analyze the stack performance. The distributions of
polarization losses and EMF are shown in Fig. 7 (d, e) further sup-
ports the conclusion made in Fig. 7 (a, b) that robust internal
reforming reaction rate (in the case of F5) causes highest total
overpotential. However, relatively higher EMF of F5 is observed at
Y > 0.04 m, compared to F3 and F4 because of low temperature and
gradual increase of H2 and CO mole fractions. The comparison of
the stack performance for all 5 fuel cases is also made to determine
which fuel can deliver the highest performance under same oper-
ating conditions. It is cleared in Fig. 7 (f) that fuel F1 and F2 have
attained similar performance, so as fuel F3 and F4. However, F5
exhibited the least performance for present operating conditions.
From the above discussion, it is evident that uniform temperature
and current density distributions in the stack can be achieved by
controlling the pre-reforming rate. Fuel F1 and F2 are the fuels
which exhibited the highest stack performance and most uniform
temperature and current density distributions. However, larger air
flow rates or bigger HX's are required to dissipate the excess heat
that is harmful to overall system efficiency. The results of F3 and F4
fuels are more trade-off between system efficiency and stack long-
term stable operation.
4.2. Effect of CO as electrochemical fuel on the stack performance
In the previous SOFC modeling, it was common practice to
neglect the electrochemical oxidation of CO. So, it is crucial to
determine whether its negligence has any effect on the stack per-
formance. In this section, the stack performance is analyzed by
comparing the results with and without electro-oxidation of CO.
When ECR of CO is not considered, total current density obtained is
only related to the H2 fuel, whereas CO is assumed to utilize in
WGSR. And when both H2 and CO are considered as ECR species,
total current density is a summation of both individual current
densities of H2 and CO. Fig. 8 (a) provides a comparison of perfor-
mance curves for both the modeling. It is evident that there is a
drop in the stack performance when ECR of CO is not considered. In
Fig. 8 (b) this decline in stack performance is further clarified by
total stack polarization over the whole active surface. The model
without the ECR of CO has shown higher polarizations at all the
regions along the fuel streamwise direction.
It is also demonstrated in Fig. S3 (b) that relatively lower current
density distributions are obtained in comparison to Fig. S3 (a). It is
also noticed that whether ECR of CO is included in the modeling or
not, specie distribution of H2 (Fig. 9 (a, b)) and rate of MSR reaction
683
(Fig. 8 (d)) is not changed significantly. However, numerical results
of the model which neglect the CO ECR has extensively affected the
temperature distribution (Fig. 9 (g, h), rate of WGSR (Fig. 8(d)), and
CO molar concentration along the anode channels (Fig. 9 (c, d)). The
increase in temperature distribution is due to a couple of reasons;
first reason is that higher polarizations contribute to higher irre-
versible ECR heat, whereas, the second reason is that positive rate
of WGSR is found (without CO ECR) which is exothermic in nature
and reduces the heat sink term of internal reforming.
This positive rate of WGSR reaction is quite distinct in the
(Figs. 8 (d) and 9 (e, f)) because CO is only participating in WGSR
reaction. That is why WGSR rate and CO mole fraction along the
flow stream are higher as compared to the model which is
considering CO ECR (WGSR is mostly negative along the Y-axis for
model considering CO ECR).
4.3. Effect of inlet temperature on the stack performance
The current density and temperature distributions for two
different temperatures (1033.15 K and 1073.15 K) have been pre-
sented in Fig. 10 (a, b) to establish the effect of inlet temperature
(fuel and air) on the stack performance. It is evident that with the
increase of inlet temperature, current densities generated by both
the electrochemical species (H2 and CO) increases which signify
that higher inlet temperature is beneficial for acquiring higher
output power. It is also noticeable in Fig. 10 (a, b) that current
densities and temperature for both the fuels vary along the flow
stream which is quite clear in (Fig. S4 (e) and (f)).
A slide decline in the current densities is observed near the inlet
due to the lower temperature caused by MSR. The current density
associated with the H2 for both the temperatures is increasing
along the flow direction until Y ¼ 0.14 m then begins to decline
again due to a decrease in temperature and decrease in H2 mole
fraction (Fig. S4 (a)). However, the current density related with CO
fuel exhibited linear behavior along the flow stream with slight
decline towards the end of the active surfaces. The difference in
current densities is mainly due to the molar concentration distri-
bution in the anode electrodes. Mole fractions of both H2 and CO are
relatively higher at the inlet for 1073.15 K due to the higher MSR
rate, as a steam reforming efficiency is accelerated by higher tem-
perature. In the case of WGSR, higher temperature has an adverse
effect due to Le Chatelier principle and the strong backward reac-
tion is observed due to the build-up of CO2 (CO ECR). It is important
to note that CH4 and CO mole fractions in the reference fuel stream
are only 5 and 7%, respectively. Their contribution to the supple-
ment of H2 in the flow stream is quite low which is quite evident
from Fig. 10 (c). No distinct deviations of MSR and WGSR for both
the temperatures are observed. As a result, species distributions
shown in Fig. S4 (a, b) and reactions rates in Fig. S4 (c, d) are similar
to those of observed for 1033.15 K case. However, noticeable
improvement in the stack performance is recorded due to the in-
crease of inlet temperatures which is reflected in Fig. 10 (d). Po-
larization curves are plotted along the Y-axis (Fig. 10 (e, f)), to unveil
why there is an enhancement in the stack performance despite the
minimum effect of internal reforming. The ohmic polarization
(Fig. 10 (e)) shows steep decline along the downstream for both the
temperatures due to the rise of temperature by ECR heat (Fig. S4
(e)). Its values are lower for higher inlet temperature fuel at the
whole active surfaces. A similar effect of temperature on the other
polarizations like activation and system total losses is observed in
Fig. 10 (f). It evident that stack performance can be enhanced by
decreasing the overpotential losses by increasing the operating
temperature without increasing methane concentration in the inlet
fuel.
684 K. Rashid et al. / Energy 141 (2017) 673e691

Fig. 8. Effect of with and without CO ECR on (a) stack performance curves (b) total polarization losses (c) temperature distributions (d) MSR and WGSR rates.

4.4. Effect of oxidant utilization on the stack performance help of polarization losses especially cathode activation and total
polarization losses. Fig. 11 (d) provides a comparison of activation
It was reviewed in the introduction section that internal and total polarization losses among the different oxidant utilization
reforming is a useful tool to utilize the waste heat of ECR by used during simulation. Air stream with 50% oxidant utilization
smothering the need of excess air which ultimately increases the exhibited highest polarizations (both activation and total) among
system efficiency. However, it is numerically demonstrated that the other examined; that is the reason why it delivered the least
control of pre-reforming rate is a crucial strategy to avoid large performance (Fig. 11 (a)). It is cleared from above discussion that
thermal gradients on the stack cells which poses challenges to lower oxidant utilization is more favorable for achieving higher
stable and long-term operation. In this section, different oxidant stack performance. However, excess air supply will consume
utilizations are examined to explore optimum air flow rate which parasitic energy which results in lower system efficiency.
delivers the best stack performance. The oxidant utilizations are
varied from 20 to 50% while fuel utilization fixed at ~60%. The stack
4.5. Effect of thermal radiation modeling
performance curves have been presented in Fig. 11 (a). It is cleared
that with the decrease of oxidant utilization, stack performance
The effect of radiation heat transfer inside the stack cells and on
increases steadily, indicating excess air supply is favorable for
the outer walls of air enclosure are presented in Figs. 12 and 13. To
obtaining higher stack output power. Oxidant utilization also af-
elaborate the temperature distributions, YZ, ZX and XY planes are
fects the temperature and current density distributions which can
drawn about the stack (80 cells and air enclosure only). Fig. 12 (a)
be seen in Fig. 11 (b, c). Fig. 11 (b) shows that oxidant utilization has
illustrate the temperature distribution along the X-axis (149 mm,
an opposing effect on the temperature distribution along the flow
YZ plane) with and without the inclusion of radiative heat transfer.
stream; higher temperatures are observed for higher utilization
Noticeably, the heat transfer improvement due to RHT results in
rates. 20% oxidant utilization has the lowest temperature rise
overall reduction of the temperature distribution within the stack,
because excess air is available to carry the waste heat of the stack.
as well as the less steep temperature gradients along the cell is
In the case of current density (CD) distributions (Fig. 11 (c)),
examined. The maximum difference in temperatures between the
higher oxidant utilization has shown higher current density dis-
two cases varies from 8 to 10 K throughout the YZ plane. However,
tributions because temperature distributions are higher for higher
the temperature distribution near the side walls is relatively flat for
air utilizations. The reason for a decline in the stack performance
the RHT case. Fig. 12 (b) demonstrate the RHT modeling effect along
with the increase of oxidant utilization is further clarified with the
the Y-axis (200 mm, along the fuel stream) and it cleared that
 K. Rashid et al. / Energy 141 (2017) 673e691 685

Fig. 9. Effect of CO ECR with and without on the contours distributions.

686 K. Rashid et al. / Energy 141 (2017) 673e691

Fig. 10. Inlet temperature effect on (a) current density distributions (b) temperature variations along the flow stream (c) MSR and WGSR rates (d) stack performance curves (e)
ohmic polarization (f) activation (H2 and CO) and accumulative overpotentials distributions.

temperature difference between two cases is lowest at the entrance
of anode channels. However, the gap becomes wider towards the
end of the anode electrode because of the absorption of heat energy
by the fuel cell components due to radiation heat transfer. The
temperature rise of ~18 K from the entrance to peak region is
observed for the case without RHT whereas this rise in temperature
is about 12 K for the case with thermal radiation effects which
further endorses that radiative heat transfer helps in decreasing the
temperature gradients on the cell components. In the case of XY
plane (332 mm, along the Z-axis), the different temperature dis-
tribution is observed for both the cases. The reason for this behavior
for the case without RHT is that fuel is first supplied to the
 K. Rashid et al. / Energy 141 (2017) 673e691
Fig. 11. Effect of different oxidant utilizations on (a) Stack performance curves (b) temperature distributions (c) current density distributions (d) O2 activation and total
overpotentials.
manifold, positioned at the bottom, from there gases are distrib-
uted to the anode channels.
The first 40 cells near the manifold entrance receive higher fuel
than the last 40 cells of the stack as uniformity index of this
manifold design is 0.826 [50]. Therefore it can be seen (Fig. 12 (c))
that temperature distribution is lower for last 40 cells (without
RHT). However, the case that is considering thermal radiation ef-
fects revealed similar temperature profile for the first and last
40 cells which ascertained that radiation heat transfer is an effec-
tive mode of heat transfer in minimizing the thermal gradients
among the stack cells. Radiation heat transfer has also imparted the
positive effect on the methane steam reforming which can be seen
in Fig. 12 (d). The higher reforming rate is observed near the
entrance of anode channels for the case with RHT, however, radi-
ation heat effect diminishes around Y ¼ 0.125 m. In the case of
WGSR, the reverse reaction rate is increased further for RHT case
due to the higher mole fraction mole of CO2 than without RHT case.
A localized high-temperature region on the air enclosure wall
(facing the active surfaces of cells) is spotted in Fig. 13 (a) whereas
the decline in temperature on the side and top walls from assem-
blies 1,2 to assemblies 3,4 can be clearly seen. However, the case
with RHT revealed uniform temperature profile (Fig. 13 (b)) on the
side and top walls and there is no localized high-temperature re-
gion on other walls.
687
5. Conclusions
In this study, a comprehensive three-dimensional (3D) numer-
ical model was developed to explore the optimum operating con-
ditions for anode-supported flat tubular (FT-SOFC) 80 cells stack,
running on H2 and pre-reformed methane fuels. The CFD model
coupled the complex transport processes (mass, momentum, and
heat transfers) with electrochemical reactions (with and without
electrochemical oxidation of CO). Internal steam reforming and
water gas shift reactions were also integrated with the modeling.
The effect of radiative heat transfer on the temperature distribution
of stack components was also presented. The CFD model was
validated with experiments conducted with both the fuels sepa-
rately. The findings of the above analysis are summarized as
follows:
1) Stack running with H2 fuel exhibited the highest performance
among all the fuel considered for analysis due to higher nernst
potential and lower polarizations, i.e., 1216 W and 1123 W
respectively.
2) 5 cases of fuel with different pre-reformed methane conversion
were investigated to establish the effect of pre-reforming rate
on the stack performance. It was found that fuel with highest
pre-reforming conversion exhibited the highest stack perfor-
mance. From the in-depth analysis, it is revealed that lowers the
688 K. Rashid et al. / Energy 141 (2017) 673e691

Fig. 12. Temperature distributions on (a) YZ plane (b) ZX plane (c) XY plane (d) RHT effect on the internal steam reforming.

Fig. 13. Temperature distributions on (a) air enclosure without RHT (b) air enclosure with RHT.

pre-reforming rate, higher is the non-uniformity of temperature density distributions, however, it necessitates the use of a large
and current density distributions along the flow stream. It was excess of air to remove the irreversible heat which may reduce
also concluded that although higher pre-reformed fuel deliv- overall system efficiency.
ered higher power, relatively uniform temperature, and current
 K. Rashid et al. / Energy 141 (2017) 673e691 689

3) Modeling comparison considering with and without CO electro- DA,eff Effective diffusion coefficient of species A in gas mixture,
oxidation revealed that model without CO oxidation un- m2 s1
derestimates the stack performance due to higher polarizations. E
Standard state potential, V
It was identified that stack showed higher temperature distri- F Faraday constant, 96485.3 C mol1
butions without CO electro-oxidation. The positive rate of WGSR I Current density, Am2
was the main reason for this temperature increment which is L Length of diffusion path, m
also a weakly exothermic reaction. M Molecular weight, g/gmole
4) The stack performance was also tested by increasing the inlet P Pressure, Pa
temperature from 1033.15 K to 1073.15 K. It was found that P
Pressure at standard state, Pa
operating the stack at 1073.15 K, demonstrated higher output Ract Activation resistance, U m2
power. Higher current density distributions have been achieved Rconc Diffusion/concentration resistance, U m2
with the higher temperature. For inlet temperature analysis, Rohm Ohmic resistance, U m2
reference fuel was considered for both the temperatures. Rsr Rate of internal steam reforming reaction, mol m3 s1
However, due to lower CH4 and CO mole fractions at the Rsh Rate of water gas shift reaction, mol m3 s1
manifold inlet, the diminutive effect of temperature on the in- Sr Source term in continuity equation, Kg m3 s1
ternal reforming was observed. Sm Source term in momentum equation, Kg m2 s2
5) The performance analysis on the stack was also implemented by Sh Source term in energy equation, W m3
varying oxidant utilizations to estimate appropriate air supply Ssp Source term in species equation, Kg m3 s1
by maintaining a balance between effective heat recovery and T Temperature, K
efficient system operation. It was revealed that higher oxidant t Thickness of cell components, m
utilization exhibited lower stack performance due to the higher Vcell Cell voltage, V
polarizations. It was also found that higher oxidant utilization X Mole fraction
yields higher temperature distributions due to a lower excess air Y Mass fraction
supply. Similarly, higher oxidant utilization has shown higher
current density distributions. Greek letters
6) From the radiative heat transfer analysis, it was revealed that h Overpotential, V
inclusion of this mode of heat transfer in the modeling is critical san Electronic conductivities of anode, U1 m1
to accurately predict the temperature field of the stack. sca Electronic conductivities of anode, U1 m1
Neglecting the RHT will result in incorrect temperature distri- se Ionic conductivities of electrolyte, U1 m1
bution and hence different reaction rates and component m Fluid viscosity, Kg m1 s1
properties will be overestimated. From the temperature profiles d Interface conductivity, U1 m2
on XY, YZ and ZX planes it was learned that RHT works to lessen
the peak temperature in the stack components and it also helps
Subscripts
to decrease the temperature gradients. The maximum temper-
act Activation
ature rise was 12 K for RHT case compared to 18 K for the case
an Anode
without RHT and noticeable improvement of temperature dis-
bulk Bulk flow/flow through the anode channel
tribution on XY plane was observed for the RHT case.
conc Concentration
7) From the parametric analysis, it was established that present
cat Cathode
1 kW-class FT-SOFC stack design performed best when operated
ECR Electrochemical reaction
at 75e80% pre-reforming rate, air utilization kept around
ohm Ohmic
20e30%, fuel utilization around 60% and finally, the operating
TPB Triple phase boundary
temperature maintained at 760e780
C. The reason for limiting
* Partial pressure at TPB
fuel utilization to 60% is that afterburner is installed along with
Partial pressure in bulk flow
steam generator/reformer in the complete hotbox design. The
excess fuel and cathode air are burned to supply the necessary
heat to sustain the endothermicity of BOP components. References

[1] Chatrattanawet N, Saebea D, Authayanun S, Arpornwichanop A,

Acknowledgements
The author gratefully appreciates the financial support of the
Ministry of Trade, Industry and Energy, MOTIE (Grant No.
R0002207), and Korea Institute for Advancement of Technology
and DaeGyeong Institute for Regional Program Evaluation (DGIRPE)
through the Leading Industry Development for Economic Region.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
https://doi.org/10.1016/j.energy.2017.09.082.
Nomenclature
DAB Mass diffusivity of specie A through B, m2 s1
DKA Knudsen diffusion coefficient of A, m2 s1
Patcharavorachot Y. Performance and environmental study of a biogas-fuelled
solid oxide fuel cell with different reforming approaches. Energy 2017:1e10.
https://doi.org/10.1016/j.energy.2017.06.125.
[2] Gong W, Yan X, Liu X, Cai Z. Parameter extraction of different fuel cell models
with transferred adaptive differential evolution. Energy 2015;86:139e51.
https://doi.org/10.1016/j.energy.2015.03.117.
[3] Luo Y, Shi Y, Li W, Cai N. Comprehensive modeling of tubular solid oxide
electrolysis cell for co-electrolysis of steam and carbon dioxide. Energy
2014;70:420e34. https://doi.org/10.1016/j.energy.2014.04.019.
[4] Mushtaq U, Mehran MT, Kim SK, Lim TH, Naqvi SAA, Lee JW, et al. Evaluation
of steady-state characteristics for solid oxide carbon fuel cell short-stacks.
Appl Energy 2017;187:886e98. https://doi.org/10.1016/j.apenergy.2016.
11.015.
€ niz M. SOFC stack coupled with dry
[5] Barelli L, Bidini G, Cinti G, Gallorini F, Po
reforming. Appl Energy 2017;192:498e507. https://doi.org/10.1016/
j.apenergy.2016.08.167.
[6] Yan D, Liang L, Yang J, Zhang T, Pu J, Chi B, et al. Performance degradation and
analysis of 10-cell anode-supported SOFC stack with external manifold
structure. Energy 2017;125:663e70. https://doi.org/10.1016/j.energy.2016.
12.107.
[7] Andersson M, Yuan J, Sunde n B. SOFC cell design optimization using the finite
element method based CFD approach. Fuel Cells 2014;14:177e88. https://
doi.org/10.1002/fuce.201300160.
690 K. Rashid et al. / Energy 141 (2017) 673e691
[8] Khazaee I, Rava A. Numerical simulation of the performance of solid oxide fuel
cell with different flow channel geometries. Energy 2017;119:235e44.
https://doi.org/10.1016/j.energy.2016.12.074.
[9] Mehran MT, Lim T-H, Lee S-B, Lee J-W, Park S-J, Song R-H. Long-term per-
formance degradation study of solid oxide carbon fuel cells integrated with a
steam gasifier. Energy 2016;113:1051e61. https://doi.org/10.1016/j.energy.
2016.07.087.
[10] Xenos DP, Hofmann P, Panopoulos KD, Kakaras E. Detailed transient thermal
simulation of a planar SOFC (solid oxide fuel cell) using gPROMSTM. Energy
2015;81:84e102. https://doi.org/10.1016/j.energy.2014.11.049.
[11] Chitsaz A, Hosseinpour J, Assadi M. Effect of recycling on the thermodynamic
and thermoeconomic performances of SOFC based on trigeneration systems;
A comparative study. Energy 2017;124:613e24. https://doi.org/10.1016/
j.energy.2017.02.019.
[12] Peters R, Riensche E, Cremer P. Pre-reforming of natural gas in solid oxide
fuel-cell systems. J Power Sources 2000;86:432e41. https://doi.org/10.1016/
S0378-7753(99)00440-1.
[13] Priyakorn F, Laosiripojana N, Assabumrungrat S. Modeling of solid oxide fuel
cell with internal reforming operation fueled by natural gas. J Sustain Energy
Environ 2011;2:187e94.
[14] Liso V, Olesen AC, Nielsen MP, Kær SK. Performance comparison between
partial oxidation and methane steam reforming processes for solid oxide fuel
cell (SOFC) micro combined heat and power (CHP) system. Energy 2011;36:
4216e26. https://doi.org/10.1016/j.energy.2011.04.022.
[15] Mogensen D, Grunwaldt J-D, Hendriksen PV, Nielsen JU, Dam-Johansen K.
Methane steam reforming over an Ni-YSZ solid oxide fuel cell anode in stack
configuration. J Chem 2014;2014:1e8. https://doi.org/10.1155/2014/710391.
[16] Ishizuka M, Rudy TM, Wadekar VV, Conferences E, Yoshida H, Iwai H. Thermal
management in solid oxide fuel cell systems. Eng Conf Int 2005. CHE2005e01.
[17] Meusinger J, Riensche E, Stimming U. Reforming of natural gas in solid oxide
fuel cell systems. J Power Sources 1998;71:315e20. https://doi.org/10.1016/
S0378-7753(97)02763-8.
[18] Klein J-M, Bultel Y, Georges S, Pons M. Modeling of a SOFC fuelled by
methane: from direct internal reforming to gradual internal reforming. Chem
Eng Sci 2007;62:1636e49. https://doi.org/10.1016/j.ces.2006.11.034.
[19] Timmermann H, Sawady W, Reimert R, Ivers-Tiffe e E. Kinetics of (reversible)
internal reforming of methane in solid oxide fuel cells under stationary and
APU conditions. J Power Sources 2010;195:214e22. https://doi.org/10.1016/
j.jpowsour.2009.07.019.
[20] Mogensen D, Grunwaldt J-D, Hendriksen PV, Dam-Johansen K, Nielsen JU.
Internal steam reforming in solid oxide fuel cells: status and opportunities of
kinetic studies and their impact on modelling. J Power Sources 2011;196:
25e38. https://doi.org/10.1016/j.jpowsour.2010.06.091.
[21] Baldinelli A, Barelli L, Bidini G. Performance characterization and modelling of
syngas-fed SOFCs (solid oxide fuel cells) varying fuel composition. Energy
2015;90:2070e84. https://doi.org/10.1016/j.energy.2015.07.126.
[22] Hajimolana SA, Hussain MA, Soroush M. Wan daud WMA, chakrabarti MH.
Modeling of a tubular-SOFC: the effect of the thermal radiation of fuel com-
ponents and CO participating in the electrochemical process. Fuel Cells
2012;12:761e72. https://doi.org/10.1002/fuce.201200056.
[23] Janardhanan VM, Heuveline V, Deutschmann O. Performance analysis of a
SOFC under direct internal reforming conditions. J Power Sources 2007;172:
296e307. https://doi.org/10.1016/j.jpowsour.2007.07.008.
[24] Hussain MM, Li X. ID. Mathematical analysis of planar solid oxide fuel cells.
J Power Sources 2006;161:1012e22. https://doi.org/10.1016/j.ijhydene.2008.
02.043.
[25] Ni M. 2D thermal-fluid modeling and parametric analysis of a planar solid
oxide fuel cell. Energy Convers Manag 2010;51:714e21. https://doi.org/
10.1016/j.enconman.2009.10.028.
[26] Ho TX, Kosinski P, Hoffmann AC, Vik A. Numerical analysis of a planar anode-
supported SOFC with composite electrodes. Int J Hydrogen Energy 2009;34:
3488e99. https://doi.org/10.1016/j.ijhydene.2009.02.016.
[27] Aguiar P, Adjiman CS, Brandon NP. Anode-supported intermediate tempera-
ture direct internal reforming solid oxide fuel cell. I: model-based steady-state
performance. J Power Sources 2004;138:120e36. https://doi.org/10.1016/
j.jpowsour.2004.06.040.
[28] Amedi HR, Bazooyar B, Pishvaie MR. Control of anode supported SOFCs (solid
oxide fuel cells): Part I. mathematical modeling and state estimation within
one cell. Energy 2015;90:605e21. https://doi.org/10.1016/j.energy.2015.
07.095.
[29] Li PW, Chyu MK. Simulation of the chemical/electrochemical reactions and
heat/mass transfer for a tubular SOFC in a stack. J Power Sources 2003;124:
487e98. https://doi.org/10.1016/j.jpowsour.2003.06.001.
[30] Haberman BA, Young JB. Three-dimensional simulation of chemically reacting
gas flows in the porous support structure of an integrated-planar solid oxide
fuel cell. Int J Heat Mass Transf 2004;47:3617e29. https://doi.org/10.1016/
j.ijheatmasstransfer.2004.04.010.
[31] Hofmann P, Panopoulos KD, Fryda LE, Kakaras E. Comparison between two
methane reforming models applied to a quasi-two-dimensional planar solid
oxide fuel cell model. Energy 2009;34:2151e7. https://doi.org/10.1016/
j.energy.2008.09.015.
[32] Petruzzi L, Cocchi S, Fineschi F. A global thermo-electrochemical model for
SOFC systems design and engineering. J Power Sources 2003;118:96e107.
https://doi.org/10.1016/S0378-7753(03)00067-3.
[33] Suwanwarangkul R, Croiset E, Entchev E, Charojrochkul S, Pritzker MD,
Fowler MW, et al. Experimental and modeling study of solid oxide fuel cell
operating with syngas fuel. J Power Sources 2006;161:308e22. https://
doi.org/10.1016/j.jpowsour.2006.03.080.
[34] Iwai H, Yamamoto Y, Saito M, Yoshida H. Numerical simulation of
intermediate-temperature direct-internal-reforming planar solid oxide fuel
cell. Energy 2011;36:2225e34. https://doi.org/10.1016/j.energy.2010.03.058.
[35] Ni M. Modeling of SOFC running on partially pre-reformed gas mixture. Int J
Hydrogen Energy 2012;37:1731e45. https://doi.org/10.1016/
j.ijhydene.2011.10.042.
[36] Andersson M, Yuan J, Sunde n B. SOFC modeling considering hydrogen and
carbon monoxide as electrochemical reactants. J Power Sources 2013;232:
42e54. https://doi.org/10.1016/j.jpowsour.2012.12.122.
[37] Lin B, Shi Y, Ni M, Cai N. Numerical investigation on impacts on fuel velocity
distribution nonuniformity among solid oxide fuel cell unit channels. Int J
Hydrogen Energy 2015;40:3035e47. https://doi.org/10.1016/
j.ijhydene.2014.12.088.
[38] Lin B, Shi Y, Cai N. Numerical simulation of cell-to-cell performance variation
within a syngas-fuelled planar solid oxide fuel cell stack. Appl Therm Eng
2017;114:653e62. https://doi.org/10.1016/j.applthermaleng.2016.12.014.
[39] Izzo JR, Peracchio AA, Chiu WKS. Modeling of gas transport through a tubular
solid oxide fuel cell and the porous anode layer. J Power Sources 2008;176:
200e6. https://doi.org/10.1016/j.jpowsour.2007.10.075.
[40] Jin X, Xue X. Mathematical modeling analysis of regenerative solid oxide fuel
cells in switching mode conditions. J Power Sources 2010;195:6652e8.
https://doi.org/10.1016/j.jpowsour.2010.04.018.
[41] Zhu H, Kee RJ. A general mathematical model for analyzing the performance
of fuel-cell membrane-electrode assemblies, vol. 117; 2003. p. 61e74. https://
doi.org/10.1016/S0378-7753(03)00358-6.
[42] Shi Y, Cai N, Li C, Bao C, Croiset E, Qian J, et al. Modeling of an anode-supported
Ni-YSZjNi-ScSZjScSZjLSM-ScSZ multiple layers SOFC cell. Part I. Experiments,
model development and validation. J Power Sources 2007;172:235e45.
https://doi.org/10.1016/j.jpowsour.2007.04.037.
[43] Lee S, Park M, Kim H, Yoon KJ, Son JW, Lee JH, et al. Thermal conditions and
heat transfer characteristics of high-temperature solid oxide fuel cells inves-
tigated by three-dimensional numerical simulations. Energy 2017;120:
293e305. https://doi.org/10.1016/j.energy.2016.11.084.
[44] Murthy S, Fedorov AG. Radiation heat transfer analysis of the monolith type
solid oxide fuel cell. J Power Sources 2003;124:453e8. https://doi.org/
10.1016/S0378-7753(03)00732-8.
[45] VanderSteen JDJ, Pharoah JG. Modeling radiation heat transfer with partici-
pating media in solid oxide fuel cells. J Fuel Cell Sci Technol 2006;3:62.
https://doi.org/10.1115/1.2134738.
[46] Mejri I, Mahmoudi A, Abbassi MA, Omri A. Radiation heat transfer effect in
solid oxide fuel Cell . Appl Lattice Boltzmann Method 2014;8:642e6.
[47] Wang Y, Zeng H, Shi Y, Cao T, Cai N, Ye X, et al. Power and heat co-generation
by micro-tubular flame fuel cell on a porous media burner. Energy 2016;109:
117e23. https://doi.org/10.1016/j.energy.2016.04.095.
[48] Modest MF. Radiative heat transfer. second ed. New York, USA: Academic
press; 2003.
[49] Bessette II Norman F, Wepfer William J. Prediction of on-design and off-design
performance for a solid oxide fuel cell power module. Energy Convers Manag
1996;37:281e93.
[50] Rashid K, Keun S, Ali R, Hwan S. Optimization of manifold design for 1 kW-
class fl at-tubular solid oxide fuel cell stack operating on reformed natural
gas. J Power Sources 2016;327:638e52. https://doi.org/10.1016/
j.jpowsour.2016.07.077.
[51] Lehnert W, Meusinger J, Thom F. Modelling of gas transport phenomenan in
SOFC anodes. J Power Sources 2000;87:57e63.
[52] Nagata S, Momma A, Kato T, Kasuga Y. Numerical analysis of output charac-
teristics of tubular SOFC with internal reformer. J Power Sources 2001;101:
60e71. https://doi.org/10.1016/S0378-7753(01)00547-X.
[53] Arpino F, Carotenuto A, Massarotti N, Nithiarasu P. A robust model and nu-
merical approach for solving solid oxide fuel cell (SOFC) problems. Int J Numer
Methods Heat Fluid Flow 2008;18:811e34. https://doi.org/10.1108/
09615530810898971.
[54] Welty JR, Wicks CE, Wilson RE, Rorrer GL. Fundamentals of momentum, heat,
and mass transfer. fifth ed. John Wiley Sons, Inc.; 2007.
[55] Suwanwarangkul R, Croiset E, Fowler MW, Douglas PL, Entchev E. Douglas M
a. Performance comparison of Fick's, dusty-gas and Stefan-Maxwell models to
predict the concentration overpotential of a SOFC anode. J Power Sources
2003;122:9e18. https://doi.org/10.1016/S0378-7753(02)00724-3.
[56] Costamagna P, Selimovic A, Del Borghi M, Agnew G. Electrochemical model of
the integrated planar solid oxide fuel cell (IP-SOFC). Chem Eng J 2004;102:
61e9. https://doi.org/10.1016/j.cej.2004.02.005.
[57] Achenbach E. Three-dimensional and time-dependent simulation of a planar
solid oxide fuel cell stack, vol. 49; 1994. p. 333e48.
[58] Damm DL, Fedorov AG. Radiation heat transfer in SOFC materials and com-
ponents. J Power Sources 2005;143:158e65. https://doi.org/10.1016/
j.jpowsour.2004.11.063.
[59] David L, Damm AGF. Spectral radiative heat transfer analysis of the planar
SOFC. In: ASME 2004 int. Mech. Eng. Congr. Expo. Anaheim, California, USA:
ASME; 2004. p. 399e405. https://doi.org/10.1115/IMECE2004-60142.
[60] Hayre RPO, Prinz FB. Fuel cell fundamentals. John Wiley Sons; 2006.
[61] Guide AFT. Ansys Fluent Tutorial Guide. Ansys INC n.d.;15.0.
[62] Fischer K, Seume JR. Location and magnitude of heat sources in solid oxide
 K. Rashid et al. / Energy 141 (2017) 673e691
fuel cells. J Fuel Cell Sci Technol 2009;6:11002. https://doi.org/10.1115/
1.2971042.
[63] Navasa M, Andersson M, Yuan J, Sund n B. Transport phenomena in solid
oxide fuel cell electrodes focusing on heat transfer related to chemical re-
actions. J Phys Conf Ser 2012:395. https://doi.org/10.1088/1742-6596/395/1/
012086.
[64] Andersson M, Yuan J, Sundn B. SOFC modeling considering electrochemical
reactions at the active three phase boundaries. Int J Heat Mass Transf 2012;55:
773e88. https://doi.org/10.1016/j.ijheatmasstransfer.2011.10.032.
691
[65] Tanaka T, Inui Y, Urata A, Kanno T. Three dimensional analysis of planar solid
oxide fuel cell stack considering radiation. Energy Convers Manag 2007;48:
1491e8. https://doi.org/10.1016/j.enconman.2006.11.019.
[66] Autissier N, Larrain D, Van Herle J, Favrat D. CFD simulation tool for solid oxide
fuel cells. J Power Sources 2004;131:313e9. https://doi.org/10.1016/
j.jpowsour.2003.11.089.
[67] Sanjay Tushar C. Thermodynamic assessment of SOFC-ICGT hybrid cycle:
energy analysis and entropy generation minimization. Energy 2017;134:
1013e28. https://doi.org/10.1016/j.ijhydene.2017.04.277.