Energy 116 (2016) 1131e1144

Contents lists available at ScienceDirect

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

PEM fuel cell model and simulation in MatlabeSimulink based on

physical parameters
Z. Abdin, C.J. Webb, E.MacA. Gray*
Queensland Micro- and Nanotechnology Centre, Griffith University, Nathan 4111, Australia

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

Article history: An advanced PEM fuel cell mathematical model is described and realised in four ancillaries in the Matlab
Received 22 March 2016 eSimulink environment. Where possible, the model is based on parameters with direct physical
Received in revised form meaning, with the aim of going beyond empirically describing the characteristics of the fuel cell. The
4 October 2016
model can therefore be used to predict enhanced performance owing to, for instance, improved electrode
Accepted 12 October 2016
materials, and to relate changes in the measured performance to internal changes affecting influential
physical parameters. Some simplifying assumptions make the model fairly light in computational de-
mand and therefore amenable to extension to simulate an entire fuel-cell stack as part of an energy
Keywords:
Fuel cell
system. Despite these assumptions, the model emulates experimental data well, especially at high cur-
Anode rent density. The influences of pressure, temperature, humidification and reactant partial pressure on cell
Cathode performance are explored. The dominating effect of membrane hydration is clearly revealed.
PEM © 2016 Elsevier Ltd. All rights reserved.
Overpotential
Modelling

1. Introduction Excessive water leads to flooding of the electrode pores and

increased electrode resistance. Hence water management, and via
The minimal generation of by-products (ideally only water for the partial vapour pressure thermal management, are critical to the
hydrogen fuel) and the high electricity generation efficiency of the good performance of PEM fuel cells.
hydrogen fuel cell make this technology attractive for a range of Better design, selection of appropriate materials and optimiza-
applications such as automobiles and small-scale stationary elec- tion of the overpotential are necessary in order to improve the PEM
tricity generation. Based on the lower heating value of hydrogen, fuel cell performance. These issues can only be addressed efficiently
fuel cells have conversion efficiencies of 40e60%, higher than most if realistic mathematical process models are available to predict the
other energy conversion systems [1]. effects of different materials, construction and so on. A realistic
The solid polymer membrane makes the PEM fuel cell simpler model provides a framework not only for analysing the overall
and less hazardous than the alkaline type, although the PEM type is performance of the PEM fuel cell, but also for studying its various
more difficult to manage because of the need for proper hydration components and their mutual influences. An effective model thus
of the membrane [2]. becomes a research tool to explore parameters which are difficult
PEM fuel cell performance is conveniently visualized via the to measure e.g. water concentration at the membraneeelectrode
different overpotentials/voltage drops, i.e. the difference between interface, water uptake into the channel, etc.
the electrode potential and the equilibrium potential. An important In the past two decades many fuel-cell models have been pre-
goal of fuel cell design is to minimise the lossy contributions to the sented in the literature. The early history of PEM fuel-cell modelling
overpotential for a given current and so maximise efficiency. work has been reviewed by Bıyıkog
lu [3]. The advantages and
Membrane hydration, achieved via humidification of the reactant disadvantages of various conceptual approaches to PEM fuel-cell
gases, plays a dominating role in PEM fuel cell performance. Too modelling have been discussed by Cheddie and Munroe [4].
little water leads to dehydration and high membrane resistance. More recently, Djilali and Lu [5] and Costamagna [6] proposed a
multi-dimensional model and more complex approaches in 3D
modelling have also been developed for the PEM fuel cell [7e13].
Sophisticated modelling approaches have been applied to under-
* Corresponding author. standing various aspects of fuel-cell behaviour. Carton and Olabi
E-mail address: e.gray@griffith.edu.au (E.MacA. Gray).

http://dx.doi.org/10.1016/j.energy.2016.10.033
0360-5442/© 2016 Elsevier Ltd. All rights reserved.
1132 Z. Abdin et al. / Energy 116 (2016) 1131e1144

Glossary P Pressure, atm

p Partial pressure, atm
A Active area of MEA, cm2 psat Saturation pressure, atm
a Water activity R Universal gas constant, 8.3144 J mol
1K
1
C Molar Concentration, mol m
3 S Stoichiometry ratio
D Diffusion coefficient, m2 s
1 T Temperature, K
E Cell potential, V X Molar fraction
Ec Activation energy, kJ mol
1 d Thickness, mm
F Faraday constant, 96485 C mol
1 r Density of water, kg m
3
G Gibbs free energy of activation, kJ mol
1 smem PEM conductivity, S m
1
I Current, A nd Electro-osmotic drag coefficient
i Current density, A cm
2 a Charge transfer coefficient
i0 Exchange current density, A cm
2 l Degree of humidification
k Reaction rate coefficient 4 Relative humidity
MEA Membrane electrode assembly g Roughness factor
N_ Molar flow rate, mol s
1 ε Porosity
n Molar flux, mol m
2-s x Tortuosity

[14] developed a 3D model based on computational fluid dynamics applications, such as relating fuel-cell performance to the operating
(CFD) for open pore cellular foam material as a flow plate and they and underlying physical parameters.
found that the reactant gases were uniformly distributed from inlet This paper continues a series in which mutually compatible
to outlet. The volume-of-fluid (VOF) method has been used to modular models of the components of an energy system, specif-
numerically investigate the two-phase flow in an anode gas chan- ically one incorporating solar-derived hydrogen [23], are presented.
~o [15]. They found that for
nel of a PEM fuel cell by Ferreira, Falca One overall objective is to link the suite of component models
hydrophilic or hydrophobic channel walls water moves as a film or together to form a whole-of-system model and simulation that can
droplets respectively. Carton, Lawlor [16] also developed a model realistically predict the behaviour of a planned energy system, as a
based on CFD and VOF to investigate water droplet movement and prerequisite to specifying the characteristics and capacities (sizes)
slug formation in PEM fuel cell mini-channels. Kang [17] developed of its components. Alternatively, such a model can simulate the
a quasi-three dimensional dynamic model to investigate the tran- behaviour of an existing energy system for research and other
sient behaviour and dynamic characteristics of a PEM fuel cell. Xing, purposes. The component models are realised in Simulink® and are
Cai [18] developed a 2D model to investigate the effects of relative fully described so they can be recreated and used by others. The
humidity, stoichiometric flow ratio and channel length, as well as first such model [24] is of a PEM electrolyser cell. The next, for the
their interactive influence, on the performance of a PEM fuel cell. PEM fuel cell, is presented here.
Salva, Iranzo [19] developed a one dimensional model to investi- A second very important objective of the work is to use the
gate the effect of relative humidity on the performance of a PEM component models as research tools to predict the effects of
fuel cell and validated it using neutron imaging. Abdollahzadeh, influential design parameters, materials and environmental con-
Pascoa [20] developed a multi-component mixture model to ditions, thus lessening the need for experimentation, which is
simulate two phase flow and transport in the cathode gas diffusion expensive and time-consuming.
layer of a PEM fuel cell. Design parameters are necessarily physical in nature. The key to
Liu, Li [21] used a numerical model to optimise flow channel achieving the second objective is therefore to place the equations
dimensions for best fuel cell performance. The latter model pro- that embody the characteristics of the various elements of the
vides a simple example of the inclusion of physical parameters so electrolyser, fuel cell etc. on a proper physical basis by embedding
that the origin of the improved performance is identifiable as a the theory of the underlying physical mechanisms. Doing this has
physical entity. In contrast, in empirical modelling the improved significant advantages. Firstly, it makes the model more generic in
performance of a real fuel cell owing to, for instance, a better its application, in contrast to the present general reliance on
electrocatalyst, does not directly associate the improvement with empirical equations, with coefficients lacking physical significance,
its physical origin, since it manifests in the model as parameters that only describe the characteristics of the particular component
that may have no direct physical meaning. Ref. [11], based on under consideration, perhaps with no applicability to the greater
empirical equations and requiring high-performance computing, class of such components. Secondly, a sound physical basis for the
exemplifies this approach. model better allows the effects of changing materials, dimensions
Generally, a multi-dimensional model is useful, perhaps neces- etc. to be predicted in the quest for enhanced performance, by
sary, in order to understand the detailed behaviour of the indi- performing sensitivity analyses, for instance, as was demonstrated
vidual elements of a fuel cell, electrolyser, etc. and to guide the in our paper on PEM electrolysers [24].
process of designing for enhanced performance. However, to model A third benefit is that a realistic model may be used to diagnose
an entire energy system in this way would be very cumbersome. problems in the modelled component or system via changes in the
Thus there is a role for realistic modelling that is not overly fitted parameters arising from, for instance, degradation of an
computation-intensive, so that a whole-of-system model becomes electrode, potentially allowing the degradation to be traced to its
feasible with commonly available computing resources. A recent physical origin.
comparison of 1D and 3D fuel-cell models by Shekhar [22] con- An example relevant to both electrolysers and fuel cells is the
cludes that 1D models are useful and cost-effective for practical atomic-level mechanisms that control membrane processes
 Z. Abdin et al. / Energy 116 (2016) 1131e1144 1133

(humidification, diffusion, conductivity). In the fuel-cell case, these Protons are transported from the anode to the cathode through
are represented in all existing models known to the authors (largely the electrolyte membrane (a polymer of persulfonic acid groups
including this one) by purely descriptive empirical equations that with a Teflon backbone) and the electrons are carried to the cath-
do not show how the underlying mechanisms control the flow of ode through the external circuit.
charge. This is unsatisfactory and further work is needed to develop The mathematical model has been developed based on the
physically-based equations for modelling membrane processes in following simplifying assumptions, without losing any essential
both fuel cells and electrolysers. contribution to the cell performance.
In line with the objectives outlined above, a new one-
dimensional mathematical model of a PEM fuel cell is described a) The model is one dimensional and the current is uniformly
and embedded in a simulation in the Simulink® environment. This distributed (i.e. I ¼ iA), so the gases are considered to be
model is distinguished from the many other PEM fuel-cell models uniformly distributed in the cell with constant pressure.
in the literature by (i) contributions to the overpotential being tied b) The humidified reactant gases (i.e. hydrogen at anode and
to physically meaningful parameters wherever possible; (ii) the oxygen at cathode) are in equilibrium with liquid water.
ability to estimate the values of physical parameters that are diffi- c) The water activity is uniform across the membrane and is in
cult to access directly by experiment; (iii) simplifications that equilibrium with the water activity at the cathode and anode
reduce the computing demand without significantly impacting the catalyst layer.
ability to simulate real fuel-cell performance, making it feasible to d) Water is present as vapour at the membraneeelectrode
embed the model as a module in the model of an entire energy interface. This is equivalent to limiting the relative humidity
system; (iv) utility in analysing the sensitivity of the overall per- to 100%. Thus the effect of cathode flooding when the partial
formance to changes in particular physical parameters; (v) utility in pressure of water exceeds the saturation vapour pressure is
relating changes in the performance of a fuel cell to the changed not predicted quantitatively. This assumption is justifiable
characteristics of an internal component owing to ageing, for for fuel cells operating at moderately elevated temperatures,
instance. Despite the simplifications listed in the next section, the above about 40  C [25] which is likely to be the case in most
influences of pressure, temperature and humidification on the cell real installations, where the cell temperature is higher than
performance are captured in the new model with good fidelity. The the coolant temperature, which is higher than the ambient
model is validated against experimental results and then used to temperature. Water may condense in the cathode channel,
explore the influence of operational parameters (pressure, tem- but as long as it is effectively removed will not compromise
perature, and humidification) on the cell polarisation curve. the electrochemical characteristics of the mem-
braneecathode interface.
2. Model development e) There is no pressure gradient between the anode and cath-
ode (i.e. H2 and O2 are fed at the same pressure), so con-
The typical fuel cell shown in Fig. 1 is broken into four ancillaries vection is not incorporated and only diffusion is considered
for modelling purposes: anode, cathode, membrane and voltage. The for gas transport.
molar balance of reactant species and their partial pressures are f) The anode, cathode and membrane are at the same
modelled in the anode and cathode ancillaries. The water transport temperature.
mechanism and water uptake into the membrane are modelled in g) The PEM is not electrically conductive and is impermeable to
the membrane ancillary. The voltage ancillary models the neutral reactant gases and so there are no internal currents
overpotential. and no fuel crossover losses.
The humidified hydrogen is fed to the anode and oxidized on the
anode according to Eq. (1). Humidified oxygen/air is fed to the 2.1. Anode ancillary
cathode and reduced on the cathode according to Eq. (2), forming
water and producing heat: The molar balance equations for the anode are given by

Anode : H2 /2Hþ þ 2e
dNH2 cons

(1) ¼ N_ H2 ;in
N_ H2 ;out
N_ H2 (3)
dt
1
Cathode : O þ 2e
þ 2Hþ /H2 O (2) an
dNH an an mem
2 2 2O
¼ N_ H2 O;in
N_ H2 O;out
N_ H2 O (4)
dt
an an
where N_ H2 ;in , N_ H2 ;out , N_ H2 O;in , N_ H2 O;out are anode the inlet and outlet
cons
molar flow rates of hydrogen and water respectively, N_ H2 is the
_ mem
molar flow rate of consumed hydrogen at the anode and NH2 O is the
molar flow rate of membrane water from anode to cathode. The
mechanism of water transport is considered later.
According to Faraday's law the molar flow rate of hydrogen at
the anode inlet is
I
N_ H2 ;in ¼ SH2  NH
cons
¼ SH2  (5)
2
2F
Because of assumption (b), the molar fraction of water at the
anode inlet is

4an psat ðTÞ

XHan2 O;in ¼ (6)
Pan
Fig. 1. Schematic of the PEM fuel cell.
1134 Z. Abdin et al. / Energy 116 (2016) 1131e1144

where 4an is the relative humidity (i.e. the ratio of the partial cat
dNH cat cat gn mem
pressure of water to the saturated vapour pressure of water) at
2O
¼ N_ H2 O;in
N_ H2 O;out þ N_ H2 O þ N_ H2 O (18)
dt
anode, psat(T) is the saturation pressure and Pan is the anode
cat cat
pressure. where N_ O2 ;in , N_ O2 ;out , N_ H2 O;in , N_ H2 O;out are the cathode inlet and
The fractional ratio of the water to hydrogen gives the molar cons
flow rate of water at the anode inlet is outlet molar flow rates of oxygen and water respectively, N_ O2 is
gn
the molar flow rate of consumed oxygen at the cathode, N_ is the H2 O
XHan2 O;in I 4an psat ðTÞ mem
molar flow rate of generated water at the cathode and N_ H2 O is the
an
N_ H2 O;in ¼  N_ H2 ;in ¼ SH2   (7)
1
XHan2 O;in 2F Pan
4an psat ðTÞ
membrane molar flow rate of water from anode to cathode through
Thus the net molar flow rate at the anode inlet can be written as the membrane.
The molar flow rates of oxygen and water at the cathode inlet


I 4an psat ðTÞ are
N_ in;an ¼ SH2  1þ (8)
2F Pan
4an psat ðTÞ
cons I
N_ O2 ;in ¼ SO2  N_ O2 ¼ SO2  (19)
where I is the current through the cell, SH2 is the stoichiometry ratio 4F
(i.e. the ratio of the inlet flow rate of hydrogen to the rate at which it
cat I 4cat Psat ðTÞ
is consumed). The molar flow rates of hydrogen and water at the N_ H2 O;in ¼ SO2   (20)
anode outlet are then 4F Pcat
4cat Psat ðTÞ
  I and so the net molar flow rate at the cathode inlet can be written as
N_ H2 ;out ¼ SH2
1  (9)
2F

I 4cat Psat ðTÞ
N_ in;cat ¼ SO2  1þ (21)
an an mem
N_ H2 O;out ¼ N_ H2 O;in
N_ H2 O (10) 4F Pcat
4cat Psat ðTÞ
mem At the cathode side the molar flow rate of water is the sum of
The molar flow rate of membrane water N_ H2 O will be described
water generated by the reaction, the water transported from anode
detail in x 2.3 and the molar flow rate at the anode outlet can be
to cathode and the water at the cathode inlet i.e.
written as
gn mem cat
an
N_ out;an ¼ N_ H2 ;out þ N_ H2 O;out (11) N_ H2 O;cat ¼ N_ H2 O þ N_ H2 O þ N_ H2 O;in (22)

The molar flux of hydrogen through the anode can be expressed The molar flow rate of generated water is
as
gn I
N_ H2 O ¼ (23)
2F
N_ H2 ;in
nH2 ¼ (12)
A and the resultant molar flux of water through the cathode is

and the molar flux of water through the anode can be expressed as gn mem cat
N_ H2 O;cat N_ H2 O þ N_ H2 O þ N_ H2 O;in
an
nH2 O;cat ¼ ¼ (24)
N_ H2 O;in A A
nan
H2 O ¼ (13)
A and the molar flux of oxygen through the cathode is
The partial pressure of hydrogen and water at the anode is given
by N_ O2 ;in
nO2 ¼ (25)
A
pi ¼ Xi Pan (14)
The partial pressure of species at the cathode can be calculated
according to Eq. (14). The molar fractions at the cathode can be
where pi is the partial pressure of the hydrogen and water, Xi is the
expressed as
molar fraction of the hydrogen and water and Pan is the pressure at
the anode. The molar fractions at the anode are nO2
XO2 ¼ (26)
nH2 nO2 þ ncat
H2 O
XH2 ¼ (15)
nH2 þ nan
H2 O
ncat
H2 O
XHcat
2O
¼ (27)
nan
H2 O
nO2 þ ncat
H2 O
XHan2 O ¼ (16)
nH2 þ nan
H2 O

2.3. Membrane ancillary

2.2. Cathode ancillary The solid electrolyte in a PEM fuel cell is typically a per-
fluorinated ionomer, Nafion®, for example. An inert but chemically
The material balance equations for the cathode are given by stable polymer backbone which is not ionically conductive provides
chemical stability and durability. Side chains of sulfonic acid
dNO2 cons bonded to the backbone enable ionic conduction. The SO
3 chains
¼ N_ O2 ;in
N_ O2 ;out
N_ O2 (17)
dt are distributed as generally isolated clusters in the membrane [30].
 Z. Abdin et al. / Energy 116 (2016) 1131e1144 1135

2.3.1. Proton transport four water transport mechanisms and defining flow from the anode
When the Nafion® structure is hydrated, the hydrophilic sul- to the cathode as positive, the net molar flow rate of water through
mem
fonic acid chains absorb water and promote proton transport by the membrane, N_ H2 O is
two mechanisms. At low hydration levels, vehicular transport oc-
mem eod bd hp temp
curs by diffusion. At high hydration levels, the Grotthuss (proton N_ H2 O ¼ N_ H2 O
N_ H2 O ±N_ H2 O ±N_ H2 O (29)
hopping) mechanism is more effective because of the enhanced
connectivity between SO
3 clusters in the highly hydrated envi- eod bd
where N_ H2 O is the molar flow rate due to electro-osmotic drag, N_ H2 O
ronment [26e28]. Membrane hydration is therefore a crucially
hp
important parameter. is due to back-diffusion, N_ H2 O is due to the pressure difference and
Membrane water content was measured by Springer, Zawod- temp
N_ H2 O is the molar flow rate of water due to thermo-osmosis. In
zinski [29] as a function of water activity (a) i.e. relative humidity
practice, electro-osmotic drag and back-diffusion are the dominant
(4) for the Nafion® 117 membrane. Water content/degree of hu-
mechanisms [33]. In the absence of a pressure difference between
midification (l) of the membrane is defined as the number of water
the anode and cathode then there is no pressure-driven water
molecules per sulfonic acid site, l ¼ SOH
2 O þ . The water uptake of a
3H transport across the membrane. Most analyses have ignored
Nafion® membrane at 30  C in contact with a gas-phase flow was
thermo-osmosis because its contribution is normally small
described empirically [29] by
compared to those of net diffusive and electro-osmotic drag transfer.
l ¼ 0:043 þ 17:184
39:8542 þ 36:043 for 0  4  1 (28) We now consider the dominant mechanisms in more detail.

Though Eq. (28) was derived for 30  C,

it has been applied to
2.3.2.1. Electro-osmotic drag
higher temperature conditions as well. In the fully humidified
The electro-osmotic drag coefficient (nd) is defined as the
condition (100% relative humidity, vapour-equilibrated water con-
number of water molecules transported (dragged) per proton.
tent), Eq. (28) implies a maximum achievable water humidification
Transport of water via this mechanism necessarily occurs from the
l z 14. In practice, however l decreases with increasing temper-
anode to the cathode. The corresponding molar flow rate of water
ature, to around 10 at 80  C. The uptake of Nafion® equilibrated
can be expressed as
with liquid water is much higher, l z 22. This difference between
the water- and vapour-equilibrated conditions is a result of eod i
Schroeder's paradox [30e32]. N_ H2 O ¼ nd  (30)
F

2.3.2. Water transport where i is the current density. In practice, nd is a function of the
Water transport across the membrane occurs via four distinct humidification (l) of the membrane. This humidification has been
processes: electro-osmotic drag, diffusion, hydraulic pressure and reported over the years with considerable variation depending on
thermo-osmosis, as illustrated in Fig. 2. Firstly, the electro-osmotic the method of measurement and data fitting. LaConti, Fragala [34]
drag is a potential-driven flow of water resulting from the attrac- measured the electro-osmotic drag by passing the current through
tion of polar water molecules to the protons passing from the anode the membrane and observed the level of a water column for water
to the cathode. Secondly, diffusion occurs as a result of the gradient equilibrated conditions. They reported that the humidification/
of water concentration across the membrane. Because the water water uptake range is 15l25, where two to three water mole-
concentration is usually higher at the cathode owing to water cules were dragged per proton, and also concluded that the drag
generation in the cathode catalyst layer, this transport occurs from coefficient decreases linearly with water content of the membrane.
the cathode to the anode and is therefore referred to as back- Springer, Zawodzinski [29] and Zawodzinski Jr, Neeman [35] used
diffusion. Thirdly, any pressure difference between the anode and the same method and measured the drag coefficient nd ¼ 2.5 for a
cathode drives water across the membrane towards the low- water-equilibrated membrane (fully hydrated) and nd ¼ 0.9 for a
pressure side. Finally, thermo-osmosis is a temperature gradient partially hydrated membrane (i.e. l¼11). The empirical expression
flow through the membrane, i.e. the flow of water across the given by Springer, Zawodzinski [29] (Eq. (31) here) is used to
membrane against the hydrostatic pressure [28]. Combining the calculate the nd as a function of membrane humidification.

l
nd ¼ 2:5  (31)
22

2.3.2.2. Back-diffusion
The combined effects of electro-osmotic drag and water gen-
eration at the cathode tend to flood the cathode and dry-out the
anode. Back-diffusion causes a counter-flow and so can partially
compensate the removal of water from the anode and so flatten the
water concentration profile through the membrane, especially for
thinner membranes [28]. The molar flow rate of water due to back-
diffusion may be calculated by integrating Fick's law through the
membrane, giving

bd ADw h cat i
N_ H2 O ¼ an
CH2 O;mem
CH 2 O;mem
(32)
dmem
where Dw is the diffusion coefficient of water in the membrane,
dmem is the membrane thickness and CHan2 O;mem , CHcat2 O;mem are the
Fig. 2. Species balance within PEM fuel cell. water concentrations at the anode and cathode sides of the
1136 Z. Abdin et al. / Energy 116 (2016) 1131e1144

membrane. These concentrations can be expressed in terms of the

water concentration in the electrode channels as follows 1:06 0:193 1:036 1:765
UD ¼ þ þ þ (40)
t0:156 expð0:476tÞ expð1:53tÞ 3:894t
an an
dan an
el nH2 O
CH2 O;mem
¼ CH 2 O;ch

an (33)
Deff sH2 þ sH2 O
sH2
H2 O ¼
2
(41)
dcat cat sO2 þ sH2 O
cat cat el nH2 O sO2
H2 O ¼
CH2 O;mem
¼ CH 2 O;ch
þ (34) 2
Dcat
eff

where Dan is the effective binary (H2eH2O) diffusion coefficient at kT

eff tH2
H2 O ¼
εH2
H2 O
the anode, Dcat
eff
is the effective binary (O2eH2O) diffusion coeffi- (42)
kT
cient at the cathode, dan cat
el , del are the thicknesses of the anode and tO2
H2 O ¼
an cat εO2
H2 O
cathode respectively and CH 2 O;ch
, CH2 O;ch
represent the water con-
centrations in the anode and cathode channels respectively. and the Lennard-Jones energies εH2
H2 O and εO2
H2 O can be
Diffusion of a molecular species with mean free path l through a expressed as
porous medium with average pore radius r occurs by two principal
mechanisms: molecular diffusion for r > > l and Knudsen diffusion pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
εH2
H2 O ¼ ε H2 ε Н 2 Ο
for r < < l, when interactions with the pore wall occur much more
(43)
frequently than collisions with other molecules. In most porous pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
εO2
H2 O ¼ εΟ2 εН2 Ο
structures, both mechanisms are significant [36]. The effective bi-
nary diffusion coefficient of water at the anode and cathode can be The values of sH2 , sO2 and sH2 O are 2.827 Å, 3.467 Å and 2.641 Å,
expressed as [36] respectively [41]. The Lennard-Jones potentials, εi/k, for H2, O2 and
2
H2 O
0 1 H2O are 59.7, 106.7 and 809.1 K, respectively. The values of DH
eff
1 ε 1 1 O2
H2 O
¼ @ H
H O þ H O; K A (35) and Deff can be obtained by solving Eqs. (35) and (43).
Dan
eff
x D 2 2 D 2 The water concentration at the anode and cathode channels can
eff eff
be expressed as
0 1
1 ε@ 1 1 A rH2 O ðTan Þ
¼ þ (36) an
CH ¼ (44)
Dcat
eff
x DO2
H2 O DH2 O; K 2 O;ch MH2 O
eff eff

2
H2 O
where ε/x is the ratio of electrode porosity to tortuosity, DH
eff
is rH2 O ðTcat Þ
cat
the effective molecular diffusion coefficient for the H2eH2O binary CH 2 O;ch
¼ (45)
MH2 O
O2
H2 O
system, Deff is the effective molecular diffusion coefficient for
where rH2 O is the water density and Tan and Tcat are the tempera-
the O2eH2O binary system and DH 2 O; K
eff
is the effective Knudsen
tures of the anode and cathode respectively.
diffusion coefficient for water. Substituting Eqs. (44) and (45) into Eqs. (33) and (34)
For Knudsen diffusion, the transport of molecules can be
modelled using kinetic theory [37,38], since the gas molecules
frequently collide with the walls of the pores, giving an rH2 O ðTan Þ dan an
el nH2 O
CH 2 O;mem
¼
(46)
MH2 O Dan
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi eff
4 8RT
DH 2 O; K
eff
¼ r (37)
3 pMH2 O
cat rH2 O ðTcat Þ dcat cat
el nH2 O
CH 2 O;mem
¼ þ (47)
where r is the mean pore radius and MH2 O is the molar weight of
MH2 O Dcat
eff
H2O. Using the Chapman-Enskog theory of the ideal gas, the Now Eq. (32) can be written as
H2
H2 O
effective molecular binary diffusion coefficients, Deff and
O2
H2 O " # " #!
Deff can be expressed as [37,39] bd ADw rH2 O ðTcat Þ dcat cat
el nH2 O rH2 O ðTan Þ dan an
el nH2 O
N_ H2 O ¼ þ


dmem MH2 O Dcat MH2 O Dan
 1=2 eff eff
H2
H2 O 1 1 T 3=2
Deff ¼ 0:00133 þ (38) (48)
MH2 MH2 O Pan s2H2
H2 O UD
The net molar rate through the membrane [Eq. (29)] becomes
 1=2 " #
1 1 T 3=2 cat
O2
H2 O
Deff ¼ 0:00133 þ (39) mem i ADw rH2 O ðTcat Þ del ncat
N_ H2 O ¼ nd 

H2 O
MO2 MH2 O Pcat s2O2
H2 O UD þ
F dmem MH2 O Dcat
eff
" an #! (49)
where MH2 and MO2 are the molar weights of H2 and O2, sH2
H2 O rH2 O ðTan Þ del nan H2 O
and sO2
H2 O are the mean molecular radii of species H2eH2O and



MH2 O Daneff
O2eH2 and UD is a dimensionless diffusion collision integral. UD,
sH2
H2 O and sO2
H2 O can be expressed analytically as [40]
 Z. Abdin et al. / Energy 116 (2016) 1131e1144 1137

2.4. Voltage ancillary where nF is the charge transferred, kf is the forward reaction rate
coefficient, kb is the backward reaction rate coefficient, COx is the
The output voltage of a fuel cell can be expressed as surface concentration of oxidation reacting species, CRe is the sur-
face concentration of reduction reacting species and a is the charge
E ¼ Eoc
Eovp (50) transfer coefficient of electrode.
At equilibrium, the resultant current density is zero and the
where Eoc is the open-circuit voltage of the cell and Eovp is the reaction proceeds in both directions at the same rate, correspond-
voltage loss or overpotential due to electro catalyst layers, electron ing to the exchange current density. If the exchange current density
migration in the bipolar plates and electrode backing and proton is high, the surface of the electrode is more active. The higher the
migration in the PEM. exchange current density, the more active the surface of the elec-
The open-circuit voltage of PEM fuel cell in the temperature trode and the lower the energy barrier to charge transfer and the
range of 23e120  C was found to be temperature dependent, with greater the current generated at a given overpotential [45]. At
values always lower than the theoretically expected values. To date, equilibrium
a quantitative explanation for open-circuit voltage behaviour has


not been provided. One explanation attributes this behaviour to H2 f
acat FðE
E0 Þ
i0 ¼ ian
0 ¼ i cat
0 ¼ nFk C
o Ox exp
crossover and/or internal current, as described by Larminie, Dicks RT
[42]. However, in reality, there are several voltage losses within a fuel
 (55)
b aan FðE
E0 Þ
cell that cause the actual cell voltage to be less than the open-circuit ¼ nFko CRe exp
RT
cell voltage. The cell voltage losses or overpotentials are composed of
activation overpotential (Eact) due to electro catalyst layers, con- Now Eq. (54) can be written as
centration overpotential (Econ) due to the mass transfer limitations at 


higher current densities and ohmic overpotential (Eohm) due to
acat FðE
E0 Þ aan FðE
E0 Þ
i ¼ i0 exp
exp (56)
electron migration in the bipolar plates and electrode backing and RT RT
proton migration in the PEM. Now Eq. (50) can be written as
i.e. the Butler
Volmer equation. The deviation of electrode poten-
E ¼ Eoc
Eact
Econ
Eohm (51) tial from the equilibrium potential value is the overpotential i.e.
E
E0¼Eact. The activation overpotential (Eact) is associated with
electrode kinetics and it plays a role at both the anode and cathode.
2.4.1. Open-circuit voltage
The activation overpotential is always defined with respect to a
The ideal cell potential/voltage is attained when the cell is in
specific reaction for which the equilibrium potential is known.
thermodynamic equilibrium. The Eoc is calculated from a modified
When the activation overpotential is large and negative at the
form of the Nernst equation:
cathode (i.e. E<E0), the first term in Eq. (56) becomes predominant
RT h pffiffiffiffiffiffiffiffi i i.e. the reduction current is predominant and now Eq. (56) can be
Eoc ¼ Erev þ ln pH2 pO2 (52) written as
2F


an 
Here Erev is the reversible voltage and it is the voltage that would be
acat FðE
E0 Þ acat FEact
i ¼ icat
0 exp ¼ icat
0 exp (57)
obtained if the Gibbs free energy could be converted directly into RT RT
electrical work without any losses. An extra term is added to ac-
Similarly, when the activation overpotential is positive at the
count for changes in reversible voltage at temperatures that deviate
anode (i.e. E>E0) the second term in Eq. (56) becomes predominant
from the standard reference temperature and now the reversible
i.e. oxidation current is predominant and now Eq. (56) can be
cell potential/voltage can be expressed empirically as [43]
written as

DSo


o
Erev ¼ Erev þ T
Tref  (53) aan FðE
E0 Þ an 
aan FEact
nF i ¼
ian
0 exp ¼
ian
0 exp (58)
RT RT
o (1.229 V) is the standard
Here Erev state or minimum theoretical
Eqs. (57) and (58) give the activation overpotential for both
voltage and DnFS (
0.9  10
3) is the
o
standard state entropy change.
anode and cathode:
!
2.4.2. Activation overpotential RT i
an
The rate of an electrochemical reaction is limited by the acti- Eact ¼ ln an (59)
aan F i0
vation energy barrier to charge transfer between the electrolyte
and electrode or vice-versa. The barrier is represented by a single !
transition state through which each charge must pass. The proba- cat RT i
bility of having been excited to the transition state is taken to be Eact ¼ ln (60)
acat F icat
0
proportional to the attempt rate and to depend exponentially on
the effective barrier height and the temperature. The reactant
consumption is proportional to the surface concentration [42]. The 2.4.3. Concentration overpotential
speed at which an electrochemical reaction proceeds on the elec- Mass transport in the electrodes and at the mem-
trode surface is the rate at which the electrons are released or braneeelectrode interfaces influences the concentrations of
consumed, i.e. the electrical current. According to transition-state hydrogen and oxygen via an inverse relationship between the
theory [44] the resultant current can be written as concentration of reactants on the electrode surface and the cur-

    rent density. In general, when the depletion rate reaches the

acat FðE
E0 Þ aan FðE
E0 Þ diffusion rate, the reactant concentration falls to zero, reducing the
i ¼ nF kf COx exp
kb CRe exp
RT RT partial pressures of the reactant gases. Thus the pressures of
(54) hydrogen and oxygen depend on the current density and on the
physical properties of the electrodes and membrane. A
1138 Z. Abdin et al. / Energy 116 (2016) 1131e1144

concentration overpotential develops when the surface coverage 2.4.4. Ohmic overpotential
of oxygen bubbles at high current density prevent the hydrogen The electrolyte, electrocatalyst layer, gas diffusion layer (GDL),
from being supplied at a higher rate, thus hindering the reaction flow fields, current collectors, interfacial contacts between the
[46]. The limiting current density is typically in the range components and the terminal connections all contribute to the
500e1500 mA/cm2[47]. ohmic overpotential in a PEM fuel cell. The ohmic overpotential can
Mass flow through the electrodes involves two-component be divided into ionic and electrical resistance. The electrical
mixtures (H2eH2O at the anode; O2eH2O at the cathode) and is component is the resistance of the electrically conductive cell
typically modelled using Fick's law, leading to effective binary components to the flow of electrons. The ionic resistance is due to
diffusion coefficients for the reactants, as outlined in x 2.3.2.2. By the proton flow through the membrane. Most of the electrical
combining Fick's law with the Nernst equation, the limiting diffu- resistance usually occurs due to the lack of proper contact between
sion rate at high current density can be calculated, which then al- the GDL, bipolar plates, cooling plates and other interconnects.
lows a calculation of the potential drop caused by excess reaction However, the ionic resistance dominates because the mobility of
products blocking access by the reactants to catalyst sites [48]. The the charge carrier in an ionic conductor is much less than in an
disparate diffusion rates of hydrogen and oxygen can then be un- electronic conductor. An ohmic overpotential Eohm arises owing to
derstood by applying this diffusion-driven approach to the cathode the ohmic resistance of the current collectors, bipolar plates and
and the anode: electrode surfaces [49]. The ohmic overpotential is (by definition
    assumed) linearly proportional to the current:
RT RT RT C1
Econ ¼ E1
Eo ¼ Eo þ ln C1
Eo þ ln C0 ¼ ln
zF zF zF C0 Eohm ¼ Rcell I (71)
(61)
where Rcell is the effective ohmic resistance of the cell:
where z is the number of electrons transferred during the reaction

and the “0” subscript denotes a reference condition. The above Eohm ¼ Rel þ Rpl þ Rmem I
equation can be applied to the anode and the cathode, giving (72)
el pl
mem
¼ Eohm þ Eohm þ Eohm
an cat
Econ ¼ Econ þ Econ (62)
Here Rel, Rpl, Rmem are the resistances of the electrodes, bipolar
mem plates and membrane respectively, and Eohm el ; E pl ; E mem are the
an RT CH2 ohm ohm
Econ ¼ ln mem (63) corresponding contributions to the ohmic overpotential.
2F CH2 ;0
The resistance to currents through the electrode and the flow
field plate have been modelled by Marr and Li [50] in terms of an
cat
mem
RT CO2 effective electrode resistivity reff (Fig. 3):
Econ ¼ ln mem (64)
4F CO2 ;0
l
Rel;pl ¼ reff (73)
where CH mem and C mem are the hydrogen and oxygen concentration
O2
A
2
at the membrane-electrode interface, respectively.
where l and A are respectively the length and cross-section of the
At the membrane-electrode interface the molar concentrations
conductor and reff can be expressed for this two-component
can be expressed as
conductor as [50]

mem ch dan
el nH2 rel
CH ¼ CH þ (65) reff ¼ (74)
2 2
Dan
eff ð1
εÞ1:5

dcat where ε is the (assumed small) electrode porosity. Marr and Li take
mem el nO2
CO ¼ COch2 þ (66) the total average conducting path length to be (wcþws)/4, where wc
2
Dcat
eff is the channel width and ws is the width of the channel support. The
The molar concentrations of oxygen and hydrogen in the average resistance per half flow channel becomes [50]
channels can be expressed as  
reff wan
c þ ws
an

Pan XH2 Ran

1 ¼ (75)
ch
CH ¼ (67) 4 L nan dan
ch el
2
RT
 
Pcat XO2 reff wcat
c þ ws
cat
ch
CO ¼ (68) Rcat
1 ¼ cat
(76)
cat
4 L nch del
2
RT
Substituting Eqs (67) and (68) into Eqs (65) and (66) gives
where L is the length of the MEA perpendicular to the plane of
Fig. 3. The total flow-channel resistance for the two halves in par-
mem Pan XH2 dan
el nH2
CH ¼ þ (69) allel is then
2
RT Daneff
"   cat #
reff wan an
c þ ws wc þ wcat
s
Rel ¼ þ (77)
mem Pcat XO2 dcat
el nO2 8L nan d an
ncat dcat
CO ¼ þ (70) ch el ch el
2
RT Dcat
eff
In Fig. 3, the resistance of the right-hand section of the flow field
plate, R2, is given for the anode and cathode by
 Z. Abdin et al. / Energy 116 (2016) 1131e1144 1139

   an 
Rpl ¼ Ran cat
s þ Rs þ R2 þ Rcat
2 (82)

Marr and Li [50] showed the polarisation curve to visualize the

contribution of different overpotentials. Ionic loss in the membrane
is the second largest source of loss and this loss becomes more
significant at high current densities. The contribution of electronic
resistance to the ohmic overpotential is very small. Bernardi and
Verbrugge [51] also showed that the ionic resistance constitutes the
major contribution to the ohmic overpotential. Springer et al.
[29,52] simply ignored the contribution of electronic resistance to
the ohmic overpotential. Laurencelle, Chahine [53] and Amphlett,
Baumert [54] demonstrated that the ohmic overpotential depends
strongly on membrane humidity and cell temperature and several
studies also showed that the ohmic overpotential is a function of
the membrane conductivity i.e.smem(l,T) [29,55]. Thus the domi-
nant contribution to ohmic overpotential (Eohm) is the ionic loss
caused by resistance to the ion flow through the membrane. This
membrane resistance can be expressed using Ohm's law in the
form i ¼ s V E, as

dmem I
Eohm ¼
Asmem
(83)
Fig. 3. Electric circuit model of the ohmic resistance of the electrode and flow field dmem i
¼
plate (after Ref. [50]). smem
where dmem is the thickness and smem is the conductivity of the
ran an
p hp
PEM. The membrane conductivity smem can be empirically
Ran
2 ¼ (78) expressed as a function of humidification/water content in the
WL
membrane [29,44] as

rcat cat " !#

p hp
DG 1 1
Rcat
2 ¼ (79) smem ¼ ð0:005139l
0:00326Þexp
(84)
WL R Tref T
where ranp , rp are the respective resistivities, hp , hp are the dis-
cat an cat

tances from the outside edge of the plate to the channel surface and where
DG/R¼1268 K and Tref¼303 K.
W is the width of the MEA. Similarly, the resistances of the channel
supports are 3. Results and discussion

rcat
p hc
cat Fig. 4 depicts the generalized PEM fuel cell Simulink model,
Rcat
s ¼ (80) shows how each of the four ancillaries discussed in the previous
ncat
chs
wcat
s L
section contributes to the simulation and maps the interactions
between them.
ran an
p hc
Ran
s ¼ an (81)
nchs wan
s L 3.1. Model parameters

where hcat an an cat

c , hc are the channel heights and nchs , nchs the numbers of To validate the model, parameters for which reliable estimates
channel supports. The total resistance of the flow-field plate is then were available in the literature were fixed. These are listed in Table 1.

Fig. 4. Simulink model of the PEM fuel cell.

1140 Z. Abdin et al. / Energy 116 (2016) 1131e1144

Table 1 and 0e2 respectively [38,62e66]. Our fitted values lie within
Fixed model parameters. these ranges. For Pt-based electrodes, the exchange current den-
Parameter Value sities for the oxygen reduction and hydrogen oxidation reactions
A 51.84 cm2
are reported as 10
9e10
12 and 10
4e10
3 A cm
2 (at 25  C and
dmem 0.0254 cm 1 atm) [67], respectively. Our fitted values lie within these ranges.
del 0.008 cm In our view, the values of parameters obtained from a well-
DGanc 29 kJ mol
1 validated physically based model may be more reliable than
DGcat 66 kJ mol
1
c those estimated from indirect experiments with many assump-
SH2 1.2
tions and uncertainties.
SO2 2
rH2 O 1 g cm
3
rel 10.6  10
6 U cm 3.2. Model validation
ran
p 16.0  10
3 U cm
rcat
p
43.1  10
6 U cm Wang, Husar [56] have explored experimentally the effects of
Dw 1.28  10
6 cm2 s
1 different operating environmental parameters. Using the
ε 0.3
x 4

Parameters that could not be reliably estimated, mostly because

they are difficult to measure directly, were obtained by fitting the
model to experimental data from Ref. [56] under standard condi-
tions (T¼25  C, humidification at 25  C, P¼1 atm). These include the
roughness factor (gM), charge transfer coefficients (a) and the ex-
ref
change current density in the reference state (i0 ). These parame-
ters were subsequently fixed for the comparison of the model with
experimental data under other conditions, and for exploring the
influences of the humidification temperature and the humidifica-
tion ratio. The fitted parameter values are listed in Table 2 and
compared to published values below.
The effective exchange current density depends on tempera-
ture and the electrode surface roughness factor, which is defined
as the effective (electrochemical) electrode area divided by
the geometrical area. If the exchange current density per
actual catalyst surface area at reference temperature and pressure
is iref
0 , then the effective exchange current density at any tem-
perature may be calculated, assuming an activation mechanism,
as [44,57]:
" !#
DGC 1 1
i0 ¼ gM exp

iref
0 (85)
R T Tref

where gM is the electrode surface roughness factor and DGC is the

activation free energy. The roughness factor depends on the
structure of the electrode surface. We tested our model using data
from Wang, Husar [56], who used Pt loadings of 0.4 mg/cm2 for
their electrodes but did not describe their microstructure. The
roughness factor can be determined experimentally [58e60] or
estimated from catalyst loading, catalyst particle density and size
[57]. Roughness factors for the oxygen reduction reaction are re-
ported as 2.7 [60], 9.2 [61], 200 [58] for Pt microdisk, Pt wire and
Pt powder electrodes respectively. Our fitted value gM ¼ 47 (at
25  C and 1 atm) lies in this range. The charge transfer coefficient
values for anode and cathode are reported to be in the range 0e1

Table 2
Fitted model parameters.

Parameter Value

iref 1  10
4 A cm
2

0;an Fig. 5. Comparison between the model predictions and experimental results [56]. (a)
ref
i0;cat 1  10
9 A cm
2 constant pressure 3 atm with constant humidification temperature 70  C, and cell
aan 0.7 temperatures 50  C and 70  C; (b) constant pressure 3 atm with humidification tem-
acat 1.7 perature equal to cell temperature at 60  C and 80  C; (c) varying pressure with
gM 47 constant humidification and cell temperature of 70  C. All model parameters were held
constant for changed environmental conditions.
 Z. Abdin et al. / Energy 116 (2016) 1131e1144 1141

parameter values listed in Tables 1 and 2, the predicted polar- fuel cell is unlikely to be operated at such low power, this
isation curve was compared to the experimental data from discrepancy can be regarded as minor.
Ref. [56]. The model was fitted to the data for one combination of The new model was also tested against three published models.
pressure, temperature and humidification and the predictions for Marr and Li [50] developed a PEM fuel cell model where they
different conditions were made with all parameters fixed. Fig. 5 incorporated the electrochemical reactions in the catalyst layers,
shows the comparison between the fitted polarisation curves voltage losses and used engineering correlation for the process of
and the experimental data in at several temperatures and pres- reactant gas transport in the flow channels and through the elec-
sures. The generally good fit to the experimental data for all trodes. Corre ^a, Farret [47] developed a dynamic electrochemical
conditions confirms that the influences of environmental condi- PEM fuel cell model based on voltage loss and power generation.
tions are well modelled. The discrepancy in the region of activa- Rowe and Li [68] developed a one dimensional non-isothermal
tion polarisation (at low current density) is believed to be due to model based on continuity, energy and the Stefan-Maxwell equa-
the reaction kinetics, as follows. In our simulation we adopted the tion. Fig. 6 shows comparisons of the new model to the published
values for activation energy and reference current density at models of the polarisation curve, made by adopting the model
standard test conditions from the literature. At low current den- parameters from the particular published models [47,50,68]. Except
sity, however, the loading of electrocatalyst on the electrodes at low current densities, the new model is at least as good as the
provides an alternative route for the reaction with a lower acti- published models.
vation energy, which means the reaction kinetics becomes faster.
As a result the activation overpotential decreases. Given that a real 3.3. Exploration of the polarisation curve

Fig. 7 shows the resulting polarisation curve and the various

Fig. 6. Benchmarking this model against published models and experimental data. (a) Fig. 7. Overpotential contributions to the polarisation curve for humidification tem-
^a, Farret [47]; (c) Rowe and Li [68].
Marr and Li [50]; (b) Corre perature equal to cell temperature of 70  C and pressure 1 atm.
1142 Z. Abdin et al. / Energy 116 (2016) 1131e1144

overpotential contributions. The activation overpotential makes the power density is achievable within the fuel cell itself, although
highest contribution and reflects the sluggish electrode kinetics, external factors such as the power needed to compress the reactant
principally due to the slow kinetics of the oxygen reduction reac- gases and shortened membrane lifetime may outweigh the benefit
tion [69]. Second highest is the ohmic overpotential, which is at the system level.
mainly caused by ionic resistance, as discussed in x 2.4.4. The Management of water in the membrane is a crucial factor and a
concentration/diffusion overpotential arises in mass diffusion lim- balance must be struck between sufficient membrane hydration
itations, and is therefore significant at high current densities and and cathode flooding, in order to maintain a high rate of ionic
low reactant concentrations. transport and low mass transport resistance [28]. If the membrane
Fig. 8 shows the separate and combined effects of pressure and is not properly hydrated (say l>10), then at high current density the
temperature on the polarisation curve of PEM fuel cell. Experi- anode
membrane interface will become dehydrated, causing
mentally it is found that temperature has a greater influence on the higher ionic resistance and, in extreme cases, irreversible damage.
cell performance than pressure [70] and Fig. 8 confirms this. It is The degree of membrane hydration is affected by water transport
apparent from Fig. 8 that under more extreme conditions of across it (see x 2.3), which is related to the humidification of the
elevated pressure and elevated temperature, significantly higher inlet gases and the operating conditions [71]. Fig. 9 illustrates the
very strong influence of PEM humidification on the polarisation
curve. For l¼5 the loss of cell voltage at high current density is very
significant.
Fig. 10 shows the polarisation curves for different relative hu-
midification levels of inlet gases at the anode (hydrogen) and
cathode (oxygen). Generally speaking, increasing humidification
enhances the fuel cell performance. A high level of anode hu-
midification is always beneficial (part (a) compared to part (b))
and for 100% anode humidification, decreased cathode humidifi-
cation (part (b)) is fairly well tolerated, even for very low values
(part (c)). On the other hand, excess cathode humidification leads
to flooding in the cathode and loss of performance. Hence, high
anode and moderate cathode humidification is the most robust
combination.

4. Summary and conclusions

A steady-state one dimensional model of a PEM fuel cell based

largely on physical parameters has been presented. This model is
structured so that it can be easily extended to a stack of fuel cells
and is modest in computing demand so that it can be incorporated
in a whole-of-system simulation. Because the model relates PEM
fuel cell performance to fundamental physical parameters, it can be
utilised in developing improved electrodes to reduce the activation
overpotential, for instance, by incorporating the measured or pre-
dicted characteristics of new electrocatalyst materials into the
relevant ancillaries.

Fig. 8. Effects of temperature and pressure on polarisation curves for constant hu-
midification temperature of 60  C. (a) constant temperature and variable pressure; (b)
constant pressure and variable temperature; (c) variable temperature and pressure, Fig. 9. Effect of PEM humidification for cell temperature of 70  C, humidification
showing the significantly improved performance predicted for more extreme oper- temperature of 60  C and pressure of 1 atm. Note the severe effect of humidification
ating conditions. ratios (l) below about 10.
 Z. Abdin et al. / Energy 116 (2016) 1131e1144 1143

Fig. 10. Effects of anode and cathode humidities for cell temperature 50  C, humidification temperature 60  C and pressure 1 atm. (a) moderately anode humidification; (b) full
anode humidification; (c) low cathode humidification; (d) full cathode humidification. Note the severe effect of combined low anode and low cathode humidification.

The mathematical model was validated in two ways. First, References

published experimental cell polarisation data were fitted very well,
with five free model parameters (charge transfer coefficients, [1] Yerramalla S, Davari A, Feliachi A, Biswas T. Modeling and simulation of the
dynamic behavior of a polymer electrolyte membrane fuel cell. J Power
roughness factor and reference exchange current densities) related Sources 2003;124(1):104e13.
to physical parameters whose values are difficult to estimate reli- [2] Ji M, Wei Z. A review of water management in polymer electrolyte membrane
ably. The fitted values of these parameters were within the ex- fuel cells. Energies 2009;2(4):1057e106.
[3] Bıyıkog
lu A. Review of proton exchange membrane fuel cell models. Int J
pected ranges. Importantly, the model also tracked changes in the Hydrogen Energy 2005;30(11):1181e212.
polarisation curve owing to changed temperature and pressure, [4] Cheddie D, Munroe N. Review and comparison of approaches to proton ex-
without altering the model parameters. Second, the model was also change membrane fuel cell modeling. J Power Sources 2005;147(1):72e84.
[5] Djilali N, Lu D. Influence of heat transfer on gas and water transport in fuel
benchmarked against three published models with varying degrees cells. Int J Therm Sci 2002;41(1):29e40.
of complexity and performed at least as well as these. Thus the [6] Costamagna P. Transport phenomena in polymeric membrane fuel cells. Chem
approach using a simplified 1D model has been demonstrated to Eng Sci 2001;56(2):323e32.
[7] Dutta S, Shimpalee S, Van Zee J. Numerical prediction of mass-exchange be-
have good functionality.
tween cathode and anode channels in a PEM fuel cell. Int J Heat Mass Transf
A parametric study in which environmental input parameters 2001;44(11):2029e42.
(temperature, pressure, and reactant gas humidity) were varied [8] Wang Y, Wang C-Y. Simulation of flow and transport phenomena in a polymer
electrolyte fuel cell under low-humidity operation. J Power Sources
demonstrated the ability of the model to reproduce the essential
2005;147(1):148e61.
characteristics of a PEM fuel cell. In particular, the model [9] Ying W, Yang T-H, Lee W-Y, Ke J, Kim C-S. Three-dimensional analysis for
showed how when the relative humidities of the feed gases effect of channel configuration on the performance of a small air-breathing
were both reduced, the membrane became dehydrated and proton exchange membrane fuel cell (PEMFC). J Power Sources
2005;145(2):572e81.
highly resistive. [10] Ying W, Sohn Y-J, Lee W-Y, Ke J, Kim C-S. Three-dimensional modeling and
A major remaining challenge is to develop a physically-based experimental investigation for an air-breathing polymer electrolyte mem-
model to calculate the water uptake of the membrane in terms brane fuel cell (PEMFC). J power sources 2005;145(2):563e71.
[11] Lum KW, McGuirk JJ. Three-dimensional model of a complete polymer elec-
of the relative humidities of the feed gases and the cell trolyte membrane fuel cellemodel formulation, validation and parametric
temperature. studies. J Power Sources 2005;143(1):103e24.
[12] Hu M, Gu A, Wang M, Zhu X, Yu L. Three dimensional, two phase flow
mathematical model for PEM fuel cell: Part I. Model development. Energy
Acknowledgement Convers Manag 2004;45(11):1861e82.
[13] Hu M, Zhu X, Wang M, Gu A, Yu L. Three dimensional, two phase flow
ZA acknowledges receipt of the Griffith University Postgraduate mathematical model for PEM fuel cell: Part II. Analysis and discussion of the
internal transport mechanisms. Energy Convers Manag 2004;45(11):
Research Scholarship and Griffith University International Post-
1883e916.
graduate Research Scholarship.
1144 Z. Abdin et al. / Energy 116 (2016) 1131e1144

[14] Carton JG, Olabi AG. Three-dimensional proton exchange membrane fuel cell isopropylacrylamide) gates. J Membr Sci 2001;192(1e2):27e39. http://
model: comparison of double channel and open pore cellular foam flow dx.doi.org/10.1016/S0376-7388(01)00464-1.
plates. Energy 2016. http://dx.doi.org/10.1016/j.energy.2016.02.010. In Press. [41] Ni M, Leung MK, Leung DY. An electrochemical model of a solid oxide steam
~o DS, Oliveira VB, Pinto AMFR. Numerical simulations of two-
[15] Ferreira RB, Falca electrolyzer for hydrogen production. Chem Eng Technol 2006;29(5):636e42.
phase flow in an anode gas channel of a proton exchange membrane fuel cell. [42] Larminie J, Dicks A, McDonald MS. Fuel cell systems explained. New York:
Energy 2015;82:619e28. http://dx.doi.org/10.1016/j.energy.2015.01.071. Wiley; 2003.
[16] Carton JG, Lawlor V, Olabi AG, Hochenauer C, Zauner G. Water droplet accu- [43] Amphlett JC, Baumert R, Mann RF, Peppley BA, Roberge PR, Harris TJ. Per-
mulation and motion in PEM (Proton Exchange Membrane) fuel cell mini- formance modeling of the Ballard Mark IV solid polymer electrolyte fuel cell I.
channels. Energy 2012;39(1):63e73. http://dx.doi.org/10.1016/j.energy.2011. Mechanistic model development. J Electrochem Soc 1995;142(1):1e8.
10.023. [44] Atkins PW, Atkins PW. The elements of physical chemistry. Oxford United
[17] Kang S. Quasi-three dimensional dynamic modeling of a proton exchange Kingdom: Oxford University Press; 1992.
membrane fuel cell with consideration of two-phase water transport through [45] Barbir F. Fuel cell basic chemistry, electrochemistry and thermodynamics.
a gas diffusion layer. Energy 2015;90:1388e400. http://dx.doi.org/10.1016/ Mini-Micro Fuel Cells. Springer; 2008. p. 13e26.
j.energy.2015.06.076. Part 2. [46] Biaku CY, Dale NV, Mann MD, Salehfar H, Peters AJ, Han T. A semiempirical
[18] Xing L, Cai Q, Xu C, Liu C, Scott K, Yan Y. Numerical study of the effect of study of the temperature dependence of the anode charge transfer coefficient
relative humidity and stoichiometric flow ratio on PEM (proton exchange of a 6 kW PEM electrolyzer. Int J Hydrogen Energy 2008;33(16):4247e54.
membrane) fuel cell performance with various channel lengths: an anode http://dx.doi.org/10.1016/j.ijhydene.2008.06.006.
partial flooding modelling. Energy 2016;106:631e45. http://dx.doi.org/ [47] Corre^a JM, Farret F, Canha LN, Simo~es MG. An electrochemical-based fuel-cell
10.1016/j.energy.2016.03.105. model suitable for electrical engineering automation approach. Ind Electron
[19] Salva JA, Iranzo A, Rosa F, Tapia E. Validation of cell voltage and water content IEEE Trans 2004;51(5):1103e12.
in a PEM (polymer electrolyte membrane) fuel cell model using neutron [48] Marangio F, Santarelli M, Cali M. Theoretical model and experimental analysis
imaging for different operating conditions. Energy 2016;101:100e12. http:// of a high pressure PEM water electrolyser for hydrogen production. Int J
dx.doi.org/10.1016/j.energy.2016.02.006. Hydrogen Energy 2009;34(3):1143e58.
[20] Abdollahzadeh M, Pascoa JC, Ranjbar AA, Esmaili Q. Analysis of PEM (Polymer [49] Hwang J, Lai L, Wu W, Chang W. Dynamic modeling of a photovoltaic
Electrolyte Membrane) fuel cell cathode two-dimensional modeling. Energy hydrogen fuel cell hybrid system. Int J Hydrogen Energy 2009;34(23):
2014;68:478e94. http://dx.doi.org/10.1016/j.energy.2014.01.075. 9531e42.
[21] Liu H, Li P, Wang K. Optimization of PEM fuel cell flow channel dimen- [50] Marr C, Li X. An engineering model of proton exchange membrane fuel cell
sionsdMathematic modeling analysis and experimental verification. Int J performance. ARI Int J Phys Eng Sci 1998;50(4):190e200.
Hydrogen Energy 2013;38(23):9835e46. [51] Bernardi DM, Verbrugge MW. Mathematical model of a gas diffusion electrode
[22] Shekhar K. An investigation into the minimum dimensionality required for bonded to a polymer electrolyte. AIChE J 1991;37(8):1151e63.
accurate simulation of proton exchange membrane fuel cells by the [52] Springer T, Wilson M, Gottesfeld S. Modeling and experimental diagnostics in
comparison between 1-and 3-dimension models. University of Cape polymer electrolyte fuel cells. J Electrochem Soc 1993;140(12):3513e26.
Town; 2013. [53] Laurencelle F, Chahine R, Hamelin J, Agbossou K, Fournier M, Bose T, et al.
[23] Abdin Z, Webb C, Gray EM. Solar hydrogen hybrid energy systems for off-grid Characterization of a Ballard MK5-E proton exchange membrane fuel cell
electricity supply: a critical review. Renew Sustain Energy Rev 2015;52: stack. Fuel Cells 2001;1(1):66e71.
1791e808. [54] Amphlett J, Baumert R, Mann R, Peppley B, Roberge P, Rodrigues A. Parametric
[24] Abdin Z, Webb C, Gray EM. Modelling and simulation of a proton exchange modelling of the performance of a 5-kW proton-exchange membrane fuel cell
membrane (PEM) electrolyser cell. Int J Hydrogen Energy 2015;40(39): stack. J Power Sources 1994;49(1):349e56.
13243e57. [55] Nguyen TV, White RE. A water and heat management model for Proton-
[25] Mennola T, Noponen M, Aronniemi M, Hottinen T, Mikkola M, Himanen O, Exchange-Membrane fuel cells. J Electrochem Soc 1993;140(8):2178e86.
et al. Mass transport in the cathode of a free-breathing polymer electrolyte [56] Wang L, Husar A, Zhou T, Liu H. A parametric study of PEM fuel cell perfor-
membrane fuel cell. J Appl Electrochem 2003;33(11):979e87. mances. Int J Hydrogen Energy 2003;28(11):1263e72.
[26] Mench MM. Fuel cell engines. John Wiley & Sons; 2008. [57] Thampan T, Malhotra S, Zhang J, Datta R. PEM fuel cell as a membrane reactor.
[27] Choi P, Jalani NH, Datta R. Thermodynamics and proton transport in nafion i. Catal Today 2001;67(1):15e32.
membrane swelling, sorption, and ion-exchange equilibrium. J Electrochem [58] Aldebert P, Novel-Cattin F, Pineri M, Millet P, Doumain C, Durand R. Prepa-
Soc 2005;152(3). E84eE9. ration and characterization of SPE composites forelectrolyzers and fuel cells.
[28] Kumbur EC, Mench MM. Fuel cells e proton-exchange membrane fuel cells j Solid State Ionics 1989;35(1):3e9.
water management A2-Garche, Jürgen. Encyclopedia of Electrochemical Po- [59] Damjanovic A, Brusic V. Electrode kinetics of oxygen reduction on oxide-free
wer Sources. Amsterdam: Elsevier; 2009. p. 828e47. platinum electrodes. Electrochim Acta 1967;12(6):615e28.
[29] Springer TE, Zawodzinski T, Gottesfeld S. Polymer electrolyte fuel cell model. [60] Beattie PD, Basura VI, Holdcroft S. Temperature and pressure dependence of
J Electrochem Soc 1991;138(8):2334e42. O2 reduction at Ptr Nafion® 117 and Ptr BAM® 407 interfaces. J Electroanal
[30] Onishi LM, Prausnitz JM, Newman J. Water-nafion equilibria. Absence of Chem 1999;468(2):180e92.
Schroeder's paradox. J Phys Chem B 2007;111(34):10166e73. [61] Parthasarathy A, Srinivasan S, Appleby AJ, Martin CR. Electrode kinetics of
[31] Vallieres C, Winkelmann D, Roizard D, Favre E, Scharfer P, Kind M. On oxygen reduction at carbon-supported and unsupported platinum micro-
Schroeder's paradox. J Membr Sci 2006;278(1):357e64. crystallite/Nafion® interfaces. J Electroanal Chem 1992;339(1):101e21.
[32] Freger V. Hydration of ionomers and Schroeder's paradox in Nafion. J Phys [62] Nguyen PT, Berning T, Djilali N. Computational model of a PEM fuel cell with
Chem B 2008;113(1):24e36. serpentine gas flow channels. J Power Sources 2004;130(1e2):149e57. http://
[33] Zawodzinski TA, Derouin C, Radzinski S, Sherman RJ, Smith VT, Springer TE, dx.doi.org/10.1016/j.jpowsour.2003.12.027.
et al. Water uptake by and transport through Nafion® 117 membranes. [63] Ju H, Meng H, Wang C-Y. A single-phase, non-isothermal model for PEM fuel
J Electrochem Soc 1993;140(4):1041e7. cells. Int J Heat Mass Transf 2005;48(7):1303e15.
[34] LaConti A, Fragala A, Boyack J. Solid polymer electrolyte electrochemical cells- [64] Cheddie DF, Munroe ND. Three dimensional modeling of high temperature
Electrode and other materials considerations. In: Conference Solid polymer PEM fuel cells. J Power Sources 2006;160(1):215e23.
electrolyte electrochemical cells-Electrode and other materials consider- [65] Santarelli MG, Torchio MF, Cochis P. Parameters estimation of a PEM fuel cell
ations, vol. 1; 1977. p. 354e74. polarization curve and analysis of their behavior with temperature. J Power
[35] Zawodzinski Jr TA, Neeman M, Sillerud LO, Gottesfeld S. Determination of Sources 2006;159(2):824e35.
water diffusion coefficients in perfluorosulfonate ionomeric membranes. [66] Walsh FC. A first course in electrochemical engineering. Electrochemical
J Phys Chem 1991;95(15):6040e4. Consultancy; 1993.
[36] Ni M, Leung MKH, Leung DYC. A modeling study on concentration over- [67] Barbir F. PEM fuel cells: theory and practice. Academic Press; 2012.
potentials of a reversible solid oxide fuel cell. J Power Sources 2006;163(1): [68] Rowe A, Li X. Mathematical modeling of proton exchange membrane fuel
460e6. http://dx.doi.org/10.1016/j.jpowsour.2006.09.024. cells. J Power Sources 2001;102(1):82e96.
[37] Herna ndez-Pacheco E, Singh D, Hutton PN, Patel N, Mann MD. A macro-level [69] Nørskov JK, Rossmeisl J, Logadottir A, Lindqvist L, Kitchin JR, Bligaard T, et al.
model for determining the performance characteristics of solid oxide fuel Origin of the overpotential for oxygen reduction at a fuel-cell cathode. J Phys
cells. J Power Sources 2004;138(1):174e86. Chem B 2004;108(46):17886e92.
[38] Zhu H, Kee RJ. A general mathematical model for analyzing the performance [70] Murthy M, Esayian M, Lee W-k, Van Zee J. The effect of temperature and
of fuel-cell membrane-electrode assemblies. J Power Sources 2003;117(1): pressure on the performance of a PEMFC exposed to transient CO concen-
61e74. trations. J Electrochem Soc 2003;150(1):A29e34.
[39] Reid RC, Prausnitz JM, Poling BE. The properties of gases and liquids. 1987. [71] Yan Q, Toghiani H, Wu J. Investigation of water transport through membrane
[40] Chu L-Y, Park S-H, Yamaguchi T, Nakao S-i. Preparation of thermo-responsive in a PEM fuel cell by water balance experiments. J Power Sources
core-shell microcapsules with a porous membrane and poly(N- 2006;158(1):316e25.