127

1

Abstract The performance of the fuel cells (FCs) are today
very appealing for industrial, transportation and home power
generation applications and their use is widely increased during
recent years. Unfortunately, in dynamic conditions the FCs
behavior can present some lacks of power, due to the transitions
between two steady-state current levels. To better design these
kind of applications, a FC model is required, to evaluate the
performance at different working conditions. The synthesis of a
dynamic model describing the FC behavior should be performed
starting form the real system configuration, including pipes, flow
and pressure regulators and valves. To this aim an automatic
measurement system for the characterization of PEM (Polymer
Electrolyte Membrane) fuel cells has been developed, and in this
paper a dynamic model of a fuel cells stack is proposed, which
has been validated through experimental measures.

I ndex Terms Fuel cells, dynamic response, power
measurements, system modeling.

I. INTRODUCTION
UEL cell is an electrochemical device that converts
chemical energy directly into electrical energy. With
respect to internal combustion engines, fuel cell has higher
energy storage capability, thus enhancing the range of
operation for automotive applications and is a cleaner source
of energy. Fuel cell also has the further advantage of using
hydrogen as fuel that could reduce worlds dependence on
nonrenewable hydrocarbon sources.
In recent years different types of technologies have been
developed [1,2], such as the: Alkaline Fuel Cell (AFC);
Proton Exchange Membrane (PEM) Fuel Cell; Phosphoric
Acid Fuel Cell (PAFC); Molten Carbonate Fuel Cell (MCFC);
Solid Oxide Fuel Cell (SOFC) and Direct Methanol Fuel Cell
(DMFC).
One of the most diffused, the PEM fuel cell, has a high proton
conductivity membrane as electrolyte. The PEM uses a thin
layer of solid organic polymer (the most commonly used is
Nafion) as electrolyte. This ion-conductive membrane is

Edoardo Fiorucci, Fabrizio Ciancetta and Giovanni Bucci are with the
Department of Electrical and Information Engineering, University of
LAquila, AQ 67040 ITALY (phone: +39.0862.434634; fax:
+39.0862.434403; e-mail: {ciancetta, fiorucci}@ing.univaq.it),
bucci@ieee.org.
Francesco Vegli is with the Department of Chemistry, Chemical
Engineering and Materials, University of LAquila, AQ 67040 ITALY
(veglio@ing.univaq.it).
coated on both sides with highly dispersed metal alloy
particles (mostly platinum) that constitute the active catalyst.
The PEM fuel cell basically requires hydrogen and oxygen as
reactants, though the oxidant may also be ambient air, and
these gasses must be humidified to prevent membrane
dehydration [3]. Each single cell produces about 0.6 V and
can be combined in a fuel cell stack to obtain the required
electrical voltage and power. The operating temperature is in
the range of 70-100 C. One of the main weak points of fuel
cell is its slow dynamics. In fact, the fuel cell dynamic is
limited by different phenomena, as the resistance variation of
the membrane due to the temperature, or the hydrogen
delivery system itself, which can introduce delays due to the
pumps, the valves and in some cases to the reforming process
[4].
II. THE MEASUREMENT SYSTEM
Many dynamic FC models are present in literature; they
describe FC function, flow gasses, reforming, etc. These
models need chemical-physic-electrical parameters that are
not usually provided by factories, in order to preserve design
patent. So, the reliability of FCs dynamic simulations depends
not only on the model accuracy but also on FC parameter
findings. To solve this problem, we propose a new approach
to define the dynamic FC model, using measurements
obtained by the FC measurement system shown in Fig.1,
developed for one PEM fuel cell stack. Its main features are:
10 cells stack, electrode area of 64 cm
2
, nominal power of
150 W, nominal voltage of 6 V, nominal current of 25 A,
reactants H
2
/air, reformate/air, H
2
/O
2
; max operating
temperature of 70 C; operating air pressure of 0-34.4 kPa;
operating H
2
pressure of 0-34.4 kPa; self-humidified stack.
In the proposed system the cell is supplied with two gases
(H
2
and N
2
) and filtered air through three digital mass flow
rate controllers (Bronkhorst DMFC Mod. F-201C-RAB-22V).
In a gas mixer N
2
can be mixed with H
2
for testing the
response to different stoichiometric values of reactants. The
electronic back-pressure controller regulates the pressure of
filtered air and H
2
-N
2
mixture. A distilled water pump is used
for the stack cooling; it requires about a 4 l/min water flow
rate, for a supplied current of 25 A at 45 C. All the described
controllers are connected to a PC using RS-485 links and a
RS-485 to RS-232 adapter.
The temperature of the stack and the reaction products are
measured by J-type thermocouples, connected to a Fluke
An Experimental Approach to the Modeling of
PEM Fuel Cells in Dynamic Conditions
Giovanni Bucci, Member, IEEE, Fabrizio Ciancetta, Student Member, IEEE,
Edoardo Fiorucci Member, IEEE and Francesco Vegli
F
127

2
Hydra 2625A data logger. The communication with the host
PC is performed via the RS-232 interface.
The electric load at the fuel cell output is implemented
using the Agilent N3301A mainframe, with a N3302A
electronic load module. The module has a current range of 0-
30 A, a voltage range of 0-60 V and a maximum power of 150
W. This system allows constant current, constant voltage,
constant resistance and transient modes to be implemented.
The electronic load allows both current and voltage to be
measured by the PC during the tests, by means of the IEEE
488 interface.
The control and measurement software has been developed
in the NI LabVIEW environment. It performs the following
tasks: i) communication with the gases controllers via a DDE
server, for both measurement and control of flow rate and
pressure of the reactants; ii) serial communication with the
Hydra data logger for the temperature measurement; iii) IEEE
488 communication with the Agilent electronic load for
control and acquisition of current and voltage data; iv)
definition of the testing procedures in terms of acquisition
time, sampling rate, current shape; v) acquired data
visualization and storage. Using this system we carried out
some measurements in both steady-state and transient
conditions.
III. THE EXPERIMENTAL RESULTS
The FC model has been obtained using an input-output
modeling technique, by applying a test input and processing
the obtained results. To this aim, different experimental
characterization tests have been carried out: firstly we
performed quasi-static run-up and run-down procedures and
then we measured the main parameters during fast current
changes.
A. Quasi-static measurements
This characterization is used to evaluate the PEM cell
performance at different load conditions. During this test the
H
2
has been set in dead-end conditions, the air flow rate has
been kept fixed, while the supplied power ranged over the
entire scale during a specific time interval.
The test has been conducted to simulate steady state
conditions, with the following test parameters: i) total
observation time of 3000 s; ii) rising and falling ramps with a
slew rate of 10 mA/s; iii) minimum current value of 0.0 A; iii)
maximum current value of 25.0 A; iv) pressure and
temperature of both air and H
2
respectively of 15 kPa and
22 C; v) constant air flow rate of 6.0 l/min; vi) dead-ended
H
2
operation.
Using this low speed of variation of the supplied electric
power, the fuel cell can be assumed to be in steady state
conditions. In the Figs. 2-3 the measured current and voltage
are reported for both the rising and falling current ramps.
When the test starts, the output current increases from 0 A
to 25 A (run up), then decreases to 0 A (run down). During
this test the fuel cell voltage decreases proportionally to the
current and reaches the minimum corresponding to the
maximum current. The FC voltage power increases
proportionally to the current and the maximum is obtained at
the maximum current.
The analysis of the first results, obtained with the proposed
system, suggests pointing out the following features: i) the
adopted PEM fuel cells have two different apparent volt-
amperometric characteristics, depending on the increasing or
decreasing of current (Fig. 4); ii) the H
2
consumption
waveform has a shape similar to the current one, with a delay
due to the quantity of H
2
present in the pipes that introduces a
smoothing effect; iii) the cells are very sensitive to the heat,
and the measurement of temperatures of both the stack and the

Fig. 1 : FC measurement system

127

3
reaction products can be used for the evaluation of the
efficiency; iv) the measurement of the delay between the
current variation and the corresponding voltage response can
be adopted for the implementation of dynamic control systems
for automotive applications.



Fig.2 Current and voltage (run up)



Fig.3 Current and voltage (run down)

4
5
6
7
8
9
10
0 5 10 15 20 25
V
o
l
t
a
g
e

[
V
]

Fig.4 Hysteresis effect

B. Dynamic measurements
The quasi-static response represents only one aspect of the
FC behaviour, because the electric load can have high
dynamic variations [2].
A dynamic change of the output current can be supplied by
the FC only if the internal chemical reaction has a subsequent
change. The problem is that the FC introduces a delay
between the required and the supplied electric power, delay
that is negligible in some applications, but is unacceptable in
others. In order to investigate about the dynamic performance,
we carried out sequential increasing and decreasing current
steps. The increasing current steps are the more critical, as
shown in Fig. 5, in which the output transient voltage for
different current step variations I
s
are reported, because of the
great voltage reductions.



Fig. 5 Experimental FC voltage for current step variations Is
IV. THE FC MODEL

The current and voltage acquired waveforms are used to
perform a system identification using MATLAB toolbox. The
transfer function P(s), such as V(s) = P(s)I(s), and its Bode
diagram are reported in (1) and Fig. 6.

( )
2
0.2618 0.03797
0.8721 0.1664
s
P s
s s

=
+ +
(1)


Fig. 6: Fuel cell transfer function Bode diagram

For a current step variation I
s
the voltage steady state value
is:
0
lim ( ) (0)
s
s
s
I
s P s I P
s

=
(2)
127

4
So, the steady-state value proportionally depends on I
s
by
P(0), that is the slope of the voltage versus current linear
characteristic.
On the other hands, the experimental static characteristic
(Fig. 6) is not linear and so the process P(s) can not properly
describe the FC steady-state behavior. Then, its necessary to
modify the model adding a block that forces the process P(s)
to the correct steady-state value given by the static
characteristic V
fc
= f(I
fc
). In general, the steady-state FC
behavior can be described by the following equation [5]:

( )
0
ln ln 1
n n
n
l
i i i i
V E R i i A B
i i
+ +
= +


(3)

where E is the no-load voltage, coefficients A and B are
respectively related to activation and polarization phenomena,
resistance R accounts for ohmic voltage drop, i
0
, i
l
and i
n
are
the exchange, limiting and internal parasitic diffusion currents
[6]. Unfortunately, the adoption of the (3) requires the
knowledge of some fuel cells parameters that usually are not
provided by the manufacturer; so we adopted an experimental
approach, starting from the polynomial interpolation of
steady-state voltage and current measurements.
The whole model of the FC is reported in Fig. 7. The value
given by equation (2) is subtracted from process P(s) and the
static one is then added. Finally the validation of the model
has been carried out for different current steps, with
satisfactory results. In Fig. 8 the comparison between
measured and simulated voltage behaviors for a current step
from 10 to 20 A is shown.




Fig. 7: Whole fuel cell model

V. THE APPLICATION OF THE MODEL FOR THE CHOICE OF
SUPERCAPACITORS

As Fig. 5 shows, the FC has a slow dynamic and the initial
transient under voltage with respect to the steady state value is
recovered in few tens of seconds. So a supercapacitor is
needed to partially or totally avoid the over drop voltage when
the load requires current step variations. To this aim we
simulated a FC system, including a supercapacitor.



Fig. 8 : Simulated and measured voltage response for a current step variation

Fig. 9 depicts the blocks diagram of the simulated system,
shown in Fig. 10 in mathematical detail. The system uder
simulation consists of: i) a programmable signal generator that
gives the required load current; ii) the fuel cell model P(s)
with its steady-state correction system; iii) the supercapacitor
transfer function SC(s), such as I
SC
(s) = SC(s) V
SC
(s), placed
on the feedback of the whole system.



Fig. 9: Blocks diagram of the simulated system

The supercapacitor model is defined by the transfer
function
( )
1
C s
SC s
R C s

=
+
, (4)
where C and R are respectively the capacitance and the
equivalent series resistance of the supercapacitor.
The simulation of a real supercapacitor has been performed by
adopting the parameters in the data sheet of a commercial
supercapacitor. The results show that the time constant RC is
slightly affected by the capacitance values. The FC model has
been utilized to evaluate the dynamic behaviour of the whole
system when a supercapacitor is parallel connected.

127

5

Fuel Cell
Model
P ( s)
P ( 0 )
Steady - State
Lookup Table
Supercapacitor
Model
Current
Load
Fuel Cell
Voltage
+
+
+
+
-
SC(s)


Fig. 10: Whole simulated system

As an example of the dynamic system response, we
performed a simulation in which the capacitance value was
undersized (C = 5 F) for the analyzed current step variation
(I
S
= 10 A), causing a transient under voltage for the fuel cell
(Fig. 11) and a consequent negative transient current of the
supercapacitor with a peak value I (Fig. 12).
Implementing an automatic routine, we performed some
simulations to correlate the fuel cell transient under voltage
and the negative current peak with the capacitance value. It
can be seen that a 50 F supercapacitor is enough to remove the
transient under voltage V, as shown in Fig. 13.


V


Fig. 11 : Fuel Cell voltage for a current step variation with an undersized
capacitance value

This value is lower than the one evaluated by means of a
simplified energetic approach, i.e. the method frequently used
to design FC systems. In such a method the capacitance is
evaluated considering that the fuel cell and the supercapacitor
share equally the increment of energy in transient conditions:

( )
2 2
1 1
2 2
S initial final
I V T C V V = (5)

I


Fig. 12 : Supercapacitor current for a current step variation with an undersized
capacitance value

where: V is the average voltage during the transient, T is the
transient time, V
initial
and V
final
are respectively the initial and
final steady-state values.
Adopting the (5) for a current I
S
of 10-20 A, a 100 F
supercapacitor is calculated.



Fig. 13 : I vs. capacitance value for different load current steps

VI. CONCLUSIONS
The PEM fuel cells can be successfully adopted for
distributed generation of electric power, UPS and electric
transportation applications. New tools for a suitable design of
FC-based systems have become extremely important; in this
paper an experimental approach to the modeling of PEM fuel
cells in dynamic conditions has been proposed.
As a first step, an automatic testing system has been
developed and both static and dynamic charaterization of a
PEM fuel cells stack have been performed. The data obtained
Capacitance ( F )
I
(A)
I
S
: 0-25 (A)
I S : 0-12.5 (A)
I S : 10-20 (A)
I
S
: 5-10 (A)
I
S
: 5-7 (A)
10 20 30 40 50 60
0
0.1
0.2
0.3
0.4
0.5
127

6
during experimetal tests has been used for the synthesis of a
dynamic numerical model of a stack of PEM fuel cells, and it
has been used to simulate a FC system with a parallel
connected supercapacitor in order to correlate the transient
under voltage and the supercapacitor parameters, when the
load requires current step variations.
In this it is possible to accurately estimate the minimum
value of capacitance needed to obtain a given transient under
voltage. Such a value is generally lower than that one
evaluated by means of a simplified energetic approach, i.e. the
method frequently utilized to design FC systems.
REFERENCES
[1] Ellis, M.W; Von Spakovsky, M.R.; Nelson, D.J.; Fuel cell
systems: efficient, flexible energy conversion for the 21st
century, Proceedings of the IEEE, Volume: 89 Issue: 12, Dec.
2001.
[2] Laughton, M.A.; Fuel cells, Power Engineering Journal [see
also Power Engineer], Volume: 16 Issue: 1, Feb. 2002.
[3] L. Guzzella, A. Sciarretta, Vehicle propulsion system, Spinger,
2005.
[4] G. Bucci, F. Ciancetta, E. Fiorucci; An Automatic Test System
for the Dynamic Characterization of PEM Fuel Cells,
Proceedings of IEEE IMTC 2004.
[5] M. Andriollo, F. Castelli Dezza, M. Tamburrino, Converter
control strategies for the power flow management in a fuel-cell
supplied vehicle for the public city transportation, Proceedings
of Symposium on Power Electronics, Electrical Drives,
Automation and Motion, Speedam 2004, Capri, IT, June, 16-18
2004.
[6] U.S. Dept. of Energy Nat.l Energy Techn. Lab., Fuel Cell
Handbook (5
th
ed.), October 2000.


Giovanni Bucci received a degree in electrical engineering in 1985 from
the University of L'Aquila in Italy. He is now a Full Professor in Electrical
Measurement at the University of L'Aquila. His current research interests
include ADC and wireless device testing, multiprocessor-based measuring
systems, digital algorithms for real-time measuring instruments, power
measurements and Fuel Cell. He has authored more than 100 scientific papers.
Prof. Bucci is a member of IEEE Instrumentation and Measurement Society.
Fabrizio Ciancetta was born in Pescara, Italy, in 1977 and he received
the degree of electronic engineering in 2003 from the University of L'Aquila
in Italy. In 2003 he joined the Department of Electrical Engineering of the
University of L'Aquila as a part-time researcher on a research project aimed at
the development of digital and distributed measurement system and
measurement software optimization. He is now a Ph.D. student in electrical
and information engineering at the same university.
Edoardo Fiorucci was born in Popoli (PE), Italy, in 1974.
He received the degree in electrical engineering in 2000 from the University
of LAquila (Italy). In 2004 he received the Ph.D. degree in Electrical and
Information Engineering from the same university. He is currently an assistant
professor in Measurements and Instrumentation in the Department of
Electrical and Information Engineering at the University of L'Aquila; his
current research interests include measurement systems for power quality
evaluation, fuel cells systems and smart web sensors. Dr. Fiorucci is a member
of IEEE Instrumentation and Measurement Society.
Francesco Vegli is a Full Professor in the Department of Chemistry,
Chemical Engineering and Materials of Theory and Development of Chemical
Processes. He has more than 100 papers on international journals, more than
80 communication to International Congresses, extended abstract e poster and
2 patents (1 national patent; 1 EU patent);