ww.sciencedirect.com
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 6
Available online at w
ScienceDirect
journal homepage: www.elsevier .com/locate/he
Design and testing of a 9.5 kWe proton exchangemembrane fuel cellesupercapacitor passive hybridsystem
Billy Wu a,b,*, Michael A. Parkes b, Vladimir Yufit b, Luca De Benedetti a,Sven Veismann a, Christian Wirsching a, Felix Vesper a,Ricardo F. Martinez-Botas a, Andrew J. Marquis a, Gregory J. Offer a,Nigel P. Brandon b
aDepartment of Mechanical Engineering, Imperial College London, SW7 2AZ, UKbDepartment of Earth Science and Engineering, Imperial College London, SW7 2AZ, UK
a r t i c l e i n f o
Article history:
Received 4 October 2013
Received in revised form
23 February 2014
Accepted 10 March 2014
Available online 13 April 2014
Keywords:
Proton exchange membrane fuel cell
Supercapacitors
Passive system
Balance of plant
Electric vehicle
* Corresponding author. Department of EarthE-mail address: [email protected]
http://dx.doi.org/10.1016/j.ijhydene.2014.03.00360-3199/Copyright ª 2014, Hydrogen Ener
a b s t r a c t
The design and test of a 9.5 kWe proton exchange membrane fuel cell passively coupled
with a 33 � 1500 F supercapacitor pack is presented. Experimental results showed that the
system reduced dynamic loads on the fuel cell without the need for additional DC/DC
converters. Fuel efficiency gains of approximately 5% were achieved by passive hybrid-
isation in addition to addressing two main operational degradation mechanisms: no-load
idling and rapid load cycling.
Electrochemical Impedance Spectroscopy measurements indicated that the super-
capacitor capacitance dropped with decreasing cell voltage and suggested that operation
below 1.3 V is not recommended. Knee-frequency measurements suggested little benefit
was gained in using passive systems with load cycles that have frequency components
above 0.19 Hz. Analysis of system sizing suggested using the minimum number of
supercapacitors to match the open circuit voltage of the fuel cell to maximise load
buffering.
Copyright ª 2014, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights
reserved.
Introduction
Fuel Cells (FC) are electrochemical energy conversion devices
that turn a chemical fuel into electrical energy at an efficiency
typically higher than direct combustion. Of the different
Science and Engineeringk (B. Wu).
83gy Publications, LLC. Publ
varieties of FCs, low temperature proton exchangemembrane
fuel cells (PEMFC) fuelled by hydrogen have received the
greatest attention with respect to their automotive applica-
tions as a possible replacement for the Internal Combustion
Engine (ICE) [1]. Barriers to mainstream adoption of Fuel Cell
Vehicles (FCV) include cost, durability and lack of hydrogen
, Imperial College London, SW7 2AZ, UK.
ished by Elsevier Ltd. All rights reserved.
i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 67886
refuelling infrastructure. Addressing the issue of degradation,
the United States Department of Energy (US DOE) stated that
in order for FC systems to compete with ICEs, lifetimes of
more than 5000 h with less than a 10% loss in performance
needs to be achieved [2]. Under automotive applications, the
three main causes of PEMFC degradation have often been
cited as being: no-load idling, rapid power cycling and start-
up/shut-down cycles [3].
No-load idling leads to accelerated rates of carbon corro-
sion, especially so on the Cathode Catalyst Layer (CCL). This
corrosion is mainly associated with the electrochemical
oxidation of the carbon particles supporting the catalyst, and
can lead to a loss of catalytic activity [4,5]. It is accelerated by
the FC running at elevated potentials which exceed the carbon
oxidation potential of 0.518 V. Two mechanisms of carbon
corrosion are shown in Eqns. (1) and (2). Carbon corrosion can
also be accelerated under conditions where the potential
changes rapidly through load cycling or through fuel starva-
tion leading to cell reversal [6].
Cþ 2H2O/CO2 þ 4Hþ þ 4e� E0 ¼ 0:207VRHE 25 �C (1)
CþH2O/COþ 2Hþ þ 2e� E0 ¼ 0:518VRHE 25 �C (2)
Another process affecting FC durability is platinum cata-
lyst dissolution. This leads to the decrease of effective catalyst
surface area and a loss of electrochemical performance. Plat-
inum dissolution may proceed through a number of mecha-
nisms including; coarsening of the platinum particles in the
catalyst layer [7] and Oswald ripening [8], diffusion of plat-
inum into the membrane [9] and a thermodynamically
favourable particle agglomeration [10]. While the actual
mechanism is still debated, platinum dissolution rates have
been linked to FC operating conditions. Mathias et al. [11]
showed that platinum dissolution could be reduced by
lowering the operating voltage of a FC at low relative humid-
ity, while Borup et al. [12] found that load cycling significantly
accelerated platinum dissolution in comparison to steady
state operation.
During start-up/shut-down events, a hydrogen/air bound-
ary forms which can result in large internal currents; accel-
erating the degradation of FC electrodes. Maranzana et al. [13]
showed that these internal currents can be as high a 1 A/cm2
and vary depending on the flow rate of gases during start-up.
Rapid load cycling of PEMFCs also leads to inefficiencies
due to increased mass transport losses. Gas manifolding ef-
fects, compounded with compressor inertia can result in a
delay time on the order of seconds before a requested increase
in the mass flow rate of air is provided. By comparison, the
rate of electrochemical reaction of the Oxygen Reduction Re-
action (ORR) is several orders of magnitude faster. Therefore,
unless the system is operated with a high oxygen stoichiom-
etry, or the load ramp rate is limited, rapid increases in the
load will result in increased mass transport losses.
To alleviate these detrimental effects and allow the
downsizing of the FC stack, hybridisation with batteries and/
or supercapacitors has been suggested. Yufit and Brandon [14]
developed a FCebattery hybrid system which was actively
controlled using a DC/DC converter where the lithium-ion
battery was used to meet peak power requirements whilst
the DC/DC converter maintained the FC under a constant
current load. Hosseinzadeh et al. [15] investigated the theo-
retical optimisation of a FCelead acid battery system for a
forklift truck powertrain. An optimal design was found to give
theoretical reductions in hydrogen consumption by up to 26%
while Garcia et al. [16] modelled the performance of a
FCebatteryeSC powertrain. Using lithium-ion batteries and
supercapacitors connected via DC/DC converters, the hybrid
powertrain was found to smooth load cycling and enable the
FC to operate at conditions close to steady state. Efficiencies of
the FC were predicted at 60.9% while the hybrid vehicle had a
system efficiency of 55.6%. Hwang et al. [17] investigated the
performance of a FCelithium-ion battery hybrid for use in a
light electric vehicle. This powertrain was found to have an
efficiency of 46%. Hwang also reported on the performance of
a FCelithium-ion battery generator [18]. This system was able
to smooth rapid changes in load while achieving FC effi-
ciencies between 52 and 58%.
The addition of the DC/DC converters in all examples,
however, introduces an extra level of power conversion,
which typically ranges from between 90 and 99%, depending
on the load and construction of the converter [19,20]. The
additional cost of this extra component can also be significant,
with 2010 costs in the region of 25.9 $/kW�1 [21] and predicted
2030 costs of 6e19 $/kW�1 [22]. The concept of passive
hybridisation has therefore been proposed as a means of
gaining the benefits of hybridisation without the addition of
large costly power electronics. In this configuration, the FC
and secondary energy storage device are connected in parallel
directly. During operation the voltages of both devices are
therefore intrinsically balanced.
Several studies have previously examined the effect of
passive hybrid systems. Kuperman et al. [23] used a frequency
domain based analysis to investigate the potential benefits of
a batteryesupercapacitor passive hybrid. It was shown that
supercapacitors act as a low-pass filter for the battery under
pulse loading, leading to reduced battery stress. The disad-
vantage of passive hybridisation was highlighted as being the
loss of active power regulation. Bernard et al. [24] showed that
active pressure control of a FC anode and cathode can be used
to regulate power flows in a FCebattery passive hybrid sys-
tem. Nishizawa et al. [25] developed a FCebattery passive
hybrid system for aircraft applications and showed that there
was an increase in efficiency compared to equivalent systems
using DC/DC converters and that the lower impedance
response of the battery resulted in more mild transient loads
for the FC. Keranen et al. [26] developed a FCesupercapaci-
torebattery triple hybrid and showed that benefits of passive
hybridisation could be achieved with a relatively small
supercapacitor pack and that system voltage should be near
the maximum voltage of the supercapacitor pack due to small
voltage variations.
Within the academic literature, there have been no sys-
tems that have studied passive hybridisation between a FC
and supercapacitor. Therefore, this paper presents electro-
chemical characterisation of supercapacitor cells and FC
stacks to provide an unique insight into the coupled perfor-
mance of a FCesupercapacitor passive hybrid system and
propose a sizing methodology that can be used in automotive
applications.
Fig. 1 e Schematic of balance of plant system for a 9.5 kWe PEMFC stack. Adapted from Wu et al. [27].
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 6 7887
Fuel cell system
The FC stack used in this study was a Nedstack P9.5-75, which
comprised of 75 cells (Johnson Matthey), each with an active
area of 200 cm2. The stack of 75 cells was rated to supply
9.5 kWe of power.
Balance of plant
A schematic of the supporting Balance of Plant (BOP) system,
designed and built in-house for the 9.5 kWe Nedstack, is
shown in Fig. 1.
The air subsystem draws in ambient air using a 24 VDC low
pressure blower followed by a membrane humidifier before
entering the cathode inlet of the stack. Oxygen depleted air
from the cathode outlet was then channelled back into the
humidifier which recovered heat and water. The air was then
released to the atmosphere through a back pressure regulator.
Air stoichiometry was maintained during normal operation at
a target value of 2 through a feedforward controller.
Hydrogen was supplied at 1.5 bar at the inlet. The dry
hydrogen entered into an in-house designed bubble humidi-
fier, through de-ionised water, that was also used to recover
heat from the cooling loop through a heat exchanger. In order
to prevent an uneven distribution of hydrogen across the in-
dividual cells, the exhaust hydrogen was recirculated back to
the inlet with a diaphragm pump to increase the working
stoichiometry to >1.5. Uneven hydrogen distribution across a
single cell can occur as result of the flow plate design and
reduction in partial pressure of hydrogen due to the con-
sumption of reactants. The anode gas loopwas also connected
to a purge valve which periodically removed the mixture of
nitrogen crossover, water and hydrogen, minimising the
cause of localised cell flooding.
In order to minimise the effect of internal currents, caused
by the hydrogen front propagation across the flow channels
during start-up events, a nitrogen supplywas used to purge air
in the anode manifolds and Gas Diffusion Layer (GDL). In
addition, the nitrogen supply was used to remove excess
hydrogen in the anode after use.
To remove excess heat and maintain the FC stack at a
constant operating temperature of 65 �C, a cooling loop which
consisted of a centrifugal pump, radiator and PID-controlled
fan was implemented.
Control and monitoring of the BOP components was real-
ised with a National Instruments (NI) compactRIO (cRIO)
running LabVIEW�. A detailed description of the sensors and
equipment used for the FC can be found in Wu et al. [27] and
Cordner et al. [28].
Fuel cell stack test
The currentevoltage characteristics of the FC stack will
essentially define the operating region of the supercapacitors
in a passive hybrid system, since the potentials are linked. For
a FC, the voltage is a function of temperature, hydration and a
range of other external conditions. The optimum operating
temperature of a conventional PEMFC is generally regarded as
being in the range of 60e80 �C [29]. Higher temperatures risk
dehydration of the Nafion� membrane due to water evapo-
ration, subsequently increasing the likelihood of mechanical
failure of the membrane and pin-hole formation [30]. Lower
temperature operation results in slower electrochemical ki-
netics and increased likelihood of flooding due to the reduced
capacity of the air to remove produced water. Fig. 2(a) shows a
low temperature (30 �C) polarisation curve for the Nedstack
P9.5-75 stack before the membrane was hydrated (left idle for
>1 week). The stoichiometry of the cathode and anode were
maintained at 2 and 1.5 respectively, with the relative hu-
midity of the cathode and anode at >99% and 70% respec-
tively. This represents the typical operating conditions of the
designed system. The achieved peak power was only 4 kWe
due to lower membrane conductivity (low hydration) and
slower rates of electrochemical kinetics (low temperature). At
an operating temperature of 65 �C and fully humidified con-
ditions, the peak power reached 8 kWe. The full 9.5 kWe power
was not achieved due to the low pressure operating
Fig. 2 e Comparison of ambient temperature/dry start and
operational temperature polarisation curves (a) and stack
resistance values (b).
Fig. 3 e Bode plot of total capacitance measured for a 1500 F
Maxwell supercapacitor at different OCP at a temperature
of 20 �C.
i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 67888
limitations of the blower used,whichwas not able to achieve a
pressure greater than 1.2 bar (absolute). This limited both the
anode and cathode operating pressure.
The resistance of the FC stack varies with load current.
Fig. 2(b) shows the resistance of the FC stack which was
calculated based on Eqn. (3), where R, V and I represents the
resistance, voltage and current, respectively. Here VOCV is the
no-load voltage of the FC stack.
R ¼ DVDI
¼��VOCV � Vworking
��jIj (3)
The resistance of the FC and supercapacitor will define the
instantaneous power distribution between the devices in a
coupled system. At higher currents, the stack resistance de-
creases due to a combination of increased water production at
higher currents, leading to higher membrane conductivity,
and lower charge transfer resistance as suggested by the
exponential component in the ButlereVolmer equation
shown in Eqn. (4) for a 1 electron reaction. Here i represents
the current density, i0 the exchange current density, aa/ac the
anodic and cathodic charge transfer coefficients, F is Faraday’s
constant, R is the universal gas constant, T is temperature and
h is the overpotential. Operation at even higher currents
showed an increase in resistance, which is likely due to a
combination mass transport effects and dehydration of the
anode due to osmotic drag effects.
i ¼ i0
�exp
�aaFhRT
�� exp
�acFhRT
��(4)
Supercapacitor characterisation
Direct coupling necessitates the balancing between the po-
tentials of a supercapacitor pack and the FC stack. Therefore,
in order to appropriately size the system and understand its’
limitations, the dynamic performance of supercapacitors, in
terms of impedance and capacitance behaviour at different
frequencies and State-Of-Charges (SOC), were characterised.
Supercapacitors store energy via the formation of a
charged double layer at the interface between its porous car-
bon electrodes and an electrolyte. Due to the absence of the
charge transfer process, supercapacitors are characterised by
relatively long lifetime, low internal impedance and fast
response time. However, their energy density is limited, as is
their low operating potential (2.7e2.85 V) due to the possible
onset of electrolyte decomposition.
The supercapacitor pack constructed for this study con-
sisted of 33 � 1500 F Maxwell supercapacitors connected in
series giving a maximum operating potential of 89.1 V and
rated capacitance of 45.45 F. It should be noted that the
maximum pack voltage was never reached during operation
because of the limitation imposed by open circuit voltage
(OCV) of the FC which was 72 V. Redundancy was added to
compensate for the absence of a cell balancing system.
In order to characterise a single supercapacitor cell, Elec-
trochemical Impedance Spectroscopy (EIS) measurements at
different OCP were recorded in galvanostatic mode with a 1 A
current amplitude on a Biologic VSP multichannel potentio-
stat/FRA equippedwith 5 A booster. Prior to themeasurement,
the cell was charged or discharged at constant current to a
target potential and then held at this potential until the cur-
rent dropped to <1mA, after which a 30min settling time was
allowed before the measurement was taken. By extracting the
Fig. 4 e (a) Nyquist plot of impedance and (b) fitted series,
pores, total polarisation resistances and DL capacitance for
a 1500 F Maxwell supercapacitor at 20 �C.
Fig. 5 e Bode plot of imaginary capacitance for a 1500 F
Maxwell supercapacitor at different cell voltages. Inset
shows the knee-frequency dependency with cell voltage at
20 �C.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 6 7889
real (ZRe) and imaginary (ZIm) components of the total
impedance (jZj) from the EIS measurements, it was possible to
describe the complex capacitance as shown in Eqns. (5)e(7)
[31]. Where Ctot, CRe and CIm are the total complex, real and
imaginary parts of the capacitance respectively.
Ctot ¼ CRe � jCIm (5)
CRe ¼ �ZIm
ujZj2 (6)
CIm ¼ ZRe
ujZj2 (7)
A bode plot of the total capacitance at different OCP values
is shown in Fig. 3. Capacitance was observed as being fre-
quency dependent; at higher frequencies under AC current
oscillations, ions do not have sufficient time to diffuse into
mesopores of the electrode, thus not utilising the available
surface area for ion adsorption resulting in a lower accessible
capacitance. At lower frequency, there is sufficient time for
ions to diffuse into the mesopores and access the full surface
area [32]. At higher potentials, the measured capacitance is
greater as result of several physical phenomena such as
reduction of the solvent layer thickness, increase of the
solvent dielectric constant [33], increase in the electronic state
density in the carbon pore walls [34] and higher ion penetra-
tion in mesopores [35]. The available capacitance will directly
impact the dynamic response of the coupled FCesupercapa-
citor system.
The impedance of the cell also varies with the cell voltage.
Fig. 4(a) shows the Nyquist plot of the impedance and Fig. 4(b)
shows the fitted EIS results according to the equivalent circuit
presented in the inset. The equivalent circuit contains high
frequency elements L (induction), Rs (series resistance) and an
impedance ZP of porous electrode as described by de Levie
equation as shown in Eqn. (8) [36]. The inductive element was
added to give a better fit at high frequencies rather than infer
any physical meaning, as this was a function of the experi-
mental set-up.
ZP ¼ffiffiffiffiffiffiffiffiffiRiZi
pcoth
ffiffiffiffiffiffiffiffiffiffiffiRi=Zi
p(8)
In Eqn. (8), Ri represents ionic resistance inside the pores
and Zi is the interfacial resistance which, in absence of any
Faradaic process, is simply a capacitance of double layer (DL):
Zi ¼ 1/(juCDL). It then follows that the equation can be re-
arranged to give Eqns. (9) and (10)
ZP ¼ Ricoth
ffiffiffiffiffiffiffijus
pffiffiffiffiffiffiffijus
p (9)
s ¼ RiCDL (10)
The total resistance (Rs þ Ri) of a single 1500 F super-
capacitor was therefore found to vary between 0.68 and
0.92 mU as SOC increased, resulting in a pack resistance of
22.4e30.4 mU, neglecting contact resistances between
different cells.
Although DL capacitance does not vary linearly with
voltage, it appears there are regions when the variation can be
approximated as linear: 0.5e1.3 V, 1.3e2.1 V and 2.1e2.7 V.
The capacitanceevoltage gradients within these regions were
134, 42 and 360 F/V respectively. Therefore, moving the
Fig. 6 e Nyquist plot of complex capacitance for a 1500 F
Maxwell supercapacitor for different potentials at 20 �C.
Fig. 7 e Wiring diagram for powertrain components in the
9.5 kWe PEMFCesupercapacitor passive hybrid system.
i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 67890
average operating point between 1.3 and 2.1 V will have little
impact on the available capacitance, but varying it between
2.1 and 2.7 V will have more significant improvements in
performance, as a larger capacitance will enable slower load
dynamics for the FC. As the available low frequency capaci-
tance drops off further below 1.3 V, the load buffering capa-
bilities of the supercapacitors decreased. System designers
should therefore aim to size their systems such that the
lowest operating voltage of supercapacitor pack will be below
that at which mass transport limitations occurs in the FC
stack because highly transient loads on the FC during mass
transport limited operation can lead to accelerated
degradation.
The frequency dependency of the total complex capaci-
tance was also analysed by considering the bode plot of its
imaginary part as shown in Fig. 5. The peak value in the
imaginary capacitance occurs at a frequency known as the
knee frequency which represents the transition point above
which the available capacitance becomes frequency inde-
pendent. This frequency also represents the 50% point of the
charge storage efficiency [37]. The inset in Fig. 5 shows the
voltage dependency of the knee frequency. Due to the discrete
measurement frequency of the EIS measurements, fitting of
the curve allowed for a more accurate estimate of the knee
frequency. The exact value for the knee frequency was esti-
mated by fitting the natural log of the frequency-imaginary
capacitance data to a normal distribution.
As the potential of the cell decreased, the knee frequency
increased. This was likely to be caused by a combination of
processes mentioned previously. It also became apparent that
above the knee frequency, the imaginary capacitances
become independent of cell voltage. Again, the measured
knee-frequency dependency is non-linear over the whole
range of operation, however it can be considered as linear
within the operating ranges of 0.5e1.3 V, 1.3e2 V and 2e2.7 V
with knee frequencyevoltage gradients of �29.5, �13.4 and
�57.6 mHz/V respectively. Important aspects to consider, for
system designers, are that the load buffering capabilities of
the supercapacitors are reduced if the applied load has fre-
quency components oscillating above the knee frequency. It is
therefore advised that, for highly dynamic load cycles, the
knee frequency should be the upper limit of load oscillation.
Little benefit is achieved in using a FCesupercapacitor passive
hybrid system for loads cycles with frequency components
above the knee frequency.
As both the real and imaginary components of the capac-
itancewere available, it is also possible to present the complex
capacitance data in the form of a Nyquist plot in a similar
fashion to Nyquist plots of impedance. This is shown in Fig. 6
for different cell voltages. The same trend of increasing
capacitance with voltage can be concluded; however, the
relationship between the complex variations of capacitance is
now more evident, as is the low frequency capacitance which
is the point at which the semi-depressed circle would inter-
cept the real axis at an infinitely small frequency. This
potentially offers an alternative means of estimating the low
frequency capacitance of a supercapacitor without having to
measure to the extremely low frequencies.
Combined FCesupercapacitor passive hybridsystem
The wiring diagram for the combined FCesupercapacitor
passive hybrid system is shown in Fig. 7. Contactors between
the FC, supercapacitor and load provided electrical isolation
for each component. A 310 A fuse, in-line with the FC and
supercapacitors, provided additional passive protection in the
case of large in-rush currents. A diode was placed in front of
the FC to prevent possible charging of the FC stack from the
supercapacitors in case the thermodynamic OCP of the stack
dropped below that of the supercapacitors.
All BOP components were powered by a 24 VDC supply via
a low voltage bus. During start-up/shut-down events, the BOP
was powered by two 12 V lead acid batteries connected in
series. During normal operation, two parallel DC/DC con-
verters delivered power the low voltage bus.
Prior to connecting the FC and supercapacitors to the main
power bus, the supercapacitors were pre-charged to just
below the FC OCP via an external charging circuit to prevent
large in-rush current from the FC to the supercapacitors when
the two devices were initially connected.
Table 1 e Component list for fuel cellesupercapacitorhybrid system.
Product Manufacturer Product number/Name
DC/DC converters RS components 491-257
Sealed lead acid
battery
RS components 727-0391
Lead acid battery
charger
Ideal Power AC 0724A
Blower Domel 497.3.265
Water pump Jabsco 50870 Series
Fan Comex Europe Axial Fan 24V
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 6 7891
A resistor was wired in parallel to the FC and connected
through a contactor. This resistor was used during the shut-
down procedure under autonomous operation. Upon shut-
down, the hydrogen supply was closed however residual
amounts of reagents in the GDLs and flow channels gave rise
to an elevated OCP until the reagents were consumed. As
elevated potentials were detrimental for carbon corrosion, the
resistor was used to apply a load to the FC to facilitate the
consumption of residual hydrogen.
A Computer Aided Design (CAD) drawing of the FCesu-
percapacitor rig is shown in Fig. 8(a) with a photograph of the
rig shown in Fig. 8(b). Table 1 lists the components used.
Fig. 8 e (a) CAD drawing of fuel cellesupercapacitor passive
hybrid test system and (b) photograph of completed
system.
Recirculation pump Thomas 7015Z DC
Contactor 500 A TE Connectivity LEV200A5ANA
Contactor 20 A Finder 22.23.9.024.4000
Diode 190 A IXYS MMD172-08N1
Diode 120 A ON semiconductor STMSTPS24045TVG
80 A battery fuse Pudenz 153.5631.5801
Resistor 100 U
300 W
Farnell 1768254
Fuel cell Nedstack P9.5-75
Supercapacitors Maxwell BCAP 1500
Passive hybrid system test
The concept of the FCesupercapacitor passive hybrid system
is that the supercapacitors act as a low-pass filter to high
frequency oscillating loads which are detrimental to FC effi-
ciency and durability. EIS measurements indicated that the
total polarisation resistance of the 33 cell supercapacitor pack
was 22.4e30.4 mU. FC stack resistance was shown to range
from 1000 to 100 mU depending on the load current. Since the
supercapacitor resistance is an order of magnitude smaller
than the FC, it will initially handle the majority of the load.
Fig. 9 shows the dynamic response of the FCesupercapa-
citor passive hybrid to a 100 A pulse load from no-load con-
ditions. Upon application of the step load, the supercapacitors
initially handled all the load, as the thermodynamic OCP of
both devices was the same under steady state conditions.
Gradually, as the OCP of the supercapacitor dropped due to the
discharge, the operating potential also drops, meaning that
the FC started to handle a higher proportion of the load. When
Fig. 9 e Response of the FCesupercapacitor passive hybrid
system to a 100 A pulse load from no-load conditions.
i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 67892
the external load was removed, the FC charged the super-
capacitors as the OCP of the supercapacitors was lower than
that of the FC.
Smoothing of the transient load, imposed on the FC,
effectively translates to reduced losses and increased lifetime.
Fig. 10 compares the dynamic response of the pure FC and
FCesupercapacitor passive hybrid system to step loads of 75 A
from cold start conditions. In pure FC mode, the step loads
lead to an observed voltage drop followed by recovery. This
was due to the finite time required to overcome the blower
impeller inertia as well as gas manifolding effects. The rate of
electrochemical reaction was orders of magnitude faster than
the blower inertia and manifolding dynamics. As a result,
there was a brief period, in the order of seconds, when the
local operating stoichiometry of air was lower than the target
value of 2, causing additional mass transport losses and
reduced operating potentials. Under no-load conditions, the
stack voltage returned to OCP. The gradual recovery was due
to a combination of double layer charging and hydrogen
manifolding effects.
When the same load was applied to the passive hybrid
system, the supercapacitor pack met the peak load before the
Fig. 10 e Comparison between (a) pure FC and (b)
FCesupercapacitor passive hybrid under 75 A step loads
with a frequency of 0.2 Hz. Positive currents represent
discharge of the device.
FC. Having reached a dynamic equilibrium, the supercapacitor
operated around an average current of zero as it removed the
peaks from the load cycle. Consequently, the FC met the
average loadwith sufficient time for the air blower to adjust to
the transient loads. The result was therefore a removal of the
voltage drop caused by the blower inertia and gas
manifolding.
Comparison between the efficiency of the pure FC and
FCesupercapacitor hybrid under a 0.2 Hz square wave load of
varying amplitude can be seen in Fig. 11. The efficiency of the
FC was taken as the energy per mol of hydrogen consumed
against the enthalpy of formation assuming the Higher
Heating Value (HHV) for thewater generation reaction. In both
cases, the efficiency of the FC decreased with increasing pulse
load amplitudes due to the increased losses associated with
higher currents. Comparison between the pure FC and FCeSC
passive hybrid system showed that a 5% efficiency gain was
achieved. This was mainly as a result of operation at a lower
average current, and also reduced transient mass transport
losses associated with blower inertia and manifolding effects.
Pulse loading conditions, however, do not sufficiently
represent real load conditions. Therefore, the system was run
under a 1/10th scale (with respect to power) Highway Fuel
Economy Test (HWFET) drive cycle which is representative of
highway driving [38,39]. A power profile, assuming a vehicle
with a 2000 kg mass, accounting for inertia, aerodynamic and
rolling resistance losses was generated based on the govern-
ing equations given by Ehsani et al. [40]. A constant motor
efficiency of 95% was assumed. Other vehicle parameters
were taken to be the same as the values quoted by Baptista
et al. [41], which represents typical parameters for a London
taxi. Fig. 12 shows the response of the FCesupercapacitor
system to the HWFET drive cycle. Before the cycle started, the
systemwas allowed to equilibrate to 69 Vwhich represented a
charging current of less than 1 A from the FC to the super-
capacitors. After the load cycle, the system was again equili-
brated to the same point to allow for direct comparison
between the FCesupercapacitor results and the pure FCmode.
Fig. 11 e Comparison of fuel cell efficiency assuming HHV
between pure FC and FCeSC passive hybrid system with
the fuel cell operating with an air stoichiometry of 2 and
neglecting BOP power losses.
Fig. 12 e Fuel cellesupercapacitor passive hybrid power
distribution between devices with inset zoomed view.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 6 7893
It can be seen that the FC followed the load profile, however, it
did not experience most of the higher frequency loads which
were buffered by the supercapacitors.
A breakdown of the operating current distribution for the
FC and supercapacitor pack under the HWFET drive cycle can
be seen in Fig. 13(a). It becomes apparent that in the FCesu-
percapacitor hybrid configuration, the peak FC load was
reduced from 76 A to 62 A, resulting in an 18.4% decrease.
There was also a 2% decrease in the average load current of
the FC from 28.5 A for the applied load to 27.9 A. The super-
capacitors operated at an average load of 0 A with a normal
distribution and standard deviation of 11.2 A.
The reduced operation time at no/low load conditions for
the FC, resulted in a lower average operating cell potential for
the FCeSC hybrid system as shown in Fig. 13(b). This therefore
gave an indication of the possible improved durability of the
system from a FC side (near-OCP working potential of FC
corresponds to a higher carbon corrosion rate [11]).
The apparent advantage of this system was that it
addressed two of the main PEMFC degradation modes asso-
ciated with a vehicle operation: rapid power cycling, which
leads to catalyst dissolution, and no-load idling, which in-
duces accelerated rate of carbon corrosion. From an efficiency
point of view, the hybridisation reduced both the transient
voltage drops when operating in pure FC mode, upon the
application of large step loads, and BOP related losses.
Fig. 13 e Histograms of (a) currents and (b) single FCpotentials for a 1/10th scale HWFET drive cycle for the
FCesupercapacitor passive hybrid system.
Estimating buffering effect of the supercapacitorsQuantitatively, the buffering effect of the supercapacitor pack
can be estimated by considering the fundamental equation for
a capacitor as shown in Eqn. (11), where ISC, VSC and C repre-
sents the supercapacitor current, voltage and capacitance
respectively.
ISC ¼ CdVSC
dt(11)
In a passive system, if interconnection losses are neglec-
ted, the voltage of the supercapacitor pack and that of the FC
stack (VFC) are equal i.e. VSC ¼ VFC (see Fig. 10(b)). Applying
Kirchoff’s current law allowed for the expression of the
supercapacitor current in terms of the FC (IFC) and load (ILoad)
currents (ISC ¼ ILoad � IFC) as shown in Eqn. (12).
ILoad � IFC ¼ CdVFC
dt(12)
Assuming that the FC is operated in the ohmic dominated
region, and the FC resistance is largely invariant with current
i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 67894
over a certain range (for example, for currents between 40 and
80 A at 30 �C the resistance of the FC is 0.29 U � 0.03 U), it can
be assumed that the FC voltage is related to the FC current as:
VFC ¼ VOCP � IFCROhmic � IFCRCTzVOCP � IFCRFC (13)
where RFC ¼ RCT þ ROhmic.
Here, VOCP represents OCP of FC, RCT is the charge transfer
resistance of the FC, ROhmic is the ohmic resistance of the FC
and RFC is the combined ohmic and charge transfer resistance
of the FC. Assuming that the OCP of the FC and ohmic resis-
tance are time invariant, the current is sufficiently large such
that the charge transfer polarisation losses do not vary
significantly with current, and can therefore be combined
with RFC, and the capacitance is constant at the potential
defined by the average load current and FC polarisation curve,
results in Eqn. (14).
Iload � IFC ¼ �CRFCdIFCdt
(14)
When compared to experimental data, this simple
approximation showed good agreement as demonstrated by
Fig. 14(a). Note that the first pulse and settling period exhibit
Fig. 14 e (a) Comparison between experimental and
simulated data for the FCesupercapacitor passive hybrid
system under a 75 A 0.2 Hz step load and (b) ratio of
supercapacitor to fuel cell resistance under cold/dry start
and nominal operating conditions.
the highest deviation from simulated data due to the larger
non-linear variation of charge transfer resistance at low cur-
rent densities. The validity of assuming a constant charge
transfer resistance when assessing the power split in a pas-
sive hybrid system can be highlighted by Fig. 14(b), which
shows the ratio of supercapacitor to FC resistance. It becomes
apparent that, over the whole operating range of the FC, the
supercapacitor impedance is consistently lower. For a passive
hybrid system, this dictates that variations in the FC resis-
tance will not significantly affect the power split between the
devices due to the dominant effect of the lower supercapacitor
resistance. This therefore allows for the assumption of con-
stant FC resistance in the simplified approximation presented.
It was therefore possible to compare the impact of the
passive load filtering offered for different numbers of super-
capacitors in the pack. Fig. 15 shows the rate of change of the
FC current plotted against the applied system load assuming
the FC current starts at zero. A lower rate of change was
preferable in order to reduce the level of dynamic loading on
the FC. It can be seen that at relatively low currents, the
FCesupercapacitor hybrid offers a good level of load buffering
and is largely independent of the number of supercapacitors.
However, as the load current increases, it becomes apparent
that having more cells in series is detrimental to the load
buffering.
The reason for this is evident by considering the voltage
and current dependencies between the FC and super-
capacitors as shown in Eqn. (15). Neglecting losses, the FC
voltage and supercapacitor voltage are equal. The rate of
change of supercapacitor voltage is the same as the rate of
change of the FC voltage. The resulting change in FC voltage
would therefore result in a proportional change in the rate of
change of FC current:
dVSC
dt¼ dVFC
dtfdIFCdt
(15)
By considering Fig. 16, it can be seen that if the same cur-
rent flows through the supercapacitors, increasing the
Fig. 15 e Comparison of different step loads and
supercapacitor pack size with the response time of the FC
current assuming the same 1500 F supercapacitors used in
this study.
Fig. 16 e Diagram showing how increasing the number of
supercapacitors in series increases the rate of change of
voltage for the supercapacitor pack. The faster the rate of
change of voltage for the supercapacitor pack results in
faster rate of change of load experienced by the FC.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 6 7895
number of supercapacitors in series will increase the rate of
change of voltage of the entire pack. This translates to an
increased rate of change of FC voltage and FC current. Hence,
the maximum buffering effect of the FCesupercapacitor sys-
tem was achieved when the number of supercapacitors in
seriesmatched themaximumOCP of the FC stack tominimise
the number of cells used. This reduced the rate of change of
voltage of the supercapacitor pack for a given current.
Conclusion
Rapid load cycling has been identified as being one of themain
causes of FC inefficiencies as a consequence of compressor
inertia and gas manifolding effects. Passive coupling of a FC
with a supercapacitor pack allowed for the reduction of dy-
namic loading experienced by the FC.
EIS measurements of a 1500 F Maxwell supercapacitor
indicated a proportional relationship between the DL capaci-
tance and cell voltage with 3 pseudo-linear regions. Operation
at potentials lower than 1.3 V resulted in the greatest decrease
in DL capacitance and therefore offered reduced buffering.
The knee frequency was identified as being inversely pro-
portional to cell voltage and ranged from 0.19 to 0.26 Hz for the
given supercapacitor cell. Load cycles with frequency com-
ponents greater than the knee frequency are not suitable for
the FCesupercapacitor passive hybrid system.
Efficiency gains achieved through passive hybridisation
under step loads have been demonstrated to be approximately
5% with respect to fuel utilisation. This was mainly attributed
to reduced transient mass transport losses and a shift in the
average operating current to a more efficient region. Under a
1/10th scaled HWFET drive cycle, it was shown that the
operating current of the supercapacitor has a normal distri-
bution with an average around zero indicating that it is purely
buffering loads. The FC load showed a 18.4% decrease in peak
loads with the average 2% lower. A shift in the average oper-
ating potential suggested lower rates of carbon catalyst sup-
ports will be afforded in the FCesupercapacitor passive hybrid
system.
Analysis showed that in order to gain themost benefit from
the passive system, the number of supercapacitors used
should be kept to a minimumwhilst ensuring the FC does not
overcharge the supercapacitors under no-load conditions.
Derivation of the system equations shows that the FC buff-
ering is determined by the product of the supercapacitor pack
capacitance and FC resistance.
Acknowledgements
The authors would like to thank Richard Silversides for
assistance with advice on the electric systems and previous
fuel cell development team members: Mardit Matian, Ralph
Clague, Mark Cordner, Sam Tippets, Ed Spofforth-Jones, Till
Hanten, Alana Johnson, Laura Harito, Charles Banner-Martin,
Akash Agrawal, Matthew Wong, Dan-Fung Chan, Tanya
Chong, Omar Al Fakir, Nicolas Lee, Robert Bilinski, Nicolas
Higginson, Rebecca Nelson, Michael Squire, Ashwin Suguna-
Balan, Ryan Williams, Jignesh Patel, Olivia Tillbert, Adya Jha,
Xin Miao, Nasrin Shahed-Khah and Sze Li.
The authors would also like to acknowledge the Engi-
neering and Physical Sciences Research Council for funding of
this work, through both a Career Acceleration Fellowship for
Gregory Offer, award number EP/I00422X/1. As well as the in-
kind contributions from: Johnson Matthey, Nedstack, BOC,
Domel, National Instruments, Swagelok and RS.
r e f e r e n c e s
[1] Pollet BG, Staffell I, Shang JL. Current status of hybrid, batteryand fuel cell electric vehicles: from electrochemistry tomarket prospects. Electrochim Acta 2012;84:235e49.
[2] US Drive. Fuel cell technical team roadmap. p. 2. http://www1.eere.energy.gov/vehiclesandfuels/pdfs/program/fctt_roadmap_june2013.pdf; 2013 [last date accessed 12.02.14].
[3] Shimoi R, Aoyama T, Iiyama A. Development of fuel cellstack durability based on actual vehicle test data: currentstatus and future work. SAE Int J; 2009:4970.
[4] Dhanushkodi SR, Tam M, Kundu S, Fowler MW, Pritzker MD.Carbon corrosion fingerprint development and de-convolution of performance loss according to degradationmechanism in PEM fuel cells. J Power Sources2013;240:114e21.
[5] Maass S, Finsterwalder F, Frank G, Hartmann R, Merten C.Carbon support oxidation in PEM fuel cell cathodes. J PowerSources 2008;176:444e51.
[6] Avasarala B, Moore R, Haldar P. Surface oxidation of carbonsupports due to potential cycling under PEM fuel cellconditions. Electrochim Acta 2010;55:4765e71.
[7] Wu J, Yuan XZ, Martin JJ, Wang H, Zhang J, Shen J, et al. Areview of PEM fuel cell durability: degradation mechanismsand mitigation strategies. J Power Sources 2008;184:104e19.
[8] Watanabe M. Activity and stability of ordered and disorderedCoePt alloys for phosphoric acid fuel cells. J Electrochem Soc1994;141:2659.
[9] Akita T, Taniguchi A, Maekawa J, Siroma Z, Tanaka K,Kohyama M, et al. Analytical TEM study of Pt particledeposition in the proton-exchange membrane of amembrane-electrode-assembly. J Power Sources2006;159:461e7.
[10] Zhai Y, Zhang H, Xing D, Shao Z-G. The stability of Pt/Ccatalyst in H3PO4/PBI PEMFC during high temperature lifetest. J Power Sources 2007;164:126e33.
i n t e rn a t i o n a l j o u r n a l o f h y d r o g e n en e r g y 3 9 ( 2 0 1 4 ) 7 8 8 5e7 8 9 67896
[11] Mathias MF, Makharia R, Gasteiger HA, Conley JJ, Fuller TJ,Gittleman CJ, et al. Two fuel cell cars in every garage.Electrochem Soc Interface 2005;14:24e35.
[12] Borup RL, Davey JR, Garzon FH, Wood DL, Inbody MA. PEMfuel cell electrocatalyst durability measurements. J PowerSources 2006;163:76e81.
[13] Maranzana G, Moyne C, Dillet J, Didierjean S, Lottin O. Aboutinternal currents during start-up in proton exchangemembrane fuel cell. J Power Sources 2010;195:5990e5.
[14] Yufit V, Brandon NP. Development and application of anactively controlled hybrid proton exchange membrane fuelcelldlithium-ion battery laboratory test-bed based on off-the-shelf components. J Power Sources 2011;196:801e7.
[15] Hosseinzadeh E, Rokni M, Advani SG, Prasad AK.Performance simulation and analysis of a fuel cell/batteryhybrid forklift truck. Int J Hydrogen Energy 2013;38:4241e9.
[16] Garcıa P, Torreglosa JP, Fernandez LM, Jurado F. Viabilitystudy of a FC-battery-SC tramway controlled by equivalentconsumption minimization strategy. Int J Hydrogen Energy2012;37:9368e82.
[17] Hwang JJ, ChangWR. Characteristic study on fuel cell/batteryhybrid power system on a light electric vehicle. J PowerSources 2012;207:111e9.
[18] Hwang JJ. Transient power characteristic measurement of aproton exchange membrane fuel cell generator. Int JHydrogen Energy 2013;38:3727e40.
[19] Kim M-J, Peng H. Power management and designoptimization of fuel cell/battery hybrid vehicles. J PowerSources 2007;165:819e32.
[20] Garrigos A, Blanes JM, Carrasco JA, Lizan JL, Beneito R,Molina JA. 5 kW DC/DC converter for hydrogen generationfrom photovoltaic sources. Int J Hydrogen Energy2010;35:6123e30.
[21] Van Vliet OPR, Kruithof T, Turkenburg WC, Faaij APC.Techno-economic comparison of series hybrid, plug-inhybrid, fuel cell and regular cars. J Power Sources2010;195:6570e85.
[22] Contestabile M, Offer GJ, Slade R, Jaeger F, Thoennes M.Battery electric vehicles, hydrogen fuel cells and biofuels.Which will be the winner? Energy Environ Sci 2011;4:3754.
[23] Kuperman A, Aharon I, Kara A, Malki S. A frequency domainapproach to analyzing passive batteryeultracapacitorhybrids supplying periodic pulsed current loads. EnergyConvers Manag 2011;52:3433e8.
[24] Bernard J, Hofer M, Hannesen U, Toth A, Tsukada A,Buchi FN, et al. Fuel cell/battery passive hybrid power sourcefor electric powertrains. J Power Sources 2011;196:5867e72.
[25] Nishizawa A, Kallo J, Garrot O, Weiss-Ungethum J. Fuel celland Li-ion battery direct hybridization system for aircraftapplications. J Power Sources 2013;222:294e300.
[26] Keranen TM, Karimaki H, Viitakangas J, Vallet J, Ihonen J,Hyotyla P, et al. Development of integrated fuel cell hybrid
power source for electric forklift. J Power Sources2011;196:9058e68.
[27] Wu B, Matian M, Offer GJ. Hydrogen PEMFC system forautomotive applications. Int J Low-carbon Technol2012;7:28e37.
[28] Cordner M, Matian M, Offer GJ, Hanten T, Spofforth-Jones E,Tippetts S, et al. Designing, building, testing and racing alow-cost fuel cell range extender for a motorsportapplication. J Power Sources 2010;195:7838e48.
[29] Kandlikar SG, Lu Z. Thermal management issues in a PEMFCstack e a brief review of current status. Appl Therm Eng2009;29:1276e80.
[30] Kreitmeier S, Michiardi M, Wokaun A, Buchi FN. Factorsdetermining the gas crossover through pinholes in polymerelectrolyte fuel cell membranes. Electrochim Acta2012;80:240e7.
[31] Taberna PL, Simon P, Fauvarque JF. Electrochemicalcharacteristics and impedance spectroscopy studies ofcarbonecarbon supercapacitors. J Electrochem Soc2003;150:A292.
[32] Kaus M, Kowal J, Sauer DU. Modelling the effects of chargeredistribution during self-discharge of supercapacitors.Electrochim Acta 2010;55:7516e23.
[33] Gualous H, Gallay R. Supercapacitor module sizing and heatmanagement under electric, thermal, and agingconstraintsIn Supercapacitors. Wiley-VCH Verlag GmbH &Co. KGaA; 2013. pp. 373e436.
[34] Hahn M, Barbieri O, Campana FP, Kotz R, Gallay R. Carbonbased double layer capacitors with aprotic electrolytesolutions: the possible role of intercalation/insertionprocesses. Appl Phys A 2005;82:633e8.
[35] Salitra G, Soffer A, Eliad L, Cohen Y, Aurbach D. Carbonelectrodes for double-layer capacitors I. Relations betweenion and pore dimensions. J Electrochem Soc 2000;147:2486.
[36] De Levie R. On porous electrodes in electrolyte solution e IV.Electrochem Acta 1964;9.
[37] Taberna P-L, Simon P. Electrochemical techniquesInSupercapacitors. Wiley-VCH Verlag GmbH & Co. KGaA; 2013.pp. 111e30.
[38] United State Environmental Protection Agency. EPAdynamometer drive schedules. URL: http://www.epa.gov/nvfel/testing/dynamometer.htm [last date accessed12.02.14].
[39] Mahlia TMI, Tohno S, Tezuka T. A review on fuel economytest procedure for automobiles: implementation possibilitiesin Malaysia and lessons for other countries. Renew SustainEnergy Rev 2012;16:4029e46.
[40] Ehsani M, Gao Y, Emadi A. Modern electric, hybrid electric,and fuel cell vehicles: fundamentals, theory and design. BocaRaton: CRC Press; 2010.
[41] Baptista P, Ribau J, Bravo J, Silva C, Adcock P, Kells A. Fuel cellhybrid taxi life cycle analysis. Energy Policy 2011;39:4683e91.