Energy Conversion and Management 177 (2018) 519–534

Contents lists available at ScienceDirect

Energy Conversion and Management

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

Numerical study of anode side CO contamination effects on PEM fuel cell T

performance; and mitigation methods
A. Moradi Bilondia, M. Abdollahzadeha,b,c, M.J. Kermania, , H. Heidarya, P. Havaeja
⁎

a
Mechanical Engineering Dept., Amirkabir University of Technology, Tehran, Iran
b
Departamento de Engenharia Electromecanica, C-MAST – Center for Mechanical and Aerospace Sciences and Technologies, Universidade da Beira Interior, Covilha,
Portugal
c
Department of Mechanical Engineering, University of Guilan, Rasht, Iran

ARTICLE INFO ABSTRACT

Keywords: In the present work, a two-dimensional, transient, isothermal, two-phase, multicomponent transport model was
Anode side of polymer electrolyte membrane considered for the anode-side electrode of a PEMFC. The governing equations of two-phase flow in the PEM fuel
fuel cell cell were discretized by finite volume method, and the SIMPLE algorithm was used to handle the pressure-
Two-phase flow velocity coupling. The discretized governing equations of the model were numerically solved on a non-uniform
Numerical simulation
grid with an in-house developed code. The simulation was performed for velocity, pressure, concentration of
CO poisoning
Air bleeding
species, and liquid water saturation in the anode side of the PEMFC. At first, the steady-state and transient effects
Platinum–ruthenium of introducing the CO-contaminated hydrogen on the cell performance were investigated. Then, a comprehen-
sive investigation of the commonly used mitigation techniques including the effect of air or oxygen bleeding,
elevation of temperature and the effect of using CO-tolerant catalysts (PtRu/C), was conducted. The numerical
results of the model were compared and validated with the experimental data. The results indicated that even
using a low CO concentration, leads to significant degradation of the fuel cell output current density (about 30%
of the output current was lost within 30 min when the hydrogen is pre-mixed with 15 ppm of CO as the fuel).
Injecting a small amount of air into the anode stream, resulted in fast recovery of the lost current density (by
injecting about 5% air into the fuel, 80% of the output current was recovered within 2 min at 53 ppm CO).
Higher air bleeding ratio only resulted in minor improvement of the cell performance. Increasing the cell
temperature; also using PtRu/C instead of Pt/C (at low temperatures) led to improving the cell performance. The
use of PtRu/C at a high operating temperatures only resulted in minor improvement of the cell performance.

1. Introduction
In recent years, the excessive consumption of fossil fuels, followed
by air pollution, global warming and depleting petroleum resources,
has increased the demand for new technologies to produce clean and
alternative energy [1,2]. One of these technologies is polymer electro-
lyte membrane (PEM) fuel cells. These fuel cells have been considered
as the main candidate of future clean power generation devices for
stationary and transport applications due to their high power density
and low operating temperature [3,4]. Thus, it is very important to en-
hance their performance, durability, and reduce their cost for their
commercial applications; however, there are some barriers to the
commercialization of PEMFCs. One of these barriers is contamination of
the anode fuel stream by impurities, which leads to significant de-
gradation in cell performance [5]. Actually, due to the technical and
practical limitations for using pure hydrogen as anode fuel (related to
production and storage of pure hydrogen), reformate of hydrocarbon
fuels is usually used, which have many impurities (such as carbon
monoxide). Carbon monoxide is one of these impurities that has a
significant effect on cell performance. The presence of CO-impurities
even at very low concentrations can still hinder the cell performance
[6]. Actually, when CO is introduced into the anode gas channel, it is
absorbed on the catalyst surfaces. The bond between CO and the pla-
tinum sites on the surface of the catalyst is very strong and prevents the
hydrogen from reaching the catalyst sites and therefore, the efficiency
of hydrogen oxidation reaction is reduced.
The CO poisoning and its effects on fuel cell performance have been
the focus of several previous researches in the literature [7–9]. Oetjen
et al. [10] carried out an experimental measurement to compare the
effects of CO poisoning on three different anode catalysts. They showed

⁎
Corresponding author at: Mechanical Engineering Dept., Amirkabir University of Technology (Tehran Polytechnic), 424 Hafez Avenue, P.O. Box: 15875-4413,
Tehran, Iran.
E-mail address: mkermani@aut.ac.ir (M.J. Kermani).

https://doi.org/10.1016/j.enconman.2018.09.076
Received 22 June 2018; Received in revised form 21 September 2018; Accepted 25 September 2018
0196-8904/ © 2018 Elsevier Ltd. All rights reserved.
A. Moradi Bilondi et al.
Nomenclature
C molar concentration (mol m−3)
D diffusivity coefficient (m2 s−1)
i current density (A m−2)
K permeability of electrodes (m2)
n order of reaction
P pressure (Pa)
S source term of equations
s saturation of liquid water
T cell temperature (K)
V voltage (V)
X mass fraction (kg kg−1)
Y mole fraction (mol mol−1)
Greek letters
α transfer coefficient
β net water transport coefficient per proton
that the cell performance depends strongly on the CO concentration and
the anode catalyst used. Springer et al. [11,12] developed a pseudo-
two-dimensional steady-state kinetic model for CO poisoning (which is
one of the first models to account for the negative effect of CO on the
fuel cell performance). It was reported that the presence of a trace
amount of CO in the anode fuel stream causes a significant drop in cell
performance. Zhang et al. [13] experimentally and numerically studied
the effect of anode flow rate and cathode oxygen partial pressure on the
cell performance of a fuel cell operated under fuel containing hydrogen
and CO. They found that the anode flow rate strongly influences the
fuel cell performance. It was reported that the cell voltage decreases
dramatically with increasing the anode flow rate at a constant current
density. Zhou and Liu [14] presented a steady state 3-D mathematical
model to study the CO poisoning effect as well as the effect of hydrogen
dilution due to the inert gases. They investigated the cell performance
under various operating and design conditions. Most of the previous
works had focused on the steady-state CO poisoning process while, CO
poisoning phenomenon occurs slowly over time; through a dynamic
process [5]. Bhatia and Wang [15] studied the transient CO poisoning
by simplifying the kinetics proposed by Springer et al. [11] neglecting
the transport phenomena and restricting their attention to the anode
catalyst layer (ACL). Chu et al. [16] presented a one-dimensional
transient model based on the CO kinetic model developed by Springer
et al. [11] to analyze how CO poisoning affects the performance of a
PEMFC. It was reported that as time elapses, due to the adsorption of
CO on the catalyst surfaces, the hydrogen coverage decreases. Also,
they indicated that a higher CO concentration results in fewer available
catalyst sites for hydrogen electro-oxidation and a significant decrease
in the response time to reach steady state. Wang and Chu [17] devel-
oped a one-dimensional, two-phase, transient mathematical model to
simulate the carbon monoxide poisoning effect on the performance of a
PEM fuel cell. They investigated both vapor and liquid water transport
inside the cell. Their theoretical results revealed that a higher CO
concentration results in less hydrogen coverage and a large drop in the
time to reach the steady state.
There are several methods to reduce the destructive CO poisoning
effects. Some of the most commonly used mitigation techniques are
including oxygen bleeding, increasing the cell operating temperature
and the use of alternative catalysts or CO-tolerant catalysts. The oxygen
bleeding method was first introduced for PEMFC application by
Gottesfeld and Pafford [18]. Baschuk and Li [19] presented a one-di-
mensional steady-state kinetic model to simulate the CO poisoning and
oxygen bleeding. In their study, the adsorption, desorption and electro-
oxidation of carbon monoxide and hydrogen were modeled by the
520
Energy Conversion and Management 177 (2018) 519–534
θ contact surface angle (°), coverage fraction of species
μ dynamic viscosity (kg m−1 s−1)
ρ density (kg m−3)
σ surface tension (N m−1)
Subscripts
a anode
act activation
cond condensation
lq liquid
mem membrane
oc open circuit
ss steady state
Superscripts
ads adsorption
eff effective
reactant-pair and Tafel-Volmer mechanisms, respectively. Also, oxygen
bleeding was modeled by the heterogeneous oxidation of carbon
monoxide and hydrogen through a Langmuir-Hinshelwood mechanism.
Murthy et al. [20,21] carried out the steady-state and transient ex-
perimental investigations to understand the effect of CO poisoning (at
relatively high CO level) as well as CO mitigation methods such as O2/
air bleeding and elevation of operating temperature and pressure, on
the cell performance. They showed the strong effect of air bleeding in
the recovery of lost cell performance. Also, their measurements sug-
gested that increasing the pressure and temperature significantly re-
duces the CO poisoning rate. Zamel and Li [5] presented a transient
model to investigate the CO poisoning and oxygen bleeding effects on
the cell performance. They used a model similar to the model proposed
by Baschuk and Li [19]. They investigated the effects of oxygen
bleeding and operating temperature and pressure on the recovery of the
lost fuel cell performance.
Another way to mitigate the destructive effects of CO poisoning on
the cell performance is using CO-tolerant catalysts. One of the most
widely researched CO-tolerant catalysts is the platinum/ruthenium
catalysts. Investigation on PtRu electrodes has been extensively carried
out from as early as the 1960s [6]. Lee et al. [22] investigated the
electrocatalysis of CO tolerance in the hydrogen oxidation reaction for
Pt/C, PtSn/C and PtRu/C electrocatalysts in the PEMFC. Both half and
single cell polarization characteristics were studied at several tem-
peratures and CO concentrations. It was proposed that the CO adsorp-
tion step occurs predominantly through a displacement path for PtSn/C
and through a free site attack path for CO on both Pt/C and PtRu/C. The
data were more consistent with the participation of linear and bridged
bonded adsorbed CO on Pt/C and PtRu/C. They showed that oxidation
of CO by the alloy catalysts is not the only contributor to CO tolerance.
Changes in the thermodynamics and the kinetics of the CO adsorption
process, induced by the alloy catalysts, also contribute to CO tolerance.
Shah et al. [23] presented a transient, fully two-phase, non-isothermal
model of CO poisoning and oxygen bleeding in MEA of a PEMFC. Also,
they investigated the effects of operating temperature as well as using
PtRu catalyst as a CO-tolerant catalyst on cell performance. They in-
corporated a bi-functional mechanism for carbon monoxide oxidation
on PtRu catalysts. Sung et al. [24] experimentally and numerically in-
vestigated the dynamic behavior of a PEMFC under CO poisoning and
the effect of air bleeding on the recovery ratio on PtRu catalyst sites.
Their experimental results suggested that even using a CO-tolerant
catalyst, CO-poisoning cannot be avoided with the operating conditions
in their study (about 80% of the output current was lost within 20 min).
Also, it was reported that injecting less than 5% air into the
A. Moradi Bilondi et al.
contaminated fuel have a significant improvement in recovery of lost
current density.
Although there is some research on CO poisoning effect on fuel cell
performance and the mitigation methods, there is a lack for a com-
prehensive validated dynamic model to simulate the CO poisoning
process and the main mitigation techniques. Therefore, in the present
work, a comprehensive study has been conducted to investigate the
effect of CO poisoning phenomenon and the most influential mitigation
techniques on reducing the destructive effects of CO poisoning. Both
steady state and transient behavior of the cell under CO poisoning effect
as well as validation of these results against the published experimental
data, is reported for the first time. In addition, the effect of most in-
fluential mitigation techniques including O2 or air bleeding, elevation
of operating temperature, and alternative catalyst or CO-tolerant cata-
lyst, were simulated and all the results of these mitigation methods
were validated against the experimental data. Further, the comparison
these mitigation techniques as well as combination of them (such as
elevation of temperature while using CO-tolerant catalyst) are reported
for the first time.
2. Description of the mathematical model
In this section, the computational domain and the governing
equations for the two-phase flow and mass transport along with the
kinetic reactions in the anode side of the PEMFC are presented. Then,
the model assumptions and the auxiliary equations used in modeling, as
well as the boundary conditions are described.
A schematic view of the anode-side of a PEMFC with a straight
channel is shown in Fig. 1 that includes gas channel, gas diffusion layer
(GDL), and catalyst layer (CL). Reactants are transported by diffusion
and advection from the gas channel into the GDL and then to the CL
where the reactants participate in the electrochemical reaction.
The main assumptions of the mathematical model are as follows:
(1) The reactants and products are assumed to be ideal gas mixtures.
(2) Due to the low velocity of the fluid in the fuel cell, the gas flow in
the channels is considered laminar (Re < 40).
(3) The electrode is treated as isotropic and homogeneous porous
zones, and the properties, such as porosity and permeability are
constants [25,26].
(4) The membrane is impermeable to the gas phase.
(5) Ohmic losses in the current collector (or bipolar plate) are ne-
glected.
(6) The modeling domain is assumed isothermal [27].
(7) The gas and liquid phases in the fuel cell exist as continuous phases.
2.1. Kinetic reactions in the anode catalyst layer of PEMFC
2.1.1. Kinetics of the hydrogen (electro-) oxidation reaction (HOR)
When pure hydrogen is introduced to the gas channel of the anode,
without the presence of impurities, the reactions shown in Eqs. (1) and
(2) occur in the anode catalyst layer that is referred to as the hydrogen
Fig.1. Geometry and schematic view of 2-D half-cell model.
521
Energy Conversion and Management 177 (2018) 519–534
adsorption/desorption reaction (Tafel mechanism) and electro-oxida-
tion reaction (Volmer mechanism), respectively [28].
kfh
H2 + 2Pt 2(Pt H)
bfh kfh (1)
k eh
(Pt H) H+ + e +Pt (2)
2.1.2. Kinetic of CO poisoning
The carbon monoxide that may exist in the anode fuel stream is
quickly absorbed on the platinum covered surface of the catalyst layer
(reactions (3)) and create an obstacle for the hydrogen molecules to
reach the reaction sites [6]. The adsorption reaction of CO on the pla-
tinum sites decreases the available surface area for the hydrogen oxi-
dation reaction, and thus, the output current density of the fuel cell is
reduced. At higher overpotentials (0.6–0.9 V) the electrochemical oxi-
dation of CO to CO2 (shown in Eq. (4)) begins to provide more hy-
drogen sites to increase the current density [12]. However, most of the
electrical current density is produced by the electrochemical oxidation
of the adsorbed hydrogen (reactions (2)). The adsorption/desorption
and electro-oxidation of CO are expressed by the following reaction pair
mechanisms [11,19].
kfc
CO + Pt (Pt CO)
bfc kfc (3)
k ec
H2 O+ (Pt CO) Pt + CO2 + H+ + e (4)
2.1.3. Kinetic of oxygen or air bleeding
Introducing a few percents of oxygen into the anode fuel stream,
known as oxygen bleeding, will significantly reduce the poisoning effect
of CO on the anode catalyst layer. The O2 may react with the adsorbed
CO to form CO2 or with the adsorbed H atoms to form H2O. These
reactions are expressed as follows [19].
bfc kfc
O2 + 2Pt 2(Pt O)
bfo kfo (5)
(Pt O) + (Pt CO)
k oc
2Pt + CO2 (6)
(Pt O) + 2(Pt H)
k oh
3Pt + H2 O (7)
Eq. (6) illustrates that oxygen mitigates poisoning by removing the
absorbed CO through heterogeneous oxidation. However, oxygen can
also react with the adsorbed hydrogen, which results in the reduction of
the overall hydrogen available for reaction. Note that reactions (6) and
(7) are assumed to proceed in one direction only.
2.1.4. CO-tolerant catalysts
As it's known, even trace amounts of CO (ppm level) in the anode
feeds can severely decrease the performance of PEMFC which uses a Pt
catalyst at the anode side [29]. Platinum is the most widely used cat-
alyst in PEMFCs. At low temperatures, the adsorption probability of CO
onto platinum is very high. Thus, in order to alleviate the destructive
effects of CO poisoning, alternative catalysts (CO tolerant electro-cat-
alysts) were investigated in the literature. CO tolerant catalysts are
usually a combination of platinum with noble metals, such as Ru, Sn,
Co, Cr, Fe, Ni, Pd, Os, Au, W, Mo, and Mn or a combination of these
elements. Among these bimetallic catalysts, the PtRu/C exhibits the
highest CO tolerance. It is believed that combining platinum with these
noble metals encourages the bi-functional mechanism [6]. A bi-func-
tional mechanism and ligand effects are usually proposed as the origin
of improved CO tolerance on PtRu catalysts. The bi-functional me-
chanism is attributed to the facilitation of adsorbed CO oxidation on Pt
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

in the presence of Ru by producing oxygen-containing species at low
potentials. The ligand effect suggests that the electronic structure of Pt
is modified by forming an alloy with Ru, which weakens CO adsorption
on Pt [29].
Several studies have been performed to understand the mechanism
of the electrooxidation of CO and hydrogen on the surface of CO-tol-
erant PtRu/C catalyst [6]. In the present work, the effect of using PtRu/
C as a CO tolerant catalyst is studied using a dualistic approach that is
presented in [30]. In this approach, it is assumed that the CO molecules
adsorb mainly onto the Pt surfaces and the water molecules adsorb
mainly onto the Ru sites of the anode catalyst surface, allowing for the
formation of carbon dioxide as shown by the following set of reactions
[23]:
k1
H2 O+ M (M OH) + H+ + e
k 1
k2
(M OH) + (Pt CO) Pt + M+ CO2 + H+ + e
k 2
where M is representing the Ru noble metal. Adsorption of water onto
the Pt sites is considered negligible because adsorption of water takes
place at a significantly lower potential on pure Ru sites than on Pt [31].
Also, since the electro-oxidation of hydrogen onto the Ru sites is two
orders of magnitude smaller than onto the Pt sites [32], it is considered
negligible. Adsorption of carbon monoxide on Ru surfaces is also con-
sidered small in this approximation [30].
2.2. Modeling of CO poisoning and O2 bleeding in a CO-tolerant catalyst
(PtRu alloy)
According to the dualistic approach [30], the surface coverage of H,
CO, and O are assumed to be valid on the Pt sites only (Eq. (10)). The
Ru sites are assumed to be occupied by OH, which has surface coverage
OH , or else free, with surface coverage M (Eqs. (11)). Assuming the
Ru Ru
oxygen molecules do not react with ruthenium molecules on the cata-
lyst surface [33], the kinetic equations in the anode of PEMFC are ex-
pressed with Eqs. (12)–(15). Also, the Eqs. (16)–(24) represent the
adsorption, desorption and electro-oxidation rate of species on the
catalyst surface. In Eqs. (16) and (17), the parameters kfc , bfc and kfh are
(8)
expressed in the form of functions of the CO site coverage as Eqs.
(25)–(27) in which, kfc0 , bfc0 and kfh0 denote the values of kfc , bfc and kfh
when CO is equal to zero and they can be calculated from Eqs. (29) and
(9)
(30). Also, ( EH ) represents the change in the activation energy for
hydrogen dissociative adsorption near CO occupied site. In Eq. (25), is
a symmetry factor and has a value between 0 and 1 and r is an inter-
action parameter that represents the effect of lateral-interaction [34]. In
this study, a value of 0.1 has been chosen for and r is obtained as a
function of temperature from Eq. (28).

Table 1
Species coverage equations for the anode side of CO-poisoned PEMFC.
Description No. Equation Ref.

Free surface coverage on Pt site (10) M = 1 H CO O [23]

Free surface coverage on Ru site (11) Ru
M =1
Ru
OH
[23]
H coverage (12) H
= 2qH , ads qH , ox 2qH [5,19]
pt t O, ox
CO coverage (13) CO
= qCO, ads qCO, ox qCO qCO [23]
pt t O, ox OH , ox
O coverage (14) O
= 2qO, ads qH O, ox qCO O,ox [5,19]
pt t
OH coverage (15) Ru
OH
[23]
Ru = qH2 O,ads qCO OH , ox
t
H2 adsorption on Pt (16) qH , ads = kfh P H2 (1 CO O H ) n bfh kfh ( H ) n [19]
CO adsorption on Pt (17) qCO, ads = kfc PCO (1 CO O H) bfc kfc CO [19]
O2 adsorption on Pt (18) qO, ads = kfo PO2 (1 CO O H)
n bfo kfo ( O)
n [19]
H2O adsorption on Ru (19) m RT Ru
F a
Ru
F a [23]
qH2 O, ads = k2 (1 OH ) e 2RT k 2 OH e 2RT
EW
H2 oxidation (20) a [19]
qH , ox = 2 H k eh sinh
bh
CO oxidation (21) [19]
qCO, ox = k ec CO exp b ( )
a
co
H2 oxidation by O2 (22) qH O, ox = koh H n O [19]
CO oxidation by O2 (23) qCO O, ox = koc O CO [19]
CO oxidation by OH (24) Ru
F a
Ru
F a [23]
qCO OH , ox = k1 OH CO e 2RT k 1 (1 OH )(1 H CO O ) PCO2 e 2RT

Rate constant for CO adsorption (25) [19]

kfc = kfc 0exp ( ) r CO
RT
with = 0.1
Rate constant for CO desorption (26) [19]
exp ( ) r CO
bfc = bfc 0
RT
Rate constant for Hydrogen adsorption (27) [11]
kfh = kfh0exp
( EH )
RT
1 exp ( p CO
CO 1 )
with ( EH ) = 13212.4 J mol 1, p = 5.0
Interaction parameter (28) 39.7 kJ mol 1 T < 373.15 (K) [19]
r = 41.4 kJ mol 1 373.15 T 388.15 (K)
56.5 kJ mol 1 T > 388.15 (K)
Arrhenius law for temperature dependency of the rate constants (29) Eik [5,19]
ki = kiref exp
RT

(30) Eib [11,35]

bi = biref exp
RT

(31) Fitted
k1 = k1ref exp ( ) E k1
RT
(32) bh =
RT
, bco =
RT [11]
hF co F

522
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

To properly model the effect of temperature, the temperature de-
pendency of the reaction rate constants in Eqs. (16)–(18) and Eqs.
(20)–(23) are determined using the Arrhenius law as Eqs. (29) and (30).
Further, the temperature dependency of the kinetic parameters related
to the PtRu catalyst, i.e. k1, k−1, k2, and k−2 must be specified. Among
these kinetic parameters (k1, k−1, k2, and k−2), k1 is dominant, and any
change of that will have a more significant effect on numerical results.
Therefore, only this parameter is considered as a function of tempera-
ture (Eq. (31)), and it is assumed that the three other parameters are
constant. The temperature dependency of k1 is obtained from fitting the
model predictions to the experimental measurement [22]. Table 1 lists
the species coverage equations for the anode side of CO-poisoned
PEMFC.
water in the porous electrode. Also, the effective diffusion coefficient of
the species for flow in the porous electrodes is determined by Brugge-
man's model as Eq. (41).
The last terms on the right side of Eqs. (33) and (35) denote the
volumetric source terms due to the phase change, water transport from
the anode to cathode and electrochemical reactions in the CL, and they
are zero in other parts of the computational domain including GDL and
gas channel. These terms are given as Eqs. (42)–(44), in which, the
active area of Pt and Ru sites are derived from Eqs. (45)–(47). In the
Eqs. (45)–(47), rPt Ru is the Pt-Ru ratio and aeff
tot
is the combined active
area of Pt and Ru. The relative permeability for the gas and liquid
phases (in Eqs. (34) and (37)) are also defined as Eq. (45). Table 2 lists
all the governing equations of two-phase flow used in the simulation.

2.3. Governing equations of two-phase flow and mass in the anode of 2.4. Polarization characteristics
PEMFC
When the electrical energy is drawn from the cell, the cell potential
The equations governing the transport of liquid water and gaseous is dropped due to irreversible losses. Table 3 presents the equations of
species are generally expressed by the principles of conservation of these irreversible losses as well as the output current densities equa-
mass, momentum, and chemical species written for each phase (see Eqs. tions. Eqs. (49)–(51) illustrate voltage and overpotentials of the cell.
(33)–(37)) [36]. The last term on the right side of Eq. (34) represents Here, only the anode side of PEMFC is simulated, and the cathode
Darcy’s drag force applied by the pore walls on the fluid within the overpotential is calculated via Eq. (50). The total current density in the
pores, which usually results in a significant pressure drop across the anode catalyst layer, ia , is a sum of the current generated during the
porous medium. The two-phase model presented in this paper is one of electro-oxidation reaction of hydrogen, carbon monoxide and water
the two-fluid flow models [37]. In this model, each phase is introduced and is expressed by Eq. (56).
by a set of conservative equations, and these two phases (gas and li-
quid) are coupled in the saturation level. The basic assumption in the 2.5. Boundary conditions
two-fluid model is that the liquid water in the porous regions is driven
by the capillary pressure [38], which is defined as the difference be- At the gas channel inlet, uniform velocity and mass fraction of the
tween the pressure of the liquid phase and the gas phase (Eq. (39)). The species are defined (Dirichlet B.C.). These boundary conditions take the
interaction between gas and liquid phase is accounted through the following form [53]:
volumetric mass transport rate of water due to the evaporation and
u = ua . in, v=0, T = Tcell, s=0
condensation processes (Eq. (40)). When water vapor pressure exceeds
the saturation level, condensation begins, and so there will be liquid Xi = Xiin , i = H2 , H2 O , O2, CO (57)

Table 2
Governing equations of two-phase flow in the anode side.
Description No. Equation Ref.

Mass conservation equation of gas mixture (33) ( g (1 s )) [39,40]

+ ·( g ug ) = Sm
t
Momentum conservation equation of gas mixture (34) 1 ( g ug ) 1 1 µg [38,41,42]
+ 2 (1 ·( g ug ug ) = Pg + ·(µg ug ) ug
(1 s) t s )2 (1 s) KKrg
Species conservation equation of gas mixture (35) ( g (1 s ) Xi ) [38,39]
+ ·( g ug Xi ) = ·( g Dieff Xi ) + Si
t
i = H2, H2 O, CO, O2
Mass conservation equation of liquid water (36) ( l s) [43]
+ ·( l ul) = (Sg l) M H2 O
t
Momentum conservation equation of liquid water (37) ul =
KKrl
Pl [38]
µl
Liquid pressure (38) Pl = Pg Pc [44,45]
Capillary pressure (39) Pc = cos [46,47]
c J (s )
K
Phase change term (40) k c (1 s )(Pv Psat ) Y H2 O |Pv Psat | ke s l |Pv Psat | [41,43]
Sg l = 1+ + 1
2RT Pv Psat 2MH2 O Pv Psat
Bruggeman's model (41) Dieff = Di [ (1 s)]1.5 [48]
Source terms in the species equation of gas mixture (42) [38]
( ) M (a
qi, ads
Si = Pt
F i eff )
i = H2, CO, O2
Water source term in the species equation of gas mixture (43) [38]
( )M
q H2 O, ads
S H2 O = Ru
F H2 O (aeff )

( ia
H2 O F )M Pt
H2 O (aeff ) (Sg l ) MH2 O
Source term in mass equation (44) Sm = Si, i = H2, H2 O, CO, O2 [38]
i
Active area of Pt sites (45) Pt =
aeff
mPt As [23,49]
lcl
Active area of Ru sites (46) Ru
aeff = tot
aeff Pt
aeff [5,25]
Combined active area of Pt and Ru (47) a Pt
eff
[5,25]
tot =
aeff
rPt Ru
Relative permeability (48) Krg = (1 s )3, Krl = s 3 [44,50]

523
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

Table 3
Voltage losses and output current densities equations.
Description No. Equation Ref.

Cell voltage (49) Vcell = Voc a c ohmic [50]

Cathode overpotential (50) [15,51]
c =
RT
cF
ln ( )( )
1
1
s
ia
ioc
Ohmic loss (51) ohmic = ia R ohmic [52]
Cathode exchange current density (52) ref
ioc = ioc exp 7900 ( 1
T
1
353.15 )
H2 Current density (53) a [6]
iH = 2 H k eh sinh
bh
CO current density (54) a [6]
iCO = k ec CO e bco
OH current density (55) Ru
F a
Ru
F a
iOH = k1 OH CO e 2RT k 1 (1 OH )(1 H CO O ) PCO2 e 2RT
Total current density (56) ia = iH + iCO + iOH

The temperature at the inlet to the channel is set equal to the op- system of algebraic equations arising from the discretization of the
erating temperature of the cell (Tcell ). Also, the inlet velocity of the gas transport equations and pressure correction equation. In order to check
mixture to the channel (ua . in ) is obtained as a function of the reactant the convergence of the numerical solution at each time step, the relative
stoichiometric flow ratio [54]: error for each dependent variable between two consecutive iterations
has been monitored to become less than 10−6.
Iref AMEA Mh2 Iref AMEA 1
ua . in = or ua . in = Also, a grid independence test was carried out for different grid
2F Ach hin2 2F Ach Chin2 (58) systems, and it was found that the grid system of 63 (length) × 30
(height) is appropriate for the present study. The gas channel is divided
Pa Mh2 in
in
= Y h2 into 12 control volumes in the y-direction, the GDL 12, and catalyst
h2
RT (59)
layer 6. Hence, all the subsequent calculations in the present study are
where Iref is the reference current density, AMEA the geometrical area of performed using this grid system.
the MEA, Ach the cross-section area of the gas channel, and Pa is the
operating pressure of the anode side. At the exit plane, fully developed 4. Results and discussion
conditions were enforced as below [41].
In this section, in order to validate the numerical results, a com-
V Xi parison between the polarization curve of a PEMFC obtained from the
=0, = 0, p = pout
x x (60) presented two-phase model with experimental data are first performed
At the wall boundaries, the no-slip and impermeability conditions (in the simulation of this section, pure hydrogen has been used as fuel
are implemented in the normal direction to the wall (n) [53,55,56]. feed to the anode). Then the steady-state and transient response of the
cell under the CO poisoning is investigated. Finally, the effects of mi-
Xi S tigation methods on the recovery of cell performance loss are presented,
V = 0, = 0, =0
n n (61) and their results are discussed.
Also, the pressure values at the gas channel inlet and wall bound-
aries are determined by extrapolation from the interior field. 4.1. Model validation

3. Numerical procedure The fuel cell polarization curve obtained from the simulation as well
as the experimental results of Ticianelli et al. [58] are compared in
The governing equations on the fluid flow and mass transport of
PEMFC were discretized using the finite volume approach, and the
SIMPLE algorithm was used to handle the pressure-velocity coupling.
The discretized governing equations of the model were numerically
solved on a non-uniform grid with an In-House developed code based
on FORTRAN programming language. The numerical method employed
a fully collocated grid (viz., all flow variables are stored at the same
location). This grid is known to provoke chequerboard oscillations in
the pressure field, reflecting a state of pressure–velocity decoupling. To
avoid this, the widely used method of Rhie and Chow [57] was adopted
to non-linearly interpolate the cell face velocities from the nodal values
(at the center of the cells). The interpolation scheme essentially in-
troduces a fourth-order pressure diffusion correction that smooths out
the pressure if it oscillates rapidly. In addition, the nonlinear ODEs
equations of the species coverage fraction equations were solved im-
plicitly at each time step.
The simulation is performed for velocity, pressure, and concentra-
tion of species and also liquid water saturation distribution on the
anode side of the PEM fuel cell. Since all governing equations are
coupled, they should be solved simultaneously using an iterative
method. A Tri-Diagonal Matrix Algorithm (TDMA) was applied itera- Fig. 2. The predicted polarization curve of the cell and the experimental data
tively, in a line-by-line fashion. This algorithm was utilized to solve the from Ref. [58].

524
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

Table 4
Cell geometric parameters as well as the physicochemical parameters used in the model [58].
Parameter Description Quantity Dimension

L Gas channel length 0.07112 (m)

HCH Gas channel height 7.62 × 10−4 (m)
HGDL GDL thickness 2.54 × 10−4 (m)
HCL CL thickness 2.87 × 10−5 (m)
K gdl GDL permeability 1.76 × 10−11 (m2)
K cl CL permeability 1 × 10−14 (m2)
GDL GDL porosity 0.4
CL CL porosity 0.28
Tcell Cell temperature 323.15 (K)
Pa Anode operating pressure 3 (atm)
DHref2 H2 diffusion coefficient 0.915 × 10−4 (m2 s 1), (1 atm, 307 K)
−4
DHref2o H2O diffusion coefficient 0.256 × 10 (m2 s 1), (1 atm, 307 K)
a, c Anode and cathode transfer coefficient 0.5
R ohmic Ohmic resistance 0.5 × 10−4 ( m2)
ref
ioc Cathode exchange current density 63.8 (A m 2)
Voc Open circuit voltage 1.04 (V)
Fuel stoichiometric flow ratio 3
µH2 H2 dynamic viscosity 0.993 × 10−5 (Pa s)
µH 2o H2O dynamic viscosity 3.5 × 10−5 (Pa s)
c, GDL GDL contact angle 110 (°)
c, CL CL contact angle 95 (°)
ke Evaporation rate constant 1 × 10−4 (Pa 1 s 1)

kc Condensation rate constant 5000 (s 1)

Surface tension 0.0625 (Nm 1)

Fig. 2. Their experiments were carried out in single cells made with
carbon end plates. The cells were equipped with a reversible hydrogen
reference electrode. The graphite endplates contain gas feed inlets and
outlets, ribbed channels for gas flow behind the porous gas diffusion
electrodes, and holes for cartridge heaters and a thermocouple. The
active electrode area (geometric) in this cell was 5 cm2. The polariza-
tion curve in Fig. 2 is the result of experimental conditions such as 4%
Nation-impregnated Prototech electrodes (0.35 mg Pt/cm2) and H2/air
fuel at 3/5 atm at a temperature of 50 °C [58]. The geometric para-
meters as well as the physical and chemical parameters used in mod-
eling are listed in Table 4. According to the results, a good agreement
with experimental results was observed at lower current densities;
while at the high current densities (> 0.4 A cm−2), deviation from
experimental results was observed for the predicted results (numerical
results predicted a higher current density than experimental measure-
ments). The main reason for this deviation is related to this fact that in
the present work, the simulation is performed only for the anode side of
the PEMFC and the cathode side is not considered. At high current
densities, a significant amount of water is produced at the cathode
which leads to cell voltage drop (concentration losses). Furthermore, a
significant reduction in oxygen concentration leads to an increase in the
cathode activation overpotential and thus, the cell voltage is reduced.
Not considering these two changes led to the deviation of the numerical
results from the experimental results.
4.2. Effects of CO poisoning on the cell performance
With the presence of CO in the anode fuel (even at very low con-
centrations), the fuel cell performance is strongly affected. In this sec-
tion, the steady-state and dynamic responses of PEMFC under CO poi-
soning are investigated.
Fig. 3 presents the semi-log plots of the hydrogen electrode polar-
ization as a function of the current density with various concentration
of CO in the anode gas stream, as well as the experimental results of Lee

Table 5
The geometric and physicochemical parameters used in the modeling of CO poisoning and air (or O2) bleeding in the anode side of PEMFC.
Parameter Description Case 1 Case 2
[24] [15]

R ohmic ( m2) Ohmic resistance 0.1 × 10−4 0.3 × 10−4

ref
ioc (A m 2) Reference exchange current density for cathode 0.32 9
Voc (V) Open circuit voltage 1.24 1.2
Pa (atm) Anode operating pressure 1 3
Tcell (°C) Cell temperature 65 80
Fuel stoichiometric flow ratio 1.5 1.5
c Cathode transfer coefficient 1 0.5
a Anode transfer coefficient 0.36 0.5
RHa (% ) Relative humidity of inlet fuel 100 100
pt (C m
2) Molar area density of catalyst sites times Faraday constant 2.5 × 102 0.6 × 102
L (m) Gas channel length 0.07 0.07112 (Assumed)
HCH (m) Gas channel height 1 × 10−4 7.62 × 10−4 (Assumed)
HGDL (m) GDL thickness 4 × 10−4 2.54 × 10−4 (Assumed)
HCL (m) CL thickness 2.87 × 10−5 (Assumed) 2.87 × 10−5 (Assumed)

525
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

Fig. 4. Effect of CO concentration on cell performance.

Fig. 3. Anode overpotential as a function of the current density with various
concentration of CO in the anode gas stream for Pt/C catalyst and the operating
were containing 0.4 mg cm−2 platinum (20 wt% Pt on Vulcan XC-72).
temperature of 85 °C [22].
All the electrodes were impregnated with 1.9 mg cm−2 (dry amount
basis) of Nafion (Aldrich Chemical Co.) [22]. It can be noticed from the
Table 6 Fig. 3 that the anode overpotential increases with an increment of CO
Kinetic parameters used in the present model [19,23,30]. concentration in the anode gas streams. As shown in Fig. 3, a good
Reaction rate constants Activation energy agreement is observed between numerical and experimental results.
Tables 6 and 7 list a complete set of the model parameters used in this
steady-state simulation.
k = 10.4(kJ mol 1)
k ref
fh = 238.53 (A m
2 Pa 1) E fh
ref
keh = 2.96 × 106 (A m 2 ) k = 16.7(kJ mol 1)
Eeh An analysis of the results shown in Fig. 3, reveals that there is a
ref
k fc = 47214.59(A m 2 Pa 1) k = 47.3(kJ mol 1)
E fc critical current density regime (the current density which indicates the
ref k
location of the steep inflections in Fig. 3) at high concentrations of CO
kec = 243.6 × 1014 (A m 2 ) Eec = 127(kJ mol 1)
(higher than 20 ppm) so that, beyond this critical current density, the
k ref
fo = 0.876 ( ) (A m
F
RT
2 Pa 1) k = 14.2(kJ mol 1)
E fo
anode catalyst layer is strongly affected by CO poisoning [59]. At the
ref
koh = 9.41 × 1011 (F )(A m 2 ) k = 65.9(kJ mol 1)
Eoh current density lower than the critical regime (from 0 to 0.4 A cm−2 for
ref
koc = 8.3 × 1014 (F ) (A m 2 ) k
Eoc = 90(kJ mol 1) 20 ppm and from 0 to 0.3 A cm−2 for 50 ppm and from 0 to 0.2 A cm−2
ref
bfh = 5.08 × 1017 (Pa) E bfh = 87.9(kJ mol 1) for 100 ppm), a relatively small increase in the anode overpotential is
observed. When the output current density becomes greater than
bref = 1.9 × 1012 (Pa) E bfc = 100(kJ mol 1)
fc
0.4 A cm−2 for 20 ppm and greater than 0.3 A cm−2 for 50 ppm and
bref E bfo = 250(kJ mol 1)
greater than 0.2 A cm−2 for 100 ppm the increase in anode over-
36
fo = 1.36 × 10 (RT ) (Pa)

k1ref = 79.124 × 106 (A m 2) Ek1 = 78.1(kJ mol 1) (fitted) potential is significant which results in a sharp drop in the fuel cell
k2 = 0.1353 × 10 7 (A m 2Pa 1) k 2 = 86.8383 × 10 4 (A m 2 ) voltage. This significant raise in the anode overpotential leads to a great
k 1 = 0.03383 × 10 9 (A m 2Pa 1) acceleration in the CO electro-oxidation rate, which rapidly removes
the CO from the catalyst surface [59]. When the current density further
increases (more than critical current density), the slope of the change is
et al. [22]. In their experiments, anode electrodes were prepared using reduced, and the anode overpotential increases with a slight gradient.
catalyst powders (20 wt% Pt and PtRu (a/o 1:1) supported on Vulcan The effect of CO concentration on the dynamic response of the cell
XC-72 carbon). A brushing/rolling technique made the electrodes with current density is shown in Fig. 4. The parameters used in the modeling
a total catalyst loading of approximately 0.4 mg cm−2. The cathodes are listed in Tables 6 and 7.

Table 7
Electrochemical parameters used in the simulation [22].
Parameter Definition Value Dimension

Pt Molar area density of catalyst sites (for Pt/C) 9.65 C m−2

Pt Molar area density of catalyst sites (for PtRu/C) 8.68 C m−2
Ru Molar area density of catalyst sites (for PtRu/C) Pt /2 C m−2
co Transfer coefficient of CO 0.15 –
a, c, h Anode, cathode and H2 transfer coefficient 0.5 –
mem Membrane dry density 1980 kg m−3
EW Membrane dry equivalent weight 1100 kg mol−1
Interaction parameter 0.1 –
Voc Open circuit voltage 1.2 V
R ohmic Ohmic resistance 0.3 × 104 Ω m−2
ref
ioc Reference exchange current density for cathode 7 A m−2
n Order of reaction 2 –

526
A. Moradi Bilondi et al.
Fig. 5. Variation of output current density vs. time at a voltage of 0.6 V by removing CO from the CO-poisoned half-cell and compared with the experimental data
[24].
When CO is introduced into the H2 fuel stream, the current density
decreases sharply (a drop in the current density from 1026 to
702 mA cm−2 at 15 ppm), this dramatic drop shows that the presence of
very low levels of CO has a significant detrimental effect on the cell
performance. After a long time, the current density eventually reached
to a steady state. By increasing the CO concentration, the time required
to reach a steady state current density is decreased. For example, at
15 ppm, the steady state time is about 33 min, while at 135 ppm, this
steady state time is reduced to 7 min. CO concentration also affects the
steady-state current density. As the CO concentration is increased, the
steady-state current density decreases. For example, the steady-state
current density decreases from 702 to 299 mA cm−2 when the CO level
is increased from 15 to 135 ppm. Although increasing in CO con-
centration decreases the steady-state current density, the amount of
these changes decreases with increasing CO level, so that the change in
the steady state current density at 120–135 ppm is very small.
Therefore, the steady-state current density is more sensitive to changes
in CO concentration at low levels of CO.
Fig. 5 plots the variation of output current density over time at
53 ppm CO. The simulation was performed for two different cathode
fuel supplying (air and O2). The modeling results are seen to be in
reasonable agreement with the experimental data of Sung et al. [24]. In
the experiments of Lee et al. [24] a single cell was used for all of the
experiments. The membrane electrode assembly (MEA) used Gore
PRIMEA® 5621 with an electrode active area of 25 × 10−4 m2
(5 cm × 5 cm). The electrodes catalyst loading for the anode and
cathode sides were 0.45 mg cm−2 Pt-Ru and 0.6 mg cm−2 Pt, respec-
tively. Pure hydrogen and hydrogen pre-mixed with 53 ppm CO were
used as fuels for the anode in their experiments. Air and oxygen were
used as cathode feeds. Nitrogen was used for purging the system [24].
The parameters used in the modeling are listed in Table 5 (case 1). The
amount of bfcref is considered equal to 58.4 × 1012 Pa as a fitting para-
meter to the experimental data. In Fig. 5, the anode fuel stream is
changed from pure H2 to H2/CO and then from H2/CO back to pure H2.
At t = 15 min, 53 ppm CO is introduced into the pure H2 feed stream,
resulting in a dramatic drop in output current density over 10 min
(t = 25 min). When the oxygen is supplied at the cathode, the current
density is decreased about 87% (a drop in current from 1925 to
250 mA cm−2) with a degradation rate equal to 168 mA cm−2 min−1.
When the air is supplied at the cathode, the current density decreased
by about 85% (a drop in the current density from 1050 to
150 mA cm−2), with a degradation rate equal to 90 mA cm−2 min−1.
527
Energy Conversion and Management 177 (2018) 519–534
After about 13 min from the start of contamination (t = 28 min), the
current density reached to a steady state condition, which is equal to
243 mA cm−2 for O2 and 135 mA cm−2 for air. This steady-state current
density represents the balance between CO absorption and desorption.
When the fuel changes from H2/CO back to pure H2 (t = 52 min), the
output current density is increased gradually. The recovery rate is slow,
so that after a long time (about 60 min) the current density is reached
90% of the initial value, indicating that the effect of CO poisoning has
not completely disappeared. When the anode feed stream is changed
from H2/CO back to pure H2, due to the absorption of CO molecules on
the surface of the anode catalyst, these CO absorbed molecules remain
on the catalyst surface.
4.3. Effect of Pt-Ru alloys
Using an alternative catalyst (or CO-tolerant catalyst) is one of the
mitigation methods to lessen the CO poisoning effects. Here, platinum/
ruthenium (PtRu/C) electrodes are considered, and the performance of
a fuel cell under CO poisoning using this type of electrodes is modeled
numerically. The kinetic reaction parameters used in the simulations
are listed in Table 6, which were obtained from fitting the model pre-
dictions to the experimental measurements reported in [22]. Fig. 6
shows the comparisons between model predictions and experimental
data [22] of the anode polarization for PtRu/C electrode at the oper-
ating temperature of 58 °C and different CO concentration level in the
supplied hydrogen. This figure illustrates a good agreement between
the simulated results and the experimental data. The results indicate
that an increase in the CO levels results in increasing the anode over-
potential and degradation of the anode performance.
In Fig. 7, the anode polarization as a function of current density has
been plotted for Pt/C and PtRu/C. As the figure clearly illustrates, the
trend of the results for Pt/C and PtRu/C are the same, but the anode
overpotential obtained from the simulation of PtRu/C is less than Pt/C
results, which suggests a better cell performance. Also, at low current
densities, there is no difference between the results of Pt/C and PtRu/C,
but as the current density increases, the anode overpotential for Pt/C
takes larger values than PtRu/C. Furthermore, when the CO con-
centration decreases, the difference between the results becomes in-
significant, so that at 5 ppm CO the difference between the results is
negligible.
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

4.4. Effects of air/O2 bleeding on the cell performance

A known method for preventing the degradation of the fuel cell

Fig. 6. Semi-log plot of polarization of the hydrogen electrode for PtRu/C at
several CO concentrations and comparison between modeling results and ex-
perimental data [22].
Fig. 7. Comparison between the polarization of the hydrogen electrode for Pt/C
and PtRu/C at different CO concentrations.
performance in the presence of CO is to inject air or pure O2 into the
anode fuel stream, the so-called O2 bleeding method. By adding a small
amount of oxygen to the fuel, it is provided an additional pathway for
CO oxidation on the Pt surfaces, and the output current density is re-
covered and increased immediately. Fig. 8 shows the variation of the
current density at a voltage of 0.6 V over time. In this figure, a com-
parison between the experimental results [24] and the results of nu-
merical modeling has been made. Although the simulation does not
match the experimental data exactly for all cases, it does predicts the
general trend of transient CO poisoning and air bleeding, and demon-
strates the combined effects that air and CO can have on the cell per-
formance. The cathode is fed with pure O2 during the entire simulation.
At t = 1.5 min, 53 ppm CO is introduced into the fuel, which leads to
reduction of the output current density (from 1790 to 240 mA cm−2)
over 10 min. At t = 27.5 min, different levels of air (2%, 3%, 5%, 20%
air) are introduced into the CO-poisoned cell, and as the results in-
dicate, the output current density is increased quickly. The reason for
this fast recovery is that oxygen molecules react and oxidize the ad-
sorbed CO molecules, which results in the mitigation of CO and pro-
duction of CO2. Removing CO from the surface catalyst leads to an
increase in the available surfaces for HOR reaction and then, more H2 is
oxidized, and eventually, the output current density is increased.
When 2% air is injected into hydrogen fuel, the output current
density rises rapidly from 240 to 910 mA cm−2 (about 50% of the initial
current density is recovered). When 3% air is injected, the current
density is increased to 1275 mA cm−2 (about 71% of the initial current
density is recovered), when 5% air is injected, the current density is
increased to 1442 mA cm−2 (about 81% of the initial current density is
recovered), and when 20% air is injected, the current density reached
1646 mA cm−2 (about 92% of the initial current density is recovered).
In the low air bleeding region (0–3%), the current density is sig-
nificantly increased and recovered as the air bleeding concentration is
increased. In the high air bleeding region (> 5%), only a minor im-
provement of the output current density is seen as the air bleeding
concentration is increased. This suggests that air bleeding at a low air
concentration is more effective than that at a high air concentration. At
a high air concentration level, nitrogen may dilute the concentration of
hydrogen and part of the hydrogen may also be oxidized by the oxygen
in the air.
By injecting air into a poisoned fuel (in order to mitigate the CO

Fig. 8. Effect of anode air bleeding on the output current density and compared with the experimental data [24].
528
A. Moradi Bilondi et al.
Fig. 9. Effect of hydrogen dilution on CO poisoning [15].
poisoning effects), the nitrogen molecules are also injected into the fuel,
which results in dilution the hydrogen fuel. It is assumed, the nitrogen
molecules do not participate in the surface reactions and only dilute the
hydrogen fuel. According to this, the effect of hydrogen dilution on the
performance of the cell is studied, and the numerical results are com-
pared with the experimental results of Bhaita et al. [15]. The experi-
mental tests were performed in a 50 × 10−4 m2 titanium fuel cell. Both
anode and cathode flow fields consisted of 6 parallel channels following
a serpentine path. The membrane electrode assemblies (MEAs) were
composed of 40% platinum on carbon, Nafion® 112, with a platinum
catalyst loading of 0.5 mg cm−2 on each side. Although a platinum-
ruthenium combination is known to be more CO-tolerant, pure pla-
tinum was chosen as the anode catalyst in order to undertake a basic
study of the CO poisoning mechanism. Fig. 9 plots the variations of
current density over time at a voltage of 0.6 V for different anode feed
gas compositions. In this figure, a comparison between the experi-
mental results [15] and the results of numerical modeling has been
made. Although the simulation does not match the experimental data
exactly for all cases, it does predict the general trend of transient poi-
soning and demonstrates the combined effects that CO and hydrogen
dilution on the performance of fuel cell. The parameters used in the
modeling are listed in Table 5 (case 2). In the simulation of this section,
the amount of k fhref
is considered equal to 97.5 A m 2 Pa 1 as a fitting
parameter to the experimental data. According to the obtained results,
when the CO level is increased from 10 ppm to 100 ppm, the drop in the
current density is also increased from 160 mA cm−2 to 412 mA cm−2.
Therefore, increasing the concentration of CO leads to an increase in the
loss of current density. Also, hydrogen dilution increases the drop in the
current density. For example, at 100 ppm CO in a 100% hydrogen feed
stream, the drop in the current density is equal to 412 mA cm−2 (about
69% of the initial current density has been dropped) and with 40%
hydrogen, the drop in the current density is equal to 519 mA cm−2
(about 87% of the initial current density has been dropped). Hydrogen
dilution causes less hydrogen to reach the catalyst level, and the HOR
reaction will be reduced, and thus less current will be generated.
Fig. 10, shows the comparisons between the effect of air bleeding
and oxygen bleeding on the recovery of the cell performance at an
operating temperature of 65 °C. The vertical axis of this graph, re-
presents the ratio of the steady-state current density after air or oxygen
bleeding to the initial current density before contamination of the hy-
drogen fuel stream by CO (iss/iinitial). An increase in this ratio indicates
the increase in the recovery current density and, consequently, an im-
provement in the cell performance. Fig. 10 clearly illustrates that the
oxygen bleeding has a much greater effect than air bleeding on the
529
Energy Conversion and Management 177 (2018) 519–534
Fig. 10. Comparison between the effect of oxygen bleeding and air bleeding on
the cell performance at the operating temperature of 65 °C.
recovery of the lost current density, so that injection of 20% air is
equivalent to the injection of 5% oxygen and both has the same effect
on the cell performance. The reason is that when the air is injected into
the anode, nitrogen in the air is also injected into the fuel and dilutes
the concentration of the hydrogen fuel and reduces the cell perfor-
mance.
Fig. 11a and b plot the current density ratio, iss/iinitial, as a function
of CO concentration for different air and oxygen bleeding concentra-
tions (2% to 40%) and at the operating temperature of 65 °C. Fig. 11
shows that an increase in the concentration of CO decreases the current
density ratio. In other words, increasing the CO concentration reduces
the air or oxygen bleeding effects. Although an increase in the CO
concentration reduces the current density ratio, the amount of these
changes decreases with increasing air or oxygen bleeding concentra-
tion. In fact, increasing the air or oxygen bleeding concentration, re-
sulted in a nonlinear decrease in the slope of the change in such way
that this slope reaches nearly to zero when the air bleeding con-
centration is greater than 40% or oxygen bleeding concentration is
greater than 20%. In this condition, an increase in the CO levels has a
negligible effect on the current density ratio, and the current density
ratio is not affected by CO concentration. The reason is that in this
condition, the concentration of oxygen in the fuel is high and an in-
crease in CO level results in a minor reduction of O2 molecules in the
fuel stream (through the oxidation reaction of CO by oxygen).
With further scrutiny in the results of Fig. 11, it is observed that by
increasing the air or oxygen bleeding concentration, the current density
ratio is increased, but the amount of these changes decreases with in-
creasing the air or oxygen bleeding concentration, so that injection of
more than 30% air or injection of more than 20% oxygen not only does
not increase the current density ratio, but also leads to a loss of current
density and reduction of the current density ratio. This is associated
with the oxidation of CO and hydrogen by oxygen reactions (Eqs. (6)
and (7)). When a high oxygen concentration is injected into the fuel
stream, the excessive air or oxygen bleeding results in a reduction of the
overall hydrogen available for reaction from the catalyst surface which
reduces the cell performance. Also, electro oxidation of hydrogen
generates heat on the surface of the anode catalyst, which it can result
in destruction of the anode catalyst layer.
The effect of oxygen bleeding on the reduction of the destructive
effects of CO poisoning for Pt/C and PtRu/C catalysts are compared in
Fig. 12. When 2% oxygen is introduced into the fuel, in both Pt and
PtRu catalysts, the corresponding anode polarizations are dropped to
the same level. This can be explained by the kinetic reactions occur on
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

(a) (b)
Fig. 11. Plots of iss/iinitial versus CO concentration under various anode air (and O2) bleeding conditions.

Fig. 12. Effect of O2 bleeding for PtRu/C and compare the results with O2 Fig. 13. Semi-log plot of the simulated polarization curves for the hydrogen Pt/
bleeding for Pt/C. C electrode with 20 ppm CO at several operating temperatures compared with
the experimental data [22].
the catalyst surface. Actually, in this study, it is assumed that the re-
action between oxygen molecules and ruthenium surfaces is insignif-
icant compared to O2 adsorption on Pt [33].

4.5. Effect of operating temperature on the performance of CO-poisoned

PEMFC

As explained earlier, alleviation of temperature is one of the miti-

gation techniques which is investigated in this section. It is worth
mentioning that increasing the operating pressure is also another way
that leads to enhancing the cell performance through increasing the
hydrogen concentration on the catalyst surfaces. Since the effect of
pressure has been presented briefly in some previous works in the lit-
erature [5,19], it is wrapped up with just presenting in the Appendix A
and in following, just the effect of temperature is presented. At first, the
effect of operating temperature on the steady-state response of fuel cell
is investigated. Fig. 13 shows a semi-log plot of polarization obtained
for the hydrogen Pt/C electrode with 20 ppm CO at several operating
temperatures. In addition, a comparison between the modeling results
and the experimental data of Lee et al. [22] is presented in this figure
which shows a reasonable agreement between the experimental data Fig. 14. Effect of the cell temperature on oxygen bleeding.

530
A. Moradi Bilondi et al.
Fig. 15. Comparison of the simulated semi-log plot of the polarization of the
hydrogen PtRu/C electrode with 20 ppm CO at several temperatures and ex-
perimental data [22].
and the modeling results. Tables 6 and 7 list a complete set of the model
parameters used in this steady-state simulation. The results indicate
that an increase in the operating temperature of the fuel cell leads to
reduction of the anode overpotential and enhancement of the fuel cell
performance. Therefore, increasing the operating temperature can be
assumed as a mitigation technique in order to reduce the CO poisoning
effects. Although increasing in temperature decreases the anode over-
potential, the amount of these changes decreases with increasing tem-
perature, so that the change in the anode overpotential at 100–115 °C is
not significant.
Fig. 14, illustrates the effect of different oxygen bleeding con-
centration (2–40%) on the recovery of the lost current density at the
operating temperatures of 65 °C and 80 °C. With scrutiny in the results
of Fig. 14, it is observed a limiting oxygen bleeding concentration in
which, injection of oxygen more than this critical value, reduces the
current density ratio and the cell performance. When the operating
temperature is increased from 65 °C to 80 °C, the current density ratio is
decreased for the oxygen bleeding concentration less than the critical
oxygen bleeding concentration. As the temperature is increased, both
the initial and steady-state current densities (iinitial and iss) increase.
Fig. 16. Effect of temperature on dynamic behavior of CO coverage at the
voltage of 0.5 V for Pt/C and PtRu/C and CO concentration of 50 ppm.
531
Energy Conversion and Management 177 (2018) 519–534
Fig. 17. Effect of temperature on dynamic behavior of the current density at the
voltage of 0.5 V for Pt/C and PtRu/C and CO concentration of 50 ppm.
However, the increase in the initial current density is higher than the
steady-state current density, and thus the current density ratio is de-
creased. Also, an increase in the temperature reduces the critical
oxygen bleeding concentration (the critical oxygen bleeding con-
centrations at the operating temperatures of 65 °C and 80 °C are about
20% and 13%, respectively). This creates a further limitation to in-
crease the current density ratio and recovery of the lost cell perfor-
mance.
The effect of operating temperature on steady-state response of fuel
cell with the use of PtRu alloy catalysts is shown in Fig. 15. This figure
shows the semi-log plot of polarization of the hydrogen PtRu/C elec-
trode with 20 ppm CO at several temperatures and also shows the
comparison between the modeling results and the experimental data of
Lee et al. [22]. The simulation results present a good agreement with
the experimental data for the studied operating conditions. Similar to
the results obtained for Pt/C, these results indicate that an increase in
the operating temperature of the fuel cell will lead to a reduction of the
anode overpotential and improvement of the fuel cell performance.
The effect of temperature on dynamic behavior of cell performance
at the voltage of 0.5 V for Pt/C and PtRu/C and CO concentration of
50 ppm has been investigated. Tables 6 and 7 list the complete set of the
model parameters used in the simulations. Figs. 16 and 17 show the
variation of the CO coverage and current density over time for different
operating temperatures (50 °C, 60 °C, 70 °C, 80 °C and 90 °C). It is ob-
served that by increasing the cell operating temperature, the time re-
quired to reach a steady state is shorter. This can be explained through
Eqs. (29) and (30), where an increase in the temperature leads to an
increase in the reaction rates. Therefore, by increasing the adsorption
and desorption rate of species (H2 and CO), the steady-state time be-
comes shorter. Also, an increase in the adsorption and desorption rate
of CO results in reduction of the overall bond of CO on the Pt and PtRu
catalyst sites. Fig. 16 shows that by increasing the temperature from
50 °C to 90 °C, the steady state CO coverage is decreased from 0.66 to
0.43 for Pt/C and is decreased from 0.55 to 0.42 for PtRu/C.
For both Pt and PtRu catalysts, by increasing the operating tem-
perature, the initial current density is increased (at the cell tempera-
tures of 50 °C, 60 °C, 70 °C, 80 °C and 90 °C the initial current density is
about 0.59, 0.65, 0.72, 0.78 and 0.84 A cm2, respectively). Also, an
increase in the temperature leads to a higher steady state current
density. At the cell temperatures of 50 °C, 60 °C, 70 °C, 80 °C and 90 °C
the steady state current density is about 0.08, 0.138, 0.232, 0.378 and
0.566 A cm2 for Pt/C and 0.128, 0.19, 0.28, 0.407 and 0.578 A cm2 for
PtRu/C, respectively. So, it can be concluded that the fuel cell with the
A. Moradi Bilondi et al. Energy Conversion and Management 177 (2018) 519–534

PtRu catalysts has a higher steady state current density than Pt/C.
Furthermore, by increasing the temperature, the difference between the
steady state current density for Pt and PtRu catalysts becomes smaller.
At the operating temperature of 50 °C the difference is about
0.05 A cm−2 but at the temperature of 90 °C the difference is decreased
to 0.015 A cm−2. This suggests that using PtRu/C instead of Pt/C at
high cell temperature is not reasonable, i.e., at high temperature, using
these two mitigation techniques simultaneously is not economically
reasonable (due to the higher cost of PtRu/C than Pt/C).
5. Conclusion
In this study, the transient and steady-state effects of hydrogen fuel
contamination on the cell performance and the mitigation techniques
for reducing the effect of CO poisoning were investigated. The results of
this research are as follows:
• The CO poisoning process and its detrimental effects occur over a
significant period of time (in the range of hours for cases presented
in this paper), and the case with higher CO level has a faster drop in
the current density, and it takes less time to reach to the steady state
condition.
• The presence of very low levels of carbon monoxide has a significant
detrimental effect on the cell performance (about 30% of the output
current was lost within 30 min when the hydrogen was pre-mixed
with 15 ppm of CO as the fuel).
• When the CO concentration in the anode inlet feed is increased, the
sharp voltage drop (critical current density) begins at lower current
densities.
• The steady-state current density is more sensitive to change in CO
concentration at low levels of CO.
• Injection a small amount of air (or oxygen) into the contaminated
anode feed stream leads to a fast and significant performance re-
covery (by injecting about 5% air into the fuel, 80% of the output
current was recovered within 3 min at 53 ppm CO). Higher air
bleeding ratio only resulted in minor improvement of the cell per-
formance (injection between 5% and 30% air). Injection a sig-
nificant amount of air, does not increase the current density, but
leads to reduction of the output current density (injection more than
30% air).
• Hydrogen dilution in the presence of CO worsens the loss of the
output current density (an increase in the loss of the current density
from 69% to 87% due to the dilution of the fuel at a concentration of
100 ppm CO).
• An increase in the operating temperature of the fuel cell leads to
increase in the reaction rates and reduction in the anode over-
potential and subsequently, enhancement of the dynamic and
steady-state cell performance for both Pt/C and PtRu/C.
• Using PtRu/C instead of Pt/C as a CO-tolerant catalyst (at low op-
erating temperature) results in improving the fuel cell performance.
Using PtRu/C at a high operating temperature only resulted in
minor improvement of the cell performance.
Acknowledgment
Financial supports from Alexander von Humboldt Foundation
(Germany) and Iran’s National Elites Foundation (Iran) are acknowl-
edged.

Appendix A

The effect of operating pressure on the cell performance has been investigated and the numerical results has been presented in this section. Figs.
A.1 and A.2 illustrate the steady state and transient effect of pressure on the output current density.
According to the results, when the operating pressure is increased, the effect of CO poisoning is mitigated and the output current density is
increased and thus the cell performance is improved. Because increasing the pressure results in an increase in hydrogen concentration (see Eq. (16)),
so the performance is increased. Increasing pressure improves the cell performance; however, the benefits from increased pressure diminish as
pressure increases. Actually, an increase in pressure will increase the H2 concentration more than the CO concentration (see Eqs. (16) and (17)).
However, the sticking probability of CO on platinum is much more that of H2. CO coverage of the catalyst benefits much more than hydrogen
coverage if the concentration is increased. Therefore, the benefits of an increased H2 concentration is balanced by the increase in CO concentration
and the decrease in anode overpotential due to increased pressure diminishes at high pressures.
Also, Fig. A.2 shows that the effect of using PtRu/C as a CO-tolerant catalyst is diminished by increasing the pressure, so that at pressure higher
than 3 atm it is resulted in minor improvement of the cell performance

Fig. A.1. The effect of operating pressure on the anode overpotential in the anode gas stream for Pt/C catalyst with 100 ppm CO and the operating temperature of
80 °C.

532
A. Moradi Bilondi et al.
Fig. A.2. The effect of operating pressure on dynamic behavior of the current density at the voltage of 0.5 V for Pt/C and PtRu/C with 10 ppm CO and the operating
temperature of 80 °C.
References
[1] Heidary H, Abbassi A, Kermani MJ. Enhanced heat transfer with corrugated flow
channel in anode side of direct methanol fuel cells. Energy Convers Manage
2013;75:748–60.
[2] Thanaa F, Eskander MN, El-Hagry MT. Energy flow and management of a hybrid
wind/PV/fuel cell generation system. Energy Convers Manage 2006;47:1264–80.
[3] Hsieh S-S, Yang S-H, Kuo J-K, Huang C-F, Tsai H-H. Study of operational parameters
on the performance of micro PEMFCs with different flow fields. Energy Convers
Manage 2006;47:1868–78.
[4] Haddadi AH, Ghomian A, Yosefi SH, Torabi F. Design and thermoeconomic analysis
of a PEMFC-TEC cogeneration plant. In: 2nd Conf. Hydrog. Fuel Cell; 2012.
[5] Zamel N, Li X. Transient analysis of carbon monoxide poisoning and oxygen
bleeding in a PEM fuel cell anode catalyst layer. Int J Hydrogen Energy
2008;33:1335–44.
[6] Zamel N, Li X. Effect of contaminants on polymer electrolyte membrane fuel cells.
Prog Energy Combust Sci 2011;37:292–329.
[7] Baschuk JJ, Li X. Carbon monoxide poisoning of proton exchange membrane fuel
cells. Int J Energy Res 2001;25:695–713.
[8] Bellows RJ, Marucchi-Soos EP, Buckley DT. Analysis of reaction kinetics for carbon
monoxide and carbon dioxide on polycrystalline platinum relative to fuel cell op-
eration. Ind Eng Chem Res 1996;35:1235–42.
[9] Divisek J, Oetjen H-F, Peinecke V, Schmidt VM, Stimming U. Components for PEM
fuel cell systems using hydrogen and CO containing fuels. Electrochim Acta
1998;43:3811–5.
[10] Oetjen H-F, Schmidt VM, Stimming U, Trila F. Performance data of a proton ex-
change membrane fuel cell using H2/CO as fuel gas. J Electrochem Soc
1996;143:3838–42.
[11] Springer TE, Rockward T, Zawodzinski TA, Gottesfeld S. Model for polymer elec-
trolyte fuel cell operation on reformate feed: effects of CO, H; 2 dilution, and high
fuel utilization. J Electrochem Soc 2001;148:A11–23.
[12] Springer T, Zawodzinski T, Gottesfeld S. Modeling of polymer electrolyte fuel cell
performance with reformate feed streams: effects of low levels of CO in hydrogen.
Electrochem Soc Proc Ser 1997:97–113.
[13] Zhang J, Thampan T, Datta R. Influence of anode flow rate and cathode oxygen
pressure on CO poisoning of proton exchange membrane fuel cells. J Electrochem
Soc 2002;149:A765–72.
[14] Zhou T, Liu H. A 3D model for PEM fuel cells operated on reformate. J Power
Sources 2004;138:101–10.
[15] Bhatia KK, Wang C-Y. Transient carbon monoxide poisoning of a polymer electro-
lyte fuel cell operating on diluted hydrogen feed. Electrochim Acta
2004;49:2333–41.
[16] Chu HS, Wang CP, Liao WC, Yan WM. Transient behavior of CO poisoning of the
anode catalyst layer of a PEM fuel cell. J Power Sources 2006;159:1071–7.
[17] Wang C-P, Chu H-S. Transient analysis of multicomponent transport with carbon
monoxide poisoning effect of a PEM fuel cell. J Power Sources 2006;159:1025–33.
[18] Gottesfeld S, Pafford J. A new approach to the problem of carbon monoxide poi-
soning in fuel cells operating at low temperatures. J Electrochem Soc
1988;135:2651–2.
[19] Baschuk JJ, Li X. Modelling CO poisoning and O2 bleeding in a PEM fuel cell anode.
Int J Energy Res 2003;27:1095–116.
[20] Murthy M, Esayian M, Hobson A, MacKenzie S, Lee W, Van Zee JW. Performance of
a polymer electrolyte membrane fuel cell exposed to transient CO concentrations. J
Electrochem Soc 2001;148:A1141–7.
[21] Murthy M, Esayian M, Lee W, Van Zee JW. The effect of temperature and pressure
on the performance of a PEMFC exposed to transient CO concentrations. J
Electrochem Soc 2003;150:A29–34.
533
Energy Conversion and Management 177 (2018) 519–534
[22] Lee SJ, Mukerjee S, Ticianelli EA, McBreen J. Electrocatalysis of CO tolerance in
hydrogen oxidation reaction in PEM fuel cells. Electrochim Acta 1999;44:3283–93.
[23] Shah AA, Sui PC, Kim G-S, Ye S. A transient PEMFC model with CO poisoning and
mitigation by O2 bleeding and Ru-containing catalyst. J Power Sources
2007;166:1–21.
[24] Sung L-Y, Hwang B-J, Hsueh K-L, Tsau F-H. Effects of anode air bleeding on the
performance of CO-poisoned proton-exchange membrane fuel cells. J Power
Sources 2010;195:1630–9.
[25] Heidary H, Kermani MJ, Dabir B. Influences of bipolar plate channel blockages on
PEM fuel cell performances. Energy Convers Manage 2016;124:51–60.
[26] Takalloo PK, Nia ES, Ghazikhani M. Numerical and experimental investigation on
effects of inlet humidity and fuel flow rate and oxidant on the performance on
polymer fuel cell. Energy Convers Manage 2016;114:290–302.
[27] Esmaili Q, Ranjbar AA, Abdollahzadeh M. Numerical simulation of a direct me-
thanol fuel cell through a 1D+ 1D approach. Int J Green Energy 2013;10:190–204.
[28] Ahmed NA, Miyatake M, Al-Othman AK. Power fluctuations suppression of stand-
alone hybrid generation combining solar photovoltaic/wind turbine and fuel cell
systems. Energy Convers Manage 2008;49:2711–9.
[29] Wang Q, Wang G, Tao H, Li Z, Han L. Highly CO tolerant PtRu/PtNi/C catalyst for
polymer electrolyte membrane fuel cell. RSC Adv 2017;7:8453–9.
[30] Enbäck S, Lindbergh G. Experimentally validated model for CO oxidation on PtRu/
C in a porous PEFC electrode. J Electrochem Soc 2005;152:A23–31.
[31] Gasteiger HA, Markovic NM, Ross Jr PN. H2 and CO electrooxidation on well-
characterized Pt, Ru, and Pt-Ru. 1. Rotating disk electrode studies of the pure gases
including temperature effects. J Phys Chem 1995;99:8290–301.
[32] Gasteiger HA, Marković N, Ross PN, Cairns EJ. Temperature-dependent methanol
electro-oxidation on well-characterized Pt-Ru alloys. J Electrochem Soc
1994;141:1795–803.
[33] Taylor KC. Determination of ruthenium surface areas by hydrogen and oxygen
chemisorption. J Catal 1975;38:299–306. https://doi.org/10.1016/0021-9517(75)
90091-3.
[34] Gileadi E, Kirowa-Eisner E, Penciner J. Interfacial electrochemistry: an experi-
mental approach. Addison-Wesley Pub. Co. Advanced Book Program; 1975.
[35] Hafttananian M, Ramiar A, Ranjbar AA. Novel techniques of oxygen bleeding for
polymer electrolyte fuel cells under impure anode feeding and poisoning condition:
a computational study using OpenFOAM®. Energy Convers Manage
2016;122:564–79.
[36] Yaghoubi A, Gheibi S, Torabi F. Numerical Simulation of Electrochemical Processes
in Polymeric Membrane Fuel Cells n.d.
[37] Ishii M, Mishima K. Two-fluid model and hydrodynamic constitutive relations. Nucl
Eng Des 1984;82:107–26.
[38] Liu X, Lou G, Wen Z. Three-dimensional two-phase flow model of proton exchange
membrane fuel cell with parallel gas distributors. J Power Sources
2010;195:2764–73.
[39] Havaej P, Kermani MJ, Abdollahzadeh M, Heidary H, Moradi A. A numerical
modeling study on the influence of catalyst loading distribution on the performance
of polymer electrolyte membrane fuel cell. Int J Hydrogen Energy 2018. https://
doi.org/10.1016/j.ijhydene.2018.04.063.
[40] Abdollahzadeh M, Ribeirinha P, Boaventura M, Mendes A. Three-dimensional
modeling of PEMFC with contaminated anode fuel. Energy 2018;152:939–59.
[41] Meng H. A two-phase non-isothermal mixed-domain PEM fuel cell model and its
application to two-dimensional simulations. J Power Sources 2007;168:218–28.
[42] Meng H. Multi-dimensional liquid water transport in the cathode of a PEM fuel cell
with consideration of the micro-porous layer (MPL). Int J Hydrogen Energy
2009;34:5488–97.
[43] Meng H. Numerical investigation of transient responses of a PEM fuel cell using a
two-phase non-isothermal mixed-domain model. J Power Sources
2007;171:738–46.
A. Moradi Bilondi et al.
[44] Ramiar A, Mahmoudi AH, Esmaili Q, Abdollahzadeh M. Influence of cathode flow
pulsation on performance of proton exchange membrane fuel cell with inter-
digitated gas distributors. Energy 2016;94:206–17.
[45] Abdollahzadeh M, Ranjbar AA, Esmaili Q. (1D+ 1D) approach mathematical
modeling of two phase multicomponent transport flow in PEMFC. Russ J
Electrochem 2012;48:1187–96.
[46] Wang X-D, Yan W-M, Duan Y-Y, Weng F-B, Jung G-B, Lee C-Y. Numerical study on
channel size effect for proton exchange membrane fuel cell with serpentine flow
field. Energy Convers Manage 2010;51:959–68.
[47] Mahmoudi AH, Ramiar A, Esmaili Q. Effect of inhomogeneous compression of gas
diffusion layer on the performance of PEMFC with interdigitated flow field. Energy
Convers Manage 2016;110:78–89.
[48] Berning T, Lu DM, Djilali N. Three-dimensional computational analysis of transport
phenomena in a PEM fuel cell. J Power Sources 2002;106:284–94.
[49] Khajeh-Hosseini-Dalasm N, Kermani MJ, Moghaddam DG, Stockie JM. A parametric
study of cathode catalyst layer structural parameters on the performance of a PEM
fuel cell. Int J Hydrogen Energy 2010;35:2417–27. https://doi.org/10.1016/j.
ijhydene.2009.12.111.
[50] Hu M, Gu A, Wang M, Zhu X, Yu L. Three dimensional, two phase flow mathe-
matical model for PEM fuel cell: Part I. Model development. Energy Convers
Manage 2004;45:1861–82.
[51] Ghanbarian A, Kermani MJ. Enhancement of PEM fuel cell performance by flow
channel indentation. Energy Convers Manag 2016;110:356–66.
534
Energy Conversion and Management 177 (2018) 519–534
[52] Zehtabiyan-Rezaie N, Arefian A, Kermani MJ, Noughabi AK, Abdollahzadeh M.
Effect of flow field with converging and diverging channels on proton exchange
membrane fuel cell performance. Energy Convers Manage 2017;152:31–44. https://
doi.org/10.1016/j.enconman.2017.09.009.
[53] Baboli MK, Kermani MJ. A two-dimensional, transient, compressible isothermal and
two-phase model for the air-side electrode of PEM fuel cells. Electrochim Acta
2008;53:7644–54.
[54] Min C-H. A novel three-dimensional, two-phase and non-isothermal numerical
model for proton exchange membrane fuel cell. J Power Sources 2010;195:1880–7.
[55] Perng S-W, Wu H-W, Wang R-H. Effect of modified flow field on non-isothermal
transport characteristics and cell performance of a PEMFC. Energy Convers Manage
2014;80:87–96. https://doi.org/10.1016/j.enconman.2013.12.044.
[56] Ghanbarian A, Kermani MJ, Scholta J, Abdollahzadeh M. Polymer electrolyte
membrane fuel cell flow field design criteria–application to parallel serpentine flow
patterns. Energy Convers Manage 2018;166:281–96.
[57] Lien FS, Yee E, Cheng Y. Simulation of mean flow and turbulence over a 2D building
array using high-resolution CFD and a distributed drag force approach. J Wind Eng
Ind Aerodyn 2004;92:117–58.
[58] Ticianelli EA, Derouin CR, Redondo A, Srinivasan S. Methods to advance tech-
nology of proton exchange membrane fuel cells. J Electrochem Soc
1988;135:2209–14.
[59] Ju H, Lee K-S, Um S. Multi-dimensional modeling of CO poisoning effects on proton
exchange membrane fuel cells (PEMFCs). J Mech Sci Technol 2008;22:991–8.