Transient heat transfer modeling of a solid oxide fuel cell

operating with humidied hydrogen

C. Ozgur Colpan
a,
*, Feridun Hamdullahpur
b
, Ibrahim Dincer
c
a
Mechanical and Aerospace Engineering Department, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario, Canada K1S 5B6
b
Mechanical and Mechatronics Engineering Department, University of Waterloo, 200 University Avenue West, Waterloo, Ontario,
Canada N2L 3G1
c
Faculty of Engineering and Applied Science, University of Ontario Institute of Technology, 2000 Simcoe Street North, Oshawa,
Ontario, Canada L1H 7L7
a r t i c l e i n f o
Article history:
Received 16 August 2010
Received in revised form
6 November 2010
Accepted 26 November 2010
Available online 3 January 2011
Keywords:
Efciency
Solid oxide fuel cell
Hydrogen
Transient heat transfer
Fuel utilization
Finite difference method
a b s t r a c t
This paper presents the development of a new transient heat transfer model of a planar
solid oxide fuel cell (SOFC) operating with humidied hydrogen. The model is rst validated
with some benchmark test data and then used to simulate the transient behavior of the co-
and counter-ow SOFCs at the heat-up and start-up stages. In addition, a parametric study
including the effects of Reynolds number at the fuel channel inlet and excess air coefcient
on the output parameters is conducted. The model predictions are found to be in very good
agreement with the data published in the literature. The transient simulations show that
counter-ow SOFC yields higher performance, e.g. power density and electrical efciency,
but it needs slightly more time to reach the steady-state conditions. The results of the
parametric study point out that taking the Reynolds number low and excess air coefcient
high gives higher electrical efciencies for both of the congurations. For the given input
data, it is found that the counter-ow conguration has a higher electrical efciency for
low Reynolds numbers, e.g. 0.67 and all possible excess air coefcients.
Copyright 2010, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights
reserved.
1. Introduction
The solid oxide fuel cell is one of the emerging energy tech-
nologies that is primarily used to generate electrical power
from the movement of electrons produced by electrochemical
reactions in the cell. In addition, the heating value of gas
streams exiting the SOFC is high enough to be recovered using
bottoming cycles to produce additional heat and/or electricity.
SOFCs have advantages over other fuel cell types. These
advantages include [1]: a) no water management issues, since
only solid and gas phases exist, b) cheaper materials used for
manufacturing electrocatalysts, c) ability to utilize a variety of
fuels including hydrocarbons, methanol and biomass
produced gas, d) internal reforming of gases, and e) thermal
integration with bottoming cycles, e.g. gas turbine and gasi-
cation systems. Possible disadvantages of the SOFC over
other fuel cells are: a) challenges for construction and dura-
bility due to its high temperature, and b) a carbon deposition
problem when a fuel such as methane or syngas is used. The
main application of the SOFC is stationary power and heat
generation. Other areas include transportation, military and
portable applications.
SOFC models can be developed at cell, stack and system
levels. When developing a SOFC model in cell and stack levels,
* Corresponding author.
E-mail address: cocolpan@connect.carleton.ca (C.O. Colpan).
Avai l abl e at www. sci encedi r ect . com
j our nal homepage: www. el sevi er . com/ l ocat e/ he
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9
0360-3199/$ e see front matter Copyright 2010, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2010.11.127
different considerations may be taken into account according
tothepurposeandneedsof themodel. For example, 0-D, 1-D, 2-
Dand 3-Dmodeling approaches can be used depending on the
necessity for knowledge of the output parameters, such as
temperature andcurrent density distributions [2e4]. Transient
modeling is used when one, or a combination of the following
stages, needs to be simulated: heat-up, start-up, shut-down
and load change [5e8]. Thermomechanical modeling provides
us withinformationtopredict the stresses occurring inside the
fuel cell [9]. The problems related to carbon deposition due to
using a fuel containing carbon can also be analyzed through
numerical studies [10e13]. In system level modeling, inte-
grated SOFC systems can be assessed using several methods
including energy, exergy and thermoeconomic analyses
[14e17]. More information on different types of SOFC models
can be found in the paper by Colpan et al. [1].
The transient behavior of the cell during the heat-up and
start-up stages plays an important role in the successful
operation of the fuel cell. In the literature, a few studies that
model these stages to nd the characteristics of a SOFC can be
found. For example, Colpan et al. [18] modeled a direct
internal reforming SOFC operating with a multi-gas mixture
including CH
4
, CO, CO
2
, H
2
, H
2
O, and N
2
. The effects of the
reforming process on the performance of the cell were
included in the model. The heat-up, start-up, and steady state
behaviors of the cell were investigated. In addition, the rst
principal thermal stresses were calculated to nd the proba-
bility of failure of the cell during its operation. Ki and Kim [19]
developed a model to predict the thermal dynamics of planar
SOFCs. Two methods for stack-heating (hot air through
cathode channels and electric heating inside a furnace) were
investigated. Barzi et al. [20] developed a 2-D transient model
to analyze the start-up behavior of a tubular SOFC. Their
model gives the cell voltage, the electromotive force, and the
variables such as pressure, temperature and species concen-
tration during the start-up as the output. In addition, the cell
heat-up rate for hot inlet gases as well as the start-up time of
the SOFC were calculated.
The main objective of the current study is to develop
a comprehensive model for a planar SOFC operating with
humidied hydrogen taking into account all three heat
transfer mechanisms, (i.e. conduction, convection and radia-
tion), and all polarization nodes, (i.e. ohmic, activation and
concentration). This study differs from the other transient
modeling studies in the literature in terms of the modeling
equations and strategy used in the heat-up and start-up
stages, and the selection of the input and output parameters.
This model is used for the simulation of the transient behavior
of the co- and counter-ow SOFCs that operate using the
humidied hydrogen. In addition, a parametric study is con-
ducted to assess the effects of Reynolds number at the fuel
channel inlet and excess air coefcient on the output
parameters.
2. Modeling
This section describes the modeling approach, the formula-
tion of the SOFC including the continuity and heat transfer
equations, and the validation of the model.
2.1. Modeling approach
The rst step in the modeling of a SOFC is the formulation of
the system considered together with the specication of the
control volumes, and the coordinates. For this reason, the
repeat element of a SOFCfound in the middle of a stack, which
is shown in Fig. 1, is divided into ve control volumes: anode
interconnect, fuel channel, PEN (consisting of anode, electro-
lyteandcathode), air channel andcathodeinterconnect. Dueto
the symmetrical conditions of the considered repeat element
in the stack, the solid structure has adiabatic boundary
conditions at the exterior surfaces [9]. The Cartesian coordi-
nate system is selected for all the control volumes given their
specic geometry. Then, the general laws, e.g. conservation of
mass, energy and momentum, and the particular laws, e.g. the
relation between the cell voltage and polarizations, and the
initial and boundary conditions are written for each of these
control volumes. In modeling, instead of solving the conser-
vation of momentum, some simplications are made
assuming fully developed laminar ow conditions. This
assumption is well justied since the gases ow with low
velocity, whichis requiredtoobtainahighfuel utilizationratio.
Under these ow conditions, the Nusselt number becomes
a single function of the aspect ratio for rectangular ducts. This
derivation is based on solutions of the differential momentum
and energy equations for different boundary conditions [21].
There is a discrepancy in the literature about how some of
the input and output parameters of SOFC models are selected.
Parameters such as average current density, fuel utilization
ratio, cell voltage and mass owrate of the channel inlets may
be chosen as input or output according to the purpose of the
model. In ourmodel, the cell voltage, which is assumed to be
equal at the top and bottom surfaces of the interconnect, the
Reynolds number at the fuel channel inlet that controls the
fuel mass ow rate, and the excess air coefcient that deter-
mines the mass ow rate at the air channel inlet are taken as
input parameters. Other input parameters selected in this
study are: the cell geometry, the properties of materials, the
ambient temperature, the molar composition at the fuel and
air channel inlets, the mass ow rate of air for the heat-up
stage, and the cell pressure. The expected outcome parame-
ters of the model are: the heat-up and start-up time, the fuel
utilization ratio, the current density, the temperature and
molar gas composition distributions, and the power output
and electrical efciency of the cell.
In this study, the strategy followed for the modeling of the
heat-up and start-up stages is as follows. In modeling the
Fig. 1 e Schematic of the repeat element of a SOFC.
i nt e r na t i o na l j our na l o f hy d r og e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11489
heat-up period, only the heat transfer equations are solved
since there is no fuel ow taking place in the fuel channel. At
this stage, the temperature of the air channel is controlled so
as not to cause excessive thermomechanical stresses. The
minimum solid temperature of the cell is calculated for each
time step, and the air channel inlet temperature is set to
T
min,solid
100

C for the subsequent time step [22]. In
modeling the start-up period, the temperatures of the air and
fuel channel inlets are kept constant, and mass balances are
rst solved for air and fuel channels for each time step. The
molar ow rates and the composition of the gas species, and
current density distribution through the gas channels are
determined after the mass balances are obtained. Using these
data and the temperature distribution obtained from the
previous time step, heat transfer equations are applied to each
control volume. Considered in the heat transfer equations are
conduction between PEN and interconnects, natural convec-
tion in the heat-up stage and forced convection in the start-up
stage, surface-to-surface radiation between the PEN and
interconnects. Hence, the temperature distribution in a given
time step is calculated, and the iterations are repeated until
the absolute temperature difference between the two
consecutive time steps for each node becomes less than the
threshold value. This value is chosen as 10
4
in this study.
Among the different numerical solution methods, the
nite difference method is used in this study because it is
straightforward for orthogonal grids. In applying this method,
spatial and temporal domains are divided into several
sections, which is called meshing. After generating the mesh,
nite difference approximations are substituted for the
derivatives to convert the partial differential equations to an
algebraic form. Then, a computer code that is capable of
solving the system of equations in an efcient way for
different input parameters is developed. In this study, the
code was developed in Matlab.
The nal step of modeling is validation. In this study,
because of the lack of experimental results in the literature,
the results of the SOFC benchmark test [23] and Brauns model
[24] are used for validating the model.
2.2. Formulation of the SOFC
Continuity equations and transient heat transfer equations
are applied to the control volumes enclosing the components
of the cell, e.g. gas channels, PEN, and interconnects. These
equations are shown below for the co-ow conguration.
In the continuity equations, the source terms are derived
from the reaction of oxidation of hydrogen. Based on this
reaction and neglecting the transient terms, the continuity
equations at the fuel channel are shown in Eqs. (1) and (2), and
these equations at the air channel are shownin Eqs. (3) and (4).
d _ n
00
H
2
dx

_ r
00
el
t
fc
(1)
d _ n
00
H
2
O
dx

_ r
00
el
t
fc
(2)
d _ n
00
O
2
dx

_ r
00
el
=2
t
ac
(3)
d _ n
00
N
2
dx
0 (4)
where the rate of conversion for electrochemical reaction
becomes _ r
00
el

i
2F
.
The local current density is found by solving the relation
between the Nernst voltage and the polarizations using
V V
N
V
ohm
V
act
V
con
(5)
where the Nernst voltage and the polarizations canbe givenas
[25,26]:
V
N

Dg

r
2F

RT
2F
$ln
_
P
H
2
O
P
H
2
$

P
O
2
=P
+
_
_
V
ohm

_

k
r
el;k
$t
k
_
$i
V
act

RT
F
$sinh
1
_
i
2i
o;a
_

RT
F
$sinh
1
_
i
2i
o;c
_
V
conc

RT
s
2F
ln
_
1
RT
2F
$
s
a
t
a
D
a
V
va
P
b
H
2
i
_

RT
s
2F
ln
_
1
RT
2F
$
s
a
t
a
D
a
V
va
P
b
H
2
O
i
_

RT
s
4F
ln
_

_
P
b
O
2
P
_
PP
b
O
2
_
exp
_
RT
4F
$
s
c
t
c
D
c
V
vc
P
i
_
_

_
The electrical resistivity of the cell components, r, can be
written as functions of the solid temperature as given in [27].
The current density, the molar ow rate and molar composi-
tion of the gas species through the gas channels are found by
solving the equations given above.
The transient heat transfer equation for the air channel
becomes
r
ac
$c
p;ac
$
vT
vt

v
vx
_
_ n
00
O
2
h
O
2
_ n
00
N
2
h
N
2
_

h
c;a
T
PEN
T
a
h
c;a
T
ci
T
a

_
_ r
00
el
=2
_
$h
O
2
$w
solid
=w
gas
t
ac
(6)
with the following boundary and initial conditions:
x 00
T f t Heat up
T T
w ac
Start up
; t 00T T
o
100
o
C
Here, it should be noted that we have used a 1-D modeling
approach for the air and fuel channels. Hence, the convection
heat transfer and enthalpy inow/outow terms appear as
a source term in these equations.
The 2-D transient heat diffusion equation for the PEN is
given as
v
2
T
vx
2

v
2
T
vy
2

_ r
00
el
$
_
h
H
2
h
O
2
=2 h
H
2
O
_
i$V
cell
k
PEN
$t
PEN

1
a
PEN
$
vT
vt
(7)
with the following boundary and initial conditions:
x 0 and x L0
vT
vx
0
y t
ci
t
ac
0k
PEN
$
vT
vy

w
gas
w
solid
$h
c;a
$T
PEN
T
a
h
r;a
$T
PEN
T
ci

_
1
w
gas
w
solid
_
$k
ci
$
T
PEN
T
ci

t
ac
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11490
y t
ci
t
ac
t
PEN
0k
PEN
$
vT
vy

w
gas
w
solid
$
_
h
c;f
$
_
T
PEN
T
f
_
h
r;f
$T
PEN
T
ai

_
1
w
gas
w
solid
_
$k
ai
$
T
PEN
T
ai

t
fc
t 00T T
o
It should be noted that we have neglected the effects of the
ow in the porous media in the equation shown above.
The transient heat transfer equation for the fuel channel is
r
fc
$c
p;fc
$
vT
vt

v
vx
_
_ n
00
H
2
h
H
2
_ n
00
H
2
O
h
H
2
O
_

h
c;f
_
T
ai
T
f
_
h
c;f
_
T
PEN
T
f
_

_
_ r
00
el
$
_
h
H
2
h
H
2
O
_
$w
solid
=w
gas
t
fc
(8)
with the following boundary and initial conditions:
x 00T T
w fc
Start up; t 00T T
o
The 2-D transient heat diffusion equation for the cathode
and anode interconnects is
v
2
T
vx
2

v
2
T
vy
2

1
a
i
$
vT
vt
(9)
with the following boundary and initial conditions:
x 0 and x L0
vT
vx
0; y 0 and y t
ci
t
ac
t
PEN
t
fc
t
ai
0
vT
vy
0
y t
ci
and y t
ci
t
ac
t
PEN
t
fc
0k
i
$
vT
vy

w
gas
w
solid
$
_
h
c;g
$
_
T
i
T
g
_
h
r;g
$T
i
T
PEN

_
1
w
gas
w
solid
_
$k
i
$
T
i
T
PEN

t
gc
t 00T T
o
Here, g should be replaced with f and a; i should be replaced
with ai and ci; and gc should be replaced with fc and ac in Eq.
(9), when modeling the anode and cathode interconnects,
respectively. In the above equations, h
c
is the heat transfer
coefcient. It represents the natural convection and forced
convection in the heat-up and start-up stages, respectively.
The correlations for such heat transfer coefcients are avail-
able elsewhere [21,28].
Dimensionless numbers: The Reynolds number for the fuel
channel inlet based on the hydraulic diameter of the rectan-
gular channel cross section is shown in Eq. (10). In the model,
this number is considered as one of the input parameters.
Hence, usingthisnumber, themassowrateof thegasmixture
per cross section of the fuel channel at the inlet can be found.
Re
D
h

_ m
00
fi
$
_
2$t
fc
$w
gas
_
m
mix
$
_
t
fc
w
gas
_ (10)
Eq. (10) can also be written in terms of molar owrate of the
gas species at the fuel channel inlet as follows:
Re
D
h

_ n
00
k;fi
$M
mix
$
_
2$t
fc
$w
gas
_
x
k;fi
$m
mix
$
_
t
fc
w
gas
_ (11)
where k denotes H
2
and H
2
O.
The excess air coefcient used as an input parameter in the
model is dened as the amount of the oxygen in the inlet
stream divided by the amount of oxygen that is needed for
a stoichiometric reaction. This coefcient, as given in Eq. (12),
is used to calculate the molar ow rate of the gas species per
cross sectional area of the air channel at the inlet of the cell.
l
air

_ n
00
O
2
;ai
_
_ n
00
H
2
;fi
=2
_$
t
ac
t
fc
(12)
Output parameters: The main output parameters of the
model are: the fuel utilization ratio, the power density, the
power output, and the electrical efciency of the cell.
The fuel utilization ratio is dened as the amount of
hydrogen that is electrochemically reacted to the amount of
hydrogen in the inlet stream as follows:
U
F

m
i2
_
_ r
00
el
_
i
$Dx$w
solid

_ n
00
H
2
;fi
$w
gas
$t
fc
(13)
where m denotes the number of nodes in ow direction.
(1,1)
(1,p)
(1,p+1)
(1,p+2)
(1,p+3)
(1,r)
(1,r+1)
(1,r+2)
(1,r+3)
(1,s)
(m,1)
(m,p)
(m,r)
(m,s)
(m,p+1)
(m,p+2)
(m,p+3)
(m,r+1)
(m,r+2)
(m,r+3)
Cathode Interconnect
Air channel
PEN
Fuel channel
Anode Interconnect
Lcell
y
x
Fig. 2 e Numbering scheme for the nite difference
solution of the repeat element of a SOFC.
Table 1 e Input data used in the benchmark test.
Variable Value
Cell geometry
Active area [mm
2
] 100 100
Anode thickness [m] 50 10
6
Cathode thickness [m] 50 10
6
Electrolyte thickness [m] 150 10
6
Channel width [mm] 3
Channel height [mm] 1
Rib width [mm] 2.42
Total thickness (with ribs) [mm] 2.5
Material properties
Density of PEN and interconnects [g/cm
3
] 6.6
Specic heat of PEN and interconnects [J/gK] 0.4
Thermal conductivity of PEN
and interconnects [W/cmK]
0.02
Operating parameters
Temperature at the fuel channel inlet [K] 1173
Temperature at the air channel inlet [K] 1173
Pressure of the cell [bar] 1
Excess air coefcient 7
Fuel utilization 0.85
Mean current density [A/m
2
] 3000
Gas composition at the air channel inlet 21% O
2
, 79% N
2
Gas composition at the fuel channel inlet 90% H
2
, 10% H
2
O
i nt e r na t i o na l j our na l o f hy d r og e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11491
The power density and power output of the SOFC are given
below:
_
W
00
SOFC
i
c;ave
$V
cell
(14)
_
W
SOFC

_
W
00
SOFC
$L
cell
$w
solid
(15)
The primary purpose of using SOFC is to generate electricity
and its performance can be assessed through electrical ef-
ciency of the cell as follows:
h
el

_
W
SOFC
LHV$
P
2
k1
_ n
00
k;fi
$t
fc
$w
gas
(16)
where k denotes the types of gases found at the fuel channel
inlet, i.e. H
2
and H
2
O.
2.3. Numerical solution
The schematic of the 2-Dcross sectionof a repeat element and
the numbering scheme for the nodes used in the numerical
solution are showninFig. 2. As canbe seenfromthe gure, the
repeat element has ms nodes. The lengthof the cell, and the
thicknesses of the cathode interconnect, PEN, and anode
interconnect aredividedintom-1, p-1, r-p-3, ands-r-3parts. An
implicit nite different scheme is used for the solution of the
heat transfer equations. The nite difference equations for the
boundaryconditions aretakenas secondorder accurate. These
equations are derived by considering an imaginary node
outside the control volume and eliminating this node between
the general equation for interior nodes and the boundary
equation. The details of this approachcanbe found in[29]. The
set of equations are linearized by using the lagging properties
by one time step method [29], and solved using the Gauss
elimination method. Iterations are repeated until the solid
temperaturereachesacertainvaluefor theheat-upperiod, and
the system reaches steady state for the start-up period.
2.4. Validation
The results of the benchmark test, which was conducted at
a workshop organized by the International Energy Agency in
1994 [23], was used to validate the model. In this benchmark
test, nine institutions modeled planar SOFCs with the same
operating data. These institutions were: KFA-Julich (Germany),
ISTIC, University of Genova (Italy), ECN Petten (Holland), Riso,
National Laboratory (Denmark), Eniricerche (Italy), Dornier
(Germany), Statoil (Norway), Ife-Kjeller (Norway) and Siemens
(Germany). The mainassumption used inthe test was to accept
each of the polarizations in the anode and cathode as equal to
the ohmic loss of the electrolyte. These models were developed
under steady state conditions. The input data for the bench-
mark test are given in Table 1. In another study, Braun [24]
developed a steady state model using the same input data
Table 2 e Cell voltage for the benchmark Test-I.
Company/Institution Co-ow [V] Counter-ow [V]
Dornier, D 0.684 0.689
ECN Petten, NL 0.704 N.A.
Eniricerche, I 0.722 0.730
Inst. For Energiteknikk Kjeller, N 0.710 0.710
KFA-Julich, D 0.706 0.712
Siemens, D 0.712 0.716
Statoil, N 0.702 0.709
Riso, DK 0.703 0.710
Table 3 e Validation of maximum and minimum values
of current density.
Company/Institution Co-ow
(max/min) (A/m
2
)
Counter-ow
(max/min) (A/m
2
)
Dornier, D 3636/1686 7192/1297
ECN Petten, NL 3614/1211 N.A.
Eniricerche, I 3840/1020 8970/1080
Inst. For Energiteknikk
Kjeller, N
3933/1191 7862/1113
KFA-Julich, D 3725/1237 7910/1163
Siemens, D 3863/1236 8513/1135
Statoil, N 3956/1366 7391/1235
Riso, DK 3739/1296 7107/1187
Brauns model 3799/1211 7393/1152
Model-V1 3760/1187 7564/1202
Model-V2 5175/1175 5530/1586
Table 4 e Validation of maximum and minimum values
of solid temperature.
Company/Institution Co-ow (max/min)
(

C)
Counter-ow
(max/min) (

C)
Dornier, D 1070/928 1085/914
ECN Petten, NL 1082/899 N.A.
Eniricerche, I 1069/916 1083/906
Inst. For Energiteknikk
Kjeller, N
1058/930 1084/912
KFA-Julich, D 1059/913 1073/906
Siemens, D 1049/909 1062/904
Statoil, N 1098/970 1082/913
Riso, DK 1061/924 1075/910
Brauns model 1059/924 1073/910
Model-V1 1049/903 1056/904
Model-V2 1043/907 1054/906
Table 5 e Validation of air and fuel channel outlet
temperatures.
Company/Institution Co-ow (air/fuel)
(

C)
Counter-ow
(air/fuel) (

C)
Dornier, D 1068/1070 1080/914
ECN Petten, NL 1082/1082 N.A.
Eniricerche, I 1068/1068 1080/906
Inst. For Energiteknikk
Kjeller, N
1055/1058 1073/912
KFA-Julich, D 1059/1059 1070/906
Siemens, D 1048/1048 1061/1064
Statoil, N 1067/1067 1082/914
Riso, DK 1059/1061 1070/910
Brauns model 1058/1059 1068/910
Model-V1 1048/1047 1051/900
Model-V2 1042/1043 1051/900
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11492
and same assumptions as the benchmark test. In addition to
the data given in Table 1, we have considered the data needed
for calculating the activation and concentration polarizations
in the model as follows. Exchange current densities of anode
and cathode are taken as 0.53 A/cm
2
and 0.2 A/cm
2
, respec-
tively; diffusivity of the anode and cathode are taken as
0.91 cm
2
/s and 0.22 cm
2
/s, respectively; porosity of the anode
and cathode are both taken as 0.5; and tortuosity of the anode
and cathode are both taken as 4.
3. Results and discussion
3.1. Validation
In this study, two models, using different assumptions, have
been developed for a co-ow and counter-ow SOFC. A tran-
sient heat transfer model was rst developed using the same
assumption for polarizations as the benchmark tests (i.e.
activation and concentration polarizations were included
with a simplication). This model is called Model-V1. In the
second model, the assumption used in Model-V1 is altered in
that different analytical equations are considered for ohmic,
activation and concentration polarizations, as givenin Section
2.2. This model is called Model-V2. There are some differences
in the input parameters of this model and the benchmark test.
Unlike the input parameters used in the benchmark test, fuel
utilization and mean current density are taken as output
parameters, but the cell voltage and Reynolds number are
taken as input parameters in the present models, i.e. Model-
V1 and Model-V2. Since the results of the benchmark tests are
given in steady state condition, the model is validated for this
condition.
The heat transfer model is simulated for the benchmark
test-1 conditions. Anodal analysis is rst carriedout to ndthe
number of nodes that will make the results independent from
the grid size. It was found that taking 750 nodes in the ow
direction, 15 nodes inthe thickness directionandthe time step
as 1 s is sufcient to obtain grid-independent results.
For validating the present models, the input parameters
were rst calibrated. As discussed before, cell voltage is
considered as an input parameter in the present models and
not in the benchmark tests. The results for the cell voltage for
the benchmark test are given in Table 2. From these results,
we chose the cell voltage as 0.7 V for the co-owand 0.71 V for
the counter-ow case. Average current density and fuel utili-
zation are input parameters in the benchmark tests and their
values are given as 0.3 A/cm
2
and 0.85, respectively. To get
results closer to these values, the Reynolds number is found to
be 0.67 in Model-V1. The same value for Reynolds number is
used in Model-V2.
Maximum and minimum values for the current density,
solid temperature and air and fuel channel outlet tempera-
tures are given in Tables 3e5, respectively. FromTable 3, it can
be seenthat the current density, found by different companies
and institutions, is between 1020 A/m
2
and 3956 A/m
2
for the
co-ow case, and 1080 A/m
2
and 8970 A/m
2
for the counter-
ow case. It can be seen from this table that the results for
Model-V1 are between these values. When we take the
average of the maximum and minimum current densities
found by the companies and institutions that participated in
the benchmark test, and compare these average values with
the results of Model-V1, it was found that the relative error for
l
l
l
l
l
l
Fig. 3 e Comparison of current density distribution found
using the Model-V1 and Model-V2 with the benchmark test
(ECNs data).
l
fl
l
l fl
fl
fl
fl
Fig. 4 e Comparison of temperature distribution in the fuel channel found using Model-V1 and Model-V2 with the
benchmark test (ECNs data).
i nt e r na t i o na l j our na l o f hy d r og e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11493
the maximumcurrent density is 0.78% and 3.02%, and that for
the minimum current density is 7.22% and 2.64% for co-ow
and counter-ow cases, respectively. The same procedure is
followed for the solid temperature and air and fuel channel
outlet temperatures, which are given in Tables 4 and 5,
respectively. It was found that only the maximum solid
temperature for the counter-owcase is not inthe range given
in Table 4. It is 0.57% lower than the bottom limit for the
maximum solid temperature. This result is mainly due to the
difference in modeling between Model-V1 and the benchmark
test. In Model-V1 for counter-ow conguration, the outlet
temperature for the fuel channel and the inlet temperature for
the air channel are xed to obtain a uniform temperature
distribution. The inlet temperature of the fuel channel and the
outlet temperature for the air channel were calculated.
However, it is not clear how the inlet and outlet temperatures
for the gas channels were calculated in the models by the
companies and institutions that participated in the bench-
mark test. The same procedure has also been followed for the
maximum and minimum solid temperature, and air and fuel
channel outlet temperature. It is found the relative error is less
than 2.3% for all cases.
When we checked the results for Model-V2 from Tables
3e5, it was seen that except for the current density
l
l
l
l
l
l
Fig. 5 e Comparison of molar hydrogen fraction
distribution in the fuel channel found using the Model-V1
and Model-V2 with the benchmark test (ECNs data).
Fig. 6 e Temperature distribution for co-ow conguration at different time: (a) 397 s, (b) 794 s (end of heat-up stage),
(c) 1503 s, (d) 4143 s (end of start-up).
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11494
distribution, the results are comparable with the results of the
benchmark test and Model-V1. The difference in the results
for current density distribution between Model-V1 and Model-
V2 is as expected since the models in the benchmark tests
were developed using an assumption on polarizations, as
discussed in Section2.4. However, this assumptionis not valid
today. Detailed correlations have been published on the acti-
vation and concentration polarizations in the literature, [e.g.
[25,26]]. However, the temperature distribution is still
comparable between Model-V2 and the benchmark test-1. For
this model, we have also checked the relative error for the
maximum and minimum solid temperature, and air and fuel
channel temperature. It is found that this error is less than
2.4% for all cases.
The distributions of current density, fuel channel temper-
ature and molar hydrogen fraction in the fuel channel, found
by using Model-V1 and Model-V2 for the co-ow case, are also
validated with the data published by ECN, which is an insti-
tute that participated in the benchmark test. This validation is
shown in Figs. 3e5. The distributions for the counter-ow
case, found by the companies participated in the benchmark
test, are not available in the literature, but the distributions,
found by using the present models, are added to these gures
for comparison. As can be seen from Fig. 3, current density
trends for Model-V1, and the model developed by ECN, are
Fig. 7 e Temperature distribution for counter-ow conguration at different time: (a) 397 s, (b) 794 s (end of heat-up stage),
(c) 1503 s, (d) 4143 s (end of start-up).
I I
I
I
I
I
Fig. 8 e Change of fuel utilization and current density with
time for the SOFC fueled with humidied hydrogen.
i nt e r na t i o na l j our na l o f hy d r og e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11495
similar except that the current density for Model-V1 is slightly
higher at the rst half of the cell. Model-V2 has a different
trend for both co-ow and counter-ow cases because of the
different correlations for activation and concentration polar-
izations in this model. However, when we calculate the
average current densities for Model-V1 and Model-V2, it was
found that the values are very close to the average current
density of the model developed by ECN, which is 0.3 A/cm
2
.
The average current densities for the co-ow case are 0.304
A/cm
2
and 0.294 A/cm
2
for the Model-V1 and Model-V2,
respectively; whereas, those for the counter-ow case are
0.299 A/cm
2
and 0.301 A/cm
2
for the Model-V1 and Model-V2,
respectively. When we compare the temperature distribution
in the fuel channel found by Model-V1 and Model-V2 with the
results of ECN, as shownin Fig. 4, it can be seenthat the trends
are similar. The temperature at the fuel channel exit was
found to be higher for ECN. However, when we check Table 5,
it may be seen that this temperature is comparatively higher
for ECN than for that of the other companies and institutions.
From Fig. 5, it can be seen that molar hydrogen fraction has
almost the same trend as ECN.
3.2. Transient behavior of the cell
After validating the model, the transient behaviors of the cell
at the heat-up and start-up stages are simulated. These
simulations give the change of temperature, fuel utilization,
average current density, electrical efciency, power density
and molar fraction of hydrogen with time. These simulations
are conducted for both co-ow and counter-ow cases.
In Fig. 6, temperature distributions for the co-ow case for
Model-V2 are given for the heat-up and start-up stages. In the
heat-up period, temperature at the air channel inlet is
controlled due to thermomechanical considerations, as dis-
cussed in Section 2.1. This temperature increases by 100

C
more than the minimum solid temperature at each time step.
At this stage, forced convection at the air channel, natural
convection at the fuel channel, and radiation and conduction
between the solid parts affect the temperature distribution.
The heat-up period ends when the minimum solid tempera-
ture reaches a prescribed value, which was chosen as 700

C in
this study. At this temperature, the resistivity of the electro-
lyte and the ohmic polarization become low enough to
produce meaningful amount of power. In Fig. 6, we can see
Fig. 9 e Change of electrical efciency and power density
with time for the SOFC fueled with humidied hydrogen.
a
b
Fig. 10 e Change of molar fraction of hydrogen with time
for the SOFC fueled with humidied hydrogen for
(a) co-ow case, (b) counter-ow case.
Fig. 11 e Effect of Reynolds number on the fuel utilization
and average current density.
Fig. 12 e Effect of Reynolds number on the electrical
efciency and power density.
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11496
that the temperature drops, at the x and y directions at the end
of the heat-up period, i.e. t 794 s, are approximately 5.5

C/
cm and 11.2

C/cm for an air ow rate of 0.0712 g/s. In the
start-up period, the temperatures at the air and fuel channel
inlets are xed. There is a temperature rise through the
channel length because of the heat generation due to polari-
zations; however some of this heat is carried away by the
excess air sent through the air channel. The temperature
gradients in the x and y directions at the end of the start-up
period are approximately 13

C/cm and 2.9

C/cm, respec-
tively. We compared these values with the results of our
previous study in which a multi-gas mixture was used as
a fuel in a direct internal reforming SOFC (DIR-SOFC) (15.6

C/
cm in the x direction, and 1.03

C/cm in the y direction [18]).
The results show that the temperature gradients are close to
each other for both cases.
Fig. 7 shows the temperature distribution of the counter-
ow case for Model-V2 for the heat-up and start-up stages.
In the counter-ow case, air enters the cell from the side
opposite to that of the co-ow case. Hence, the temperature
distributions for the heat-up stage, as shown in Fig. 7a and b,
are symmetrical to those shown in Fig. 6a and b. The temp-
erature gradients in the x and y directions, at the end of the
start-up period, are approximately 14.6

C/cm and 1.25

C/cm,
respectively. Whenthese values are comparedwiththe results
of the counter-ow DIR-SOFC model (7.48

C/cm in the
x direction and 1.01

C/cm in the y direction [18]), the
temperature gradient in the x direction for humidied
hydrogen case is found to be higher but that in the y direction
is found to be comparable for both cases.
It follows from Figs. 8 and 9 that the output parameters are
zero in the heat-up period since there is no ow in the fuel
channel. For the co-ow case, average current density, fuel
utilization, power density and electrical efciency increase
from 0.19 to 0.3 A/cm
2
, 0.53 to 0.83, 0.13 to 0.21 W/cm
2
, and
0.29 to 0.47, respectively, during the start-up period. The
molar ow rate of hydrogen at the exit of the fuel channel is
higher at the beginning of the start-up period compared with
the steady state condition, as can be seen in Fig. 10, because of
the higher fuel utilization of hydrogen at the beginning of the
start-up period due to the lower operating temperature.
Figs. 8e10 show that the transient behaviors for co- and
counter-ow congurations do not differ signicantly. They
show a similar trend, but the counter-ow conguration
yields higher performance and it takes slightly more time to
reach the steady state condition.
3.3. Parametric studies
The Reynolds number at the fuel channel inlet and excess air
coefcient are the key input parameters affecting the perfor-
mance of the cell. Due to this reason, the effects of these input
parameters on the output parameters such as fuel utilization,
average current density, electrical efciency, and power
density are investigated.
Figs. 11 and 12 show the effect of the Reynolds number on
the output parameters. Reynolds number is directly propor-
tional to the mass ow rate of the fuel, which is shown on the
second horizontal axis of these gures. As it can be seen from
these gures, Reynolds number should be greater than
a certain value to get any meaningful results. If we choose this
number very low, the computer code gives us imaginary
numbers as the output. From Fig. 11, it is seen that as the
Reynolds number increases, fuel utilization decreases,
whereas average current density increases, which can be
explained as follows: As the Reynolds number increase, both
molar ow rate of hydrogen and molar ow rate of hydrogen
that is utilized increase, which in turn increases the average
current density. However, since the increase in molar ow
rate of hydrogen is more than the molar ow rate of hydrogen
utilized, fuel utilization decreases. Power density has the
same trend with current density, as shown in Fig. 12; because
the cell voltage is assumed to be constant in the modeling. It
can be easily shown that electrical efciency is directly
Fig. 13 e Effect of excess air coefcient on the air channel
outlet temperature.
Fig. 14 e Effect of excess air coefcient on the fuel
utilization and average current density.
Fig. 15 e Effect of excess air coefcient on the electrical
efciency and power density.
i nt e r na t i o na l j our na l o f hy d r og e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11497
proportional to the fuel utilization; hence it has the same
trend with fuel utilization as shown in this gure. These
gures also show that counter-ow conguration has a better
performance, e.g. electrical efciency, for low Reynolds
numbers that we obtain meaningful amount of fuel utiliza-
tion, e.g. fuel utilization of 0.85. For example, for Reynolds
number 0.67, electrical efciency is 46.5% and 48.3%, for co-
ow and counter-ow congurations, respectively.
Excess air coefcient, which controls the mass ow rate of
air at the inlet of the air channel, is another important oper-
ating variable because it controls the current density and the
temperature of the fuel cell, which in turn affects the perfor-
mance of the cell. If low amounts of air is sent through the air
channel, the temperature of the exit increases, as shown in
Fig. 13. Therefore, the excess air coefcient should be carefully
selected not to cause a thermomechanical problem. It can be
seen fromFigs. 14 and 15 that the excess air coefcient should
be takenhighenoughto get a better performance fromthe cell,
whichcanbe explained as follows: As the excess air coefcient
increases, temperature of the fuel cell decreases. This decrease
causes an increase in the Nernst voltage, and decrease in the
activation andconcentrationpolarizations. Hence, the current
density and the performance of the cell increase. However, the
blower power requirement and the operation cost also
increase with an increase in the excess air coefcient. In
addition, higher exit temperature from the channels, which
necessitates lower excess air coefcient, is generally required
for the integrated SOFC systems. Hence, an optimum excess
air coefcient should be selected depending on the application
and taking into account the performance and economics.
When we compare the co-ow and counter-ow congura-
tions, Fig. 15 shows that counter-ow conguration has
a higher electrical efciencythanco-owcongurationfor any
given excess air coefcients.
4. Conclusions
A new transient, 2-D heat transfer model of a SOFC operating
with humidied hydrogen has been developed. This model
takes into account all the polarizations, (i.e. ohmic, activation
and concentration), and heat transfer mechanisms, (i.e.
conduction, convection and radiation). The relations based on
electrochemistry are coupled with the heat transfer equa-
tions. For a validation, a model using the same polarization
assumption with the benchmark test is rst developed. Then,
the model is further improved by altering this assumption and
using updated electrochemical relations on polarizations. It is
found that the results are in very good agreement with those
of the benchmark test. It is also shown that the activation and
concentration polarizations affect the current density distri-
bution signicantly. After validation, transient behaviors of
co- and counter-ow SOFC are simulated. It is found that
counter-ow SOFC has a higher fuel utilization, average
current density, power density and electrical efciency when
the system reaches steady state condition. However, this type
of SOFC needs slightly more time to reach the steady state
condition. The effects of Reynolds number at the fuel channel
inlet and excess air coefcient on the output parameters are
investigated. It is shown that the Reynolds number should be
taken low; whereas the excess air coefcient should be taken
high to get better electrical efciencies.
Acknowledgement
The nancial and technical support of an Ontario Premiers
Research Excellence Award, the Natural Sciences and Engi-
neering Research Council of Canada, Carleton University and
University of Ontario and Institute of Technology is gratefully
acknowledged.
Nomenclature
c
p
specic heat at constant pressure, J/gK
D diffusivity, cm
2
/s
D
h
hydraulic diameter, m
F Faraday constant, C
h heat transfer coefcient, W/cm
2
K
h specic molar enthalpy, J/mole
_
H enthalpy ow rate, W
i current density, A/cm
2
i
o
exchange current density, A/cm
2
k thermal conductivity, W/cm-K
L length of the cell, cm
LHV lower heating value, J/mole
M molecular weight, g/mole
_ m mass ow rate, g/s
_ n
}
molar ow rate per cross section, mole/cm
2
s
P pressure, bar
_ q heat transfer rate, W
_ r
}
conversion rate, mole/cm
2
s
R universal gas constant, J/moleK
Re
D
h
Reynolds number in an internal ow
t time, s; thickness, cm
T temperature, K
U
F
fuel utilization ratio
V voltage, V
V
v
Porosity
w width, cm
_
W power output, W
_
W
}
power density, W/cm
2
x molar concentration
Greek letters
r mass density, g/cm
3
r
el
electrical resistivity of cell components, U cm
h
el
electrical efciency
l
air
excess air coefcient
s tortuosity
m viscosity, g/s-cm
a thermal diffusivity, cm
2
/s
Subscripts
a anode; air
ac air channel
act activation
ai anode interconnect
ave average
i nt e r na t i ona l j o ur na l o f hy d r o g e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11498
c cathode; convection
ci cathode interconnect
conc concentration
e electrolyte
el electrochemical; electrical
fc fuel channel
fuel channel inlet
g gas
gc gas channel
i interconnect
ohm ohmic
mix mixture
N Nernst
PEN positive/electrolyte/negative
r reaction; radiation
s solid structure
w wall
Superscripts
b bulk
o standard state
r e f e r e n c e s
[1] Colpan CO, Dincer I, Hamdullahpur F. A review on macro-
level modeling of planar solid oxide fuel cells. Int J Energy
Res 2008;32:336e55.
[2] Colpan CO, Dincer I, Hamdullahpur F. Thermodynamic
modeling of direct internal reforming solid oxide fuel cells
operating with syngas. Int J Hydrogen Energy 2007;32:787e95.
[3] Campanari S, Iora P. Comparison of nite volume SOFC
models for the simulation of a planar cell geometry. Fuel
Cells 2005;5(1):34e51.
[4] Mandin P, Bernay C, Tran-Dac S, Broto A, Abes D, Cassir M.
SOFC modelling and numerical simulation of performances.
Fuel Cells 2006;6(1):71e8.
[5] Damm DL, Fedorov AG. Reduced-order transient thermal
modeling for SOFC heating and cooling. J Power Sources
2006;159:956e67.
[6] Apfel H, Rzepka M, Tu H, Stimming U. Thermal start-up
behaviour and thermal management of SOFCs. J Power
Sources 2006;154:370e8.
[7] Bhattacharyya D, Rengaswamy R, Finnerty C. Dynamic
modeling and validation studies of a tubular solid oxide fuel
cell. Chem Eng Sci 2009;64:2158e72.
[8] Thorud B., Stiller C., Weydahl T., Bolland O., and Karoliussen
H., Part-load and load change simulation of tubular SOFC
systems, Proc. The 6th European Solid Oxide Fuel Cell Forum,
Lucerne, Switzerland; 2004, p. 716e729.
[9] Yakabe H, Ogiwara T, Hishinuma M, Yasuda I. 3-D model
calculation for planar SOFC. J Power Sources 2001;102:
144e54.
[10] Farhad S, Hamdullahpur F. Developing fuel map to predict
the effect of fuel composition on the maximum voltage of
solid oxide fuel cells. J Power Sources 2009;191:407e16.
[11] Singh D, Hernandez-Pacheco E, Hutton PN, Patel N,
Mann MD. Carbon deposition in an SOFC fueled by tar-laden
biomass gas: a thermodynamic analysis. J Power Sources
2005;142:194e9.
[12] Koh J, Kang B, Lim CH, Yoo Y. Thermodynamic analysis of
carbon deposition and electrochemical oxidation of methane
for SOFC anodes. Electrochem Solid-State Lett 2001;4(2):
A12e5.
[13] Sasaki K, Teraoka Y. Equilibria in fuel cell gases I.
Equilibrium compositions and reforming conditions.
J Electrochem Soc 2003;150(7):A878e84.
[14] Colpan CO, Hamdullahpur F, Dincer I, Yoo Y. Effect of
gasication agent on the performance of solid oxide fuel cell
and biomass gasication systems. Int J Hydrogen Energy
2010;35:5001e9.
[15] Colpan CO, Yoo Y, Dincer I, Hamdullahpur F. Thermal
modeling and simulation of an integrated solid oxide fuel
cell and charcoal gasication system. Environ Prog Sustain
Energy 2009;28(3):380e5.
[16] Al-Sulaiman FA, Dincer I, Hamdullahpur F. Energy analysis of
a trigeneration plant based onsolid oxide fuel cell and organic
rankine cycle. Int J Hydrogen Energy 2010;35:5104e13.
[17] Farhad S, Yoo Y, Hamdullahpur F. Performance evaluation of
different congurations of biogas-fuelled SOFC micro-CHP
systems for residential applications. Int J Hydrogen Energy
2010;35:3758e68.
[18] Colpan CO, Hamdullahpur F, Dincer I. Heat-up and start-up
modeling of direct internal reforming solid oxide fuel cells.
J Power Sources 2010;195:3579e89.
[19] Ki J, Kim D. Computational model to predict thermal
dynamics of planar solid oxide fuel cell stack during start-up
process. J Power Sources 2010;195:3186e200.
[20] Barzi YM, Ghassemi M, Hamedi MH. Numerical analysis of
start-up operation of a tubular solid oxide fuel cell. Int
J Hydrogen Energy 2009;34:2015e25.
[21] Shah RK. Laminar ow forced convection in ducts: a source
book for compact heat exchanger analytical data. New York:
Academic Press; 1978.
[22] Selimovic A, Kemm M, Torisson T, Assadi M. Steady state
and transient thermal stress analysis in planar solid oxide
fuel cells. J Power Sources 2005;145:463e9.
[23] Achenbach E.,SOFC stack modelling, Final Report of Activity
A2, Annex II: Modelling and Evaluation of Advanced Solid
Oxide Fuel Cells, International Energy Agency Programme on
R, D&D on Advanced Fuel Cells, Juelich, Germany; 1996.
[24] Braun R.J., Optimal design and operation of solid oxide fuel
cell systems for small-scale stationary applications, PhD
thesis, University of Wisconsin-Madison, 2002.
[25] Kim J, Virkar AV, Fung K, Mehta K, Singhal SC. Polarization
effects in intermediate temperature, anode-supported solid
oxide fuel cells. J Electrochem Soc 1999;146(1):69e78.
[26] Chan SH, Xia ZT. Polarization effects in electrolyte/electrode-
supported solid oxide fuel cells. J Appl Electrochem 2002;32:
339e47.
[27] Bossel UG. Final report on SOFC data facts and gures. Berne,
CH: Swiss Federal Ofce of Energy; 1992.
[28] Incropera FP, Dewitt DP. Fundamentals of heat and mass
transfer. 4th ed. John Wiley & Sons; 1996.
[29] Ozisik N. Finite difference methods in heat transfer. U.S.A.:
CRC-Press; 1994.
i nt e r na t i o na l j our na l o f hy d r og e n e ne r g y 3 6 ( 2 0 1 1 ) 1 1 4 8 8 e1 1 4 9 9 11499