Journal of Power Sources 297 (2015) 23e32

Contents lists available at ScienceDirect

Journal of Power Sources

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

Modeling the performance of hydrogeneoxygen unitized regenerative

proton exchange membrane fuel cells for energy storage
Massimo Guarnieri*, Piergiorgio Alotto, Federico Moro
Department of Industrial Engineering, University of Padua, Via Gradenigo 6/a, 35131 Padova, Italy

h i g h l i g h t s

We developed a comprehensive FC þ EL performance model of unified regenerative PEMFCs.

The model describes both EL and FC modes of a PEMFC-based energy storage system.

Dependence on temperature, pressure and humidification is taken into account.

The model allows the exploration of the sensitivity to several working parameters.

Critical issues for improving the concept competitiveness can thus be identified.

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

Article history: Thanks to the independent sizing of power and energy, hydrogen-based energy storage is one of the very
Received 12 June 2015 few technologies capable of providing long operational times in addition to the other advantages offered
Received in revised form by electrochemical energy storage, for example scalability, site versatility, and mobile service. The typical
13 July 2015
design consists of an electrolyzer in charge mode and a separate fuel cell in discharge mode. Instead, a
Accepted 22 July 2015
Available online xxx
unitized regenerative fuel cell (URFC) is a single device performing both energy conversions, achieving a
higher compactness and power-to-weight ratio. This paper presents a performance model of a URFC
based on a proton exchange membrane (PEM) electrolyte and working on hydrogen and oxygen, which
Keywords:
Regenerative fuel cell
can provide high energy and power densities (>0.7 W cm2). It provides voltage, power, and efficiency at
Fuel cell varying load conditions as functions of the controlling physical quantities: temperature, pressure, con-
PEMFC model centration, and humidification. The model constitutes a tool for designing the interface and control sub-
Electrolyzer model system as well as for exploring optimized cell/stack designs and operational conditions. To date, only a
Hydrogen energy storage few of such analyses have been carried out and more research is needed in order to explore the true
Performance model potential of URFCs.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction factor for the widespread adoption of electric vehicles (EVs). The
commercial success of early models such as the Toyota Prius
Renewable sources like wind and solar present two major (hybrid EV, HEV), Chevrolet Volt/Opel Ampera (plug-in hybrid EV,
drawbacks with respect to conventional fossil- and fissile-power PHEV), Nissan Leaf, and Tesla cars (battery EVs, BEVs) reveals the
plants: they are more expensive and intermittent according to potential of this market.
time and climatic conditions. While the former can be overcome Storage technologies capable of providing the needed perfor-
with the development of advanced devices based on innovative mance for both stationary and mobile applications are expected to
materials and configurations, the latter calls for suitable energy increase dramatically in the near future [2]. Several surveys indi-
storage technologies [1]. Efficient energy storage is also funda- cate that electrochemical energy storage (ECES) technologies for
mental for mobile applications and in the near future it will be a key stationary applications can be the solution of choice for many
storage services with short to long discharge time and kilowatt to
megawatt power [3]. Their advantages include wide scalability due
to intrinsic modularity, site versatility, limited or null environ-
* Corresponding author.
E-mail addresses: [email protected] (M. Guarnieri), piergiorgio. mental impact, and ease of operation [4]. In the case of mobile
[email protected] (P. Alotto), [email protected] (F. Moro). applications, no practical alternative to ECES exists. However, some

http://dx.doi.org/10.1016/j.jpowsour.2015.07.067
0378-7753/© 2015 Elsevier B.V. All rights reserved.
24 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32

major disadvantages must be addressed to take them to full suc- devices, such as supercapacitors or lithium polymer batteries, thus
cess, including cost, discharge duration, round-trip efficiency, life configuring the vehicle as a hybrid FC-BEV. A unitized regenerative
cycles, and calendar life. fuel cell (URFC) is an interesting alternative consisting in a device
Almost all ECES technologies are hampered by a major draw- capable of performing both EL and FC functions.
back: the same device provides power conversion and energy
storage, thus binding together power rating and discharge time.
2. PEM unitized regenerative fuel cells
Two ECES technologies are free of this constraint: redox flow bat-
teries (RFBs) [5] and hydrogen-based energy storage systems
A PEM (proton exchange membrane or polymer electrolyte
(HBESs) [6]. The typical architecture of the latter, forming a
membrane) is a convenient electrolyte technology for both ELs and
regenerative fuel cell (RFC), consists of an electrolyzer (EL) that
FCs, since it allows for low operational temperature, quick start, fast
converts electric energy into hydrogen by water electrolysis, a fuel
response, and high power and energy densities. In fact, PEMELs and
cell (FC) for extracting back electricity from hydrogen, and a tank
PEMFCs essentially use the same reversible core, the membrane
for storing the gas [7]. The tank can be sized according to the energy
electrode assembly (MEA). This is a sandwich structure made of
rating requirements, independently of the power rating. Several
two porous layers (the diffusion layers, DLs, also called gas diffusion
RFC technologies, based on different chemistries, are under devel-
layers, GDLs, 101e102 mm thick), each bearing a thinner electro-
opment. Some chemistries exhibit low power density but high
catalyst layer (CL, 100e101 mm thick, devoted to enhancing the
round-trip efficiency, similarly to RFBs, which makes them attrac-
electrochemical reactions) on the face in contact with the inter-
tive for stationary service, particularly in the case of long-term
posed PEM (101 mm thick). Specific bi-functional electrocatalysts
storage (e.g. seasonal applications) [8]. Other chemistries present
are being developed for PEM-URFC (or UR-PEMFC) [11]. A proper
lower efficiency, but energy densities as high as 1300 Wh kg1
design of the gas flow management, namely with bipolar plates
make them very attractive for electrical vehicles, where they can
(BPs), whose flow channels distribute/draw the fluids to/from the
overcome the range limitations of present batteries [9]. In fact,
DLs, pumps, back-pressure valves, and separators, which split the
mass production of H2-EVs has just started with the Hyundai ix35,
outlet fluids, allows the same MEA to perform both FC and EL
Toyota Mirai FCV, and Honda FCX Clarity. They are also extensively
functions, resulting in a UR-PEMFC (Fig. 1). The simplest architec-
studied for stationary applications: the UE FCH2-JU 2014e2020
ture solution for mobile terrestrial systems consists of breathable
program, with funding of 1.3 GV, is devoted to the development of
cells fed with atmospheric oxygen. Conversely, underwater and
researches on hydrogen technologies for both transport and sta-
space systems store and reuse oxygen, beside hydrogen, requiring
tionary applications. Conventional RFC systems consisting of
higher capital costs, but also resulting in higher efficiency and po-
separated and optimized ELs and FCs are profitable for long term-
wer density [12].
stationary applications [10]. However, this architecture is too
Devices of this kind are already on the market, but their real
cumbersome for electric mobility, where the usual solution for
success can only come after system improvements in terms of
recovering braking energy into electricity resorts to the use of other
costs, durability, and efficiency, which require extensive research

DC bus
power
icharge<0 management unit idischarge>0

system
UR-PEMFC supervisor
MEA
e e
+

O2
+
H2 H2 H O2
H2O
H2O

bipolar plates diffusion

and electrodes layers PEM calalyst layers separator

Fig. 1. Schematic of a H2eO2 UR-PEMFC system, showing ideally a single cell as a power converter. Three tanks store fluids (H2O, H2, and O2) and energy independently of the R-
PEMFC power rating. O2 storage allows the R-PEMFC to work on pure oxygen, gaining higher power and efficiency at a higher capital cost, but for terrestrial applications, cheaper
breathable devices are usually preferred, which get rid of the O2 tank. The power management unit and system supervisor control the electrical performance and the related fluid-
dynamic parameters (temperature, pressure, flow, hydration).
 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32 25

and development in materials and technology. Such de- 3.1. Electromotive force (reversible voltage)
velopments can greatly benefit from accurate numerical
modeling, simulation, and optimization. The vast literature on According to the Nernst equation, the emf (electromotive force,
PEMFC modeling deals with single sub-components (PEM, DLs, BP and also reversible voltage in the electrochemical taxonomy) can be
flow channels, e.g. Ref. [13]), single cells (e.g. Ref. [14]), stacks written as [30]:
made of several cells put in series through BPs, ancillary sub-
systems, and the whole PEMFC system (e.g. Ref. [15]). Several ar-
E ¼ E0 þ DEs þ DEc (2)
ticles have been published on PEMEL experimental research and
modeling, some examples being given in Refs. [16e22], and on R- where E0 is the emf value at standard temperature and pressure:
PEMFC with separate FC and EL, for example [23]. On the other
hand, very few articles have been published on physical models of
UR-PEMFC dedicated to hydrogenechlorine (H2eCl2) chemistry E0 ¼ 1:229 V at T 0 ¼ 298:15 K; p0 ¼ 100 kPa (3)
[24] and hydrogen-bromine (H2eB2) chemistry [25,26]. The DEs is the entropic variation from standard conditions
literature on modeling of H2eO2 UR-PEMFC is extremely limited
and very few papers presenting unified multiphysics models have
ZT
been published to date [27,28]. This paper presents a zero- 1 DT X
dimensional model aimed at analyzing the performance of a UR- DEs ¼ Dbs ðTÞdT ¼ Db
s ðTi Þ (4)
nF nF i
PEMFC in terms of power and round-trip efficiency under T0
different physical conditions (temperature, pressure, flow, hy-
dration). It can be used to compare the potential performance of where F ¼ 96,485 C mol1 is the Faraday constant and, according to
UR-PEMFCs with alternative concepts for both stationary and (1), the number of charge carriers per reaction is n ¼ 2. The sum
mobile applications in order to estimate their competitiveness allows an accurate computation of the integral for large tempera-
and to address the optimal design of ancillary sub-systems, upon ture variations, by using tabulated thermodynamic values of the
which the competitiveness of the system relies [29]. molar reaction entropy Db s ðTi Þ. Since this model aims at exploring
Section 3 describes the multiphysics zero-dimensional model of the variation of performance with temperature, (4) has been
a UR-PEMFC based on a unified system of equations for both preferred to the often-used linearized expression ðDbs =nFÞðT  T 0 Þ.
working modes, namely the charge (EL-mode) and discharge (FC- DEc in (2) is the variation from standard conditions related to
mode) operations. It consists of physical-chemical equations which concentration and hence to pressure:
yield the UR-PEMFC polarization curve under different physical
conditions (temperature, gas pressure, and PEM hydration) under
2 ! !0:5 3
the control of the system supervisor. As usual, the polarization RT 4 cH2 cO2 5
DEc ¼ ln (5)
curve expresses a single cell voltage as a function of the cross- nF c0H2 c0O2
sectional current density. The practical system power rating is
attainable by a proper sizing of the cell active cross-section A and of where R ¼ 8.314 J mol1 K1 is the gas constant and cx =c0x is the x-
the number of cells put in series to form a stack. The model does not gas activity (namely, its concentration normalized to standard
include the power management unit, which in real systems adapts conditions). The “surface” concentrations cx at the CLs change in
the cell voltage and current to the dc bus while controlling the operation with respect to the “bulk” (undisturbed) values cx at
electric power flux. The model numerical implementation is pre- which the gases are fed into the flow channels of the BPs, because of
sented in Section 4, while some results are presented and discussed flow-induced gradients. Equation (5) can be rewritten as the sum of
in Section 5. two terms. The first one is the variation from standard conditions in
rest (no-load) conditions, namely for cx ¼ cx :

2 ! !0:5 3
3. Multiphysics model T 4 cH2 cO2
DEco ¼ ln 5 (6)
f c0H2 c0O2
The two half-reactions occurring at the negative and positive
CLs, respectively, of a UR-PEMFC are:
where f ¼ nF/R ¼ 23,210 K V1 has been introduced. The second
charge term of the sum is related to the concentration gradients, which
H2 O ) * 1 O2 þ 2Hþ þ 2e appear in load conditions (cx scx ):
discharge 2
(1)
charge "  0:5 #
þ  * H2
2H þ 2e ) T cH2 cO2
discharge DEcl ¼ ln (7)
f cH2 cO2
In the charge mode (EL mode), water is split into oxygen, pro-
tons, and electrons at the positive CL (anode). Electrons, which flow By substituting (4) and (6) in (2) the open circuit emf is
in an external circuit, and protons, which flow through the PEM, obtained:
recombine at the negative CL (cathode), forming hydrogen. In the
discharge mode (FC mode), the inverse reactions occur, with the 2 ! !0:5 3
ZT
negative CL working as the anode and the positive CL as the cath- 1 T 4 cH2 cO2 5
EOC ¼E þ 0
Db
s ðTÞdT þ ln (8)
ode. The PEM electrolyte segregates the half-reactions, allowing the nF f c0H2 c0O2
T0
flow of protons while forcing the electrons into the external circuit.
In the following, we will use the terms anode and cathode according EOC differs from the open circuit voltage (OCV) because of
to the fuel cell terminology, that is, in discharge condition, when hydrogen crossover (Subsection 3.8). Instead, the load-gradient-
the anode performs the hydrogen oxidation reaction (HOR) and the dependent term (7) is accounted for as concentration losses
cathode the oxygen reduction reaction (ORR). (Subsection 3.6).
26 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32

3.2. Load voltage  

cR af Dva =T cP ð1aÞf Dva =T
jt ¼ j0 e  e (13)
In load conditions (electric current density js0), the cell voltage cR cP
v differs from EOC of (8), due to losses, which cause voltage drops Dv
(Dv > 0 in discharge mode, i.e. a voltage decrease, and Dv < 0 in where the total equivalent current density jt appears in place of the
charge mode, i.e. a voltage increase. These voltage drops are electric current density j, to account for the effect of hydrogen
referred to as overpotentials and are denoted by h in electro- crossover on the overpotentials (Subsection 3.8).
chemistry [31]):

vðjÞ ¼ EOC  Dva  Dvaþ  Dvc  Dvcþ  Dvm (9) 3.4. Exchange current density
Subscripts ae and aþ indicate the activation losses occurring at
the negative and positive CLs respectively, ce and cþ indicate the Since we are interested in evaluating the cell performance at
concentration losses occurring at the negative and positive DLs different temperatures, we take into account the dependence of j0
respectively, and m refers to the ohmic losses in the PEM. on T by means of an Arrhenius-like relation [32]:
  
Wj 1 1
j0 ¼ j*0 exp  (14)
3.3. Activation losses R T* T

Voltage drops Dva and Dvaþ are due to the electrochemical where Wj is the activation barrier of the half-reaction and j*0 is the
kinetics of the half-reactions (1) occurring at the active electro- value of j0 at the temperature T* (with undisturbed concentrations
catalyst sites TPBs (triple phase boundaries), namely the nano- c ¼ c). The values of j0 strongly affect the performance and round-
surfaces which are spread inside the CLs, where PEM and DLs trip efficiency of the device (Subsection 3.9), since at a given jt the
merge and where reactants arrive and products depart. Each acti- higher j0 is, the lower Dva is, according to (13). Equation (14) allows
vation voltage drop is related to the rate of charge density sepa- to assess the effects of operational measures which can increase j0,
ration vtse (proton and electron creation at the anode) and like higher temperatures T and higher concentrations c. But it also
recombination (at the cathode) produced in load condition by the allows to explore the effect of structural measures, such as lowering
electrochemical half-reactions (1) according to the ButlereVolmer Wj (i.e. achieving higher activity by means of efficient catalysts) and
equation (vt represents the partial time derivative): increasing ATPB (by means of highly nanostructured catalyst layers).
  The exponential dependence of jt on Dva and the low operating
cRx ax f Dvax =T cPx ð1ax Þf Dvax =T temperature of PEMs make activation losses the main causes of the
vt sex ¼ j0TPB x e  e (10)
cRx cPx cell voltage drop at low current densities. Since Dvaþ is typically one
Here subscript x can be either minus (negative CL, where order of magnitude larger than Dva, the positive electrode losses
hydrogen reacts) or plus (positive CL, where oxygen reacts) and is remain the dominant effect in determining the behavior of the UR-
omitted for simplicity in the following, assuming that if the elec- PEMFC in the charge/discharge operation at low current density,
trode is not specified, symbols and equations hold for both of them. close to the OCV condition.
Consequently, in FC mode subscript Rx stands for Re (i.e. H2) and
Rþ (i.e. O2), and in practice Px only for Pþ (i.e. H2O). The ratios c=c
are the reagent (R) and product (P) surface concentrations at TPBs 3.5. Concentration gradients
normalized to the bulk values as in (7). Here again, these ratios
differ from 1 when concentration gradients arise in load conditions. In load operation, convective mass flow in the flow channels of
a is the transfer coefficient of each half-reaction and j0TPB is its the BPs and diffusive mass flow in the DLs sustain the electric
exchange current density, namely the charge rate of the direct and current generation at the CLs. Mass flows create gradients in the
inverse reactions in open circuit conditions, which balance each reagent and product concentrations and pressures, making the
other with zero net charge flow. Since half-reactions occur at the values c and p at the CLs different from the inlet values c and p,
TPB surfaces, the charge separation/combination rate vtse has di- which, in no-load conditions, constitute the bulk undisturbed
mensions [C s1 cm2], like a current density. In general conditions, values. In FC mode, reacting gases at both DLs are mixed with water
the surface continuity equation of each charge yields vapor which, at high current densities, can condensate and two-
V$jTPB þ vt s ¼ vt se , with vts being the rate of charge density accu- phase flows can arise [33]. However, in order to avoid burdening
mulation. Since a steady state condition with no charge accumu- the model of the whole device, single-phase mixed-gas transport
lation (i.e. vts ¼ 0) has been assumed, the surface divergence at the has been assumed in all FC-mode load conditions. Thus, Fick's first
TPBs of the current density vector jTPB (i.e. V$jTPB ¼ jTPB ) yields: law N ¼ eDVc has been used to relate the gas concentration
gradient Vc to the gas molar flow vector N through the medium
vt se ¼ jTPB (11) diffusivity D, and thus to the current density vector:

Due to the nanostructured geometry of the TPBs inside the j ¼ nF N (15)

porous structure of the CLs, the total active area ATPB where the
exchange current density is produced is much larger than the cross- Assuming a steady-state linear variation through the DLs, the
section A of the device. As in the case of standard fuel cells, we concentrations at the CLs of the reacting and produced gases are:
assume ATPB/A values of the order of 103 in the UR-PEMFC model.
With the aim of modeling the system performance, it is preferable cR ¼ cR  jt =ðn F KÞ
(16)
to refer the exchange current density to A: cP ¼ cP þ jt =ðn F KÞ
According to the model zero-dimensional assumption, in (16), c
j0 ¼ ðATPB =AÞj0TPB (12)
represents the average inlet concentration in the BP flow channels,
Combining Equations (10)e(12) yields the cell current density whereas the charge carriers are n ¼ 2 for hydrogen and n ¼ 4 for
expression oxygen. The mass transfer coefficient K
 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32 27

!1 range 2e3. Viscosities of fluids have been expressed as:

dfc d
K¼ þ dl (17) 247:8
Sh Dfc Deff
dl
mw ¼ 2:414$105 $10T140 ½Pa$s
 1:5 (21)
8:46$103 T
takes into account both the diffusion effects in the DLs and the mO2 ¼ ½Pa$s
convective flow in the flow channels of the BPs (Fig. 2). dfc and ddl T þ 127 292:25
are the flow channel hydraulic diameter and the diffusion layer
eff where T is in K, as usual. Consequently, in the EL mode fluid pres-
thickness, respectively, and Dfc and Ddl are their diffusion co-
sure jumps at the positive DL result (neglecting gravitational ef-
efficients. According to (17), diffusion in the flow channels depends
fects, which depend on the geometry and position of the cell):
on the Sherwood number Sh, namely the ratio between convective
and diffusive mass transport at the flow channel interface [34].
mM ddl
Instead, the dependence of diffusion on porosity ε and tortuosity t pw ¼ pw  jt
in the DLs has been accounted for by mean of the effective diffusion
εkrnF
 1=2 (22)
coefficient expressed with the Bruggeman relation Deff dl
¼ Ddl ε=t ¼ 2mR T ddl
Ddl ε1:5 (assuming t ¼ ε0:5 ) [35,36]. The dependence of the diffu- pO2 ¼ pO2  jt
εknF
sion coefficients on pressure and temperature has been expressed
as with n ¼ 2 for water, n ¼ 4 for oxygen. M is the water molar mass
     2:33 and r its density. Since this model aims at analyzing the steady-
D ¼ D* p* p T T * (18) state performance of a UR-PEMFC, dynamic mass transfer effects
have not been taken into account.
where D* is the value of D at the temperature T* and pressure p*. In
an experimental set-up, inlet pressures are preferably enforced
rather than concentrations. In order to express all concentration-
3.6. Concentration losses and limit conditions
related equations in term of pressure, the ideal gas law p ¼ cRT
has been used, thus obtaining for both reactants and products
Gradients Vc and Vp affect the on-load emf variation (Nernst
  losses) and activation losses (reaction losses), reducing the emf Eoc
c ¼ p=RT ; c ¼ p=RT ; c=c* ¼ pT * ðpT Þ (19)
and increasing the activation voltage drop Dva. These effects, which
constitute concentration losses and dominate the cell performance
with the last expression allowing for temperature variations as in
at high current densities, are evaluated with (7) and (13). Actually,
(14). In EL mode, liquid water percolation in the positive DL toward
in order to account for secondary concentration-related losses, (7)
the CL, as well as oxygen percolation from it, have been modeled
is increased by a dimensionless coefficient k somewhat larger than
with Darcy's equation q ¼ (k/m)Vp, which relates the mass flux
unity:
vector q [m s1] and the pressure gradient Vp via the through-plane
medium permeability k and the fluid viscosity m. The former is "   #
expressed by means of the CarmaneKozeny absolute permeability T cH2 cO2 0:5
DEcl ¼ k ln (23)
kabs corrected by the relative permeability kr in order to account for f cH2 cO2
the presence of both fluids inside the pores [35,37]:
Equation (21) can be split into two concentration voltage drops
f 2 ε3  m h i at the positive and negative electrodes, allowing for two values for
kw ¼ kabs kr;w ¼ 2
1  sO 2 m2 k:
180ð1  εÞ
(20)    
f2 ε3  m h i T cH2 T cO2
kO2 ¼ kabs kr;O2 ¼ sO2 m2 Dvc þ Dvcþ ¼ k ln þ kþ ln ¼ DEcl (24)
180ð1  εÞ2 f cH2 2f cO2
In FC (discharge) mode, starvation-related working limits occur
where f is the mean pore diameter, sO2 the saturation of oxygen
when reagent gases at the CLs drop to cR ¼ 0 (i.e. pR ¼ 0). The cor-
(namely, its volume fraction in the pores), and m, a coefficient in the
responding limit current densities of each electrode are obtained
from (16):

DL CL PEM nFK
jLd ¼ p (25)
R T R
In fact, each electrode has its own limit, but (23) provides a
smaller value of the limit current density at the positive electrode,
jLdþ, due to the lower diffusivity of oxygen compared to hydrogen.
Thus, jLdþ constitutes the FC-mode cell current density limit.
H2 H+ Conversely, in EL (charge) mode, the limit current density depends
on liquid water feed at the positive electrode and can be deduced
from (22):

n F kw ε r
jLcþ ¼ p (26)
mw M ddl R

Fig. 2. Scheme of fluid mass transport occurring at flow channel and diffusion layer in
However, due to the higher density and faster transport of liquid
the negative electrode, with convective (in BP) and diffusive (in DL) hydrogen flows, water, the limit falls outside practicable current densities, and
respectively in FC-mode operation (in EL-mode flow arrows reverse). therefore it does not appear in the simulation results of Section 5.
28 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32

3.7. PEM ohmic losses

jcr ¼ jcrd þ jcre (35)
When properly hydrated, a Nafion PEM becomes a protonic The two terms jcrd and jcre making up jcr are each due to a
conductor with ohmic behavior, so that Dvm of (9) is proportional to different cause: diffusion and electro-osmotic drag, respectively.
the ionic current density j in the membrane: Diffusive crossover depends on the hydrogen concentration
gradient inside the PEM and can be modeled according to Fick's
Dvm ¼ j dm =g (27)
first law:
where dm is the PEM thickness and g its conductivity, which de-   
jcrd ¼ n F DmH2 dm cH2 (36)
pends on temperature and hydration l ¼ cw/csa, that is, the ratio
between water cw and sulfonic acid csa concentrations, which can DmH2 is the hydrogen diffusivity inside the PEM and cH2 is its
range between 0 and 22. The latter dependence has been modeled concentration at the negative CL as in (18) (assuming zero value at
by adapting an often-used empirical model [38], corrected so as to the positive CL). An Arrhenius-like dependence of DmH2 on tem-
avoid negative values: perature has been considered:
  
gðT; lÞ ¼ go ðTÞ B l (28) WmH2 1 1
DmH2 ¼ D*mH2 exp  (37)
R T* T
where B ¼ 8.7172 is a dimensionless parameter. The temperature
dependence has been expressed according to the Electro-osmotic drag consists of hydrogen molecules H2 drawn
VogeleTammaneFulcher model [39]: by protons Hþ, which move through the PEM to produce the pro-
tonic (i.e. electric) current density j. The term jcre accounts for
  

go ðTÞ ¼ A0 T 0:5 exp Wm R Tg  T (29) persistent crossover effects when cH2 tends to vanish near the limit
current density. In order to estimate it, a model similar to that often
where A0 ¼ 0.048 K0.5 S/cm, Tg ¼ 157 K is the glass transition proposed for water electro-osmotic drag, which increases with
temperature, and Wm ¼ 1.86 kJ mol1 is the activation barrier. The PEM hydration, has been used [41]:
hydration profile along the PEM thickness depends on two major
effects: back-diffusion and electro-osmotic drag [40]. However, a jcre ¼ n z l j (38)
simple linear interpolation between border values lþ and l has
where z is a dimensionless electro-osmotic drag coefficient. The
been assumed in this model, so their average value is used in (25).
resulting equivalent crossover current density depends on
lþ and l are defined by the water activities aw and awþ of the
hydrogen concentration, hydration, and PEM protonic current
reagent gases at the CLs, according to the empirical interpolation
density:
formula [38]:
  
 jcr ¼ n F DmH2 dm cH2 þ n z l j (39)
l¼ 0:043 þ 17:81 aw  39:85 a2w þ 36:0 a3w 0  aw  1
14 þ 1:4ðaw  1Þ 1  aw  3 Crossover hydrogen reaching the positive CL reacts with oxygen,
(30) so that both half-reactions (1) occur here increasing the positive
activation overpotential Dvaþ but without producing electric cur-
Assuming high gas flow rates, water activities aw and awþ (i.e. rent and power conversion. Moreover, since gas molar flows at both
the relative humidity values expressed in absolute terms) at the DLs sustain this reaction of crossover hydrogen, concentration
electrodes have been considered to be actively controlled, namely losses Dvc and Dvcþ at both DLs are also increased. Consequently,
equal to the relative humidity of inlet gases: in computing (13), (16), and (22), the total equivalent current
density jt is considered:
aw ¼ RH=100 ¼ pw =pws ðTÞ (31)
jt ¼ j þ jcr (40)
where pw is the water vapor partial pressure and pws the water
vapor saturation pressure [31]: Crossover hydrogen also produces a loss of stored energy that
reduces round-trip efficiency. Since a concentration gradient
pws ¼ 10ð2:1794þ0:2953T9:1837$10 T 2 þ1:4454$107 T 3 Þ
5
(32) through the PEM exists in OC conditions (i.e. at no-load j ¼ 0), losses
due to diffusive crossover (36) are identified as the main cause of
The pressures of the inlet mixtures pIx and the water activities
the difference between the OC emf of (8) and the observed OCV,
awx define the gas partial pressures and concentrations at the
namely the cell voltage v(0) of (9) in no-load conditions [42].
negative (H2) and positive (O2) electrodes:

p ¼ ðpI  aw pws Þ (33)

3.9. Round-trip efficiency

c ¼ ðpI  aw pws Þ=R T (34) A major figure of merit of URFCs is the round-trip energy effi-
ciency h, i.e. the net output electrical energy Wch divided by overall
input electrical energy Wdch when the URFC system performs a
complete cycle, being first charged (consuming Wch) and then
3.8. Hydrogen crossover
discharged (delivering Wdch) so as to go back to initial conditions:
A major side-effect that impacts on the device performance and Wdch ðj; T; p; aw Þ
efficiency is fuel crossover, namely hydrogen that migrates through h¼ (41)
Wch ðj; T; p; aw Þ
the PEM towards the positive electrode without reacting at the
negative one. To account for it, an equivalent current density cor- Both energy exchanges depend on the currentevoltage time
responding to the migrating hydrogen mass flow as in (15) has been profile v(j) and on the physical operating conditions (temperature T,
considered: gas pressure p, and water activity aw). The overall input and net
 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32 29

output energies include the energy Wan required to operate ancil- current density as a function of the activation voltage drop, a
laries (air pump, gas compressors, cooling, power management tabulated-dataset approach has been exploited, that allows to
system, system control unit, …): Wch ¼ WFCch þ Wan and obtain the activation losses from the current density by resampling,
Wdch ¼ WFCdch  Wan , whereas WFCch;dch are the energies converted i.e. first the code computes the jt(Dva) table from (13) and then
by the fuel cell stack during charge/discharge. obtains the required values Dva(jt) through a polynomial interpo-
In order to derive a simple parametric expression for h, the lation. As already stated, the model is intended as a tool for
charge and discharge phases can be assumed to occur in electrical exploring the potential performance of a storage system based on
steady state conditions and at fixed physical parameters, so as to H2eO2 UR-PEMFC and to address the optimal design of its ancillary
write WFCch;dch ¼ vch;dch Qch;dch , with Q the exchanged electric sub-systems, as usually needed in FC system design [44]. Although
charge. This allows expressing the round-trip energy efficiency as: the model has not been developed for matching existing devices, its
parameters (Table 1) have been identified from the performance of
h ¼ hv hga hst han (42) a test single-cell Nafion PEMFC operated in a test chamber under
tight controlled pressure, humidity, and temperature conditions.
where product terms are ancillary (han), stoichiometric (hst), On the other hand, this cell could only operate in FC-mode, so that
galvanic (hga), and voltage (hv) round-trip efficiency, only FC-mode experimental points could be obtained. The param-
respectively. The ancillary round-trip efficiency eters identification has been carried out by means of a stochastic
han ¼ ð1  Wan =WFCdch Þ=ð1 þ Wan =WFCch Þ largely depends on the optimization procedure [43] that is capable of automatically iden-
design of sub-systems and on the adopted technologies (e.g. gas tifying the model parameters which allow the model performance
storage at high pressure or as solid hydrides), which are out of the to match the experimental data in a proper number of iterations
scope of this work. The second and third terms synthesize the ratio avoiding to fall into local minima. Typically, the search space for
between electric charges reacting during discharge and charge: each parameter ranges from one-tenth of an initial reference value
hga hst ¼ Qdch =Qch . The latter factor, hst, is the reciprocal of the fuel taken from the literature to a tenfold higher value, i.e. on a range
stoichiometric factor (usually labeled with l in the literature, the spanning over two orders of magnitude. If tighter bounds are given,
symbol here used for hydration), which is kept greater than one in the procedure can automatically expand the search space for each
operational conditions to prevent cell starvation. However, a proper parameter according to user preference. Fig. 3 shows the agreement
recovery system of the unburned fuel can ensure hst close to unity. between the model outputs and experimental data in FC-mode. The
The galvanic efficiency hga depends on fuel crossover. According to accuracy in extrapolating the model performance in the EL-mode
the model presented in Subsection 3.8, the crossover equivalent was checked against some experimental working points of a PEM
electric current density is in the order of a few thousandths of the electrolyzed (CINEL® AD-1020 H2 generator) based on a similar
full-load protonic electric current density produced by the MEA, but obtained in less controlled conditions. Nevertheless, nu-
consumed hydrogen. However, being crossover independent of the merical simulations were found to be consistent with such exper-
load and occurring also in OC conditions, its time-integral effect on imental data.
hga can be higher, typically in the order of some percent, depending
on the duty cycle of the system. We are here more interested in the
voltage round-trip efficiency, that can be far from unity since it
Table 1
strongly depends on the electrochemistry of the cell and, conse-
Numerical parameters used in the FC-mode validation.
quently, on the working points along the v(j) curve in both EL and
FC modes. Maintaining the assumption of steady state charge and Parameter Symbol Unit Value

discharge operations, the voltage round-trip efficiency can be Std reversible voltage E 0
V 1.229
expressed as the ratio between discharge and charge voltages Std temperature T0 K 298.15
Std pressure p0 kPa 100
E vdch vdch Neg. exchange current density j0 A cm2 1.13,102
hv ¼ ¼ (43) Pos. exchange current density j0þ A cm2 9.7,105
vch E vch j0 activation barrier Wj J mol1 1951
j0þ activation barrier Wjþ J mol1 2261
and consequently the voltage efficiency depends on the points Neg. transfer coefficient a e 0.52
chosen along the polarization curve for charging and discharging Pos. transfer coefficient aþ e 0.47
Flow-channel hydraulic diameter dfc cm 0.1
(i.e. on j) and on the physical operation conditions (i.e. on T, p, aw).
Sherwood number Sh e 2.69
Due to these features, it affects the overall efficiency much more H2 flow-channel diffusion coefficient D*fcH2 cm2 s1 0.69 (40  C, 1 bar)
than hga, since at high current densities (i.e. at high device exploi- O2 flow-channel diffusion coefficient D*fcO2 cm2 s1 0.11 (40  C, 1 bar)
tation) it is intrinsically far from unity. The model allows investi- Diffusion layer porosity ε e 0.4
gating such dependence on the charge and discharge operating Mean pore diameter f m 3$106
Water molar mass M g mol1 18.0153
points, in order to ensure proper exploitation of both device (capex) Relative permeability parameter m e 2.8
and fuel (running costs). Oxygen pore saturation sO2 e 0.0e0.57
Water density r g m3 1$106
Neg. and Pos. DL thickness ddl mm 200 (100)
4. Numerical implementation Deff cm2 s1 0.18 (40  C, 1 bar)
H2 DL diffusion coefficient H2 *
O2 DL diffusion coefficient eff
DO2 * cm2 s1 0.028 (40  C, 1 bar)
The multi-physical model presented in Section 3 has been Neg. concentration loss coefficient k e 2.61
implemented in a Matlab numerical code in order to simulate the Pos. concentration loss coefficient kþ e 1.47
steady-state performance of a reference UR-PEMFC. The numerical PEM thickness dm mm 150
PEM conductivity parameter B e 8.7172
code avoids introducing approximations in order to obtain
PEM conductivity parameter A0 K0.5S cm1 0.048
consistent analytical expressions for all the input/output relation- PEM glass trans. temperature Tg K 157
ships. The independent variable of the model is the electric current PEM conductivity activation barrier Wm J mol1 1863
density j, that is used to compute the cell voltage terms included in H2 diffusion coefficient in PEM DmH2 cm2 s1 5,104
(9). In order to overcome the difficulties arising from the non- H2 diffusion activation barrier WmH2 J mol1 3517
H2 electro-osmotic drag coefficient z e 1.78,104
invertible ButlereVolmer Equation (13), which expresses the
30 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32

Fig. 3. FC-mode operation: polarization curve validated against experimental data

from a test single-cell Nafion PEMFC (T ¼ 40  C, p ¼ 1.03 bar, aw ¼ awþ ¼ 1). Fig. 5. Cell voltage v and power density P versus PEM current density j at four (positive
electrode) water activities in the range of 0.1e1.0 and at T ¼ 80  C, p ¼ 2 bar.

5. Results

The model is designed for assessing the steady-state electrical

performance of UR-PEMFCs as a function of the main physical pa-
rameters which are under the control of the system supervisor:
temperature, gas humidification, and fluid inlet pressures (i.e.
concentrations). Since this research is aimed at creating a general-
purpose H2eO2 UR-PEMFC model, capable of investigating both EL
and FC modes of a potential high-performance H2eO2 UR-PEMFC
based on up-to-date materials, some of the parameter values
used in the dual-mode simulations have been updated with respect
to those obtained in the validation phase, in order to reflect the
state-of-the-art with data taken from the literature as reported in
the References (such values are indicated in parentheses in Table 1).
Figs. 4e6 show the cell voltage v [V] and the power density P ¼ vj
[W cm2] as functions of the electric current density j [A cm2] in
the PEM in both EL (j < 0) and FC (j > 0) modes for four tempera-
tures in the range 60e120  C, four water activities in the range Fig. 6. Cell voltage v and power density P versus PEM current density j at three inlet
gas pressures p in the range 1e3 bar and at T ¼ 80  C, aw ¼ 1.
0.1e1.0, and three gas inlet pressures p in the range 1e3 bar. These
ranges correspond to proper operating conditions of existing PEM
ELs and FCs based on Nafion® membranes. The simulated perfor-
mance shows that lower water activities, allowing for higher gas benefit of operating at higher T appears in both modes. It can also
partial pressures, can provide better performance at high load be noted that, although the PEM conductivity strongly depends on
conditions, both in the FC-mode and in the EL-mode. The inlet gas temperature and hydration, such effects are not so strong on the
pressure dramatically influences the performance in the FC mode overall device performance, provided that a well-functionalized
and higher pressure can significantly extend the discharge current polymer is used for the PEM. The beneficial effect in both EL and
density range, but it has a weak effect in the EL-mode, where the FC modes of adopting thinner layers (DLs and PEM) also emerges
reactant is liquid water. Conversely, temperature affects the work- from the simulations, according to the current trend of commercial
ing condition more in the EL mode than in the FC mode, even if the FCs. At present many laboratories are engaged in an effort for
developing high temperature PEMFCs (H-PEMFCs) based on ma-
terials such as PBI (polybenzimidazole), which could allow opera-
tions at 150e180  C. The model here presented is not suitable for
such devices, because a membrane model different from that pre-
sented in Sections 3.7 and 3.8 should be used for PBI, but it suggests
that even more beneficial effects from such new materials could
result in the EL operation of the stack.
The voltage round-trip efficiency (43) varies with the working
points on v(j) and with the operating conditions (T, p, aw). Fig. 7
shows how the voltage efficiencies (E/v in EL mode and v/E in FC
mode) are related to the power density P in both modes at a
reference condition (T ¼ 80  C, p ¼ 2 bar, and aw ¼ 0.7) and how
they vary when temperature and pressure are changed to
T ¼ 100  C and p ¼ 3 bar, respectively. In the FC mode, a voltage
efficiency as high as 0.8 can be achieved at 25% of Pmax, but it drops
to 0.45 when the device is operated at maximum power Pmax.
Referring to the FC mode, the low-efficiency operating points
Fig. 4. Cell voltage v and power density P versus PEM current density j at four tem- represent unfavorable conditions, since, for a given power density,
peratures in the range of 60e120  C and at p ¼ 2 bar, aw ¼ 1. higher efficiencies are found on the sections of the curves on the
 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32 31

Acknowledgments

The authors wish to thank Prof. Vito Di Noto and Dr. Enrico
Negro for providing the experimental data used to validate the
model. This work was supported by the University of Padua under
the 2011 strategic project MAESTRA, grant STPD11XNRY.

References

[1] World Energy Outlook 2014, International Energy Agency, 2014 (Retrieved
June 2nd, 2015), www.worldenergyoutlook.org/publications/weo-2014/.
[2] C. Pieper, H. Rubel, Revisiting Energy Storage. There is a Business Case, Boston
Consulting Group Report, Feb 2011.
[3] B. Dunn, H. Kamath, J. Tarascon, Electrical energy storage for the grid: a bat-
tery of choices, Science 334 (2011) 928e935.
[4] Z. Yang, J. Zhang, M.C.W. Kintner-Meyer, X. Lu, D. Choi, J.P. Lemmon, J. Liu,
Electrochemical energy storage for green grid, Chem. Rev. 111 (2011)
3577e3613.
[5] Z. Weber, M.M. Mench, J.P. Meyers, P.N. Ross, J.T. Gostick, Q. Liu, Redox flow
batteries: a review, J. Appl. Electrochem. 41 (2011) 1137e1164.
[6] W. Smith, The role of fuel cells in energy storage, J. Power Sources 86 (2000)
74e83.
[7] G.L. Soloveichik, Regenerative fuel cells for energy storage, Proc. IEEE 102
(2011) 964e974.
[8] O.A. Converse, Seasonal energy storage in a renewable energy system, Proc.
IEEE 100 (2012) 401e409.
[9] C.E. Thomas, Fuel cell and battery electric vehicles compared, Int. J. Hydrogen
Fig. 7. Voltage efficiencies in both FC and EL modes (the product of which gives the Energy 34 (2009) 6005e6020.
round-trip voltage efficiency) versus power density P for three different physical [10] D. Steward, G. Saur, M. Penev, T. Ramsden, Lifecycle Cost Analysis of Hydrogen
conditions. Note that power density scales are different in the FC (positive) and EL Versus Other Technologies for Electrical Energy Storage, National Renewable
(negative) modes. Energy Laboratory, Golden, CO, USA, 2009. NREL/TP-560e46719.
[11] S.A. Grigoriev, P. Millet, K.A. Dzhus, H. Middleton, T.O. Saetre, V.N. Fateev,
Design and characterization of bi-functional electrocatalytic layers for appli-
cation in PEM unitized regenerative fuel cells, Int. J. Hydrogen Energy 3 (2010)
right of the dotted line. This region of the Cartesian plane therefore 5070e5076.
represents the Pareto front of the problem, that is, those points for [12] F. Barbir, T. Molter, L. Dalton, Efficiency and weight trade-off analysis of
which an increase of one objective (e.g. power density) results in a regenerative fuel cells as energy storage for aerospace applications, Int. J.
Hydrogen Energy 30 (2005) 351e357.
decrease of another one (e.g. efficiency). Optimal conditions can be [13] P. Alotto, M. Guarnieri, F. Moro, A fully coupled three-dimensional dynamic
chosen depending on additional parameters, for example, of eco- model of polymeric membranes for fuel cells, IEEE Trans. Magn. 46 (2010)
nomic nature, which have not been included in the model. High 3257e3260.
[14] P. Alotto, M. Guarnieri, F. Moro, A. Stella, Multi-physic 3D dynamic modelling
round-trip efficiency is directly related to running-cost effective-
of polymer membranes with a proper generalized decomposition model
ness, whereas power density is related to capex effectiveness. In reduction approach, Electrochim. Acta 57 (2011) 250e256.
any real storage system the cell will operate over a rather wide [15] V. Di Noto, M. Guarnieri, F. Moro, A dynamic circuit model of a small direct
range of current densities, depending on the load conditions. At the methanol fuel cell for portable electronic devices, IEEE Trans. Ind. Electron 57
(2010) 1865e1873.
same time, each loss effect gives a different relative contribution to [16] M. Carmo, D.L. Fritz, J. Mergel, D. Stolten, A comprehensive review on PEM
the overall efficiency depending on the load condition. Similarly, water electrolysis, Int. J. Hydrogen Energy 38 (2013) 4901e4934.
other sets of objectives, even more than two, and any combinations [17] S.A. Grigoriev, V.I. Porembsky, V.N. Fateev, Pure hydrogen production by PEM
electrolysis for hydrogen energy, Int. J. Hydrogen Energy 31 (2006) 171e175.
of parameters can be studied (e.g. related to the combined opti- [18] F. Barbir, PEM electrolysis for production of hydrogen from renewable energy
mization of the power management system or to optimal control by sources, Sol. Energy 78 (2005) 661e669.
the system supervisor, which are usual aims in designing FC sys- [19] P. Millet, R. Ngameni, S.A. Grigoriev, N. Mbemba, F. Brisset, A. Ranjbari,
C. Etievant, PEM water electrolyzers: from electrocatalysis to stack develop-
tems [44,45]). In this view, the model can constitute an effective ment, Int. J. Hydrogen Energy 35 (2010) 5043e5052.
tool in addressing the design of new devices suited for a specific [20] A. Awasthi, Keith Scott, S. Basu, Dynamic modeling and simulation of a proton
service, if it is coupled with a multi-objective optimization proce- exchange membrane electrolyzer for hydrogen production, Int. J. Hydrogen
Energy 36 (2) (Nov. 2011) 14779e14786.
dure aimed at determining the ideal constructive and operational [21] M. Chandesris, V. Me deau, N. Guillet, S. Chelghoum, D. Thoby, F. Fouda-Onana,
parameters. However, these analyses are out of the scope of this Membrane degradation in PEM water electrolyzer: numerical modeling and
paper. experimental evidence of the influence of temperature and current density,
Int. J. Hydrogen Energy 40 (2015) 1353e1366.
[22] L.F.L. Oliveira, S. Laref, E. Mayousse, C. Jallutwb, A.A. Franco, A multiscale
6. Conclusions physical model for the transient analysis of PEM water electrolyzer anodes,
Phys. Chem. Chem. Phys. 14 (2012) 10215e10224.
[23] C. Wang, M.H. Nehrir, Power management of a stand-alone wind/photovol-
The goal of this paper is to demonstrate that a zero-dimensional taic/fuel cell energy system, IEEE Trans. Energy Convers. 23 (2008) 957e967.
steady-state model developed for simulating the behavior of [24] J. Rugolo, B. Huskinson, M.J. Aziz, Model of performance of a regenerative
PEMFCs as electric power generators can be extended to analyze hydrogen chlorine fuel cell for grid-scale electrical energy storage,
J. Electrochem. Soc. 159 (2012) B133eB144.
the behavior of a UR-PEMFC capable of working in both charge and [25] B. Huskinson, M.J. Aziz, Performance model of a regenerative hydrogen
discharge modes. A proper tuning of the parameters used in the bromine fuel cell for grid-scale energy storage, Energy Sci. Technol. 5 (2013)
non-linear equations is needed to calibrate the model on a specific 1e16.
[26] H. Kreutzer, V. Yarlagadda, T. Van Nguyen, Performance evaluation of a
device. Such an issue can be addressed by means of numerical
regenerative hydrogen-bromine fuel cell, J. Electrochem. Soc. 159 (2012)
optimization and identification techniques applied to large sets of F331eF337.
experimental data taken from prototype UR-PEMFCs. The further [27] A. Doddathimmaiah, J. Andrews, Theory, modelling and performance mea-
development of similar models will be instrumental in forecasting surement of unitised regenerative fuel cells, Int. J. Hydrogen Energy 34 (2009)
8157e8170.
the performance of hydrogen-based energy storage systems for [28] P. Alotto, M. Guarnieri, Multi-physics model for regenerative PEM fuel cell
both stationary and mobile electric applications. energy storage, in: Proc. IEEE Int. Conf. Industrial Technology (ICIT 2013), Cape
32 M. Guarnieri et al. / Journal of Power Sources 297 (2015) 23e32

Town (SA), 746e751, 2013, pp. 25e27. f: reaction parameter (nF/R ¼ 23,210 K V1)
[29] A. Giustiniani, G. Petrone, G. Spagnuolo, M. Vitelli, Low-frequency current j: electric current density (A m2)
oscillations and maximum power point tracking in grid-connected fuel-cell- j: electric current density vector
based systems, IEEE Trans. Ind. Electron 57 (2010) 2042e2053. j0: sub-reaction exchange current density (A m2)
[30] A.J. Bard, L.R. Faulkner, Electrochemical Methods, Fundamentals and Appli- jt: total equivalent current density (A m2)
cations, John Wiley & Sons, New York, 2001. K: mass transfer coefficient (m s1)
[31] R. O'Hayre, S.W. Cha, W. Coltella, F.B. Prinz, Fuel Cell Fundamentals, John k: permeability (m2)
Wiley & Sons, New York, 2009, pp. 195e202. M: water molar mass (18.01529 g mol1)
[32] A. Parthasarathy, S. Srinivasan, A.J. Appleby, C.R. Martin, Temperature m: relative permeability parameter (dimensionless)
dependence of the electrode kinetics of oxygen reduction at the platinum/ n: charge carriers per reaction
Nafion® interface e a microelectrode investigation, J. Electrochem. Soc. 139 N: gas molar flow vector
(1992) 2530e2537. N: gas molar flow (mol m2 s1)
[33] L. You, H. Liu, A two-phase flow and transport model for the cathode of PEM p: pressure (Pa)
fuel cells, Int. J. Heat. Mass Transf. 45 (2002) 2277e2287. q: fluid mass flux vector
[34] W.M. Kays, M.E. Crawford, Convective Heat and Mass Transfer, McGraw-Hill, q: fluid mass flux (m s1)
New York, 1993. R: gas constant (8.314 J mol1 K1)
[35] M.M. Mench, Fuel Cell Engines, Wiley, Hoboken, 2008. s: fluid pore saturation (dimensionless)
[36] Z.H. Wang, C.Y. Wang, K.S. Chen, Two-phase flow and transport in the air Sh: Sherwood number (dimensionless)
cathode of the proton exchange membrane of fuel cells, J. Power Sources 94 T: temperature (K)
(2001) 40e50. v: cell voltage (V)
[37] W.L. McCabe, J.C. Smith, P. Harriot, Unit Operations of Chemical Engineering, W: activation barrier (J mol1)
seventh ed., McGraw-Hill, New York, 2005, pp. 163e165. a: transfer coefficient
[38] T.E. Springer, T.A. Zawodzinski, S. Gottesfeld, Polymer electrolyte fuel cell В: PEM conductivity parameter (dimensionless)
models, J. Electrochem. Soc. 138 (1991) 2334e2341. g: PEM conductivity (S m1)
[39] V. Di Noto, M. Piga, G. Pace, E. Negro, S. Lavina, Dielectric relaxations and DE: emf variation from standard and no-load condition (V)
conductivity mechanism of Nafion: studies based on broadband dielectric Dv: cell voltage variation (V)
spectroscopy, ECS Trans. 16 (2008) 1183e1193. Db s ðTi Þ: molar entropy variation with temperature
[40] S. Tsushima, S. Hirai, In situ diagnostics for water transport in proton ex- ε: porosity (dimensionless)
change membrane fuel cells, Prog. Energy Combust. Sci. 37 (2011) 204e220. z: electro-osmotic drag coefficient (dimensionless)
[41] K.D. Baik, I.M. Kong, B.K. Hong, S.H. Kim, M.S. Kim, Local measurements of h: efficiency (dimensionless)
hydrogen crossover rate in polymer electrolyte membrane fuel cells, Appl. k: concentration losses coefficient (dimensionless)
Energy 101 (2013) 560e566. l: hydration (dimensionless)
[42] S.A. Vilekar, R. Datta, The effect of hydrogen crossover on open-circuit voltage m: viscosity (Pa s)
in polymer electrolyte membrane fuel cells, J. Power Sources 195 (2010) r: water density (106 g m3)
2241e2247. s: accumulating surface charge density (C m2)
[43] P. Alotto, M. Guarnieri, Stochastic methods for parameter estimation of se: reacting surface charge density (C m2)
multiphysics models of fuel cells, IEEE Trans. Magn. 50 (2014). #7017304. t: tortuosity (dimensionless)
[44] L. Boulon, D. Hissel, A. Bouscayrol, M.C. Pe ra, From modeling to control of a f: mean pore diameter in the porous DL
PEM fuel cell using energetic macroscopic representation, IEEE Trans. Ind. V: nabla (del gradient vector operator)
Electron 57 (2010) 1882e1891.
[45] C.A. Ramos-Paja, R. Giral, L. Martinez-Salamero, J. Romano, A. Romero, Subscripts and superscripts
G. Spagnuolo, A PEM fuel-cell model featuring oxygen-excess-ratio estimation
and power-electronics interaction, IEEE Trans. Ind. Electron 57 (2010)
1914e1924. c, p: values of c and p at the bipolar plate flow channels
e: negative electrode and half-reaction
*: reference value
Nomenclature þ: positive electrode and half-reaction
a: activation losses
abs: absolute
an: ancillary
Acronyms
c: concentration losses
ch: charge mode
BEV: battery electric vehicle cl: on-load concentration losses
BP: bipolar plate co: open-circuit concentration losses
CL: catalyst layer cr: crossover
DL: diffusion layer crd: diffusive crossover
ECES: electrochemical energy storage cre: electro-osmotic crossover
EL: electrolyzer d: diffusion
EV: electric vehicle dch: discharge mode
FC: fuel cell dl: diffusion layer
HBES: hydrogen based e: electro-osmotic
HOR: hydrogen oxidation reaction eff: effective (of diffusion coefficients)
MEA: membrane electrode assembly fc: flow channel
ORR: oxygen reduction reaction FC: fuel cell
PEM: proton exchange membrane g: glass transition
PEMEL: PEM EL ga: galvanic
PEMFC: PEM FC I: inlet
RFB: redox flow battery l: on-load
RFC: regenerative FC L: limit
TPB: triple phase boundaries Lc: limit during charge (EL mode)
UR-PEMFC: unitized regenerative PEMFC Ld: limit during discharge (FC mode)
URFC: unitized regenerative FC m: membrane, proton exchange membrane (PEM)
OC: open circuit (no-load)
Symbols Px, P: product of the x half-reaction
r: relative
Rx, R: reactant of the x half-reaction
A: cell active cross-sectional area (m2) sa: sulfonic acid
a: activity (of water, dimensionless) st: stoichiometric
A0: PEM conductivity parameter (0.048 K0.5 S/cm) TPB: triple phase boundaries (half-reaction sites)
c: concentration (mol m3) v: voltage
d: thickness, diameter (m) w: water
D: diffusion coefficient Wj: activation barrier of the half-reaction kinetics (J/mol)
E: electromotive force e emf - reversible voltage (V) ws: water saturation
E0: emf in standard conditions (V) x: x sub-reaction
F: Faraday constant (96,485 C mol1 )