International Journal of Heat and Mass Transfer 77 (2014) 489500

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Elucidating the constant power, current and voltage cold start modes of
proton exchange membrane fuel cell
Yueqi Luo, Kui Jiao , Bin Jia
State Key Laboratory of Engines, Tianjin University, 92 Weijin Rd, Tianjin 300072, China

a r t i c l e

i n f o

Article history:
Received 23 June 2013
Received in revised form 8 February 2014
Accepted 28 May 2014
Available online 16 June 2014
Keywords:
Proton exchange membrane (PEM) fuel cell
stack
Constant power
Cold start
Start-up mode

a b s t r a c t
Constant power cold start mode of proton exchange membrane (PEM) fuel cell is essential in practical
applications, however, unlike the constant current and voltage cold start modes, the constant power
mode was largely ignored in previous fundamental modeling and experimental studies. In this study, a
PEM fuel cell stack cold start model for analyzing the constant power cold start process is developed,
and the fundamental differences among the various start-up modes are elucidated. In the constant power
cold start mode, the start-up process may fail before the ice fully covers the cathode catalyst layers (CLs),
because the stack may not be able to supply the required power output, which is different from the constant current and voltage start-up modes. The initial water content and start-up temperature could limit
the power output signicantly. The constant power can be controlled at higher current (CPHC) or lower
current (CPLC). In the CPHC mode, the current density decreases and the stack voltage increases during
the cold start process, while it is reversed in the CPLC mode. Generally, the CPLC mode produces less heat
during cold start process than the other start-up modes, because the current density is often kept at low
level.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Since the rst practical application of hydrogenoxygen fuel
cell stack in 1960s [1], the development of modern fuel cell has
attracted considerable attention. Proton exchange membrane
(PEM) fuel cell is a very promising type among the various types
of fuel cell. The high power density, zero/low emission and low
operating temperature are its major advantages over other kinds
of fuel cell and traditional internal combustion engine [2]. According to Mench [3], a bus powered by a PEM fuel cell stack has almost
twice the fuel efciency as a bus powered by a diesel engine. After
decades of development, PEM fuel cell stacks are applied in many
elds, such as space shuttles, automobiles and sub-water devices
[4].
Environmental adaptability is a vital issue for PEM fuel cell
stack to be widely applied in transportation and stationary applications, and start-up in cold environment is one of the problems to be
solved. Successful cold start from 20 C has been achieved in
commercialized fuel cell vehicles in recent years [5]. Even more
strict targets of cold start ability for commercial PEM fuel cell stack
have been set, e.g. successful start-up from 30 C in 60 s set by
Corresponding author. Tel.: +86 22 27404460; fax: +86 22 27383362.
E-mail address: kjiao@tju.edu.cn (K. Jiao).
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.05.050
0017-9310/ 2014 Elsevier Ltd. All rights reserved.

Ballard [6]. The target of cold start ability may be achieved through
the assist of accessory equipment in fuel cell system, and many
patents explained the various assisting methods in details [79].
The fundamental research for this problem is also important to
understand the transport processes and to reduce the power consumption and system complexity required for the successful cold
start targets.
Fundamental experimental studies for PEM fuel cell cold start
include the visualization of water freezing processes [1016], degradation tests of different components [17,18], and analysis of relationship between performance and operational parameters [19
22]. Moreover, mathematical modeling can provide insights into
the details and mechanisms of the phenomena during the cold
start processes. Mathematical modeling studies for PEM fuel cell
cold start usually concentrate on the temperature issues [23,24],
water transport and ice formation issues [2528], effect of design
parameters and operating methods [2934], and system characteristics [35]. Generally, the cold start performance predictions are
often based on the evolutions of output voltage, current density
and temperature.
Although the analysis of PEM fuel cell cold start has developed
in various aspects as mentioned above, the start-up modes adopted
in these studies are mainly constant current mode and constant
voltage mode, and models were rarely developed for constant

490

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

Nomenclature
a
A
ASR
c
Cp
D
EW
F
h
I
j
k
K
M
MEA
p
Q_
R
s
S
t
T
T0
V
CPHC
CPLC
CC
CV

water activity
cell geometric area, m2
area specic resistance, X cm2
mole concentration, mol m3
specic heat, J kg1 K1
mass diffusivity, m2 s1
equivalent weight of membrane, 1.1 kg mol1
Faradays constant, 96487 C mol1
surrounding heat transfer coefcient,
W m2 K1
current density, A cm2
exchange current density, A cm2
thermal conductivity, W m1 K1
permeability, m2
molecular weight, kg mol1
membrane electrode assembly
pressure, Pa
heat transfer rate, W
universal gas constant, J mol1 K1,
volume fraction
source terms
time, s
temperature, K
standard temperature, 298 K
voltage
constant power higher current
constant power lower current
constant current
constant voltage

Greek letters
a
transfer coefcient
e
porosity
f
water transfer rate, s1
j
electric conductivity, S m1
k
water content in ionomer
l
dynamic viscosity, kg m1 s1
n
stoichiometry ratio
q
density, kg m3

power mode. The constant power start-up mode is a traditional

and essential operating mode in practical applications, which can
be implemented through DCDC converters [3638]. In fact, PEM
fuel cell stacks are often used to power vehicles through a hybrid
system, in which the fuel cell works at steady power to operate a
generator, because fuel cell is often favored to work under a constant load for optimum performance and efciency [39,40], and
simultaneously the demand of peak power output can also be satised by a secondary system with peak power sources (e.g. batteries [41] and supercapacitors [42]). Since fuel cell vehicles are
supposed to satisfy various road conditions, PEM fuel cell stack is
required to work in composite modes, which imply the combination of different elementary modes, basically including the constant power, current and voltage modes. However, most of the
previous studies for PEM fuel cell constant power operation can
only be found for steady-state operating conditions [41,42]. Few
experimental studies related to PEM fuel cell cold start in constant
power mode were carried out. Oszcipok et al. [43] analyzed the
cold start problem of a portable PEM fuel cell through experiment
with the constant power operation involved. Datta et al. [44] studied a PEM fuel cell based power generator used in Antarctica. It is
mentioned that the DCDC converter is required to get constant

x
d

volume fraction of ionomer in catalyst layer

thickness, m

Subscripts and superscripts

a
anode
act
activation
atm
atmosphere
BP
bipolar plate
c
cathode, capillary
Cell
cell characteristic
channel ow channel
CL
catalyst layer
conc
concentration
eff
effective
env
environment
eq
equilibrium
f
frozen

uid phase
fmw
frozen membrane water
g
gas
GDL
gas diffusion layer
H2O
water
i
the ith components
ice
ice
lq
liquid water
mem
membrane
N
cell number
nernst
Nernst
nf
non-frozen
nmw
non-frozen membrane water
ohmic
ohmic
out
outlet, output
P
power
sat
saturation
sl
solid phase
Stack
stack characteristic
T
energy (for source term)
vp
water vapor
wall
surrounding wall of the stack

power output, and the constant power mode is good for charging
other devices. On the other hand, investigations on the fundamental transport phenomena during the constant power cold start processes of PEM fuel cell were not found by the authors. Therefore, to
understand the fundamental difference in the heat and mass transfer among the various cold start modes, it is necessary to develop a
cold start model for PEM fuel cell stack with the ability for the simulations of constant power, current and voltage modes.
In this study, a PEM fuel cell stack cold start model is developed
and described in details; the differences among the constant
power, current and voltage cold start modes are analyzed and elucidated; and the constant power cold start processes are comprehensively investigated under various operating conditions.
2. Model development
The physical problems considered in the PEM fuel cell stack
model are illustrated in Fig. 1, and the parameters of the stack
design and operating condition are listed in Table 1. The stacks
include certain numbers of single cells and are clamped with two
end plates. In each single cell, the components considered include
the bipolar plate (BP), gas diffusion layer (GDL), catalyst layer (CL)

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

491

(a)
Liquid water conservation in porous media
Anode CL,GDL

BPa
Energy conservation in all zones

Cathode CL,GDL

MEA

BPc

Membrane water conservation

(b)
Fig. 1. Schematic diagram of the PEM fuel cell stack model. (a) Major transport phenomenon; (b) conservation equations describing the transport phenomena.

and PEM. In the membrane electrolyte, free water and frozen water
may present, and the free water may diffuse through the membrane and be dragged by the proton transport. In the pores of CL
and GDL, ice and liquid water may appear according to the temperature and pressure. Heat is generated or consumed during operation from various sources, including the reversible heat due to
the entropy change, activational heat from the activational loss,

ohmic heat caused by the proton and electron transport, and latent
heat from the water phase change.
2.1. Governing equations
The various transport processes are simplied into one-dimension (through-plane direction, normal to the membrane) to achieve

492

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

Table 1
PEM fuel cell stack design parameters and operating conditions.
No.

Parameters

Value

1
2
3
4
5

Thickness of BP
Depth of channel; width of channel/rib
Thicknesses of membrane; CL; GDL
Effective reaction area
Densities of membrane; CL; GDL; BP

2 mm
1; 1 mm
0.03; 0.01; 0.2 mm
100 cm2
qmen;CL;GDL;BP 1980; 1000; 1000; 1000 kg m3

Specic heat capacities of membrane; CL; GDL; BP

C p mem;CL;GDL;BP 833; 3300; 568; 1580 J kg

Thermal conductivities of membrane; CL; GDL; BP

kmem;CL;GDL;BP 0:95; 1:0; 1:0; 20 Wm

Electric conductivities of CL; GDL; BP

9
10
11
12

Volume fraction of ionomer (x) in CL

Porosities (e) of CL; GDL
Stoichiometry ratio
Relative humidities of inlet gases

jCL;GDL;BP 300; 300; 20000 S m1

x = 0.3; 0.35; 0.4
eCL;GDL 0:3; 0:6

13

Inlet gas and surrounding temperatures


T in
a;c T env 20; 15; 30 C

14
15

Initial stack temperatures

Pressure at outlets

T init 20; 15; 30  C

pout
a;c 1 atm

16
17
18
19
20

Heat transfer coefcients on endplate surfaces

Heat transfer coefcients on side surfaces
Initial ice volume fraction
Initial non-frozen water content in ionomer
Initial frozen water content in ionomer

h = 100 W m2 K1

h = 2 W m2 K1
0
3, 5, 7
0

1

1

K 1

1

na,c = 2.0
RHin
a;c 0

high calculation efciency and meanwhile keep the integrality of

the major transport processes. In the electrode along the
through-plane direction, the convective transport can be neglected,
because it is only signicant along the in-plane directions caused
by strong cross ow. Fig. 1 shows the one-dimensional computational domain. Different governing equations are discretized and
solved within a certain number of grid points in the corresponding
zones. The governing equations and analytical formulations are
given in Table 2, and the corresponding transport parameters are
presented in Table 3.
The membrane water conservation equation is solved in membrane electrode assembly (MEA) (No. 1 in Table 2), the variable k
represents water content, and t (s) is time. The term on the left
hand side of this equation is the transient term, in which x is
the ionomer volume fraction in CL. The rst term on the right hand
side is the diffusion term, and the diffusion coefcient, Dnmw
(m s1), is calculated based on the correlation [45] in Table 3. It
is also corrected with the ionomer fraction in CL. The last term is
the source term, representing the water generation, electroosmotic drag effect and phase change [25]. The frozen water in
membrane (No. 2 in Table 2) appears when the total water content

is higher than the saturated value, as shown in Table 3. The conservation equation of liquid water (No. 3 in Table 2) is solved in CL and
GDL. The diffusion coefcients are calculated with the permeability
and capillary pressure [46], as presented in Table 3.
The ice volume fraction in CL or GDL is dened as the ratio of ice
volume to pore volume, and when it becomes 1, the CL or GDL is
fully blocked by ice, the reactant cannot participate in the electrochemical reaction, which is widely recognized as the major cause
of the failure of a cold start process [1116]. In this model, the
average ice volume fractions in CL and GDL are calculated directly
at each time step. In this equation (No. 4 in Table 2), Dt (s) represents the time step size, f (s1) is the phase change coefcient. As
described in the equation, ice is produced when the water content
in CL reaches saturation (given in Table 3), and the phase change
rate is also inuenced by the pore volume corrected by the accumulation of ice or liquid.
Similar simplication is made for the calculation of water vapor,
as shown in Table 2 (No. 5). Water vapor molar concentration at
the center of CL and at the center of GDL is calculated at each time
step. For the formulation of CL, the second term on the right hand
side is the phase change rate from membrane water to water

Table 2
Governing equations.
No.

Property

Conservation equation

Membrane water

qmem @xknf

Frozen water in membrane

qmem @xkf

Liquid water

Ice

Water vapor

EW

@t

mem
qEW


@ 2 x1:5 Dnmw knf

@x2

@t
EW
@eqlq slq
@ 2 qlq Dlq slq

Slq
@t
@x2
qmem
t
tDt
kCL 
sice sice f EW
Dt
ctv p;CL ct
v p;CL

tDt
Dt
Dt
ksat
 st
CL eCL 1  slq
ice M H2 O q

ice

qmem
fkCL  keq
CL EW Dt 

Dt
tDt
ct
cGDL
Deff
CL
CLGDL

Dt
ctv p;GDL ct
v p;GDL

Polarization

tDt
tDt
cCL
cGDL
Deff
Dt
CLGDL
dCL =2dGDL =2 dCL eCL

dCL =2dGDL =2

Dt
dCL eCL

Dt
ct
cchannel Deff
Dt
GDL
GDL
dGDL =2
dGDL eGDL

pH pO

2
2
0
V out V nernst V act V conc V ohmic V nernst 1:23  0:9  103  T  T 0 RT
2F ln pH O
2
0
1
!
h
i
I
I
@
AV ohmic  ASRBP ASRGDL ASRmem I ASRCL I
V act  RT
V conc RT
4Fch
111=n0:21pc
0:5
3
aF  ln
aF  ln 1 

1sice slq

Snmw

Sfmw

Energy
@
@t



 @2

qC p eff
T
fl;sl

eff

kfl;sl T

@x2

j dCL 


ST

2RTcref

eff
eff
dGDL =D
0:5dCL =D
GDL
CL

493

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500
Table 3
Transport parameters.
No.

Property

Membrane water diffusivity [45]

Saturated membrane water content [25]

Liquid water diffusivity [25]

Correlation
8
2:69266  1010 k  2
>
>
> 10
<
1
10
 exp2416303
 1T   0:87  3  knf 2:95  knf  2; 2 < knf  3
Dnmw
10
1
1
>

exp2416
10
>
303  T   2:95  4  knf 1:642454  knf  33 < knf 6 4
>
: 10
1
 exp2416303
 1T   2:563  0:33  knf 0:0264  k2nf  0:000671  k3nf knf > 4
10
8
< 4:837 if T < 223:15K
ksat 1:304 0:01479T  3:594  105 T 2 if 223:15K  T < T N
:
> knf if T P TN

4
5

m2 s1

m2 s1

dpc
Dlq  llq ds
lq

Unit

lq

13

12

Intrinsic permeabilities of CL; GDL [25]

Capillary pressures in CL and GDL [46]

KCL = 6.2  10 ; KGDL = 6.2  10

pc 35:6  2:09e44:9slq 14:41 2:09e22:2slq 7:13 in CL
pc 2395  2431e92:36slq 52:37 2431e0:0088 slq 0:005 in GDL

m2
Pa

Effective permeability of liquid water [25]

Equivalent water content and water activity
[47]

1  sice 4:0
K lq K 0 s4:0
8 lq
< 0:043 17:81a  39:85a2 36:0a3 if 0  a < 1
X p
keq 14:0 1:4a  1 if 1 < a  3
a pv p g 2slq
sat
:

m2

Effective volumetric heat capacities

J m3 K1

Effective thermal conductivity

kfl;sl eslq klq sice kice 1  slq  sice kg  1  e  xks xkmem

qC p

eff
fl;sl

eslq qlq C p lq sice qice C p ice 1  sice  slq qg C p g  1  e  xqs Cps xqmem C p mem

eff

vapor, which is calculated based on the water content in CL and the

corresponding equivalent water content [47] (dened in Table 3).
The detailed explanation is also presented in [25]. The second term
is the amount of water transferred from CL to GDL by diffusion. The
water vapor concentration in GDL is calculated with the amount of
water transferred from the CL and the amount of water transferred
to the ow channel, and the vapor concentration in ow channel is
assumed to be 0 for simplication, because non-humidied gas is
supplied and the ow rate in channel is relatively high.
The formulation for polarization loss (No. 6 in Table 2) is
derived analytically based on the basic Tafel equation [48]. The
blockage effect of reaction sites due to ice and liquid in CL is
expressed in an exponential relationship with an exponent of 0.5,
this parameter is chosen through trial and error by comparing
the model predictions and the experimental results in [19]. The
ice/liquid resistance to gas transport in CL and GDL is also considered in this model by modifying the gas diffusivity in porous
media. The detailed derivation can be found in [48], and similar
method was also used in other modeling work [49].
The energy conservation equation is solved in all domains as
shown in Fig. 1(b) and Table 2 (No. 7). The convection term is
neglected in this model, as mentioned previously, because the convective transport is only signicant along the in-plane directions.
The parameters for the unsteady term are material density q
(kg m3) and heat capacity Cp (J kg1 K1), and the parameter for
the diffusion term is the thermal conductivity k (W m1 K1).
These parameters are updated at every time step, as shown in
Table 3 (Nos. 8 and 9). Detailed explanation of the source term
can also be found in [25].
2.2. Boundary and initial conditions
Thermal boundary condition is specied on all surrounding surfaces of the stack with the heat transfer formulation:

Q_ hAT env  T wall

where h (W m2 K) is the heat transfer coefcient, A (m2) is the surface area, Tenv (K) is the environment temperature, and Twall (K) is
the wall temperature.
As three different start-up modes are simulated in this study,
three different electric boundary conditions are used in different
cases. In constant power mode, the power output boundary condition is specied for each individual cell separately:
i

FunctionPout I I  V nernst I V act I V conc I V ohmic I

W m1 K1

where I (A cm2) is the operating current density and N is the total

cell number in the simulated stack. The Nernst voltage and three
voltage losses in this equation are given in Table 2. Due to the fact
that only the total power output of the stack can be controlled, the
following nonlinear equation system needs to be solved in order to
determine the power output boundary condition on each individual
cell:

8
>
Pout
>
>
>
>
>
>
Pout
>
>
>
<...

FunctionPout I

FunctionPout I

N
>
Pout N FunctionPout I
>
>
>
>
>
N
X
>
>
>
>
Pout i Pout stack
:

where Pout_i (W) is the power output of the ith cell in the stack, and
Pout_stack (W) is the total power output of the stack. Similarly, for
constant voltage mode, the separated cell voltage boundary condition needs to be specied on each individual cell:
i

FunctionV out I V nernst I V act I V conc I V ohmic I

8
1
>
V out 1 FunctionV out I
>
>
>
>
2
>
>
> V out 2 FunctionV out I
>
>
<......
N
>
> V out N FunctionV out I
>
>
>
>
N
>X
>
>
>
V out i V out stack
:

For constant current mode, the current density for the stack is specied as a constant value:

ICelli IStack Constant

The initial temperatures are set as the same with the ambient temperatures (30 to 15 C in different cases). The initial water content is 5 for the base case, and other values of 3 and 7 are also
used, representing a dryer membrane condition and a wetter membrane condition. The initial ice volume fractions and liquid water
volume fractions are 0.

494

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

The initial and surrounding temperatures are 20 C, and the initial water content is 5. The same initial conditions and surrounding

Stack Voltage, V
1.0

0.50

Ice
Current
Voltage

0.45

5.5

0.9

5.0

0.8

0.40

4.5

0.7

0.35

4.0

0.6

3.5

0.5

3.0

0.4

2.5

0.3

2.0

0.2

1.5

0.1

0.30
0.25
0.20
0.15
0.10

Ice Volume Fraction

The nite volume method is adopted to discretize the membrane water conservation equation, liquid water conservation
equation and energy conservation equation. As demonstrated in
Fig. 1(b), each cell is includes 250 grid points. The numbers of grid
point are different for the different layers in each cell, and the
values are 30, 20, 20 nodes for the bipolar plate (BP), ow channel
and GDL, respectively, and 30 and 50 nodes for CL and membrane.
The fourth-order RungeKutta method with explicit time discretization is adopted to implement the time discretization of the differential equations. Adaptive time step method is used with the
DormandPrince pair for error estimation, with this method, the
truncation error can be controlled within a required value, which
is set to be 109 of the relative tolerance. The nonlinear equation
systems (Eqs. (3) and (5)) are solved with the algorithm of
LevenbergMarquardt [50], and the corresponding termination
tolerance is specied as 109 to ensure the accuracy. Grid independent study has been performed to ensure that the inuence of gird
point number on the results can be neglected.

Current Density, A cm-2

2.3. Numerical procedures

0.0
0

10

15

20

25

30

35

Time, s

(a)

3. Results and discussion

200

0.1 s
2s
6s
18 s
30 s

180
160

Stack Power Output, W

In this section, simulations of a 20-cell stack based on the model

presented above are carried out. 4 cases of different start-up modes
are simulated, including constant power higher current (case 1:
CPHC), constant power lower current (case 2: CPLC), constant current (case 3: CC) and constant voltage (case 4: CV). The different
characteristics are compared and elucidated. The effects of various
operating and initial conditions on the constant power cold start
processes are analyzed. There are 2 cases for the constant power
mode, because there are often two combinations of voltage and
current to produce the same power.
The comparison between the model predictions and experimental results [19] is shown in Fig. 2. The model parameters in this
comparison are specied according to the experimental conditions
in [19]. The initial water content is set as 2.4, the initial and surrounding temperature is 20 C, and the initial ice volume fraction
is 0. The start-up processes with different current densities are
compared, and reasonable agreements are obtained in all cases.

140
120
100
80
60
40
20

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

Current Density, A cm-2

(b)

3.1. Cold start processes in different modes

18

Fig. 3 shows the performances of the CPHC cold start mode. The
total power output of the simulated 20-cell stack is set as 80 W.

0.1 s
2s
6s
18 s
30 s

16
14

1.1

Experiment results
Model predictions

1.0

Single Cell Voltage, V

Stack Voltage, V

1.2

0.9
0.8
0.7

12
10
8
6
4

0.6
2

0.5
0.4

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

-2

I = 0.04 A cm

-2

I = 0.08 A cm

0.3

Current Density, A cm-2

-2

I = 0.02 A cm

0.2

(c)

0.1
0.0
0

100

200

300

400

500

600

700

800

900

Time, s
Fig. 2. Comparison between the model predictions and the experimental results in
[19].

Fig. 3. Stack performance predictions of the constant power start-up case with
higher current (case 1: CPHC). The start-up power is set as 80 W. (a) Evolution of
current density, stack voltage and average ice volume fraction in cathode catalyst
layers; (b) evolution of powercurrent density curves; (c) evolution of voltage
current density curves.

495

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

Stack Voltage, V
1.0

0.18
0.17
0.16
0.15
0.14
0.13
0.12
0.11
0.10
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00

Ice
Current
Voltage

18

0.9

16

0.8

14

0.7

12

0.6

10

0.5

0.4

0.3

0.2

0.1

Ice Volume Fraction

Current Density, A cm-2

resistance caused by the anode dehydration (electro-osmotic drag

effect), membrane/cathode hydration (water generation in cathode) and temperature increment

0.0

10

20

30

40

50

60

70

80

90 100 110 120

Time, s

(a)

200

0.1 s
50 s
100 s
110 s
113 s

180

Stack Power Output, W

160
140
120
100
80
60
40
20

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

Current Density, A cm-2

(b)
18
16

0.1 s
50 s
100 s
110 s
113 s

14

Stack Voltage, V

temperature are also set for the other 3 cases for consistent comparison. As shown in Fig. 3(a), at the beginning of the start-up process, the current density decreases sharply and the voltage
increases correspondingly, as the power is constant. This change
is mainly caused by the increment of membrane resistance,
because the initial current density is relatively high in this case,
and the membrane water on the anode side decreases sharply
due to the electro-osmotic drag effect. The suddenly increased
resistance in anode CLs and membrane leads to the increment of
the total ohmic resistance of the stack. In the second period from
about 10 to 25 s, the current density is almost steady with slight
increment due to the balance of the electro-osmotic drag effect
(water transport from anode to cathode) and back diffusion (water
transport from cathode to anode), as well as the hydration in membrane, which reduces the ohmic resistance. In the last period, a
sudden decrement of the current density can be observed, because
the ice volume fraction in cathode CLs approaches 1. It can also be
noticed that the ice volume fraction in cathode CLs increases fast at
the beginning, and the increasing rate slows down later due to the
decreased current density.
Correspondingly, the evolutions of the powercurrent density
(PI) curves and the voltagecurrent density (VI) curves are
shown in Fig. 3(b) and (c), and the points on the curves corresponding to the different time instances during the cold start processes
are marked. Since the stack is started in constant power mode with
higher current in this case, the points are on the right hand sides of
peak values of the PI curves. The evolution path forms a horizontal line from the right hand side to the middle (peak value) on the
PI curves (Fig. 3(b)). Once the peak value is reached, the start-up
process is failed, because the required power output cannot be
supplied at the next time instance, although the ice in cathode
CLs does not fully block the pores at that moment (Fig. 3(a)). It
can be noticed that the peak powers in the curves generally
decrease during the cold start process, and the decrement is the
fastest in the rst few seconds. For the VI curves shown in
Fig. 3(c), it can also be noticed that the voltage increases and the
current density decreases during the cold start process.
Fig. 4 shows the performance predictions of the CPLC cold start
mode (case 2). The conditions are the same with case 1, except that
the lower current is chosen to produce the same power. As shown
in Fig. 4(a), the current density is relatively steady with slight
increment in the rst period. Near the end of cold start process,
the current density increases sharply. It can be noticed that the
evolutions of current density and voltage in this case are opposite
to that in case 1, and for both cases, the start-up processes are
failed when the peak powers in the PI curves are reached (Figs.
3(b) and 4(b)). The degradations of the PI and VI curves shown
in Fig. 4(b) and (c) are more moderate than that in case 1, because
the lower current results in less signicant electro-osmotic drag
effect, and ice is formed more slowly. In addition, it can also be
noticed that the current density change is less in case 2 (from left
to right) than in case 1 (from left to right).
The performance evolutions of the constant current
(0.15 A cm2) start-up process are shown in Fig. 5. The stack voltage and power decrease slowly before the sudden drop due to the
combined effect of anode drying, membrane/cathode hydration
and ice formation. At around 40 s, the ice is about to fully cover
the cathode CLs, and the sharp drops of voltage and power occur.
For the constant current mode, the start-up process is failed when
the limiting current density is lower than the output current density. Since the current density is xed, vertical lines are formed
between the curves (PI and VI curves) representing the cold start
process. For the different VI curves during the cold start process,
the activational and mass transport losses are generally increased,
because ice is continuously formed. The slope of the VI curves
changes more dramatically, reecting the change of ohmic

12
10
8
6
4
2

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

Current Density, A cm-2

(c)
Fig. 4. Stack performance predictions of the constant power start-up case with
lower current (case 2: CPLC). The start-up power is set as 80 W. (a) Evolution of
current density, stack voltage and average ice volume fraction in cathode catalyst
layers; (b) evolution of powercurrent density curves; (c) evolution of voltage
current density curves.

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500
Stack Power, W

0.7

120

0.6

12
100

0.5

10
80

0.4

60

6
4

40

20

0.3
0.2

0.25

0.20

0.15

0.10

0.1
0.0

0
0

10

15

20

25

30

35

40

45

50

0.9

160

0.8

140

0.7

120

0.6

100

0.5

80

0.4

60

0.3

40

0.2

20

0.1
0.0

10

15

20

25

30

35

40

45

Time, s

(a)

(a)
200

200

160
140
120
100
80
60
40

0.1 s
2s
5s
35 s
40 s
41 s

180
160

Stack Power Output, W

0.1 s
10 s
30 s
40 s
45 s
46 s

180

Stack Power Output, W

180

0.05

Time, s

140
120
100
80
60
40
20

20
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

Current Density, A cm-2

Current Density, A cm-2

(b)

(b)
18

18

14
12
10
8
6

0.1 s
2s
5s
35 s
40 s
41 s

16
14

Stack Voltage, V

0.1 s
10 s
30 s
40 s
45 s
46 s

16

Stack Voltage, V

Current
Power
Ice

0.30

0.8

140

14

1.0

0.9

160

Current Density, A cm-2

16

0.35

Ice Volume Fraction

Voltage
Power
Ice

18

Stack Voltage, V

Stack Power, W

1.0

20

Ice Volume Fraction

496

12
10
8
6

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

Current Density, A cm-2

Current Density, A cm-2

(c)

(c)

Fig. 5. Stack performance predictions of the constant current start-up case (case 3:
CC). The start-up current density is set as 0.15 A cm2. (a) Evolution of current
density, stack voltage and average ice volume fraction in cathode catalyst layers; (b)
evolution of powercurrent density curves; (c) evolution of voltagecurrent density
curves.

Fig. 6. Stack performance predictions of the constant voltage start-up case (case 3:
CV). The start-up stack voltage is set as 6 V (20 cells). (a) Evolution of current
density, stack voltage and average ice volume fraction in cathode catalyst layers; (b)
evolution of powercurrent density curves; (c) evolution of voltagecurrent density
curves.

Fig. 6 shows the performance evolutions of the constant voltage

(6 V for 20 cells) start-up mode. The ice formation rate is relatively
high in the early period, and it becomes lower later because the
current density is decreased. The current density and stack power

decrease fast at the beginning and near the end, while during the
middle period, the change is slower, due to the combined effect
of membrane hydration, temperature increment and ice formation.
Notable transport losses in the VI curves near the end of the cold

497

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500
Stack Voltage, V

40 W
80 W
120 W

Current Density, A cm-2

0.4

Current

0.3

Ice

0.2
0.1
0.0

Voltage
-0.1

13.5

0.9

12.0

0.8

10.5

0.7

9.0

0.6

7.5

0.5

6.0

0.4

4.5

0.3

3.0

0.2

1.5

0.1

-0.2

Ice Volume Fraction

1.0
0.5

0.0
0

8 10 12 14 16 18 20 22 24 26 28 30 32

Time, s

(a)

Fig. 7. Evolutions of stack volume averaged temperatures of the four different startup modes.

200
180

0.1 s
5s
10 s
20 s
26 s

160

Stack Power Output, W

start process can be observed, caused by the ice formation in cathode CLs.
The comparison of the stack average temperature evolutions of
the four different start-up modes is shown in Fig. 7. The temperature evolutions are similar for the constant power higher current
(CPHC) start-up, constant current start-up and constant voltage
start-up processes, because the current densities are all kept at
high levels for the 3 cold start processes. For the constant power
mode with lower current, the temperature increment rate is much
lower, and the nal temperature reached is also lower than the
other cases. It suggests that the heating effect is mainly dominated
by the current density, and keeping high current density during a
cold start process is always benecial for stack heating.

140
120
100
80
60
40
20

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

3.2. Constant power start-up mode

Current Density, A cm-2

To further understand the constant power cold start characteristics of PEM fuel cell stacks, the various start-up processes are
investigated. These cases are all based on the CPHC mode, because
the higher current mode produces more heat facilitating the cell
heating.

(b)
200

0.1 s
1s
2s
3s
5s

180

3.2.2. Effect of initial water content

Fig. 9 shows the evolutions of current density, stack voltage,
average ice volume fraction in CLs, area specic resistance and
water contents with different initial water contents of 3, 5 and 7.
The start-up power is 80 W, the initial and ambient temperatures
are 20 C, and the ionomer volume fraction in CLs is 0.35. Obviously, with higher initial water content, ice appears earlier and
the ice formation rate is higher (Fig. 9(a)), because the water in
ionomer reaches saturation more quickly with higher initial water

Stack Power Output, W

160

3.2.1. Effect of start-up power

Fig. 8 shows the comparison of 3 CPHC cases with different
start-up powers (40, 80 and 120 W). The initial and surrounding
temperatures are 20 C, the initial water content is 5, the initial
ice volume fraction is 0, and the ionomer volume fraction in CLs
is 0.35. It can be noticed that with a higher start-up power, the
voltage is higher, the current density is lower, and the ice formation is slower. This should be reversed with the CPLC mode. The
cold start process at 80 W lasts longer than at 40 W, because the
corresponding current at 80 W is lower than that at 40 W (slower
ice formation). However, the 120 W cold start process is failed in
less than 6 s. Although the CLs are not fully covered by ice at that
time instance, the stack can no longer afford the high power output, and the peak value in PI curve is achieved at 6 s.

140
120
100
80
60
40
20

0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55

Current Density, A cm-2

(c)
Fig. 8. Stack performance predictions of the constant power higher current (CPHC)
mode at 40, 80 and 120 W. (a) Evolutions of current density, stack voltage and
average ice volume fraction in cathode catalyst layers; (b) evolutions of PI curve at
40 W; (c) evolutions of PI curve at 120 W.

content, and the higher ice formation rate is a result of higher

current density with higher initial water content (lower ohmic
resistance). With higher initial water content, the current density
is initially higher, and then drops fast due to the stronger
electro-osmotic drag effect. The difference of current evolution is

498

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

caused by the area specic resistance (ASR) (Fig. 9(b)). The ASRs in
all cases increase in the rst several seconds and then decrease
until the end. The ASR of the higher initial water content case
increases more quickly due to the large variation of water
distribution, which will be discussed later with Figs. 9(c) and 10.
The superiority of lower initial water content shows the impor-

tance of the purging process in constant power cold start mode,

which is similar to the other cold start modes [51,52].
Fig. 9(c) shows the evolutions of average water contents in
cathode CL, anode CL and membrane with different initial water
contents. It can be noticed that the water contents in anode CL
drop to similar levels for all the cases, because although the elec-

(a)

(a)

(b)

(b)

14
12
Cathode CL

Water Content

10

Initial water content = 3

Initial water content = 5
Initial water content = 7

8
6
Membrane
4
2
Anode CL
0
0

10

15

20

25

30

Time, s

(c)
Fig. 9. Stack performance predictions of the constant power higher current (CPHC)
mode with initial water contents of 3, 5 and 7. (a) Evolutions of current density,
stack voltage and average ice volume fraction in cathode catalyst layers; (b)
evolutions of area specic resistance; (c) evolutions of average water content in
cathode catalyst layers, membranes and anode catalyst layers.

(c)
Fig. 10. Transient distributions of water content in membrane electrode assembly
of the constant power higher current (CPHC) mode with initial water contents of 3
(a), 5 (b) and 7 (c).

499

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

3.2.3. Effect of initial temperature

The stack performance predictions of the constant power higher
current (CPHC) mode with different initial (ambient) temperatures

(a)

Stack Voltage, V
1.0

Ionomer volume fraction = 0.4

Ionomer volume fraction = 0.3

0.55

0.50

0.45

Voltage

0.40

0.9
0.8
0.7

4
0.6

0.35

0.30

0.5

0.25

0.4

0.20

Current

0.15

0.3

-1

Ice Volume Fraction

0.60

Current Density, A cm-2

tro-osmotic drag effect with higher initial water content is more

signicant, the higher current density also leads to larger gradient
of water concentration from cathode to anode, which counteracts
the electro-osmotic drag effect. The electro-osmotic drag effect
also becomes less signicant when the local water content is
lower. In membrane, it can also be noticed that the higher the initial water content, the larger the variation occurs during the cold
start process, as mentioned previously, due to the more signicant
electro-osmotic drag and back diffusion effects.
Fig. 10 shows the evolutions of water distributions in the
membrane electrode assembly with three different initial water
contents of 3, 5 and 7. It can be noticed that the water content gradients are all formed quickly in the rst several seconds for all the
cases, mainly dominated by the electro-osmotic drag effect. After
that, when ice is formed, the change of water content becomes
slower. With lower initial water content, the water content decrement close to anode side is less, and the increment close to cathode
side is more signicant. With higher initial water content, the
decrement of water content close to anode side becomes more
noticeable. As mentioned previously, these changes are caused by
the electro-osmotic drag and back diffusion effects.

0.2

0.10

-2

Ice

0.05

-3

0.00

0.1
0.0

10

12

14

16

18

20

22

24

26

Time, s
Fig. 12. Evolutions of current density, stack voltage and average ice volume fraction
in cathode catalyst layers of the constant power higher current (CPHC) mode with
ionomer volume fractions of 0.3 and 0.4 in catalyst layers.

are shown in Fig. 11. The start-up power is 80 W for the stack, the
initial water content is 5, and the ionomer volume fraction in CLs is
0.35. It can be noticed that with lower initial temperature, the current density is lower due to the more sluggish reaction. Correspondingly, the ice formation rate is generally lower due to
lower current density. However, in the rst several seconds, the
ice formation rate is higher with lower initial temperature, because
the ionomer gets saturated more quickly at lower temperature.
3.2.4. Effect of ionomer volume fraction in catalyst layer
The effects of ionomer volume fraction in CLs are shown in
Fig. 12. The start-up power is 80 W, the initial and ambient temperature is 20 C, and the initial water content in ionomer is 5.
2 different cases with the ionomer fractions of 0.3 and 0.4 are compared. The ohmic resistance is signicantly inuenced by the ionomer fraction. With lower ionomer fraction, the current density is
much lower due to higher ohmic loss. It can also be noticed that
the cold start process with the ionomer fraction of 0.3 is failed
when the ice volume fraction reaches around 0.6, because the
start-up power of 80 W can no longer be supplied, while with
the higher ionomer fraction of 0.4, the cold start process is failed
when the ice almost fully covers the cathode CLs. This suggests
that the ohmic resistance is an important factor limiting the constant power start-up mode.
4. Conclusion
In this study, a cold start model for proton exchange membrane
(PEM) fuel cell stack has been developed to elucidate the fundamental differences among the constant power, current and voltage
cold start modes. The powercurrent and voltagecurrent curves at
different time instances are also obtained to understand the various cold start processes. The following conclusions can be made:

(b)
Fig. 11. Stack performance predictions of the constant power higher current (CPHC)
mode with initial temperatures of 15, 20 and 30 C. (a) Evolutions of current
density and stack voltage; (b) evolutions of average ice volume fractions in cathode
catalyst layers.

In the constant power cold start mode, the cold start process
may fail before the ice fully covers the cathode catalyst layers (CLs), because the stack may not be able to continuously
supply the required power output, which is often different
from the constant current and voltage start-up modes.
The constant power can be controlled at higher current
(CPHC) or lower current (CPLC). In the CPHC mode, the
current density decreases and the stack voltage increases
during the cold start process, while it is reversed in the
CPLC mode.

500

Y. Luo et al. / International Journal of Heat and Mass Transfer 77 (2014) 489500

Generally, the CPLC mode produces less heat during cold

start process than the other start-up modes, because the
current density is often kept at low level.
With higher initial water content of the CPHC mode, the
initial current density is higher and decreases fast; and
with lower initial water content, the current density and
stack voltage are more stable, and the cold start process
can last longer.
Because the required power output dominates the termination of the start-up process in the constant power mode,
the initial temperature and ionomer volume fraction in
CLs have signicant effect on the performance. The lower
reaction activity (lower temperature) and higher ohmic
loss (lower ionomer volume fraction in CLs) could limit
the power output signicantly.

Conict of interest
None declared.
Acknowledgments
This research is supported by the National Natural Science
Foundation of China (Grant No. 51276121), and the Natural Science
Foundation of Tianjin (China) (Grant No. 12JCYBJC30500).
References
[1] F. Bacon, Fuel cells: will they soon become a major source of electrical energy?,
Nature 186 (1960) 589592
[2] N.M. Sammes, Fuel Cell Technology: Reaching Towards Commercialization,
Springer-Verlag London Limited, London, 2006.
[3] M.M. Mench, Fuel Cell Engines, John Wiley & Sons, Hoboken, 2008.
[4] C.-H. Lee, J.-T. Yang, Modeling of the Ballard-Mark-V proton exchange
membrane fuel cell with power converters for applications in autonomous
underwater vehicles, J. Power Sources 196 (8) (2011) 38103823.
[5] Honda
Motor
Co.,
Ltd.
<http://www.world.honda.com/FuelCell/FCX/
FCXPK.pdf>, 2003. [Accessed 26. 03. 2013].
[6] G. Budd, <http://cute-hamburg.motum.revorm.com/presentations>, 2006.
[Accessed 26. 03. 2013].
[7] M.M. Hoch, Pulsed coolant control for improved stack cold starting, US Patent
No. 7759010, 2010.
[8] U. Gebhardt, M. Waidhas, Method for cold-starting a fuel cell battery, and fuel
cell battery suitable for this method, US Patent No. 2002/0068202 A1, 2002.
[9] K. Korytnikov, P. Novak, Method and apparatus for cold-starting a PEM fuel cell
(PEMFC), and PEM fuel cell system, US Patent No. 2005/0227126 A1, 2005.
[10] A. Santamaria, H.-Y. Tang, J.W. Park, G.-G. Park, Y.-J. Sohn, 3D neutron
tomography of a polymer electrolyte membrane fuel cell under sub-zero
conditions, Int. J. Hydrogen Energy 37 (14) (2012) 1083610843.
[11] S. Ge, C.-Y. Wang, Characteristics of subzero startup and water/ice formation
on the catalyst layer in a polymer electrolyte fuel cell, Electrochim. Acta 52
(14) (2007) 48254835.
[12] P. Oberholzer, P. Boillat, R. Siegrist, R. Perego, A. Kastner, E. Lehmann, G.G.
Scherer, A. Wokaun, Cold-start of a PEFC visualized with high resolution
dynamic in-plane neutron imaging, J. Electrochem. Soc. 159 (2) (2012) B235.
[13] S. Ge, C.-Y. Wang, In situ imaging of liquid water and ice formation in an
operating PEFC during cold start, Electrochem. Solid-State Lett. 9 (11) (2006)
A499A503.
[14] H. Mendil-Jakani, R.J. Davies, E. Dubard, A. Guillermo, G. Gebel, Water
crystallization inside fuel cell membranes probed by X-ray scattering, J.
Membr. Sci. 369 (12) (2011) 148154.
[15] Y. Ishikawa, T. Morita, K. Nakata, K. Yoshida, M. Shiozawa, Behavior of water
below the freezing point in PEFCs, J. Power Sources 163 (2) (2007) 708712.
[16] J. Li, S. Lee, J. Roberts, Ice formation and distribution in the catalyst layer
during freeze-start process CRYO-SEM investigation, Electrochim. Acta 53
(16) (2008) 53915396.
[17] G. Gavello, J. Zeng, C. Francia, U. Icardi, A. Graizzaro, S. Specchia, Experimental
studies on Naon 112 single PEM-FCs exposed to freezing conditions, Int. J.
Hydrogen Energy 36 (13) (2011) 80708081.
[18] Q. Yan, H. Toghiani, Y.-W. Lee, K. Liang, H. Causey, Effect of sub-freezing
temperatures on a PEM fuel cell performance, startup and fuel cell
components, J. Power Sources 160 (2) (2006) 12421250.
[19] Y. Tabe, M. Saito, K. Fukui, T. Chikahisa, Cold start characteristics and freezing
mechanism dependence on start-up temperature in a polymer electrolyte
membrane fuel cell, J. Power Sources 208 (2012) 366373.

[20] K. Jiao, I.E. Alaefour, G. Karimi, X. Li, Simultaneous measurement of current and
temperature distributions in a proton exchange membrane fuel cell during
cold start processes, Electrochim. Acta 56 (8) (2011) 29672982.
[21] E.L. Thompson, J. Jorne, H.A. Gasteiger, Oxygen reduction reaction kinetics
in subfreezing PEM fuel cells, J. Electrochem. Soc. 154 (8) (2007) B783B792.
[22] E. Pinton, Y. Fourneron, S. Rosini, L. Antoni, Experimental and theoretical
investigations on a proton exchange membrane fuel cell starting up at subzero
temperatures, J. Power Sources 186 (1) (2009) 8088.
[23] M. Khandelwal, S. Lee, M.M. Mench, One-dimensional thermal model of coldstart in a polymer electrolyte fuel cell stack, J. Power Sources 172 (2) (2007)
816830.
[24] M. Sundaresan, R. Moore, Polymer electrolyte fuel cell stack thermal model to
evaluate sub-freezing startup, J. Power Sources 145 (2) (2005) 534545.
[25] K. Jiao, X. Li, Three-dimensional multiphase modeling of cold start processes in
polymer electrolyte membrane fuel cells, Electrochim. Acta 54 (27) (2009)
68766891.
[26] Y. Wang, P.P. Mukherjee, J. Mishler, R. Mukundan, R.L. Borup, Cold start of
polymer electrolyte fuel cells: three-stage startup characterization,
Electrochim. Acta 55 (8) (2010) 26362644.
[27] L. Mao, C.-Y. Wang, Analysis of cold start in polymer electrolyte fuel cells, J.
Electrochem. Soc. 154 (2) (2007) B139.
[28] Y. Luo, Q. Guo, Q. Du, Y. Yin, K. Jiao, Analysis of cold start processes in proton
exchange membrane fuel cell stacks, J. Power Sources 224 (2013) 99114.
[29] J. Ko, H. Ju, Effects of cathode catalyst layer design parameters on cold start
behavior of polymer electrolyte fuel cells (PEFCs), Int. J. Hydrogen Energy 38
(11) (2013) 682691.
[30] J. Ko, W.-G. Kim, Y.-D. Lim, H. Ju, Improving the cold-start capability of polymer
electrolyte fuel cells (PEFCs) by using a dual-function micro-porous layer
(MPL): numerical simulations, Int. J. Hydrogen Energy 38 (1) (2012) 652659.
[31] Q. Guo, Y. Luo, K. Jiao, Modeling of assisted cold start processes with anode
catalytic hydrogenoxygen reaction in proton exchange membrane fuel cell,
Int. J. Hydrogen Energy 38 (2) (2012) 10041015.
[32] R.J. Balliet, J. Newman, Cold-start modeling of a polymer-electrolyte fuel cell
containing an ultrathin cathode, J. Electrochem. Soc. 158 (9) (2011) B1142.
[33] R.J. Balliet, J. Newman, Cold start of a polymer-electrolyte fuel cell III.
Optimization of operational and congurational parameters, J. Electrochem.
Soc. 158 (8) (2011) B948.
[34] J. Ko, H. Ju, Comparison of numerical simulation results and experimental data
during cold-start of polymer electrolyte fuel cells, Appl. Energy 94 (2012)
364374.
[35] E. Hosseinzadeh, M. Rokni, A. Rabbani, H.H. Mortensen, Thermal and water
management of low temperature proton exchange membrane fuel cell in forklift truck power system, Appl. Energy 104 (2013) 434444.
[36] G. Kovacevic, A. Tenconi, R. Bojoi, Advanced DCDC converter for power
conditioning in hydrogen fuel cell systems, Int. J. Hydrogen Energy 33 (12)
(2008) 32153219.
[37] F. Zenith, S. Skogestad, Control of fuel cell power output, J. Process Control 17
(4) (2007) 333347.
[38] F. Segura, J.M. Andjar, Power management based on sliding control applied to
fuel cell systems: a further step towards the hybrid control concept, Appl.
Energy 99 (2012) 213225.
[39] B.G. Pollet, I. Staffell, J.L. Shang, Current status of hybrid, battery and fuel cell
electric vehicles: from electrochemistry to market prospects, Electrochim. Acta
84 (2012) 235249.
[40] J.B. Benziger, M.B. Sattereld, W.H.J. Hogarth, J.P. Nehlsen, I.G. Kevrekidis, The
power performance curve for engineering analysis of fuel cells, J. Power
Sources 155 (2) (2006) 272285.
[41] M.J. Blackwelder, R.A. Dougal, Power coordination in a fuel cellbattery hybrid
power source using commercial power controller circuits, J. Power Sources
134 (1) (2004) 139147.
[42] P. Thounthong, S. Ral, B. Davat, Control strategy of fuel cell/supercapacitors
hybrid power sources for electric vehicle, J. Power Sources 158 (1) (2006) 806814.
[43] M. Oszcipok, M. Zedda, J. Hesselmann, M. Huppmann, M. Wodrich, M.
Junghardt, C. Hebling, Portable proton exchange membrane fuel-cell systems
for outdoor applications, J. Power Sources 157 (2) (2006) 666673.
[44] B.K. Datta, G. Velayutham, A.P. Goud, Fuel cell power source for a cold region, J.
Power Sources 106 (12) (2002) 370376.
[45] H. Ju, C.-Y. Wang, S. Cleghorn, U. Beuscher, Nonisothermal modeling of
polymer electrolyte fuel cells I. Experimental validation, J. Electrochem. Soc.
152 (8) (2005) A1645A1653.
[46] Q. Ye, T. Van Nguyen, Three-dimensional simulation of liquid water
distribution in a PEMFC with experimentally measured capillary functions, J.
Electrochem. Soc. 154 (12) (2007) B1242B1251.
[47] T.E. Springer, T. Zawodzinski, S. Gottesfeld, Polymer electrolyte fuel cell model,
J. Electrochem. Soc. 138 (8) (1991) 23342342.
[48] A.A. Kulikovsky, Analytical Modelling of Fuel Cells, Elsevier Science,
Amsterdam, 2010.
[49] S. Haji, Analytical modeling of PEM fuel cell iV curve, Renewable Energy 36
(2) (2011) 451458.
[50] D.W. Marquardt, An algorithm for least-squares estimation of nonlinear
parameters, J. Soc. Ind. Appl. Math. 11 (2) (1963) 431441.
[51] K. Jiao, X. Li, Effects of various operating and initial conditions on cold start
performance of polymer electrolyte membrane fuel cells, Int. J. Hydrogen
Energy 34 (19) (2009) 81718184.
[52] K. Jiao, X. Li, Cold start analysis of polymer electrolyte membrane fuel cells, Int.
J. Hydrogen Energy 35 (10) (2010) 50775094.