Dynamic Relaxation: A New Optimization Method for

PEM Fuel Cell Analysis

Shirin Espiari
PhD Student, School of Engineering
Kings College, University of Aberdeen
Aberdeen, United Kingdom
[email protected]
Majid Aleyaasin
Lecturer, School of Engineering
Kings College, University of Aberdeen
Aberdeen, United Kingdom
[email protected]

Abstract Sustainability of the membrane conductance in PEM
fuel cells is an important issue and determines the performance
of the cell. Herein, the relevant numerical analysis is replaced
with an optimization procedure. The objective function is energy
balance equation and the equality constraints are governing
equations of the physical model. In this paper a dynamic
relaxation method is proposed which provides a fictitious
nonlinear dynamic system model for the fuel cell. Thereafter, by
assuming appropriate initial conditions, the steady state solution
can be obtained. For any current density, the corresponding
temperature in cathode and anode, output voltage of the cell and
stack temperature are determined. It is shown that by changing
the initial conditions and also the integration method, the
convergence into the same steady state values can be guaranteed.
It is concluded that dynamic relaxation method which is well
known technique in solid mechanics can also be a powerful
optimization method for studying the equilibrium in fuel cells.
The advantages of this method compared with other techniques
are discussed and commented upon.
Keywords: PEM fuel cell, Dynamic relaxation, Water
management, Numerical solution.
TABLE I. NOMENCLATORS
Variab
le
Definition
Variab
le
Definition
A
Heat exchange area per
unit length (cm)
h
Channel width
(cm)

a Activity H
Enthalpy of overall
reaction (J/mol)
C
Concentration at
membrane surface
(mol/cm
3
)
Hvap
Enthalpy of water
evaporation
(J/mol)
Cp
Mass heat capacity
(J/gK)
kc
Condensation rate
constant (s
-1
)
d Channel height (cm) K
Heat conduction
coefficient (w/cm
k)
D

Diffusion coefficient of
water in
membrane(cm
2
/s)
L
Channel
length(cm)
e
Membrane area per unit
length (cm)
H

Molar flow (mol/s)
F
Cross section of solid
(cm
2
)
nd
Electro-osmotic
coefficient
F Faradays constant

tm(cm)
Membrane
thickness
I
Current density
I0
Exchange current
density

Ratio of water
molecules per proton
flux
Mm,dry
Membrane dry
weight (g/mol)
Variab
le
Definition
Variab
le
Definition
U
convective heat transfer
coefficient(w/cm
2
k)
p
Density (g/cm
3
)
p
Over
Potential I
0c
Open circuit
voltage (v)
TABLE II. NOMENCLATORS FOR SUBCRIPT AND SUPERCRIPT
Subscript Supercript
a anode
c cathode
inf surronding
oc open circuit
s solid
cool coolant
L liquid
Sat saturation
v vapour
avg average
w water

I. INTRODUCTION
Proton exchange membrane fuel cells as zero-emission
power generator device have gained too much attention for the
next generation. Low operating temperature, fast start up,
simplicity of its design and its potential for low weight and
volume are the most advantage features of PEM fuel cells.
They produce electricity through the electrochemical reaction.
[1]. Water and thermal management inside the cell has great
impact on fuel cell output voltage and its power density. As the
membrane conductivity is strongly dependent on its
humidification level, applying proper water management could
minimize the ohmic voltage loss during fuel cell operation
especially at high current density [2]. So it is important to study
the interaction of the heat and water transport on fuel cell
performance in both steady state and dynamic condition.
Several attempts have been made in literature to predict the
steady-state behavior of PEM fuel cell during its operation by
solving the governing equations along the length of the flow
channels. The steady state characteristics of fuel cell could be
formulated as a highly nonlinear system of equations. Neguyen
and White have investigated the effect of different
humidification design on the current density, temperature and
water distribution along the length of the flow channel in
steady-state condition. Their result showed that sufficient
humidification level at anode gas stream is necessary to
improve membrane conductivity especially at high current
densities. They used iterative methods to solve the nine
nonlinear differential equations to predict nine unknowns in
their model based on the guessed value of the cell voltage for a
2010 IEEE Electrical Power & Energy Conference
978-1-4244-8188-0/10/$26.00 2010 IEEE
given current density. In such a system the solution procedure
consists of numerous iteration until the calculated current
density converges to the specified current density [3].
Difficulties will arise when convergence is not achievable. Yi
and Nguyen developed a steady state model which predicts
temperature and water distribution along the length of the flow
channel. They also investigated the effect of heat exchanger
designs on the fuel cell operating parameters. They concluded
that humidification of the anode gas and liquid water injection
at the anode gas stream improves the membrane conductivity.
The set of twelve governing equations with twelve unknown
were solved in their model based on Newton-Raphson method
[4]. Later, Zong and Sobiesiak in another work, developed
steady-state model to predict water, temperature, reactant and
current density distribution along the length of the flow
channel. Their solution procedure again was based on iterative
method [5]. On the simulation of dynamic variation of
operating parameters along the length of the flow channels, a
transient model has been developed by Golbert and lewin, in
which their solution procedure also was based on iterative
method. In their solution the cell voltage was guessed for a
given current density, and iteratively manipulated until the
calculated current density converged to the set current density
[6]. As can be seen from the literature, iterative methods
always have been used in solving the fuel cell equations. This
is because in such systems there are ten nonlinear governing
equations for ten unknown H

H
2
, H

0
2
, H

N
2
, N

w,a
I
, N

w,c
I
, N

w,a
v
,
N

w,c
v
, I
u,
, I
c
onJ I
s
that must be solved. These unknown
variables as well as other parameters such as activity of
water (o
w
) or net water flux from anode side to the cathode ()
are all dependent to each others, so the convergence of the
iterative method which is based on initial guess value is not
easy task. In this article in order to avoid the complexity in
solving such nonlinear system of equations by iterative
methods, initially the equations which expresses heat and
energy balance is converted into a fictitious nonlinear dynamic
system and the rest of equations are assumed as equality
constraints. The steady state solution then evolved from the
integration procedure versus fictitious time. Therefore we have
changed the solution procedure to a new type of optimization
problem in which minimization achieved by dynamic
relaxation method. It should be remembered that our algorithm
in this paper is a new solution procedure and it is not related to
the optimization studies in fuel cells.
In solid mechanic the dynamic relaxation is a solution
strategy (not an optimization method) which is based on the
second-order Richardson method. In this method dynamic
transient analysis is used to find the static solution of a system
[7]. Although Dynamic Relaxation method is very popular
method in solving solid mechanics problem [8], but in our
research considered as an optimization method in which the
variables are x
1
= I
u,
x
2
= I
c,
x
3
= I
s,
and x
4
= I
ccII
. we
have shown that by integrating the fictitious dynamical
equation related to the above variables, the equilibrium (steady
state solution) can be achieved. Regardless of the type of the
integration (variable step Rosenbrock method or Runge-kutta
method) the time history solution are always approached into a
unique steady state value. This proves the effectiveness of our
new method.
II. MODEL DESCRIPTION
The simulated model herein is two dimensional which
accounts for water transport across the membrane, heat transfer
from the fluid in the channels to the solid parts, heat released
from the electrochemical reaction and water condensation-
evaporation in the modeling domain. The membrane, anode
and cathode flow channels are considered as the solid part in
this model. The model is built on the following assumption: 1-
There is not pressure drop along the channel. 2-Gaseous
specious obey the ideal gas law. 3-Liquid water exists at the
surface of the channels in the form of the small droplets and
their volumes are negligible. 4-Water enters the electrode and
membrane just in the vapor form. 5-Anode and cathode
electrodes are assumed ultra thin so the effect of diffusion in
porous layer of electrodes is neglected. 6- Due to the high
conductivity of the bipolar plates voltage loss in the flow
channel is neglected [3, 4]. The membrane electrode assembly
and flow channels are shown in Fig.1.
The rates of the hydrogen and oxygen consumption in the
anode and cathode flow channels are expressed as follows:
H

H
2
= -
hL
2P
I
ug
. (1)
H

0
2
= -
hL
4P
I
ug
. (2)
The equality constraint regarding the liquid water flow rate
in anode and cathode are given in (3) and (4). The mass of
liquid water changes in the flow channels as the result of the
water vapor condensation. When the water vapor partial
pressure is greater than the saturated pressure, water vapor will
condense into liquid form [6].
f
1
(x
1
, x
2
, x
3
) = N

w,a
I
-_
k
c
hdL
R(T
a
+273)
_
M

w,c

w,c

+M

H2
P
a
- P
w
sat
(T
a
)__ =
u. (3)
f
2
(x
1
, x
2
, x
3
) = N

w,c
I
- _
k
c
hdL
R(T
c
+273)
_
M

w,c

v,c
v
+M

O2
+M

N2
P
c
-
P
w
sat
(T
c
)_ = u. (4)

Figure 1. Schematic of model regions[3].
Similarly the corresponding equality constraint regarding
the water vapor mass flow rate at the cathode and anode are
given by (5), (6). At the cathode side (6) water vapor mass
increases as a result of the electrochemical reaction as well as
water transportation by proton (electro-osmotic drag). However
at the anode side (5) water vapor mass decreases due to the
water vapor transport to the cathode side by electro-osmotic
drag. Other terms in (5) and (6) are related to condensation and
evaporation of water in the flow channels.

f
3
(x
1
, x
2
, x
3
) = N

w,u

- _-
k
c
hdL
R(T
a
+273)
_
M

w,c

w,c

+M

H2
P
u
-
P
w
sut
(I
u
) -
hL
P
I
ug
_ = u. (5)

f
4
(x
1
, x
2
, x
3
) = N

w,c
v
- _-
k
c
hdL
R(T
c
+273)
_
M

w,c

v,c
v
+M

O2
+M

N2
P
c
-
P
w
sut
(I
c
) +
hLu
P
I
ug
+
hL
2P
I
ug
_ = u. (6)
In (5) and (6), is the ratio of water molecules per proton
flux which are transported from anode side of the membrane to
the cathode side [3, 6]. The second term on the equality
constraint (5) is the rate of water vapor mass change due to
evaporation or condensation of water vapor which depends on
the water vapor partial pressure and saturated pressure and the
third term represents the rate of water vapor mass depletion at
the anode channel due to the water transport from anode to the
cathode side by electro-osmotic drag. The third term in (6) is
the accumulation rate of water in cathode which comes from
the anode side by electro-osmotic drag and the last term is the
rate of water vapor generation in cathode due to the
electrochemical reaction [3].

the net water molecule per
proton flux is calculated as follow:
o = n
d
-
P
I
cg

w
c
w,c
-c
w,c
t
m
. (7)
Where, n
d
is the electro-osmotic drag coefficient of water
(number of water molecules carried by a proton) and D
w
is
diffusion coefficient of water.
n
d
= u.uu49 + 2.u24o
u
- 4.SSo
u
2
+4.u9o
u
3
. or o
u
1 (8)
n
d
= 1.S9 + u.1S9(o
u
- 1). or o
u
> 1 (9)

w
= n
d

0
exp j2416 [
1
303
-
1
273+1
s
[. (10)
C
w,k
=
p
m,d
H
m,d
(u.u4S + 17.8o
k
- S9.8o
k
2
+S6.uo
k
3
).
For o
k
1 (11)
C
w,k
=
p
m,drj
M
m,drj
|14 + 1.4(o
k
-1)]. For o
k
> 1 (12)
o
a
=
x
v,a
P
P
v,a
sat
= _
M

v,a
V
M

v,a
V
+M

H2
]
P
P
v,a
sat
. (13)
o
c
=
X
w,c
P
w,c
sct
P = _
M

w,c
v
M

w,c

+M

O2
+M

N2
]
P
P
w,c
sct
. (14)
log(P
w,k
sut
) = 2.9S 1u
-2
I
k
- 9.18 1u
-5
I
k
2
+1.44
1u
-7
I
k
3
- 2.18. (15)
Further equality constraints are energy balance of gas
stream in both anode and cathode flow channels. Anode and
cathode gas stream temperature changes as a result of the
convection heat transfer between the fluid in anode and cathode
flow channel and the stack, and heat gain or loss from
condensation or evaporation of water [5]. The energy balance
equation in anode and cathode is modeled according to (16)
and (17).
f
5
(x
1
, x
2
, x
3
) = (N

H2
C
p,H2
+ H

w,u

C
p,H2O
)(T
a
- T
0a
) -
|u
g
AI(I
s
- I
u
) + E
w,u
up
- H

w,u
I
| = u. (16)
f
6
(x
1
, x
2
, x
3
) = (N

N2
C
p,N2
+ H

w,c

C
p,H2O
+ N

O2
C
p,O2
) (T
c
-
T
0c
) - |u
g
AI(I
s
- I
c
) + E
w,c
up
- H

w,c
I
|. (17)
WhereC
p,w

, C
P,H2
, C
P,02
onJ C
P,N2
are calculated as follow:
C
p,w

= SS.46 +6.88 1u
-3
I
k
+ 7.6 1u
-6
I
k
2
-S.S9
1u
-9
I
k
3
. (18)
C
P,H2
= 28.84 +7.6S 1u
-5
I
k
+ S.29 1u
-6
I
k
2
-
8.7u 1u
-10
I
k
3
. (19)
C
P,02
= 29.1u +1.16 1u
-3
I
k
- 6.u8 1u
-6
I
k
2
+
1.S1 1u
-9
I
k
3
. (20)
C
P,N2
= 29.uu + 2.2u 1u
-3
I
k
+ S.72 1u
-6
I
k
2
-
2.87 1u
-9
I
k
3
. (21)
Where u
g
is the convective heat-transfer coefficient. A is
the heat transfer area per unit length of the flow channel and L
is the length of the channel. Subscript k stands for either anode
or cathode. The second term on the equality constraints (16)
and (17) accounts for the convection heat transfer between
fluid and the solid part of the flow channel. The third term
represents heat of evaporation or condensation of water which
is absorbed or released in the anode and cathode flow channels
depending on the condition in the flow channel at different
temperatures [4, 5]. The governing equation is:
E
w,k

- E
w,k
I
= 4Su7u - 41.94I
k
+ S.44 1u
-3
I
k
2
+2.S4
1u
-6
I
k
3
- 8.98 1u
-10
I
k
4
. (22)
The cell voltage as a function of current density can be
calculated as follow [9].
v
ccII
= v
oc
- -
I
cg
t
m
o
m
. (23)
Where V
oc
is the open circuit voltage. At reversible
operating condition when the net current is equal to zero, the
output voltage of the cell is 1.1 (V). In practice the output cell
voltage is less than the reversible voltage and it decreases with
increasing operating current density. The difference between
the reversible voltage and practical output voltage is due to
over-potentials.
I
cg
t
m
o
m
is the ohmic over-potential in fuel cell
which is mainly due to the ionic resistance in the electrolyte. .
is the cathode over-potential. It should be remembered that
anode over-potential is low and negligible but the cathode over
potential is [9]:
=
R(T
s
+273)
0.5F
ln_
I
cg
I
0 P
O2
]. (24)
The membrane conductivity is calculated as a function of
the membrane water content at the anode side of the membrane
and temperature of the solid part [3, 6].

m
= _u.uuS14
M
m,dy
p
m,dy
C
w,a
-u.uuS26]exp [
1
303
-
1
273+T
s
. (25)
Where C
w,u
is the membrane water concentration at the
anode side which is calculated from (11) and (12).
Finally, the objective function is expressing the energy
balance equation, which relates fuel cell stack temperature to
the several factors such as heat released as a result of the
electrochemical reaction, heat transferred by convection to the
surrounding atmosphere, heat removed by coolant from the
stack, heat transferred from the fluid in the channel to the solid
layer of the cell and heat released or consumed by
condensation and evaporation of water in the flow channels.
Equation 26 describes the stack energy balance equation. [5, 6,
10, 11]
TABLE III. VALUES FOR PARAMETERS USED IN THE MODELING
Variable Value Variable Value
Acool(cm) 0.4 H

H2in(mol/s) 1.510-5
Agas(cm) 0.4 H

O2in(mol/s) 110
-5

Uinf (w/cm
2
k) 0.25 H:

w,a,in(mol/s) 110
-5

Mm,dry(g/mol) 1100 H

l
w,a,in(mol/s) 0
Ucool (w/cm
2
k) 0.025 H

v
w,c,in(mol/s) 0.510
-6

e(cm) 0.1 H

l
w,c,in(mol/s) 0
d (cm) 0.1 Tc,in(C) 75C
f(cm
2
) 0.0106 Ta,in(C) 75C
L(cm) 10 Tinf(C) 70C
tm(cm) 0.01275 Tin,s(C) 70C
h (cm) 0.1 P
u
(ot) 1
H (J/mol) -2.4110
5
P
c
(ot) 1
p
s
(
g
cm
3
) 0.2 p
m
(
g
cm
3
) 0.2

0
(
cm
3
s
)

5.51u
-7


kc
100
v
oc
(v) 1.1 C
ps
(
[
gK
) 1

F(x
1
, x
2
, x
3
, x
4
) =
0
g
A
g
]
(I
u
+ I
c
- 2I
s
) +
0
cccl
A
cccl
]
(I
cooI
-
I
s
) -
c
]
[
H
2P
+ I
ccII
I
ug
+
1
]L
E
up
(I
s
) - (H
w,u
I
-H
w,u,n
I
+
H
w,c
I
-H
w,c,n
I
) -
20
in]
L
(I
s
-I
n]
) = u. (26)
III. FICTITIOUS DYNAMICAL FORMULATION
This stage depends on physical knowledge of the system
under investigation. In these circumstances an unsteady model
of some physical variables are constructed fictitiously and the
governing equations are converted to a multivariable fictitious
nonlinear dynamic. Herein we have converted the equations
(16) and (17) into the following dynamical form:

d1
c
dt
=
I
S
(x
1
,x
2
,x
3
)
[M

H2
C
,H2
t+C
,H2O
] M

w,c

dt
t
0

. (27)

d1
c
dt
=
I
6
(x
1
,x
2
,x
3
)
[M

N2
C
,N2
t+M

O2
C
,O2
t+C
,H2O
] M

w,c

dt
t
0

. (28)

And the objective function (26) can be converted into the
dynamical form as follows:

d1
s
dt
=
F(x
1
,x
2
,x
3
,x
4
)
p
s
C
ps
. (29)
Then instead of minimization procedure which is a
formidable task, the nonlinear dynamic equations (27)-(29) are
represented in fictitious time by considering equations (1), (2),
(7)-(15) and (18)-(25). They will be integrated by advanced
stiff integrators like Rosenbrock variable step method in
MATLAB under the command ode23s and Runge-Kutta
method under the command ode45. Any initial conditions can
be assumed and the final steady state value will be calculated.
Then cell voltage can be calculated from (23) which also
converges to the steady state value. Both integration methods
give similar steady states values and the details are discussed in
the next section.
IV. RESULT AND DISCUSSION
For determination of the anode temperature our initial guess
is 75C, in the average current densityI = u.S (Acm
2
). This is
an overestimated guess, therefore this overestimation is
recognized by dynamical equation (27) and temperature
dropped to 72.6C which is shown in Fig. 2. Moreover, Fig. 2
shows that when current density changes toI = 1 (Acm
2
) the
75C is an underestimated guess and therefore the anode
temperature rises up to 76.2C. In fact this increase is due to
the heat released from water vapor condensation as well as the
heat from the electrochemical reaction [5]. The history of
evolution is shown in Fig. 2. In each current density two
integration methods are performed the first is ode23s which is
shown by circle and star corresponding toI = u.S anu 1 (A
cm
2
). And the second is Runge-Kutta method shown by solid
and dashed line corresponding toI = u.S anu 1 (Acm
2
). As
can be seen from the figure for both methods the time history
solution is approached into a unique steady state value for
anode gas stream temperature.

Figure 2. Anode temperature approaching equiblirium after 10 second
fictitious time

Figure 3. Cathode temperature approaching equilibrium after 10 seconds
fictitious time
The cathode gas stream temperature for two operating
current load is shown in Fig. 3. Initial guess for cathode gas
temperature is 75C which is overestimated in I = u.S (A
cm
2
). Therefore it drops to 71.8C. But this is not the final
value. It takes time until it reaches to the steady state value of
72.6C. During operation at current load of 1(A/cm
2
), the
initial guess is 75C which is underestimated but surprisingly
the temperature first drops to 72.7C but finally rises to the
steady state value of 76.2C. The relevant time history is not
purely mathematical and somehow represents the unsteady
physics of the system. For example in Fig. 3, the initial
temperature drop is due to the heat transfer phenomena and the
subsequent temperature rise is because of heat released as a
result of electrochemical reaction. Fig. 4 plots the predicted
stack temperature for two operating current densities. As can
be seen from Fig. 4, stack temperature increases quickly from
initial temperature of 70C up to72.6C atI = u.S (Acm
2
).
Moreover during operation at I = 1 (Acm
2
) , stack
temperature increases sharply from initial temperature of 70C
up to 76.2C. In practice this increase is related to the high
inlet temperature of the anode and cathode gaseous stream and
heat of the electrochemical reaction [3, 10].

Figure 4. Sack temperature approaching equilibrium after 10 second
fictitous time


The output cell voltage corresponding to two operating
average current density of 0.5 and 1 (A/cm
2
) is shown in Fig. 5.
It is observed that, the output cell voltage quickly reaches to the
steady-state value of 0.73 (V) during operation at the average
current density of 0.5(A/cm
2
). And the corresponding steady
state value during operation at average current density ofI =
1 (Acm
2
), is 0.64 (V). The results show if water content of
the membrane is sufficient, the conductance of the membrane
is high enough, therefore the ohmic voltage loss is minimized.
Applying proper water management for different operating
current density keeps ohmic voltage drop ideally linear with
increasing current density [9]. By increasing in load current,
more water molecules are transported to the cathode side, if
inlet water supply is insufficient corresponding to the load
current, membrane water content at the anode side decreases.
Decrease in membrane water content results in low membrane
conductivity and high ohmic voltage loss.
V. CONCLUSION
In this research, for analysis of the PEM fuel cell
characteristics, which is a highly nonlinear phenomena, a new
optimization procedure is employed as an easy alternative to
the complicated solution procedure. For the minimization of
the objective function a new method called dynamic
relaxation is employed. This method is based on a fictitious
multivariable nonlinear dynamic system by which the steady
0 2 4 6 8 10
70
72
74
76
78
80
time sec
a
n
o
d
e

t
e
m
p
e
r
a
t
u
r
e

C


Rosenbrock,Iavg=0.5
Runge Kutta,Iavg=0.5
Rosenbrock,Iavg=1
Runge Kutta,Iavg=1
0 2 4 6 8 10
71
72
73
74
75
76
77
time sec
c
a
t
h
o
d
e

t
e
m
p
e
r
a
t
u
r
e

C


Rosenbrock,Iavg=0.5
Runge Kutta,Iavg=0.5
Rosenbrock,Iavg=1
Runge Kutta,Iavg=1
0 2 4 6 8 10
66
68
70
72
74
76
time sec
s
t
a
c
k

t
e
m
p
e
r
a
t
u
r
e

C


Rosenbrock,Iavg=0.5
Runge Kutta,Iavg=0.5
Rosenbrock,Iavg=1
Runge Kutta,Iavg=1
0 2 4 6 8 10
0.5
0.55
0.6
0.65
0.7
0.75
0.8
time sec
c
e
l
l

v
o
l
t
a
g
e

(
v
o
l
t
)


Rosenbrock,Iavg=0.5
Runge Kutta,Iavg=0.5
Rosenbrock,Iavg=1
Runge Kutta,Iavg=1
Figure 5 . Cell voltage approaching equilibrium after 2 second
fictitious time
state solution of the governing equations can be achieved. This
steady state solution is identical to the solution of the system of
nonlinear equations. Since the steady state solution can be
achieved easily, then we can eliminate the complexity of the
iterative solutions by choosing this alternative. We have
succeeded to determine the values of the stack, anode and
cathode temperatures and output cell voltage very accurately.
The minimization procedure is replaced with integration of
nonlinear ODEs. We have used two types of numerical
integrations, first one is stiff variable step, Rosenbrock
method and the 2
nd
is Runge-kutta method. It is shown that
both integrators provide the same equilibrium steady state
values. The advantage of this method is that it does not need
any iterative procedure. Instead it uses integration in fictitious
time domain which is easy and always convergent. Moreover,
satisfaction of the inequality constraints (positivity of the
physical parameters) is also embedded in the integration
procedure. However, other optimization techniques require
either direct search or gradient based search strategies which
are lengthy procedures. Since in PEM fuel cell studies the
dynamic relaxation is matched with unsteady physics of the
system, it is always convergent and powerful technique.
ACKNOWLEDGMENT
We thank The University of Aberdeen, College of Physical
Science, for providing the required facilities for this research
and their financial support.
REFERENCES
[1] S. M. Sharifi, S. Roshanzamir, M. H. Eikani, Modeling and
simulation of the steady state and dynamic behaviour of a PEM
fuel cell, Energy, Vol. 35, pp. 1633-1646, 2010.
[2] D. P. Wilkinson, H. H. Voss, K. Prater, Water management and
stack design for solid polymer fuel cell, J. of power source, Vol.
49, pp. 117-127, 1994.
[3] T. V. Neguyen and R. E. White, A Water and Heat Management
Model for Proton-Exchange-Membrane Fuel Cells, J.
Electrochem. Soc, Vol. 140, No. 8, pp. 2178-2186, Aug. 1993.
[4] J. S. Yi and T. V. Nguyen, An Along-the-Channel Model for
Proton Exchange Membrane Fuel Cells, J. Electrochem. Soc,
Vol. 145, No. 4, pp. 1149-1159, April 1998.
[5] Y. Zong, B. Zhou and A. Sobiesiak, Water and thermal
management in a single PEM fuel cell with non-uniform stack
temperature, J. of power source, Vol. 161, pp.143-159, 2006.
[6] J. Golbert, D. R. Lewin, Model-based control of fuel cells: (1)
regulatory control, J. of power source , Vol. 135, pp. 135-151,
2004.
[7] M. kadkhodayan, J. Alamatian, G. J. Turvey, A new fictitious
time for the dynamic relaxation (DXDR) method, In. J. for
numerical methods in engineering, Vol. 74, pp. 996-1018, 2007.
[8] A. S. Day, An introduction to dynamic relaxation, The
engineer, Vol. 219, pp. 218-221, 1965.
[9] N. Yousfi-steiner, Ph. Mocoteguy., D. Candusso, D.Hissel,
A.Hernandez , A. Aslanides, A review on PEM voltage
degradation associated with water management: Impacts, influent
factors and characterization. Journal of power sources, Vol. 183,
pp. 260-274, 2008.
[10] M. J. Khan and M. T. Iqbal, Modeling and Analysis of
Electrochemical, Thermal and Reactant Flow Dynamics for a
PEM Fuel Cell System, Fuel Cells, Vol. 5, No. 4, pp. 463-475,
2005.
[11] C. Berger, Handbook of Fuel Cell Technology, Prentice-Hall Inc.
Englewood Cliffs, New Jercy, 1968.