Proceedings of 2011 International Conference on
Modelling, Identification and Control, Shanghai, China, June 26-29, 2011
Linearized State Space Modeling of PEMFC for
Control Applications
Engr. M. Abbas Abbasi
abbas.abbasi@iub.edu.pk
AbstractThis paper focuses on the linearization study of the
non-linear dynamic models of Polymer Electrolyte Membrane
Fuel Cells (PEMFC). Matlab and Control Systems Toolbox have
been employed to study the model. The FC stack voltage is
dependent upon the current drawn from the cell, the change in
partial pressures and the change in the amount of inlet flow of
water vapor, hydrogen and air. A sudden change in the state
space model inputs is applied to the developed model and the
systems step response is studied. The partial pressures of H2,
O2 and H2O are selected as the state variables of the system.
Index Terms Fuel Cell, Modeling, Matlab, Control,
PEMFC, State Space, Linearization
I. INTRODUCTION
F uel cells convert chemical energy to electrical energy
based on various operating principles without producing
the byproducts typically associated with conventional power
generation methods. They are environment friendly and have
a great potential for higher efficiencies. There are numerous
types of fuel cells including Proton Exchange/Polymer
Electrolyte Membrane Fuel Cells (PEMFC), Alkaline Fuel
Cells (AFC), Direct Methanol Fuel Cells (DMFC) etc.
differing from one another by the type of fuel they use and the
operating temperature [1].
PEMFC use hydrogen as fuel and operate at low
temperatures. The energy conversion phenomenon associated
with fuel cells is very complex and needs thermodynamic as
well as chemical reaction equations to be expressed exactly.
Extensive work is being done in this field to model the fuel
cells in different perspectives and under different operating
conditions. Although numerous state space and other PEMFC
models are available, however they are too complex to be
understood and be utilized for effective control applications.
-214-
1
Engr. M. Bilal Shahid
bilal.shahid@iub.edu.pk
The fuel cell dynamics are nonlinear in nature and a large
signal state space model is very difficult to establish. However
it is very easy to linearize these models for small changes in
the state variables [2].
Such approach has been adopted and a small signal linearized
model is developed for a general PEMFC. Later this model is
converted to time domain by taking the inverse Laplace
transform of the General State Space Matrix G(s).
The model is attractive from the control point of view and can
easily be extended for other applications.
II. SYSTEM DESCRIPTION AND MODEL DEVELOPMENT
A. PEMFC Operation
Fig. 1 shows a simplified schematic of the operation of
PEMFC.
Product Gases + Water
Hydrogen as Fuel
+Unused Hydrogen
Anode
Electrolyte Load
Cathode
Product Gases + Water Air
+ Unused air
Fig. 1 PEMFC operation
The figure shows that pressurized hydrogen (fuel tank) and
oxygen taken from the air are fed to the fuel cell from anode
and cathode sides respectively. Reformed methane and other
fuels like natural gas can also be fed instead of hydrogen. The
partial pressures of both hydrogen and oxygen need to be
regulated. The quality of fuel as well as air is also of crucial
importance. Higher stack voltage can be produced by
connecting the various cells in series. In the presence of the
electrolyte catalyst, chemical reaction takes place between
 2

hydrogen and oxygen and as a result water and electricity is Ract Rcon
produced which can be summarized as [3]:
RFC

Anode : H2 2H 2e
2e
IFC +
1
Cathode : O2 2H 2e
2e H 2O Electrical Energy Heat Ccon
2
(1) Vrev VFC

B. Stack Voltage
The voltage of the fuel cell can be calculated using the -
relation [4]:
Fig. 2 Equivalent circuit model of the fuel cell voltages
VFC Vrev Vlosses
l (2)
Typical FC V-I Characteristics Curve
Vlosses Vact Vohm Vcon
c (3) 1

Where, 0.9
Vrev: Thermodynamic or reversible voltage. Also known
as Nernst Potential 0.8
VFC: Fuel Cell Stack voltage 0.7
Vlosses: Voltage losses in the fuel cell Operating Point
Vact: Activation losses due to the rate of chemical 0.6

FC voltage
reactions on electrodes surface
0.5
Vohm: Electrical resistance by the electrolyte to the flow of
electrons 0.4
Vcon: Voltage loss due to concentration reduction
0.3

The above voltages can be calculated from the following 0.2

formulae:
0.1

0
RT pH 2 pO2 0 100 200 300 400 500 600 700 800 900 1000
Vrev N Vooc ln (4) FC Current density (A/m2)
2F
2F pH 2O _ C Fig. 3 Typical FC V-I curve

NRT J Jn
Vact ln (5)
2 F Jo C. Model development
Vohm NiRFC (6) To develop the small signal state space model, the
F
perturbations in hydrogen, oxygen and water vapor pressures
di
Vcon Nce (7) ( pH 2 , pO2 , pH 2O _ C ) were selected as the state
variables of the system and small variations in the inlet flow
Where,
N: Stack cells rates of fuel injected ( FH 2 ) , air ( FO2 ) , water (
R: Universal gas constant FH 2O _ C ) and the load current ( iL ) were selected as the
T: Cell temperature
F: Faradays Constant inputs. The fuel cell voltage change ( VFFC ) was selected as
pH2, pO2, pH2O_C: partial pressures of hydrogen, oxygen and the only output of the linearized system.
water vapor from cathode side
The linearized state equations can be written as:
J: Load current density
Jo: exchange current density x A x B u
Jn: internal current density
RFC: Cell area specific resistance y C x D u
c & d: constants
pH 2
d
The equivalent circuit model [5] of the fuel cell voltages is x x , x pO2
shown in fig. 2. And fuel cell V-I characteristic curve [6] is dx
d
shown in fig. 3. pH 2 O _ C

-215-
 3

FH 2 NRT NRT NRT

C11 , C12 , C13
1
FO2 2FpH2 4FpO2 2Fp
F H 2O _ C
u , y VFC
F (8) and
FH 2O _ C
a b
iL D14 N RFC
F
J Jn J1 J Jn
While the matrices are given as [7]:
(10)
The entries defined in (10) involve various fuel cell
A1111 0 0 parameters summarized in table 1.
A 0 A22
2 0 , C C11 C12 C13
1
Table 1
0 0 A333
Fuel Cell Parameters used in the model
B11 0 0 B114
B 0 B22 B23 B24 , D 0 0 0 D114 Symbol Parameter

0 B32 B33 B334 Fan Total inlet flow into the anode (L/min)
i Operating current (A)
(9) Fcath Total inlet flow into the cathode (L/min)
And the matrices entries are defined below [8]:
kan
kan van po Anode energy constant . po is
operating pressure of fuel cell (J)
Ni Ni kcath
kcath vcath po cathode energy constant . po is
RT Fan RT Fcathh
2F 4F operating pressure of fuel cell (J)
A11 , A22
2 van 3
kan kcath Volume of anode (cm )
vcath
Ni Volume of cathode (cm3)
RT Fcathh
2F
2F 1 pH 2 XH2
Anode inlet hydrogen mole fraction
A33 , B11 RT XO2
kcath van kan X H 2 Cathode inlet oxygen mole fraction
XH2O_C
NaFC pH 2 Cathode inlet water vapor mole fraction
B14 , aFC
2 Fkan Fuel Cell active area (cm2)

pH2 Difference between fuel pressure and fuel cell operating

1 pO2 pressure (po)
B22 RT
vcathth kcathth ( X O2 X H 2O _ C ) pO2 Difference between oxygen pressure and fuel cell
operating pressure (po)
pO2
B23 RT pH2O_C Difference between water vapor pressure and fuel cell
kcathth ( X O2 X H 2O _ C ) operating pressure (po)

NaFC pO2 Matlabs Symbolic Math toolbox was used to generate the
B24 ,
4 Fkcath Y s
transfer matrix G s which equals:
U s
pH 2O _ C
B32 RT 1
kcathth ( X O2 X H 2O _ C ) G s C sI A B D.
By using this approach, the following four transfer
1 pH 2O _ C functions were obtained.
B33 RT
vcathth kcathth ( X O2 X H 2O _ C )
Y s B11C111 A11
1 t
NaFC pH 2O _ C B11C11
11e
B34 U1 s s A111
2 Fkcath
Y s B22C12
1 B32C13
1
B22C12
12 e
A22
2 t
B32C13e A333t
U2 s s A22 s A33
3

-216-
 4

Y s B23C12 B33C13
1 The response to a step perturbation in hydrogen inlet flow
B23C12
12 e
A22
2 t
B33C13e A333t 8
From: In(1)

U3 s s A22 s A333
Y s B14C11 B24C12
1 B34C13
1
7

D14
1
U4 s s A11 s A33 s A333 6

Change in stack voltage

5

A11t A22
2 t A33
3 t
D14 t B14C11e B24C12e B34C13e 4

These transfer functions basically predict the change in the 3

fuel cell output voltage when any of the controlling parameter

2
is varied while other perturbations are kept zero.
1

III. SIMULATION AND RESULTS 0

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time (sec)
Matlab and Control systems toolbox were used to investigate
Fig. 4 The response to a perturbation in fuel inlet flow
the proposed model. The parameter values used in the study
are enlisted in Table 2. The plot shows that a small perturbation in fuel flow will
cause a larger change in the stack output voltage
All the parameters were defined in the Matlab and matrices
A, B, C and D were formed. Using these matrices the
behavior of the model was studied under different input The response to a step perturbation in air inlet flow
scenarios. 2
From: In(2)

1.8
Table 2
Fuel Cell Parameter Values used in Simulation 1.6

1.4
Change in stack voltage

Parameter Value 1.2

N 6 1
F 96485 C/mol
XH2 0.91 0.8
XO2 0.189
Hydrogen inlet Flow 4.2 L/min 0.6

Air inlet flow 13.5 L/min

Cell Active area (aFC) 198 cm2 0.4

Load Current 45 A
Anode Volume 10.391 cm3 0.2

Cathode Volume 19.48 cm3

0
Operating Pressure 1 atm. 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

Operating Temperature 72C Time (sec)

RFC 0.00032
Fig. 5 The response to a perturbation in air inlet flow
a(Tafel Line Constant) 0.07 V
b (Concentration losses constant) 0.06 V When all other parameters are kept constant and the air inlet
flow is disturbed, the output voltage is also increased

Fig. 4 shows that when inlet fuel is perturbed it will cause a

similar change in the FC stack voltage. Fig. 5 shows the
response to perturbations in air flow. But these perturbations
do not affect the output as much as the fuel inelt flow
perturbations.
Similarly fig. 6 and fig 7 show the PEMFC voltage behavior
with respect the perturbations in water vapor inlet flow and
load variations. The plot shown in fig. 7 is the most important
one as it predicts the stack output voltage changes when load
of the PEMFC is varied.

-217-
 5

The response to a step perturbation in water vapor inlet flow

If we compare the model suggested here with the previous
0
From: In(3)
work, then we will find that the proposed model is simpler
and effective for small signal applications. Some of the
-0.5 models are suggested in [5] and [8]. These linearized models
involve lot of parameters and take into account the
-1 thermodynamic processes in the fuel cells which create
Change in stack voltage

complexities for common control applications. The suggested

-1.5 model is not only valid for linear operation but can also be
extended to multivariable control scenarios.
-2

-2.5

-3
V. REFERNCES

-3.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.
Time (sec)
[1] OHayre, R., S. W. Cha, W. Colella, and F.B. Prinz. Fuel Cell
Fig. 6 Step response to the variations in water vapor flow
Fundamentals. 2006. New York: John Wiley & Sons.
[2] Woonki Na and Bei Gou, Feedback Linearization based nonlinear
The transient response is of prime importance in the figures control for PEM fuel cells, IEEE Transactions on Energy Conversion,
above. The perturbations were settled down in different times. 23(1), 179-190, 2008.
[3] National Energy Technology Laboratory, Fuel Cell Handbook, 6th
For the given conditions settling times for the four input edition, Morgantown, WV, 2002.
disturbances were noted as 1.21s, 0.92s, 0.915s and 1.13s. [4] R. F. Mann, J. C. Amphlett, MA.I. Hooper, H. M.. Jansen, B.A. Peppley,
P. R. Roberge. Development and application of a generalized steady
state electrochemical model for a PEM fuel cell. Journal of Power
The response to a step perturbation in water load current Sources 2000, 86 (1-2), 173-180.
From: In(4)
-2.25 [5] C. Wang, M. H. Nehrir, and S. R. Shaw, Dynamic models and model
validation for PEM fuel cells using electrical circuits, IEEE Trans.
-2.26
Energy Conversion, vol. 20, no. 2 pp. 442, Jun, 2005.
[6] Fuel Cell Handbook, 5th ed. EGG Services Parsons Inc., U. S.
Department of Energy, Oct. 2000.
-2.27 [7] R. Methekar, V. Prasad, R. Gudi, Dynamic analysis and linear control
Change in stack voltage

strategies for proton exchange membrane fuel cell using a distributed

-2.28
parameter model, Journal of Power Sources 165 (2007) 152-170
[8] J. J. Baschuk and X. Li, Modeling of Polymer Electrolyte membrane
fuel cells with variable degrees of water flooding, J. Power Sources,
-2.29 vol 86, pp. 181-196, 2000

-2.3

-2.31

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.

Time (sec)

Fig. 7 The response to variations in the load current

IV. CONCLUSION
The model is suitable for the control applications and the
analysis can easily be extended to the combined effect of
multiple perturbations occurring in the fuel cell at the same
time. The study shows the importance of the simplicity of the
model and its usefulness in the control scenario. The model
helps for the design and optimization strategies of the control
schemes. It can also be used in the integrated systems and to
predict the systems response under different frameworks.
The simulations also indicate the accurate and logical
behavior of the PEMFC for the step inputs. It can also be
tested under various input perturbations.
The similar techniques can also be applied to analyze other
types of fuel cells.
-218-