Comparison of the series and parallel architectures

for hybrid multi-stack fuel cell - battery systems

Neigel Marx∗† , John Cardozo ∗ , Loı̈c Boulon∗ ‡ , Frédéric Gustin† Daniel Hissel† and Kodjo Agbossou∗
∗ Institut de Recherche sur l’Hydrogène, Université du Québec à Trois-Rivières, Trois-Rivières (QC), Canada.
† FCLAB, FR CNRS 3539, FEMTO-ST/Energy UMR CNRS 6174, University of Franche-Comte, Belfort, France.
‡ Corresponding author, Email: [email protected]

Abstract—Multistack fuel cell (MFC) systems provide a level strategies have been proposed throughout the literature [4],
of redundancy that is not available to single stack systems. [5]. A rule based energy management strategy designed for
Depending on the chosen architecture, this redundancy can be hybrid FC-battery systems has been chosen for this study and
used to enable degraded mode operation, act on ageing and/or
reduce fuel consumption. This paper presents a comparative adapted to MFC systems [6].
study of the series and parallel architectures in the case of an Recent studies have shown the difference between single
automotive application. Both architectures are evaluated on fuel fuel cell systems and MFC systems [7]. However, no studies
consumption and ageing. Three driving cycles representing the have been conducted to evaluate the difference in ageing or
urban, rural and highway driving styles are used to provide fuel consumption between the different MFC architectures.
results for different types of use. The simulation results are
obtained in a MATLAB - Simulink environment. The results This paper presents a comparative study of the series and
highlight the superiority in ageing of the parallel architecture parallel architectures considering an automotive application.
over the series architecture. However the results exhibits similar Both architectures are evaluated on fuel consumption and
behavior for the hydrogen consumption. ageing. Three driving cycles representing urban, rural and
Index Terms—Multistack fuel cell systems, Ageing, Fuel con- highway driving are used to provide results for different types
sumption, hybridization, battery, PEM Fuel Cell, Degraded mode
operation, State of Charge of use.
The first section presents the series and parallel MFC
I. I NTRODUCTION architectures. The second section presents the chosen energy
management strategy. The third section presents the fuel cell
A solution to the environmental crisis created by the indus- model used for the simulations. The last section presents the
try’s dependency on fossil fuels could be found in the potential comparative study and the simulation results.
of hydrogen as an energy vector. It is a clean and renewable
energy source. II. MFC ARCHITECTURES
Several technical and economical barriers prevent the large Multiple MFC architectures exist [8]. The series and parallel
penetration of fuel cells in the market. Although it is more architectures are compared in this study. They are presented
efficient than its’ competitor, the internal combustion engine, in this section. Each system is composed of N fuel cell stack.
it is more expensive, has a lower lifetime and is more often In order to limit fuel cell current transients, the system is
subject to failure. hybridized with a battery pack.
MFC systems could help mitigate these issues as they are
more efficient, more reliable and possess a longer lifetime than A. Series architecture
single stack fuel cell systems [1], [2], [3]. MFC systems are The different stacks of the system are linked in a series
composed of multiple fuel cell stacks. Different architectures configuration (figure 1). A DC-DC converter is used to adapt
can be used to link the fuel cells together. The series architec- the MFC system’s output voltage to the bus voltage.
ture is simple, provides reliability through the use of bypass The series configuration imposes the same current on all the
systems and is the less expensive of the architectures. The fuel cells. Therefore, the series architecture has an energetic
parallel architecture offers individual current control and more behavior that is close to that of the single stack fuel cell
efficient reliability. Other architectures like the series-parallel system. Its’ reliability can easily be increased with the use of
or cascaded architectures exist. They can be used to mitigate by-pass circuits to isolate fuel cell stacks when needed. This
some of the disadvantages of the aforementioned strategies. enables degraded mode operation but can also be a source of
However, it is at the cost of simplicity when compared to the stress during the reconnection of isolated stacks [9].
series architecture. And it is at the cost of modularity when
compared to the parallel architecture. B. Parallel architecture
FC and MFC systems are often coupled to energy storage Each stack possesses a power converter that enables indi-
systems in order to mitigate the damaging effect of fast vidual current control and adapts the voltage output of each
transients and increase its performance. Hybrid energy systems system. Each fuel cell system is connected to the DC bus in
require energy management strategies. Online management a parallel configuration (figure 2).

978-1-4673-7637-2/15/$31.00 ©2015 IEEE

 lowering the depth of discharge limits the battery’s ageing. The
fuel cell system possesses 3 states in this energy management
strategy:
• State 0: the fuel cell system is turned off.
• State 1: the fuel cell system provides power at its maxi-
mum efficiency point.
• State 2: the fuel cell system operates at its maximum
rated power.
The energy management strategy evolves from one state to
another using the following hysteresis behavior (figure 3):

Fig. 1: Series architecture for MFC systems

Fig. 3: Hysteresis behavior for the series architecture

The first hysteresis cycle regulates the transition between

state 0 and state 1. The transition from state 0 to state 1 is set
Fig. 2: Parallel architecture for MFC systems at a SoC of 0.5 and the transition from state 1 to state 0 is set
at a SoC of 0.6. Those values have been chosen arbitrarily in
order to limit the number of start/stop cycles (the higher the
The parallel configuration enables the use of modular
difference between the two values, the lower the number of
systems. In order to adjust the rated power of the system,
start and stop cycles).
modules can be added or subtracted without hindering the
The second hysteresis cycle regulates the transition between
other elements of the system. Degraded mode operation is
state 1 and state 2. The transition from state 1 to state 2 is set
also more efficient as the isolation and reconnection process
at a SoC of 0.425 and the transition from state 1 to state 0 is
can be conducted with minimum stress on the other fuel cell
set at a SoC of 0.45. Those values have been chosen arbitrarily
systems.
in order to limit the transitions between state 1 and 2 and have
III. E NERGY MANAGEMENT STRATEGY the fuel operate as much as possible at its maximum efficiency
point.
The systems presented in section II include an MFC system
and a battery pack. The distribution of the load’s power B. Parallel architecture
between the two sources will be handled by the following The energy management for the parallel architecture is
energy management strategy. The energy management strategy similar to that of the series architecture. Each fuel cell possess
is adapted to both the series and parallel architectures to it’s own hysteresis curve (figure 4).
account for individual control. The transition from state 0 to state 1 depends on the fuel
A. Series architecture cell. They are turned on in sequence in order to minimize the
number of active fuel cells. The transitions from state 1 to
At a given temperature and reactant pressures the fuel state 2 are kept the same as that of the series architecture in
cell system possesses one maximum efficiency point and order to minimize the operating time at the maximum rated
one maximum power point. The energy management strategy power.
introduced in this paper limits the variation of the battery State
of Charge (SoC) by operating the fuel cell at the maximum IV. F UEL CELL SYSTEM MODEL
efficiency point, maximum power point or turning/keeping the The considered fuel cells use the air-breathing technology
fuel cell off. The SoC is kept between 0.4 and 0.6 because to feed oxygen to the stack and cool it. This technology is
 TABLE I: Fuel cell model equations

Electrochemistry:

Vcell,s = En − ΔVact − ΔVconc − ΔVohm

 
ΔH − T ΔS R TF C PH2 ∗ PO2 1/2
En = − + ∗ ln
2F 2F P0 3/2
 
R TF C j + jloss
ΔVact = K1 ln
F j0
j
ΔVconc = K3 TF C ln (1 − )
jL

ΔVohm = Rm IF C
  
PO2 Ec TF C
j0 = K2 exp − 1−
P0 R TF C T0
  
tmb 1 1
Rm = K4 exp −K5 ∗ −
b11 ∗ Λmb − b12 T0 TF C
dVcell,d dI
Vcell,d + τ =K
dt dt

VF C = Ncell (Vcell,s + Vcell,d )

Ncell IF C
qH2 ,r =
2F
Fig. 4: Hysteresis behavior for the parallel architecture
Ncell IF C
qO2 ,r =
4F
mostly used for low power applications [10]. Nevertheless it qint = Ncell (ΔH/2F − Vcell ) IF C
has also been used for automotive applications [11]. In order
to have a complete model of the system, the ancillaries are
Thermal behavior:
also included. The equations of the fuel cell system model are
presented in table I. qconv,f ree = hf ree A (TF C − TAir )

A. Electrochemistry qconv,f orced = h̄ At ηtherm θb,lm

The behavior of the stack voltage is ruled by electrochemical qconv = qconv,f ree + qconv,f orced
phenomena. The presented electrochemical model includes the 
1
ohmic, activation and concentration overvoltages as well as the TF C = (qint − qconv )
CF C
dynamic effect of membrane hydration on the voltage. This
model is based on the assumption that all cells work at the Power converter:
same operation point. 
1
Iconv = (Vbus − Vconv ) dt − Iloss
B. Thermal behavior Lconv
1
The temperature of the stack is affected by the internal Vconv = VF C
1−α
heat generation, the forced convection and the free convection. 1
IF C = Iconv
The temperature provided by this model is the average stack 1−α
temperature.
Blowers:
C. Power converter
dIF AN
LF AN + RF AN IF AN = VF AN
The converters include output voltage adaptation and current dt
control. The voltage adaptation is modelled as that of a boost
E = kΩ
converter. An output inductance is used to control the output
current sent to the load. Tm = k IF AN − Tloss

D. Blower 1
Ω= (TM − Taero ) dt
JF AN
The blowers are composed of a DC motor linked to fan
blades that propel the air. The DC motor model is composed qair = ηaero kaero Ω
 
of an electromotive force, an internal resistance and an internal 1
Taero = Nb ρair Cd eb (Rb 4 − R0 4 )(Ω cos θ)2
inductance. The conversion of the blade’s rotation into air flow 8
and the resistive torque induced by the air on the motor shaft
are dependent on the blade’s geometry.
E. Battery model TABLE II: sizing results
The battery model includes the battery’s internal resistance Scenario Power (W)
through the following equation: Average power on a highway trip 30 kW
Average power on a rural mountain trip 38 kW
Highest dynamic transient 95 kW
Vbat = OCV − Ibat Rbat (1) System sizing Power (W)
The State of Charge is calculated as follows: MFC system 40 kW
Battery 100kW

1
SoC = SoCinit − Ibat dt (2)
Cbat
B. Problem formulation
F. Fuel cell model simulation results
The parameters of the model have been fitted to one of the The MFC system is composed of 4 x 10kW fuel cell systems
fuel cell of the hydrogen research institute’s fuel cell system associated in a series or parallel configuration depending on
(Fuel cell manufacturer: Horizon, model: H-500; DC-DC the chosen architecture. The model presented in section 4 is
converter manufacturer: Zahn Electronics, model: DC15036F- scaled up by a factor 25 to provide the necessary power output.
SU). The steady state efficiency curve obtained with the model The Artemis driving cycles have been chosen to compare
is presented on figure 5. both architectures. The three driving cycles reflect different
uses of automobiles. Figure 6 presents the urban, rural and
highway velocity and power profile used in simulations.
Using the energy management strategy presented in section
III, the fuel consumption and degradations are evaluated for
the three driving cycles. The ageing of fuel cells depends
on multiple phenomena like time of operation condition of
operation, number start/stop cycles and shut-down procedures.
In this study, the degradation of the system is evaluated using
the number of start and stop cycles and the operation time.
C. Fuel consumption and degradation results
The simulation results for each scenarios are summed up
in table III. The power profiles submitted to the batteries and
multistack fuel cell systems are presented on figure 7.
The power provided by the battery pack does not vary
greatly depending on the chosen multistack association due
Fig. 5: Efficiency versus power curve, power unit: 400W
to the system sizing. The battery power rating is 2.5 times
higher than the MFC system’s power rating.
V. C OMPARISON OF THE SERIES AND PARALLEL The equivalent hydrogen consumption takes into account
ARCHITECTURES both the hydrogen consumption and the variation of State of
The series and parallel architectures are compared in this Charge. It is calculated using the following equation:
section. Both fuel consumption and ageing are evaluated on
three different driving cycles. ΔSoC Ebat
Eeq = mH2 + (3)
A. System sizing ηF C,max ηbat LHV
The MFC system and the battery pack are sized according where Ebat is the total energy of the battery, ηF C,max is the
to the following criteria: maximum efficiency of the FC system, ηbat is the efficiency of
• The average power requested by the powertrain on a the battery charge at the power corresponding to the maximum
highway trip (110km/h, 0% slope) is to be supplied by efficiency of the fuel cell and LHV is the low heating value
the MFC system of hydrogen.
• The average power requested by the powertrain on a rural The energetic behavior and the operation time per FC
mountain trip (90km/h, 5% slope) is to be supplied by the system of both architectures are similar for all three driving
MFC system cycles. However, the results presented in table III show that the
• The dynamic transients requested by the powertrain on number of start and stop cycles per FC system is far lower for
any trip is to be supplied by the battery pack. the parallel architecture during urban and rural driving cycles.
The different scenarios are simulated using the ADVISOR Ageing tests indicate that that start/stop cycles play a larger
plug-in for MATLAB. A middle sized car has been chosen part in the degradation of the system than the operation time.
with a mass of 1800 kg. The results of the sizing are presented The degradation induced by a start/stop cycle equals that of
on table II. multiple hours of fuel cell operation and highly depends on
 Fig. 6: Artemis driving cycles

Fig. 7: Power provided by the battery. (from top to bottom: urban, rural and highway cycle)

TABLE III: Simulation results

Series architecture Parallel architecture
Urban Rural Highway Urban Rural Highway
Number of start/stop cycles (-) 4 8 4 1 2 4
Number of start/stop cycles per FC system (-) 1 2 1 0.25 0.5 1
Operation time (s) 156 1348 3156 485 1129 3170
Operation time per FC system (s) 39 337 789 121.25 282.25 792.5
Hydrogen consumption (g) 18.2 162.35 381.75 58.69 136.92 383.75
State of charge variation (%) 7.3 0.02 0.275 1.078 4.38 0
Equivalent hydrogen consumption (g) 67.23 162.48 383.59 65.93 166.33 383.75
the shutdown procedure. Using a simple degradation model [2] L. Palma and P. N. Enjeti, “A Modular Fuel Cell,
(equation (4)), the degradation per cycle can be established. Modular DC–DC Converter Concept for High Performance
and Enhanced Reliability,” IEEE Trans. Power Electron.,
Using the distribution of scenarios in the lifetime of a vehicle vol. 24, no. 6, pp. 1437–1443, Jun. 2009. [Online]. Available:
(69.7% urban, 27.5% rural and 2.8% highway) [12], a global http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=4939358
degradation per cycle can be calculated. [3] E. Frappe, A. De Bernardinis, G. Coquery, O. Bethoux, and
C. Marchand, “Corrective action with power converter for faulty
multiple fuel cells generator used in transportation,” in 2010 IEEE Veh.
Power Propuls. Conf. IEEE, Sep. 2010, pp. 1–6. [Online]. Available:
δtot = δstart/stop + δtime http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=5729037
= Nstart/stop Dstart/stop + ton Doperation (4) [4] M. Zandi, A. Payman, J.-p. Martin, S. Pierfederici,
B. Davat, and F. Meibody-Tabar, “Energy Management
where δtot is the total degradation per cycle, δstart/stop is of a Fuel Cell/Supercapacitor/Battery Power Source for
the part of degradation due to the start/ cycles, δtime is the Electric Vehicular Applications,” IEEE Trans. Veh. Technol.,
vol. 60, no. 2, pp. 433–443, Feb. 2011. [Online]. Available:
part of degradation due to the operation time, Nstart/stop is http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=5625923
the number of start and stop cycle per FC system, Dstart/stop [5] B. Geng, J. K. Mills, and D. Sun, “Two-Stage Energy
is the degradation per start and stop cycle, ton is the operation Management Control of Fuel Cell Plug-In Hybrid Electric
Vehicles Considering Fuel Cell Longevity,” IEEE Trans. Veh.
time per FC system and Doperation is the degradation per hour Technol., vol. 61, no. 2, pp. 498–508, Feb. 2012. [Online]. Available:
of operation. Considering the following degradation values http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6087294
based on a review of ageing tests and the results presented in [6] K. Ettihir, L. Boulon, K. Agbossou, S. Kelouwani, and
M. Hammoudi, “Design of an energy management strategy
table III, a comparison of the degradation for both architectures for PEM Fuel Cell Vehicles,” 2012 IEEE Int. Symp.
is presented on table IV: Ind. Electron., pp. 1714–1719, May 2012. [Online]. Available:
http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6237349
• Dstart/stop = 25 μV per start/stop cycle [13]
[7] J. E. Garcia, D. F. Herrera, L. Boulon, P. Sicard, and
• Doperation = 10 μV per hour of operation [14] A. Hernandez, “Power sharing for efficiency optimisation into
a multi fuel cell system,” in 2014 IEEE 23rd Int. Symp. Ind.
TABLE IV: Degradation per cycles Electron. Istanbul: IEEE, Jun. 2014, pp. 218–223. [Online]. Available:
http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=6864614
Series architecture Urban Rural Highway [8] N. Marx, L. Boulon, F. Gustin, D. Hissel, and K. Agbossou, “A review
of multi-stack and modular fuel cell systems: Interests, application
Degradation per cycle (10−2 %) 2.51 5.10 2.72 areas and on-going research activities,” Int. J. Hydrogen Energy,
Average degradation per cycle (10−2 %) 3.23 vol. 39, no. 23, pp. 12 101–12 111, Aug. 2014. [Online]. Available:
http://linkinghub.elsevier.com/retrieve/pii/S0360319914016036
Parallel architecture Urban Rural Highway [9] D. Candusso, A. De Bernardinis, M.-C. Péra, F. Harel, X. François,
Degradation per cycle (10−2 %) 0.66 1.33 2.72 D. Hissel, G. Coquery, and J.-M. Kauffmann, “Fuel cell operation
under degraded working modes and study of diode by-pass circuit
Average degradation per cycle (10−2 %) 0.90 dedicated to multi-stack association,” Energy Convers. Manag.,
vol. 49, no. 4, pp. 880–895, Apr. 2008. [Online]. Available:
These results present a 72 % decrease in average degrada- http://linkinghub.elsevier.com/retrieve/pii/S0196890407003421
[10] J. Fernández-Moreno, G. Guelbenzu, a.J. Martı́n, M. Folgado,
tion for the parallel architecture when compared to the series
P. Ferreira-Aparicio, and a.M. Chaparro, “A portable system powered
architecture. This results in a lifetime that is more than three with hydrogen and one single air-breathing PEM fuel cell,” Appl.
times as long for the parallel architecture than it is for the Energy, vol. 109, pp. 60–66, Sep. 2013. [Online]. Available:
http://linkinghub.elsevier.com/retrieve/pii/S0306261913002729
series architecture.
[11] Y. Tang, W. Yuan, M. Pan, and Z. Wan, “Experimental investigation
VI. C ONCLUSION on the dynamic performance of a hybrid PEM fuel cell/battery
system for lightweight electric vehicle application,” Appl. Energy,
The series and parallel architecture for MFC systems have vol. 88, no. 1, pp. 68–76, Jan. 2011. [Online]. Available:
been compared on three different automotive driving cycles. http://linkinghub.elsevier.com/retrieve/pii/S0306261910003028
[12] M. André, “The ARTEMIS European driving cycles for
A rule-based energy management has been designed to mini- measuring car pollutant emissions.” Sci. Total Environ.,
mize fuel consumption and degradation. The two architectures vol. 334-335, pp. 73–84, Dec. 2004. [Online]. Available:
provide similar results for the highway scenario. However the http://www.ncbi.nlm.nih.gov/pubmed/15504494
[13] J. Wu, X. Z. Yuan, J. J. Martin, H. Wang, J. Zhang, J. Shen,
parallel architecture is subject to less degradation during rural S. Wu, and W. Merida, “A review of PEM fuel cell durability:
and urban scenarios. The average degradation per cycle shows Degradation mechanisms and mitigation strategies,” J. Power Sources,
a clear advantage for the parallel architecture with a 72 % vol. 184, no. 1, pp. 104–119, Sep. 2008. [Online]. Available:
http://linkinghub.elsevier.com/retrieve/pii/S0378775308011968
average decrease in degradation compared to that of the series [14] Y. Yu, G. Wang, Z. Tu, Z. Zhan, and M. Pan, “Effect of gas
architecture. However, not all degradation phenomena have shutoff sequences on the degradation of proton exchange membrane
been taken into account and the energy management is not fuel cells with dummy load during startup and shutdown cycles,”
Electrochim. Acta, vol. 71, pp. 181–193, Jun. 2012. [Online]. Available:
optimized for the different scenarios. Therefore, further studies http://linkinghub.elsevier.com/retrieve/pii/S0013468612004938
from the authors will focus on those key aspects.
R EFERENCES
[1] D. C. Toquica Cardenas, N. Marx, L. Boulon, F. Gustin,
and D. Hissel, “Degraded Mode Operation of Multi-Stack
Fuel Cell Systems,” in 2014 IEEE Veh. Power Propuls.
Conf. IEEE, Oct. 2014, pp. 1–6. [Online]. Available:
http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7007041