1234 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO.

5, SEPTEMBER 2004

Mathematical Model and Characterization

of the Transient Behavior of a
PEM Fuel Cell
Wolfgang Friede, Member, IEEE, Stéphane Raël, and Bernard Davat, Member, IEEE

Abstract—Modeling of fuel cells is getting more and more im- ature, gas flows, pressure, and humidity. Each of them has a
portant as powerful fuel cell stacks are getting available and have strong influence on the behavior of the cell, especially on its
to be integrated into power systems. In this paper, the governing internal humidity conditions. An inappropriate setting of these
equations of the transient behavior of a proton exchange mem-
brane fuel cell are presented. They show the influence of the op- conditions can not only lower the performances and efficiency
erating conditions and the current density on internal parameters, of the fuel cell but can even lead to an irreversible degradation
especially the ohmic resistance. of internal parts as the catalysts or the membrane.
A method and specially designed electronic load are presented The measurement of the internal resistance of the fuel cell is a
for in-situ membrane resistance measurements. This resistance is
good indicator for correct operation. This is due to the strong de-
plotted in transient operation during parameter changes and cur-
rent variations. The results are compared to the simulated values pendence between the ionic conductivity of the membrane and
to explain the internal phenomena of the fuel cell. Special atten- its water content. In this work, a correlation between the ohmic
tion is given to the water management, one of the most important resistance and the electrode behavior is shown to demonstrate
challenges in low temperature fuel cell operation where drying or the influence of humidity conditions on the general humidity
flooding can decrease the performance or even lead to destruction.
state of the fuel cell.
Index Terms—Fuel cell, proton exchange membrane (PEM).
II. STATE OF THE ART
I. INTRODUCTION Hydrogen PEM fuel cells transform chemical energy into
electrical and thermal energy by the simple chemical reaction
F UEL CELLS currently attract the interest of more and more
engineers, as shown by a growing number of firms in-
vesting in research and development around this chemical re- (1)
actor, allowing production of electrical energy from hydrogen
or other fuels with high efficiency and low emission. Impor- In order to get an electric current out of this reaction, hydrogen
tant progress has been made in optimization of materials and oxidation and oxygen reduction are separated by a membrane,
catalysts for increased lifetime and efficiency as well as lower which is conducting protons from the anode to the cathode side.
weight and volume. The possible fields of use cover stationary, The semi reactions on both electrodes are
mobile and portable applications.
The use of fuel cells in electric circuits needs an under- anode
standing of internal effects for an efficient control of the system
cathode (2)
as well as an understanding of dynamic behavior in order to
design the entire energy production unit with fuel cells, con-
verters, and storage units. Especially in vehicle applications, the While the protons are transported through the membrane, elec-
power demand is rapidly changing during typical load cycles trons are carried by an electric circuit in which their energy can
including acceleration and breaking periods. To find the best be used. An energy balance shows that the electrical efficiency
suited fuel cell power train, it is necessary to know what kind of the fuel cell is about 50%. The thermodynamic (reversible)
of load changes the fuel cell is able to deliver. This information output voltage of one cell can be calculated from Gibb’s free
is necessary for the design of the energy storage unit needed to energy and its value is for a standard operating point
cover the power peaks and maybe to recover breaking energy.
The optimized operation of a fuel cell needs the control of V (3)
F
a great number of conditions, which are imposed by external
pumps, valves or compressors. The most important are temper- In a real fuel cell, one cell can provide a voltage between 0.5
and 0.8 V, due to voltage losses which will be discussed later
Manuscript received July 6, 2003; revised June 3, 2004. Recommended by on. Several cells are usually connected in series, forming a stack.
Associate Editor E. Santi. Anyway, the gas supply is provided in parallel. Existing fuel cell
The authors are with the Institut National Polytechnique de Lorraine, stacks have up to 100 single cells providing a voltage of about
Vandoeuvre-les-Nancy 54516, France (e-mail: [email protected]
nancy.fr). 60 V. For heat evacuation, an external cooling system is needed.
Digital Object Identifier 10.1109/TPEL.2004.833449 This heat can be used to provide cogeneration heating energy.
0885-8993/04$20.00 © 2004 IEEE

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.
 FRIEDE et al.: TRANSIENT BEHAVIOR OF A PEM FUEL CELL 1235

In this work, a one-dimensional (1-D) model is presented

which describes the behavior of one cell supposing that the volt-
ages of all cells in the stack are equivalent and that the current
density is homogenous. The reaction is supposed to take place at
the interfaces of the membrane and the reaction layers, three di-
mensional effects in the reaction layers are neglected. The mem-
brane is gas tight and contains liquid water while in the reaction
and diffusion layers, all water is supposed to be water vapor. The
temperature is fixed and homogenous all over the cell. This can
be assumed for a very powerful cooling system. In real cases, the
dynamic behavior has to integrate temperature transients [10].

III. DESCRIPTION OF TRANSPORT AND TRANSFER PHENOMENA

A. Channels
The design of the channels for gas distribution is very dif-
ferent from one fuel cell stack to another. Its influence on the in-
Fig. 1. Internal structure of a single fuel cell. ternal water management is important, and so every membrane
needs its own adapted flow field [11]. In the present 1-D ap-
The total efficiency of the fuel cell is calculated using the proach, an ideal gas mixture is modeled to take into account the
enthalpy difference between the reactant gases and the excess gas flow at the end of the channels. The equation de-
product water and can be determined by the following equation scribing the gas mixture is
[1]:
(5)
(4)
F
with is the partial pressure of each gas, is the total pressure,
It can be seen that the efficiency depends only on the fuel cell is the molar flow of each gas, is the total molar flow, and
voltage. For a high efficiency of the fuel cell stack, the output is the deepness of the channel. is the universal gas constant
voltage has to be as high as possible. However, this rule has to and the temperature. The indexes “in” and “cons” signify the
be applied carefully as a good operating point for the stack often inlet flow and the consumption defined as the flow toward the
needs a high gas humidity and high pressure which causes losses diffusion layers which can be positive or negative. The quantity
in the auxiliaries of the system. In this work, we use the voltage of gas delivered to the fuel cell is generally controlled to
of the stack to determine good working points. follow linearly the current value. The ratio between consumed
The difficulty in operating a fuel cell is the need of supply reactant gas and delivered gas is called utilization and represents
with gases, evacuation of product water, and cooling. In order the reciprocal value of the stoichiometry.
to find the working point with the optimum efficiency, electric
output values have to be described as a function of operating B. Diffusion Layers
conditions. The input values of a model for this purpose are the
pressure, humidity, and flow of every gas and the temperature The diffusion layers ensure the gas supply for the catalyst
of the cooling system. sites over the whole active area. As the current—voltage relation
Another important aim of modeling is the estimation of the of the fuel cell depends on the local pressure at the reaction sites,
fuel cell performances, especially in transient operation modes. the pressure drop over the diffusion layers should be known as
Especially fuel cells in vehicles are exposed to violent load a function of gas flow. These layers are characterized by the
changes. Thus, the output of the model has to provide a cur- porosity and the tortuosity of the material. Three physical
rent—voltage characteristic for stationary or transient working phenomena, known as the dusty gas model [12], are described
conditions as well as the efficiency of the fuel cell. Some in- in literature to explain the behavior of the gas molecules inside
ternal values as local gas pressure, humidity and internal resis- these layers. The counter-flow of reactant gas and product water
tance allow a better understanding of the physical phenomena vapor, in the case of an air feed within the presence of nitrogen,
inside the fuel cell. especially needs an approach that takes into account the inter-
The physical effects in the fuel cell are various and need an in- actions between the molecules of the different gases.
terdisciplinary approach. Most authors split up the cell in seven Collisions between molecules of different gases can be
zones, as presented in Fig. 1 [2], [3]. described using Stefan–Maxwell equations, collisions between
This representation is based on the work of Springer et al. [4], gas molecules and the solid part of the porous material using
Bernardi and Verbrugge [5], and Fuller and Newman [6]. These Knudsen equations. In addition to that, the gas viscosity has to
works solve the transport and transfer equations in stationary be taken into account for a complete model. For the properties
mode. Using mass balance equations, the transient behavior can of a standard fuel cell gas diffusion layer, the numerical
also be determined [7]–[9]. values of the coefficients show that considering only diffusion

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.
 1236 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 5, SEPTEMBER 2004

where the diffusion coefficient can be determined by the

limiting current which is the current at which the cell voltage
drops to zero. The effect of too much or not enough water in the
electrode, which modifies the thickness of the water layer, can
be observed in the results, but in the model, the water film is
supposed to have a constant thickness.
To describe the voltage–current relation each reaction has to
be described individually. As the anode voltage drop is much
lower than the cathode one, only the cathode side is described
here. A Tafel equation is used, linking the voltage drop to
the current density [7]. The oxygen partial pressure at the
catalyst site is included as well as the temperature

(9)
Fig. 2. Humidity conditions of the electrode.

where is the activation energy and a factor giving the

according to the Stefan–Maxwell equations with the following ration between the active surface and the geometric area of
term can make a good approximation [4], [7] the cell. represents the charge transfer coefficient and F is
Faraday’s constant.
(6)
D. Membrane
In the membrane two types of transport phenomena exist for
where is a diffusion coefficient defined for each pair of gases the transport of two molecules: water and protons. Proton trans-
depending on the properties of the gas diffusion layer of the elec- port depends directly on the current density
trode. The total pressure is constant, and, supposing
ideal gas, is constant, which leads to a system of
H (10)
equations for gases. F
To introduce the time dependence, a molar balance is intro-
Water molecules surround every proton and are dragged with it
duced, supposing that volume and temperature are constant
in the direction of proton flow from the anode to the cathode.
This is expressed by a drag coefficient which depends on the
(7)
humidity of the membrane. A humidity gradient leads to diffu-
sion of water from zones with high water content to zones with
The cited equations allow a dynamic description with a numeric
low water content. The diffusion coefficient is written . Nor-
solver as Matlab Simulink. The layer is divided into several el-
mally, due to water production at the cathode side, water diffuses
ements following the axis. For each of them, (6) and (7) are
from the cathode to the anode side of the membrane. The con-
solved. A reasonable choice of the parameters of the material,
ductivity of the membrane, relating the electrical potential to the
porosity and tortuosity , does not lead to a sufficient pressure current density is called . The conductivity depends strongly
drop in agreement with the cell voltage drop we have measured on the humidity of the membrane.
on our test bench. This shows that the mutual influences of the The transport of water and protons can be expressed in matrix
molecules of different gases are not the only diffusion phenom- form using electrochemical potential gradients [9]:
enon to be considered.

C. Electrodes F F
The electrodes are made of platinum particles which are de- (11)
posited within a carbon support on the membrane surface. As F
a result of membrane porosity, the particles are introduced into
the membrane material. For the reaction, a small water film is The electrochemical potential is linked to the water content
necessary to ensure proton transport in the membrane while too which is defined as the ratio between water molecules and
much water can constrict the pores so that gas supply is not suf- groups in the membrane by the following equations:
ficient any more (Fig. 2) [1]. In the present model, the diffusion
F
through the water layer is described by Fick’s law [13], written
here for partial pressure

(8) (12)

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.
 FRIEDE et al.: TRANSIENT BEHAVIOR OF A PEM FUEL CELL 1237

TABLE I
NUMERICAL PARAMETER VALUES AT 55 C, 10 Pa, AND  = 7

Fig. 3. Test bench with 500-W fuel cell stack.

A simplified mass balance equation gives the time derivative of

the water content Fig. 4. Electric load for test issues.

(13)
conditions, their dependency on temperature, pressure, and
Knowing the humidity distribution and the proton flow using humidity are given in [18].
(10), the equations can be solved to determine the water flow
as well as the electric potential which gives the voltage IV. EXPERIMENTAL RESULTS
drop over the membrane.
In contrast to other approaches which determine the resis- A. Test Bench
tance directly from a correlation [13]–[15], our model can deter- The considered fuel cell is a 500 W PEMFC stack installed
mine the water flow through the membrane. In addition to that, in a test bench provided with systems for gas supply and con-
a different humidity of anode and cathode can be taken into ac- trol (Fig. 3). Air is supplied from an oil-free compressor through
count which is important as flooding can take place at one elec- a humidification unit, and pure dry hydrogen from cylinders. A
trode while the other one dries up. cooling system imposes a constant stack temperature by heating
or cooling of the water circuit. A programmable controlling unit
E. Numerical Solution contains all necessary control functions such as reference set-
Each layer is divided into several elements following the ting, and measurement conditioning, and safety shutdown. The
axis. For each element, a number of state values are given as ini- test bench as well as the fuel cell have been delivered by ZSW,
tial conditions as there are partial pressures in the gas diffusion Ulm, Germany. The characteristics of the fuel cell are: 23 cells,
and reactive layer and water content in the membrane. The time active area of 100 cm , every bipolar plate water cooled.
derivative of each state vector is calculated using the above cited An electric load has been constructed locally. It is able to im-
equations. The result is integrated using a Matlab Simulink code. pose a constant current to the fuel cell as well as all kinds of
Between the different layers, continuous flow and pressure harmonics. This is especially important for diagnostic tasks by
is assumed, adjacent layers exchange a partial pressure vector methods like impedance spectroscopy which make it possible to
(from the channel to the membrane side) against a molar estimate parameters which are not simply accessible by dc mea-
flow vector (from the membrane side to the channels). To get surement at the stack terminals. The use of the linear operating
bounding conditions for the membrane water content, sorption point of a metal oxide semiconductor (MOS) transistor allows
terms are used as described in [4], [16], [17]. They give the a current source behavior of the load (Fig. 4). It follows current
water content of the membrane as a function of the water partial references of up to 2 kHz [19].
pressure in the surrounding gas. Two National Instruments data acquisition cards allow data
The numerical values for the above mentioned variables are logging and treatment while the measurement PC can impose
given in Table I. The values are given for typical operation the electrical current forms and gas flows automatically [20].

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.
 1238 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 5, SEPTEMBER 2004

Fig. 5. Small signal equivalent electric circuit of the fuel cell.

B. In-Situ Resistance Measurement

Fig. 5 shows an equivalent circuit valid for small signals. The
electrodes are represented by a capacitor which takes into account
the double layer capacity appearing because of an accumulation
of electrons and protons on both sides of the reactive layer. The
chemical reaction is represented by resistors , diffusion effects
by a Warburg impedance W [21], and finally the membrane by an
ohmic resistance . The dashed parts of the circuit—electric re-
sistance of the bipolar plates and voltage drop of the anode—are
neglected. By choosing a high frequency, the influence of the
electrodes can be separated from the one of the membrane.
For the study of the membrane resistance, a high frequency si-
nusoidal current is superposed on the load current. A frequency Fig. 6. Measured voltage and ohmic resistance during temperature changes
imposed by the cooling system. Measurement conditions: current 0.3 A cm ,
of about 2 kHz and an amplitude of 10% are sufficient to get a humidifier temperature 45 C, H / air utilization 50/25%, atmospheric pressure.
good result. This method allows to visualize the membrane re-
sistance without interrupting the load circuit.
If the membrane resistance is known as well as the reversible 1) Temperature Transients: Temperature plays an important
voltage according to (3), the membrane voltage drop (IR losses) role in the fuel cell operation. Higher temperatures are desirable
can be separated from the electrode voltage drop. The total for minimization of the cooling system and for a higher speed of
voltage can be written as the reactions, but the water management gets more difficult in
this case. A higher cell temperature leads to greater water loss
(14) to the exhaust gas streams as more water evaporates, causing
The voltage considering only electrode losses is called corrected drying of the porous media inside the cell.
Fig. 6 shows the response to stack temperature steps for a cur-
voltage and can be expressed as follows:
rent of 30 A (0.3 A cm ). The programmable logic controller
(15) of the test bench allows a setting of the cooling water input tem-
perature, represented by a dashed line in the first plot. As the
The corrected voltage in the following plots allows a good vi-
chemical reaction produces heat with the reaction, the cooling
sualization of the subdivision of the losses between membrane water output temperature is slightly higher than the one at the
and electrodes. The distance between measured and corrected input. The temperature of the bubble humidifier has been held
voltage represents the membrane state. This distance increases constant. The voltage follows the steps rapidly, but shows a slow
as the membrane dries. The state of the electrodes can be seen decrease after each step which can be attributed to liquid water
using the absolute value of the corrected voltage, knowing that accumulation in the cell. At each temperature step, less water
changes of the reversible voltage are very small in the domain is drawn out by the outlet gases, so the performances increase
of possible operating conditions of a PEM fuel cell. with a better membrane conductivity which can be observed via
the internal resistance. When liquid water appears, the voltage
C. Transient Tests
decreases while the resistance decreases, too. Thus, the slow be-
Transient tests show time constants of different phenomena havior can be attributed to the electrodes. It is important to see
and help to validate the dynamic model. Variations of tempera- that minimizing the ohmic resistance does not necessarily imply
ture, current and humidity are shown with the voltage response an optimization of the cell behavior.
as well as the evolution of the internal resistance of the fuel cell. The model response to the same operating conditions is given
This resistance is an indicator for the internal humidity condi- in Fig. 7. As most published models, this model represents well
tions of the cell. This is due to the conductivity increase of the the total voltage of the cell by explaining the voltage variations
membrane when the water content rises. by a variation of the membrane resistance with the operating

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.
 FRIEDE et al.: TRANSIENT BEHAVIOR OF A PEM FUEL CELL 1239

Fig. 7. Simulated response to temperature variations as in Fig. 6.

Fig. 8. Measurement of voltage and ohmic resistance during a current
step from 20 to 40 A (0.2 to 0.4 A cm ). Measurement conditions: stack
conditions. The total cell voltage follows well the measured temperature 55 C, humidifier temperature 45 C, H / air utilization 50/25%
(at 40 A), atmospheric pressure.
voltage shown in Fig. 6. When the ohmic voltage drop is added
to get the corrected voltage, the comparison of measurement and
simulation show that the influence on the membrane behavior is of the electrodes is still neglected while being important in the
too important while the influence on the electrode is neglected measured curves.
in the model. Nevertheless, the phenomena in the membrane are 3) Humidification Transients: The parameter with the
well predicted, their effects are overestimated but showing the strongest influence on the output conditions is the humidity of the
same behavior. inlet gases. The above presented test bench supplies the fuel cell
2) Current Transients: Current steps play a very important with dry hydrogen, but the air has to be humidified to obtain stable
role in the use of fuel cell on loads with varying power demands. operating points. To show the influence of the air humidity, dry
Unless the presence of a storage unit, the fuel cell has to react air has been temporarily delivered to the fuel cell. The response
rapidly and deliver the demanded power. We show here the be- can be seen in Fig. 10 as a function of the stack temperature.
havior of a fuel cell in operation which has already reached its These strong changes in the operating conditions lead to a
operating temperature and is already humidified by previous op- fast increase of the membrane resistance, which reaches nearly
eration. The cold start behavior is heavily dependent on the aux- twice its initial value. At higher stack temperature, the drying
iliary components of the system and is not examined here. The of the membrane is faster as more water is evacuated with the
voltage response on a current step from 20 to 40 A (0.2 to 0.4 warmer outlet air and hydrogen stream. As the humidifier tem-
A cm ) is presented in Fig. 8. The gas flows have been held perature is set to a fixed difference to the stack temperature, the
constant during this test. The voltage is decreasing before in- inlet air contains more water at higher stack temperatures. This
creasing to its final value. This effect is partially due to the mem- means, that the difference between humidified and dry operation
brane humidity which is represented by the plot of the resistance. is greater at higher temperature. At 45 C, the stack approaches
However, the greater part can be explained by the electrode be- a stable operation point, even fed with dry air. However, the op-
havior as the corrected voltage follows the form of the measured eration had to be stopped because of one single cell the voltage
voltage. The better conditions after the step are degrading prob- of which dropped below a security level of 300 mV.
ably due to liquid water accumulation after several minutes. The simulation in Fig. 11 shows similar effects. The form of
The simulation in Fig. 9 shows a similar effect. As shown the ohmic resistance curve is well reproduced. As the membrane
above for temperature changes, all effects are associated with reacts too strongly, the voltage response has a too big amplitude,
the membrane and the forms of the cell voltage and resistance too. As in the previous plots, the corrected voltage is not affected
plots can be reproduced by the model. However, the influence by the conditions.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.
 1240 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 19, NO. 5, SEPTEMBER 2004

Fig. 11. Simulation of shunted humidifier as in Fig. 10.

Fig. 9. Simulation of a current step from 20 to 40 A as in Fig. 8.

Fig. 12. Evolution of the membrane water content as a function of time and
position (x = 0 at the anode side) for the curve at 55 C in Fig. 11.

the cathode side. However, there is less water diffusion through

the membrane and the water content at the anode side is de-
creasing while dry air is fed to the cell.

V. CONCLUSION
A. Measurements
The presented measurements show that operating conditions
Fig. 10. Measurement of voltage and ohmic resistance while humidifier shunt have a very strong influence on the electrical output of a fuel
was opened and closed again. Measurement conditions: current 30 A (0.3 A cell. Two major phenomena have been described and experi-
cm ), stack temperature see legend, humidifier temperature stack temperature mentally distinguished: the humidity of the membrane and the
0 10 C, H /air utilization 50/25%, atmospheric pressure.
flooding and dry out of the electrodes. The effects on the mem-
brane have a time constant of several seconds and can be ob-
The one dimensional model uses a subdivision of the axes served by a measurement of the ohmic resistance of the fuel cell.
which is defined perpendicular to the membrane surface. The Higher time constants can be seen for the flooding of the cell,
water content of each element can be visualized as done in especially due to liquid water accumulation in the electrodes.
Fig. 12 for the simulation of the shunted humidifier at 55 C. These time constants can reach several minutes which has to be
It can be seen that the anode side, which is initially at a low hu- considered when planning an efficient control system. All op-
midity level, is less affected by the change in the air humidity as eration conditions as temperature, pressure and gas flows have

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.
 FRIEDE et al.: TRANSIENT BEHAVIOR OF A PEM FUEL CELL 1241

an effect on the humidity conditions of the fuel cell and their [10] M. D. Lukas, K. Y. Lee, H. Ghezel-Ayagh, S. G. Abens, and M. C. Cervi,
optimal choice is the greatest challenge in fuel cell operation. “Experimental transient validation of a direct fuelcell® stack model,” in
Proc. IEEE Power Eng. Soc. Summer Meeting, Vancouver, BC, Canada,
Although lots of models in literature pay much interest to the 2001, pp. 1363–1368.
membrane resistance, its influence on the fuel cell voltage is not [11] Fuel Cell Technology Handbook, G. Hoogers, Ed., CRC, Boca Raton,
as big as the influence of the electrode. The humidity of the elec- FL, 2003.
[12] E. A. Mason and A. P. Malinauskas, Gas Transport in Porous Media :
trode is very important for a proper gas and proton transport. It The Dusty-Gas Model. Amsterdam, The Netherlands: Elsevier, 1983.
has been observed that there is a strong correlation between the [13] J. C. Amphlett, R. M. Baumert, R. F. Mann, B. A. Peppley, P. R. Roberge,
and T. J. Harris, “Performance modeling of the Ballard Mark IV solid
membrane resistance and the losses in the electrode. It is then in- polymer electrolyte fuel cell -I. Mechanistic model development,” J.
teresting to supervise the ohmic resistance in order to determine a Electrochem. Soc., vol. 142, pp. 1–8, 1995.
good operating point for a given power demand. The optimal re- [14] R. F. Mann, J. C. Amphlett, M. A. I. Hooper, H. M. Jensen, B. A. Pep-
pley, and P. R. Roberge, “Development and application of a general-
sistance value has to be chosen between a too high value which ized steady-state electrochemical model for a PEM fuel cell,” J. Power
indicates a too dry operation point and a too small value ap- Sources, vol. 86, pp. 173–180, 2000.
pearing when liquid water begins to flood the electrodes. [15] J. T. Pukrushpan, A. G. Stefanopoulou, and H. Peng, “Modeling and
control for PEM fuel cell stack system,” in Proc. Amer. Control Conf.,
Anchorage, AK, 2002, pp. 3117–3122.
B. Model Limits [16] J. T. Hinatsu, M. Mizuhata, and H. Takenaka, “Water uptakeof perfluo-
rosulfonic acid membranes from liquid water and water vapor,” J. Elec-
With the presented model, it is possible to predict the dynamic trochem. Soc., vol. 141, pp. 1493–1498, 1994.
response to changes of any operating condition. Anyway, the [17] M. Legras, Y. Hirata, Q. T. Nguyen, D. Langevin, and M. Métayer,
influence of the membrane resistance on the cell voltage and “Sorption and diffusion behaviors of water in Nafion 117 membranes
with different counter ions,” Desalination, vol. 147, pp. 351–357, 2002.
the sensitivity of the membrane resistance to parameter changes [18] R. C. Reid, J. M. Prausnitz, and E. B. Poling, The Properties of Gases
is overestimated. In contrast, the influence of water, both the and Liquids. New York: McGraw-Hill, 1987.
[19] W. Friede, S. Raël, and B. Davat, “PEM fuel cell models for supply
flooding and drying of the electrodes has not yet been included. of an electric load,” in Proc. 7th Int. Conf. Modeling and Simulation
The huge number of parameters and the fact that the model does of Electric Machines, Converters, and Systems (Electrimacs’02), Mon-
not take into account the current and humidity distribution over treal, QC, Canada, 2002.
[20] W. Friede, “Modeling and characterization of a PEM fuel cell,” Ph.D.
the cell surface makes it impossible to fit the model for the great dissertation, Inst. Nat. Polytech. Lorraine, Nancy, France, 2003.
variety of operating conditions the cell can be exposed to. Un- [21] A. J. Bard and L. R. Faulkner, Electrochemical Methods. New York:
less an identification based on measured correlations is done to Wiley, 2000, pp. 368–388.
[22] G. Maggio, V. Recupero, and L. Pino, “Modeling polymer electrolyte
include liquid water influence as in [22], the necessary param- fuel cells: An innovative approach,” J. Power Sources, vol. 101, pp.
eters can not easily be found in literature. They vary between 275–286, 2001.
membranes of different producers and even between the dif-
ferent cells of a stack. The measurement conditions have an in-
Wolfgang Friede (M’00) born in Bretten, Germany,
fluence on the values, too. A membrane embedded in the fuel on February 24, 1975. He received the B.S. degree
cell behaves different from an isolated membrane, so the values in electrical engineering from the University of Karl-
of published material parameters can not always be applied to sruhe, Germany, in 2000, the M.S. degree in electrical
engineering from the Ecole Nationale Supérieure des
a fuel cell. The parameters with the biggest influence on the re- Ingénieurs Electriciens de Grenoble, France, in 2000,
sult are those describing the membrane humidity: the sorption and the Ph.D. degree from the National Polytechnic
terms giving the boundary conditions for the water content in Institute of Lorraine, Vandoeuvre-les-Nancy, France,
the membrane and the water diffusion coefficient in the mem- in 2003.
Currently, he works with ENKAT (a division of
brane. Given this difficulties in finding parameter values, the Hydrogenics Corporation), Gelsenkirchen, Germany.
presented curves give a good contribution to the understanding
and description of the internal fuel cell behavior. Stéphane Raël received the M.S. degree in electrical
engineering from the Ecole Nationale Supérieure
des Ingénieurs Electriciens de Grenoble, Grenoble,
REFERENCES France in 1992, and the Ph.D. degree from the
[1] J. Larminie and A. Dicks, Fuel Cell Systems Explained. New York: National Polytechnic Institute of Grenoble, in 1996.
Wiley, 2000, pp. 69–81. Since 1998, he has been an Assistant Professor
[2] K. Dannenberg, P. Ekdunge, and G. Lindbergh, “Mathematical model of with the National Polytechnic Institute of Lor-
the PEMFC,” J. Appl. Electrochem., vol. 12, pp. 1377–1387, Dec. 2000. raine, Vandoeuvre-les-Nancy, France. His research
[3] A. Rowe and X. Li, “Mathematical modeling of proton exchange mem- activities concern power electronic components,
brane fuel cells,” J. Power Sources, vol. 102, pp. 82–96, Dec. 2001. supercapacitors, and fuel cells.
[4] T. E. Springer, T. A. Zawodzinski, and S. Gottesfeld, “Polymer elec-
trolyte fuel cell model,” J. Electrochem. Soc., vol. 138, pp. 2334–2342,
1991. Bernard Davat (M’90) received the M.S. degree
[5] D. M. Bernardi and M. W. Verbrugge, “A mathematical model of the in electrical engineering from the Ecole Nationale
solid-polymer-electrolyte fuel cell,” J. Electrochem. Soc., vol. 139, pp.
Supérieure d’Electrotechnique, d’Electronique et
2477–2491, 1992.
d’Hydraulique de Toulouse, Toulouse, France, in
[6] T. F. Fuller and J. Newman, “Water and thermal management in
solid-polymer-electrolyte fuel cells,” J. Electrochem. Soc., vol. 140, pp. 1975, and the Ph.D. and Dr.d’Etat degrees from the
1218–1225, 1993. National Polytechnic Institute of Toulouse, in 1979
[7] M. Wöhr, Instationäres, Thermodynamisches Verhalten der Polymer- and 1984, respectively.
membran-Brennstoffzelle. Düsseldorf, Germany: VDI-Verlag, 2000. From 1980 to 1988, he was a full-time Researcher
[8] M. Wöhr, K. Bolwin, W. Schnurnberger, M. Fischer, W. Neubrand, and at the Centre National de la Recherche Scientifique,
G. Eigenberger, “Dynamic modeling and simulation of a polymer mem- Toulouse. Since 1988, he has been a Professor with
brane fuel cell including mass transfer limitation,” Int. J. Hydrogen En- the National Polytechnic Institute of Lorraine, Van-
ergy, vol. 23, pp. 213–218, 1998. doeuvre-les-Nancy, France. His research interests include static converters and
[9] W. Neubrand, Modellbildung und Simulation von Elektromembranver- their interactions with electrical machines and new electrical devices (fuel cells,
fahren. Berlin, Germany: Logos, 1999. supercapacitors).

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on July 20,2022 at 18:14:45 UTC from IEEE Xplore. Restrictions apply.