You are on page 1of 18

World Applied Programming, Vol (3), Issue (7), July 2013.

264-281
ISSN: 2222-2510
©2013 WAP journal. www.tijournals.com

264
An Intelligent, Fast and Robust Maximum Power Point
Tracking for Proton Exchange Membrane Fuel Cell

Iman Soltani Mohammad Sarvi Haniyeh Marefatjou
Faculty of Technical & Engineering
Imam Khomeini International
University
Iran
Faculty of Technical & Engineering
Imam Khomeini International
University
Iran
Faculty of Technical & Engineering
Imam Khomeini International
University
Iran
i_soltani@ikiu.ac.ir sarvi@ikiu.ac.ir h.marefat@ikiu.ac.ir

Abstract: Abstract of paper goes here. In this section, author is supposed to include three important parts. In the
first part, it talks about the purposes of writing the paper. In the second part, methodologies that are been used
in paper, should be talked. Finally, in the third part, a brief overview of results should be presented. The length
of abstracts is 1/20 of whole papers. Pay attention that it shouldn't be extended from300 words. (Font: Times
New Roman 10pt) Ability of fuel cell systems to produce power is limited so it is necessary to force the system
to operate in conditions which match up with fuel cell (FC) maximumpower point (MPP). MPP of FC is
changed with variation of its inputs and load. Therefore a MPP tracker which utilizes a maximum power point
tracking (MPPT) algorithmis necessary for tracking of MPP. This paper presents a particle swarmoptimization
(PSO) based MPPT algorithm for tracking of Proton Exchange Membrane fuel cell MPP (PEMFC) system. In
this paper, many types of PSO include simple or conventional PSO, Improved PSO (IPSO), Quantuminspired
PSO (QPSO), Vector Evaluated PSO (VEPSO) and Particle Swarm Optimization with Time Varying
Acceleration Coefficients (PSOTVAC) methods are used to track MPP of FC. The MPPT algorithmdetermine
the duty cycle of the DC/DC boost converter in order to achieve MPPT. The results of the proposed methods
are compared with the conventional perturbation and observation (P&O) method. The results show that the
proposed methods have better characteristic and performance (fast response and high accurate) in comparison
with P&O. Also among of the proposed PSO based algorithms, conventional PSO has the fast response where
PSOTVAC has the more accurate than other analyzed methods.

Keywords: Particle SwarmOptimization (PSO), Fuel Cell, MaximumPower Point Tracking, Energy storage

I. INTRODUCTION
Fuel cells (FCs) are static electric power sources that convert the chemical energy of fuel directly into the electrical
energy. FCs have advantages such as high efficiency, zero or low emission (of pollutant gases), and flexible modular
structure [1, 2]. A large number of internal parameters can impact on produced voltage of FC [2-5], but in any condition,
there is just one unique point on V-I curve which represents MPP. In this point, FC produces its maximumpower. Owing
to the limited ability of FC systems to produce power fromavailable fuel flow, it is necessary to force the FC to operate
at MPP. This can avoid excessive fuel consumption and low efficiency operation. A maximum power point tracking
(MPPT) tracks the MPP of FC using a MPPT algorithm. The MPPT algorithmdetermines duty cycle of a DC-DC
converter and a certain amount of current which corresponds to MPP is extracted fromFC. There are several methods to
search MPP of optimumvalue of a function [6-9,10].

A good study about different MPPT methods such as Hill-climbing/ Perturb and Observe (P&O), incremental
conductance, fractional open-circuit voltage, fractional short-circuit current, fuzzy logic control, neural network, ripple
correlation control , current sweep, DC-Link capacitor droop control, load current or load voltage maximization, sliding
mode control approach and other MPPT techniques for photovoltaic system may be found in [11]. Among them
Perturbation and Observation (P&O) [6] is the most commonly used method because of its simple algorithm. MPPT
methods vary in complexity, implementation hardware, popularity, convergence speed and sensed parameters [12].
Many MPPT methods have been applied to fuel cell for exacting maximum available powers fromfuel cell modules,
e.g., P&O [13-17], adaptive MPPT control [18], Moto compressor control technique [19], adaptive fuzzy logic controller
[20], MPPT algorithmbased on resistance matching between the direct methanol fuel cells internal resistance and the
tracker’s input resistance [21], voltage and current based MPPT [22], adaptive extremumseeking control [23]. The
particle swarmoptimization (PSO) method has a simple structure that can be effectively used for MPP tracking. Hence
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

265
several authors have attempted apply it; for example, in [24], PSO is applied in the constant bus voltage application.
Despite its effectiveness, the algorithmdeveloped by the authors cannot guarantee the tracking of GP for all different
conditions. In another work [25], an Adaptive Perceptive Particle SwarmOptimization (APPSO) is proposed for the
same constant bus voltage application. However, due to additional dimensional search space, the number of particles
needs to be increased compared to its original counterpart [24]. The efficiency of a fuel cell as power generating system
can be significantly improved by using optimumoperating conditions.

In this paper two areas of significant interest are: first, the optimization of the electrochemical process, characterized by
some parameters with unknown values, which must be precisely determined in order to obtain accurate simulated results;
and second, power electronic converters requires an adequate control in order to know the load applied to the fuel cell
also a fuel cell's output power have nonlinear behavior with variations voltage and current. The MPP is changed with
changing in behavior. In this paper, many types of PSO include simple or conventional PSO, Improved PSO (IPSO),
Quantuminspired PSO (QPSO), Vector Evaluated PSO (VEPSO) and Particle Swarm Optimization with Time Varying
Acceleration Coefficients (PSOTVAC) methods are used to track MPP of FC. In order to perform the accuracy and
performance of the proposed method, the well-known P&O algorithmresults are compared with the proposed MPPT
methods. Also this paper presents a detailed study of the MPPT controller to insure a high proton exchange membrane
fuel cell (PEMFC) systemperformance which can be selected for practical implementation issue. A simulation work
dealing with MPPT controller, a DC/DC boost converter feeding a resistive load is achieved. Significant extracted results
are given to verify the validity of the proposed various types of Particle SwarmOptimization scheme. In these works
different types of Particle SwarmOptimization for MPPT in PEMFC, is presented and analyzed. The rest of this paper is
organized as follows: In section 2, modeling of proton exchange membrane fuel cell is introduced. MPPT concept is
outlined in section 3. Section 4, presents a brief introduction of PSO algorithm. The proposed MPPT methods are
presented in section 5. The results are presented and discussed in section 6.

I.1. Modeling of Proton Exchange Membrane Fuel Cell
The proportional relationship between the flow of gas through a valve and the partial pressure can be stated as following:
[6]

2
2 2
2
H
H
an
H
K
M
K
P
qH
= =

(1)

and
2
2 2
2
O
O
an
O
K
M
K
P
qO
= =

(2)

Where qH
2
is molar flow of hydrogen (kmol S
-1
), PH
2
is hydrogen partial pressure (atm), is KH
2
hydrogen valve molar
constant ( ) ( )
1 ÷
atmS kmol , K
an
is anode valve constant ( )
|
|
.
|

\
|
÷1
atmS kg kmol , MH
2
is molar mass of
hydrogen |
.
|

\
|
÷1
kmol kg .
For hydrogen, the derivative of the partial pressure can be calculated by using the following perfect gas equation:

2
2 2 2
( ) 3
in out r
H H H H
an
d RT
P q q q
dt V
= ÷ ÷


(3)

Where R is the universal gas constant ((1 mol)(kmol K
-1
)), T is absolute temperature (K), V
an
is anode volume (1),
2
in
H
q
is
hydrogen input flow (kmolS
-1
),
2
out
H
q
is hydrogen output flow (kmol S
-1
),
2
r
H
q
is hydrogen flow that reacts (kmol S
-1
). The
amount of hydrogen consumed in the reaction can be calculated fromthe following electrochemical principle:
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

266
2
2 4
2
r
H r
NI
q k I
F
= =


(4)

Where N is the number of the series wound fuel cells in the stack, I is the stack current (A), F is Faraday’s constant
(C
1
kmol
÷
), K
r
is model constant
1
( ( ) ) k m o l s A
÷
.

When the output flow is replaced by Eq.1 and the Laplace transform is applied to Eqs.3 and 4, hydrogen partial pressure
can be rewritten in the s domain as:

2
2
2
2
1
( 2 ) 5
1
H in
H H r
H
k
P q k I
s 
= ÷
+


(5)

Where
2
2
an
H
H
V
k RT
 =


(6)

By the same means, the equations for oxygen partial pressure can be derived as:

2
2
2
2
1
( ) 7
1
O in
O O r
O
k
P q k I
s 
= ÷
+


(7)
Where
2
2
ca
O
O
V
k RT
 =


(8)

Eq.5 describes the relationship between stack current and hydrogen partial pressure, and Eq.7 the shows relationship
between stack current and oxygen partial pressure. As the load draws current, the reactants–hydrogen and oxygen–
become depleted in the fuel cell stack, and both partial pressures drop accordingly.

To protect the fuel cell plant fromreactants starvation, commonly excessive amounts of hydrogen and oxygen are
( ) I k q I k q
r
in
H r
in
O
2 ,
2 2
  provided for the stack.

A higher excess ratio leads to higher partial pressures, and then a higher fuel cell voltage. However, too much excess
flow is a problemas it dries out the membrane and consumes much more parasitic power. [2-10] presented a high quality
study on how to control the excess ratio. They designed several airflow controllers to regulate the input flow rate so that
it is always twice as much as the reaction rate. To focus on the MPPT control problem, this paper will not go into details
on the excess ratio control issue but simply supply the fuel cell with a constant flow rate that is sufficient for
consumption [6].

I.2. Polarization curve model
The fuel cell voltage as a function of current density in a steady state can be represented by a polarization curve, which is
influenced by such parameters as the cell temperature, oxygen partial pressure, hydrogen partial pressure and membrane
water content. When current is drawn from a fuel cell, the cell voltage
Cell
V decreases from its equilibrium
thermodynamic potential
nernst
E (open circuit voltage). This voltage drop consists of activation loss
act
 , ohmic loss
ohmic
 and concentration loss
con
 .
The basic expression for the cell voltage is:

con ohmic act nernst Cell
E V    + + + =
(9)
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

267
Reversible thermodynamic potential
nernst
E is described by the Nernst equation .With literature values for the standard-
state entropy change, the expression is [6]:

( ) ( )
2 2
5 4
ln 5 . 0 ln 10 308 . 4 15 . 298 10 5 . 8 229 . 1
O H nernst
P P T T E + × + ÷ × ÷ =
÷ ÷

(10)

Activation overvoltage
act
 is described by the Tafel equation, which can be expressed as:

I Co T T
act
ln ln
4 2 3 2 1
     + + + =
(11)

where _(i =1–4) are parametric coefficients for each cell model. CO
2
is the concentration of dissolved oxygen at the
gas/liquid interface (mol cm
−3
), which can be calculated by means of

( ) ( )
T
P
C
O
O
498
exp 10 08 . 5
6
2
2
÷
× ×
=

(12)

Ohmic overvoltage η
ohmic
results fromthe resistance of the polymer membrane in electron and proton transfers. It can be
expressed as
m ohmic
IR ÷ = 
(13)

The ohmic resistance R
m
is given by

A
t r
R
m m
m
=

(14)

where rmis membrane resistivity (Ωcm) to proton conductivity, tmmembrane thickness (cm), A cell active area (cm
2
).
Membrane resistivity depends strongly on membrane humidity and temperature, and can be described by the following
empirical expression [6] .

( ) ( ) ( )
( ) | | ( ) | |
T
T
A
I
A
I T
A
I
r
m
m
303
18 . 4 exp 3 634 . 0
303
0062 . 0 03 . 0 1 6 . 181
5 . 2 2
÷ ÷ ÷
(
¸
(

¸

÷ + +
=




(15)

where λmis the membrane water content. The membrane water content is a function of the average water activity am:

¹
´
¦
s < ÷ +
s < + ÷ +
=
3 1 ), 1 ( 4 . 1 14
1 0 , 36 85 . 39 81 . 17 043 . 0
3 2
m m
m m m m
m
a a
a a a a


(16)

The average water activity is related to the anode water vapor partial pressure Pv,an and the cathode water vapor partial
pressure
Pv,ca:
( )
sat
ca v an v
ca an m
P
P P
a a a
, ,
2
1
2
1
+
= + =

(17)

The saturation pressure of water Psat can be figured out with the following empirical expression:

3 7 2 5
10
10 4454 . 1 10 1813 . 9 02953 . 0 1794 . 2 log T T T P
sat
÷ ÷
× + × ÷ + ÷ =
(18)

Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

268
The value of λmvaries between 0 and 14, which is equivalent to the relative humidity of 0% and 100%. Under
supersaturated conditions, however, the maximumpossible value of λm can be as high as 23. In addition, λmcan also be
influenced by the membrane preparation procedure, the relative humidity of the feed gas, the stoichiometric ratio of the
feed gas, and the age of the membrane [6]. Hence, in this paper, λmis considered as an adjustable parameter with a
possible value between 0 and 23. Concentration overvoltage ηcon results fromthe concentration gradient of reactants as
they are consumed in the reaction. The equation for concentration overvoltage is shown by [6]

|
|
.
|

\
|
÷ =
A i
I
nF
RT
l
con
1 ln 

(19)

where iL is the limiting current. It denotes the maximumrate at which a reactant can be supplied to an electrode.

Parameters of a typical fuel cell are shown in Table 1. These parameters are used for analysis and simulation in this
paper. Fig.1 shows PEM fuel cell simulation model.

Table 1. Parameters of FC model [11]

Value Parameter
96484600 F(Ckmol
-1
)
8314.47 R(jkmol
-1
K)
35 N
232 A(cm
2
)
9.07×10
-8
K
r
=N/4F
4.22×10
-5
K
H2
(kmols
-1
atm)
2.11×10
-5
Ko
2
(kmols
-1
atm)
3.37 τH
2
(s)
6.47 τO
2
(s)
0.0178 T
mem
(cm)
-0.944 ζ
1

0.00354 ζ
2

7.8×10
-8
ζ
3

-1.96×10
-4
ζ
4

2.0 I
L
(Acm
-2
)


Polarization curve
model
I
P
H2
P
O2
T
cell
V
cell

Figure 1. PEM fuel cell block diagram.

II. FUEL CELL MAXIMUM POWER POINT TRACKING CONCEPT
A maximumpower point tracker, tracks the maximumpower point of FC using a MPPT algorithm. In the literature,
there are many maximumpower point tracking (MPPT) algorithms. Perturb and Observe (P&O) method is one of these
methods [12].
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

269
When a fuel cell is directly connected to an external load, its output power depends on both the internal electro-chemical
reaction and the external load impedance. The system’s operating point is at the intersection of the fuel cell’s I–P curve
and the load line. There exists a unique operating point, called the maximumpower point (MPP), at which the fuel cell
produces its maximumpower, as illustrated in Fig. 2. According to the power transfer theory, the power delivered to the
load is maximized when the fuel cell internal impedance equals the load impedance.



Figure 2. Typical fuel cell polarization and power curves.

Block diagramof the fuel cell MPPT systemis shown in Fig.3. This systemincludes of a fuel cell, a DC/DC converter,
a MPPT controller and load. With respect to dc–dc converter topologies, the boost converter is considered because of its
simplicity, low cost and high efficiency as well as increasing of the load voltage. The terminal voltage of fuel cell is
controlled by varying of DC/DC converter duty cycle. The MPPT controller determines a control signal D (duty cycle of
DC/DC converter) that maximumpower is obtained fromFC in normal condition and in the presence of disturbances
such as load and system parameter variations.
Figure 3. Block diagramof the fuel cell MPPT system.


III. PSO ALGORITHM PRINCIPLE
Kennedy and Eberhart [26], developed the PSO algorithmbased on the behavior of swarms in the nature, such as birds,
fish, and etc. The PSO has particles driven from natural swarms with communications based on evolutionary
computations. PSO combines self-experiences with social experiences. In this algorithm, a candidate solution is
presented as a particle. It uses a collection of flying particles (changing solutions) in a search area (current and possible
solutions) as well as the movement towards a promising area in order to get to a global optimum.
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

270
best best
i
j
i
j
g r c p r c wv v
j
2 2 1 1
1
+ + =
+
(20)
1 1 + +
+ =
i
j
i
j
i
j
v s s
(21)

Where, C
1
and C
2
are the balance factors between the effect of self-knowledge and social knowledge in moving the
particle towards the target. Usually the value 2 is suggested for both factors in the literature, rand is a random number
between 0 and 1, and different at each iteration, w is inertia weight,
best
p is the best position of a particle,
best
g is the
best position within the swarm,
i
j
v is the velocity of j
th
particle in i
th
iteration, and
i
j
S is the position of j
th
particle in i
th

iteration. The typical movement of particles in the optimization process is shown in Fig. 4.



Figure 4. Movement of particle in the optimization process.

Since, in the above algorithm, there is the possibility of particles movement to out of the problemspace [27], an upper
velocity bound for particle movement is specified. Flowchart of PSO algorithmis shown in Fig.5. One of the PSO
problems is its tendency to a fast and premature convergence in mid optimumpoints. A lot of effort has been made so far
to solve this problem. For instance, in [15] the best value for w in Eq. 20 is set to 0.9.

A passive swarmmodel can be added to the PSO in order to increase its efficiency, which is the PSO with passive
congregation PSOPC. This termis the same as randomly selected particle from the current swarmand particle [16]. In
PSOPC, grouping has two types: 1) Aggregation: which itself has two kinds. The first kind is passive aggregation in
which a passive group is with a physical process; like planktons swarm floating on the water that the water flow keeps
them together. The second kind is active aggregation in which the aggregation is performed by an absorbent source. The
absorbent source may be food or water. 2) Congregation: This is different fromaggregation. That is, the absorbent
supply and not external and physical factors is the group forced by it. This type is also divided in two kinds: 1) Passive
congregation in which there is an attraction from one particle to others but is not shown a social behavior. 2) Social
congregation in which there is a social behavior among the particles and they are strongly related to each other. Since in
some groups there may be a selfish behavior in information sharing, e.g. fish school, a selfish behavior may lead in
forming a passive group. Improved Particle swarmoptimizer (IPSO) is based on PSOPC. Moreover, it uses a harmony
search. It utilize a mechanismcall fly-back in order to employ the constraints [17]. Quantum-inspired version of the PSO
(QPSO) algorithmpermits all particles to have a quantumbehavior instead of the classical Newtonian dynamics assumed
so far in all versions of the PSO. Hence, instead of the Newtonian randomwalk, some sort of “quantummotion” is
imposed in the search process. When the QPSO is tested against a set of benchmarking functions, it demonstrates
superior performance as compared to the classical PSO but under the condition of large population sizes. One of the most
attractive features of the QPSO algorithmis the reduced number of control parameters. Strictly speaking, there is only
one parameter required to be tuned in the QPSO [18].Vector Evaluated PSO (VEPSO) is a novel algorithmbased on a
multi-objective interaction sort of the PSO [19].VEPSO is a co evolutionary multi-objective variant of the popular PSO.

Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

271


Figure 5. Flow chart of particle swarm optimization.

IV. PSO-TVAC CONCEPT
In the PSO, proper control of the two stochastic acceleration components: the cognitive component (c1) which
corresponds to the personal thinking of each particle and the social component (c2) which describes the collaborative
effect of the particles, to obtain the global optimal solution is very important accurately and successfully. It should be
noted that it is desirable that for cheering the particles to wander through the entire search space, without clustering
around local optima during the early stages of the swarm-based optimization. On the other hand, in order to find the
optimal solution effectively it is very important to enhancement convergence toward the global optima during the latter
processes [35]. Thus, a novel parameter automation strategy for the PSO concept called PSO with time varying
acceleration coefficients is proposed, in this paper. The motivation for using this method is enhancement the global
search in the early stage of the optimization stages and cheering the particles to converge toward the global optima at the
end of it.

All parameters including inertiaweight and acceleration coefficients are varied with time (iterations) in Equation (20).
Thus, in the PSO-TVAC the velocity is updated as follows:

{ ( ) ( )
( )
( ) }
i
j
i
j
i i f
i
j
i
j i i f
i
j
i
j
S Gbset rand
c
iter
iter
c c
S Pbest rand c
iter
iter
c c v w C v
÷ × ×
(
¸
(

¸

+ ÷ +
÷ × ×
(
¸
(

¸

+ ÷ + × =
+
()
()
2
max
2 2
1
max
1 1
1


(22)

( )
min
max
max
min max
. w
iter
iter w
w w w +
÷
÷ =

(23)

2 . 4 1 . 4 ,
4 2
2
2
s s
÷ ÷ ÷
= 
  
where C

(24)
Start
Initialize particles with random
position and velocity vectors
For each particle`s position (P)
evaluate fitness
If fitness(p) than fitness(Pbest) than
Pbest=p
Set best of Pbest as Gbest
Update particles velocity and position

Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

272
where, C is constriction factor, c1i, c1f and c2i, c2f are initial and final values of c1 and c2, respectively. Under this
situation, the inertia weight is linearly decreasing as time grows and by changing the acceleration coefficients with time
the cognitive component is reduced and the social component is increased [36]. The large and small value for cognitive
and social component at the optimization process starting is permitted the particles to move around the search space,
instead of moving toward the population best. In contrast, using a small and large cognitive and social component,
respectively the particles are permitted to converge toward the global optima in the latter part of the optimization. Thus,
PSO-TVAC is easier to understand and implement and its parameters have more straightforward effects on the
optimization performance in comparison with classic PSO. Using the above concepts, the whole PSO-TVAC algorithm
can be described as follows:

- For each particle, the position and velocity vectors will be randomly initialized with the same size as the
problemdimension within their allowable ranges.
- Evaluate the fitness of each particle (Pbest) and store the particle with the best fitness (Gbest) value.
- Update velocity and position vectors according to (21) and (22) for each particle.
- Repeat steps 2 and 3 until a termination criterion is satisfied. Comparing the classic PSO, PSO-TVAC has the
following advantages:
o Faster: PSO-TVAC can get the quality results in significantly fewer fitness evaluations and constraint
evaluations.
o Cheaper: There is need to adjust a few parameter settings for different problems than the PSO.

V. THE PROPOSED PSO BASED MAXIMUM POWER POINT TRACKING
In order to investigate the performance of proposed models, the performance of the proposed method is compared with
conventional tracking method. In the many conventional maximumpower point (MPP) tracking methods usually the
slope of P-V curve has been used. The slope is zero at the MPP, positive on the left of the MPP, and negative on the
right. Based on negative or positive slope, error has been defined to achieve zero slop. Fig 6 shows the P-I curve of fuel
cell in per unit and the positive or negative slope at the left or right side of MPP.



Figure 6. Power-Voltage characteristic fuel cell

The described PSO algorithmin preceding section is applied to realize the maximumpower point tracking control of a
FC system, where the P-V characteristic exhibits multiple local maximumpower points.

In this work, a PSO algorithmis applied to track the MPP using the Competitive technique. In order to start the
optimization, a solution vector of Voltage with k particles needs to be defined as follows in (25).

Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

273
The objective function is the generated power 'P', which is the summation of power generated by each cell. Assuming
that there is 'M' number of agents involved in the search process, the terminal voltage vector 's
k
' changes in the following
order and also computes the power 'P(s
k
)'at each stage.


(25)
...
... ...
1 1
2
1
1
3 2 1
÷ ÷ ÷ ÷
÷ ÷ ÷ ÷ ÷
+ + + k
M
k k
k
M
k k k
s s s
s s s s


This process is continued until the global optimumis reached and in reached iteration the position and velocities are
updated as per the relationships defined by Eqs. 21-22. In real time operation, the objective function 'F ' often changes
due to load or systemparameter variation. In this condition, the agents must be reinitialized to search for the new MPP
again, if the initialization process is not implemented for the change in operating point due to load variations, 'P
best
' and
'g
best
' cannot be updated automatically. As a result, the agents stop searching for new maximumpower point. The PSO
algorithmreinitializes the agents whenever the following two conditions are satisfied.


(26)


(27)
( ) ( )
( )
 (
÷
+
i
i i
s P
s P s P
1

size swarm
P P
F
i i
s s
obj
÷
=
+1


Equations 26 and 27 form the basis for convergence detection of agent and sudden change in input gas, respectively.

After finding the output of exact amount power and voltage of fuel cell can be calculated Duty Cycle. The MPP tracker
adjusts the operating point of the fuel cell to the maximumpower (
max
P ) by tuning of boost converter duty cycle.The
DC/DC boost converter transfer function is obtained by considering of its steady state operation as follows:

o
FC FC
V
i V
D
) (
1÷ =
(28)

With the optimal duty cycle, DC - DC converter is set and can reach the maximumpower point. Where d is the duty
cycle,
O
V is the DC/DC boost converter output voltage and
FC
V is the output Fuel cell voltage. The optimum duty
cycle (
opt
d ) is calculated fromEq.(28) with substitution of
mp FC
V V = The flowchart of the PSO MPPT algorithmis
shown in Fig.7.By changing the duty cycle of the boost converter appropriately we can match the source impedance
with that of the load impedance for achieving MPPT. The results are indicative of the out performance of the proposed
method. The main features of PSO based MPPT method can be summarized as:

1) MPPT with variation of temperature
2) MPPT with variation of inputs(Hydrogen gas, oxygen gas )
3) High accuracy
4) Fast response

The proposed systemis simulated in different conditions, and the results demonstrate appropriate performance of the
systemand controllers.

Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

274
Start
Initialization of particles
Evaluate fitness function(i.e FC out power) fi(t-1) and fi(t)
Finding of the pi,best:pi,best=MAX{fi(t-1),fi(t))}, i=1,2,...,N
Finding of thegbest:gbest=MAX{ pi,best i=1,2,...,N}
Updating of the velocities and position according (21) and (22)
Convergence critenia satisfied?
Output:Gbest Imp,fgbest Pmp
Vmp=Pmp/Imp
End
Yes
No


Figure 7. The flowchart of the PSO MPPT algorithm

VI. SIMULATIONS AND DISCUSSIONS
In order to investigate the performance of the proposed PSO based MPPT, a systemincludes a PEMFC, a DC/DC
converter, a resistive load and MPPT control are considered, analyzed and simulated in MATLAB/SIMULINK
environment (as shown in Fig.3 ). In order to obtain an operating point close to the MPP, the PEMFC’s systemusually
employs a DC/DC converter into the control system. Simulations are performed for three following cases:

Case 1: Constant temperature and λ
m

Case 2: Pulsed variation of λ
m

Case 3: Pulsed variation of fuel cell temperature


Case 1: Constant temperature and λ
In order to investigate the accuracy of the proposed MPPT algorithmand to compare its performance with P&O and
types of PSO MPPT method, the membrane water content (λ
m
) is considered equal to 14 and temperature is assumed
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

275
equal to 343°K. The FC output power, output current and output voltage are shown in Figs. 8, 9 and 10, respectively.
The results of conventional P&O and the proposed methods are shown in Fig.11. Comparisons of these methods are
presented in Table 2. The Fig.11 and Table 2 show that the proposed methods have better performance in comparison
with P&O. The all of proposed methods have lower rise time and settling time than the P&O method. Furthermore,
the proposed methods can track the peak power more accurate than P&O. Fig.12 shows the power-current (P-I)
characteristic of fuel cell in four different temperatures (310K, 330K, 350K, 360K). Fig.13 shows power-current (P-I)
characteristic of fuel cell in two different value of the membrane water content (λ
m
).

50 100 150 200 250 300 350 400 450 500 550 600
500
1000
1500
2000
TIME
P
O
W
E
R
(
w
)

TIME(ms)

Figure 8. Output Power of FC .

0 100 200 300 400 500 600
0
10
20
30
40
50
TIME
I
(
A
)

TIME(ms)

Figure 9. Output Current of FC.

50 100 150 200 250 300 350 400 450 500 550 600
39
40
41
42
43
TIME
V
(
v
)

TIME(ms)

Figure 10. Output Voltage of FC.



Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

276
0 100 200 300 400 500 600
1500
1550
1600
1650
1700
1750
1800
TIME
P
O
W
E
R
(
w
)

P&O
QPSO
IPSO
VEPSO
PSOTVAC
PSO

TIME(ms)




100 120 140 160 180 200 220 240 260
1620
1630
1640
1650
1660
1670
1680
1690
1700
1710
P
O
W
E
R
(
w
)


P&O
QPSO
IPSO
VEPSO
PSOTVAC
PSO

TIME(ms)


Figure 11. (a) Output power of FC with P&O and the proposed methods, (b) a zoomwindow of (a).


45 50 55 60 65 70 75 80 85 90 95
-1000
0
1000
2000
I(Current(A))
P
(
P
o
w
e
r
(
w
)
)


T=310k
T=330k
T=350k
T=360k


Figure 12. The power-current (P-I) characteristic of fuel cell in different temperatures.


(b)

Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

277
45 50 55 60 65 70 75 80 85 90 95
0
500
1000
1500
2000
2500
I(A)
P
(
w
)

Landa=16
Landa=12

Figure 13. The power-current (P-I) characteristic of Fuel cell in different
m
 .


Table 2. Comparison of P&O and application PSO results at T=343°K and 12 =  .

Applied Method PSOTVAC VEPSO IPSO QPSO PSO P&O Analytical
Average P
FC
value(w) 1719 1716 1711 1713 1710 1694 1742
Time to max(s) 3.45 3.56 3.54 3.34 3.23 21.6 -
Accuracy(%) 98.67 98.50 98.22 98.33 98.16 97.24 100


Case 2: Pulsed variation of λ
m

In this section, temperature of FC is constant but
m
 is changed. Fig.14 shows time variations of cell
m
 . Also Fig.15
shows the time evolution of P_FC under fast variation of the FC
m
 for the PSOTVAC method. Figs. 16, 17, 18, 19
and20 also show this case for PSO, VEPSO, P&O, QPSO, IPSO methods, respectively. These results show that the
VEPSO and QPSO method have more accurate and better response than P&O and conventional PSO methods and IPSO.


100 200 300 400 500 600

TIME(ms)
Figure 14. Variations of
m

.
100 150 200 250 300 350 400 450 500 550 600
1650
1700
1750
TIME
P
o
w
e
r
-
P
S
O
T
V
A
C

Figure 15. Time evolution of PEM_FC under fast
100 150 200 250 300 350 400 450 500 550 600
1650
1700
1750
TIME
P
o
w
e
r
-
P
S
O

Figure 16. Time evolution of PEM_FC under fast
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

278
variation of the fuel Cell
m

for the PSOTVAC
method.
100 150 200 250 300 350 400 450 500 550 600
1660
1680
1700
1720
TIME
P
o
w
e
r
-
V
E
P
S
O

Figure 17. The time evolution of PEM_FC under fast
variation of the Fuel Cell
m

for the VEPSO
method.

VARIATION OF THE FUEL CELL
m
 FOR THE PSO
METHOD.
500 1000 1500 2000 2500
1550
1600
1650
1700
1750
TIME
P
o
w
e
r
-
P
O

Figure 18. The time evolution of PEM_FC under fast
variation of the Fuel Cell
m

for the P&O method.

100 150 200 250 300 350 400 450 500 550 600 650
1650
1700
1750
TIME
P
o
w
e
r
-Q
P
S
O

Figure 19. The time evolution of PEM_FC under fast
variation of the Fuel Cell
m

for the QPSO method.

100 150 200 250 300 350 400 450 500 550 600
1640
1660
1680
1700
1720
TIME
P
o
w
e
r
-
I
P
S
O

Figure 20. The time evolution of PEM_FC under fast
variation of the Fuel Cell
m

for the IPSO method.




Case 3: Pulsed variation of temperature
In order to investigate the performance of the proposed method under variable the membrane water content (λ
m
), a
pulsed variation for T is considered (as shown in Fig.21). The simulation results are shown in Fig.22(a). This figure shows
the output power of FC for comparison of the conventional P&O method by the proposed methods. In Fig. 22(b) a zoom
window of Fig.22 (a) is presented. The proposed methods have lower rise time than the P&O method. Furthermore, the
proposed methods can track the peak power more accurate.

100 150 200 250 300 350 400 450 500 550 600
320
330
340
350
360
TIME
t
e
m
p
e
r
a
t
u
r
e
increase
decrease

TIME(ms)

Figure 21. Time variations of PEM FC temperature.

Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

279


0 100 200 300 400 500 600
500
1000
1500
2000
TIME
P
o
w
e
r
(
w
)


QPSO
IPSO
VEPSO
PSO
PSOTVAC

250 260 270 280 290 300 310 320 330 340
1620
1640
1660
1680
1700
1720
1740
TIME
P
o
w
e
r
(
w
)


QPSO
IPSO
VEPSO
PSO
PSOTVAC


Figure 22. (a) The time evolution of P_FC under fast variation of the Fuel Cell temperature in constant membrane water content
( ) 12 =  for both the proposed and the P&O methods; (b) a zoomwindow of (a).

VII. CONCLUSION
The main purpose of this paper is to present a PSO based MPPT controller for capturing of maximumpower fromfuel
cell. In this paper PSO, IPSO, VEPSO, PSOTVAC and QPSO based MPPT for fuel cell systems are presented and its
results are compared with conventional P&O strategy. A systemincludes a PEMFC, a DC/DC boost converter, a
resistive load and MPPT controller are considered, analyzed and simulated in MATLAB/SIMULINK environment.
Simulations are performed for several cases (constant condition and variation of λ
m
and temperature). The proposed
MPPT algorithms are accurate which have a fast dynamic response under rapid changing of the fuel cell temperature and
λ
m
. Furthermore PSOTVAC method show the more accurate among the studied methods and PSO method has the fast
response among the studied methods.









Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

280
NOMENCLATURE:
E
NERNST
NERNST VOLTAGE (V)
V
ACT

ACTIVATION VOLTAGE (V)
V
OHMIC

OHMIC VOLTAGE (V)
i

(I =1,2,3,4) PARAMETRIC COEFFICIENTS
V
CON

CONCENTRATION OVER VOLTAGE(V)
P
H2

HYDROGEN PRESSURE (PA)
P
O2

OXYGEN PRESSURE (PA)
T

TEMPERATURE(K)
C
O2
CONCENTRATION OF DISSOLVED OXYGEN(
3
. mol cm
÷
)
R
M

OHMIC RESISTANCE( O )
R
M

MEMBRANE RESISTIVITY( cm O )
A

CELL ACTIVE AREA (
2
cm )
T
M

MEMBRANE THICKNESS (cm )
TABLE I.
I
L
LIMITING CURRENT (A)
F

FARADAY CONSTANT, 96487CHARGE=MOL
m

MEMBRANE WATER CONTENT


REFERENCE

[1] Nehrir.H, Caisheng Wang, Shaw, S.R, 2006. “Fuel cells: promising devices for distributed generation”, IEEE Power and
Energy Magazine, Vol. 4, Issue 1, pp. 47-53.
[2] Caisheng Wang, Hashem Nehrir, Steven R. Shaw, 2005. “Dynamic Models and Model Validation for PEM Fuel Cells Using
Electrical Circuits”,IEEE Trans Energy Conversion, Vol. 20, No.2.
[3] J . Larminie and A. Dicks, 2001. “Fuel Cell Systems Explained” New York, Wiley.
[4] R.F.Mann, J .C. Amphlett, M.A.I. Hooper, H.M. J ensen, B.A. Peppley, P.R.Roberge, 2000. “Development and application of a
generalised steady-state electrochemical model for a PEM fuel cell”, J . Power Sources,Vol. 86, No.1–2, pp. 173–180.
[5] M.Y. El-Sharkh, A. Rahman, M.S. Alam, P.C. Byrne, A.A. Sakla, T.Thomas, 2004. “A dynamic model for a stand-alone PEM
fuel cell power plant for residential applications”, J . Power Sources, Vol. 138, No. 1–2, pp. 199–204.
[6] Zhong Zhi-dan, Huo Hai-bo, Zhu Xin-jian,Cao Guang-yi, Ren Yuan, 2008. “Adaptive maximum power point tracking control
of fuel cell power plants”, J . Power Sources, Vol. 176, pp. 259–269.
[7] O. Wasynczuck, 1983. “Dynamic Behavior of a Class of Photovoltaic Power Systems”, IEEE Trans. Apparatus and Systems,
Vol. PAS-102, No. 9, pp. 3031-3037.
[8] K. Hussein, I.Muta, T.Hoshino, and M. Osakada, 1995. “MaximumPhotovoltaic Power Tracking: An Algorithmfor Rapidly
Changing Atmospheric Conditions”, IEE Proc.-Generation, Transmission, Distribution, Vol. 142, No.1, pp. 59-64.
[9] T.Noguchi, S.Togashi, R.Nakamoto, 2002. “Short-Current Pulse-Based Maximum-Power-Point Tracking Method for Multiple
Photovoltaic and Converter Module System”, IEEE Trans. on Industrial Electronics, Vol. 49, pp. 217-223.
[10] M.Y. El-Sharkh, A. Rahman, M.S. Alam, P.C. Byrne, A.A. Sakla, T. Thomas,''A dynamic model for a stand-alone PEM fuel
cell power,” J ournal of Power Sources, vol. 138, pp.199–204. 2004.
[11] R.F. Mann, J .C. Amphlett, M.A.I. Hooper, H.M. J ensen, B.A. Peppley, P.R. Roberge, “Development and application of a
generalized steady-state electrochemical for a PEM fuel cell,” J ournal of Power Sources, vol. 86, pp.173–180, May 200.
[12] N. Femia, G. Petrone, G. Spagnuolo, and M.Vitelli,“Optimization of perturb and observe maximumpower point tracking
method,” IEEE Transaction on Power Electronics, vol. 20, pp. 963-973,2005.
Iman Soltani, et al. World Applied Programming, Vol (3), No (7), July 2013.

281
[13] L. N. Khanh, J . J . Seo, Y. S. Kim, and D. J . Won, “Power-management strategies for a grid-connected PV-FC hybrid system,”
IEEE Trans. Ind. Electron., vol. 25, no. 3, pp. 1874-1882, J ul. 2010.
[14] A. GIUSTINIANI, G. PETRONE, G. SPAGNUOLO, AND M. VITELLI," LOW-FREQUENCY CURRENT
OSCILLATIONS AND MAXIMUM POWER POINT TRACKING IN GRID-CONNECTED FUEL-CELL-BASED
SYSTEMS" IEEE TRANS. ENERG. CONV. 57:6 (2010) 2042-2053.
[15] C.A. Ramos-Paja, G. Spagnuolo, G. Petrone, R. Giral, A. Romero(2010). Fuel cell MPPT for fuel consumption optimization.
In: IEEE International Symposiumon Circuits and Systems (ISCAS 2010) Paris, France May 30 -J une 2 2010 IEEE Pag.2199-
2202 ISBN:9781424453085
[16] M. DARGAHI, M. REZANEJ AD, J . ROUHI AND M. SHAKERI," RECURSIVE ESTIMATION BASED MAXIMUM
POWER EXTRACTION TECHNIQUE FOR A FUEL CELL POWER SOURCE USED IN VEHICULAR APPLICATIONS"
IEEE INT. MULT. CONF. (2009) 33-37.
[17] L. EGIZIANO, A. GIUSTINIANI, G. PETRONE, G. SPAGNUOLO, M.VITELLI," OPTIMIZATION OF PERTURB AND
OBSERVE CONTROL OF GRID CONNECTED PEM FUEL CELLS" IEEE INT. CONF. CLEAN ELEC. POW. (2009)
775– 781.
[18] S. H. Hosseini, S. Danyali, F. Nejabatkhah, and S. A. KH. Mozafari Niapour, “Multi-input DC boost converter for grid
connected hybrid PV/FC/Battery power system,” in Proc. IEEE Electric Powe Conf., 2010, pp. 1-6.
[19] MPPT of a PEMFC based on air supply control of the motocompressor group M Becherif, D Hissel International J ournal of
Hydrogen Energy 35 (22), 12521-12530
[20] N. CHANASUT AND S. PREMRUDEEPREECHACHARN, " MAXIMUM POWER CONTROL OF GRID-CONNECTED
SOLID OXIDE FUEL CELL SYSTEM USING ADAPTIVE FUZZY LOGIC CONTROLLER"IEEE IND. APPL. SOC.
ANN. MEET. (2008)1–6.
[21] K. H. Loo, G. R. Zhu, Y. M. Lai, Chi K. Tse, “Development of a Maximum-Power-Point Tracking Algorithmfor Direct
Methanol Fuel Cell and Its Realization in a Fuel Cell/Supercapacitor Hybrid Energy System,” 8th International Conference on
Power Electronics–ECCE Asia, J eju, Korea, 30 May–3 J une 2011
[22] M. SARVI, M.M. BARATI, " VOLTAGE AND CURRENT BASED MPPT OF FUEL CELLS UNDER VARIABLE
TEMPERATURE CONDITIONS " IEEE POW. ENG. CONF. (2010) 1-4.
[23] N. Bizon,' Hybrid power source for vehicle applications operating at maximumpower point of fuel cell ''Appl. Energ. 87
(2010) 3115–3130.
[24] Miyatake M, Veerachary M, Toriumi F, Fujii N, Ko H. Maximumpower point tracking of multiple photovoltaic arrays: a
particle swarmoptimization approach. Aerosp Electron Syst IEEE Trans 2011;47:367–80.
[25] Roy Chowdhury S, Saha H. Maximumpower point tracking of partially shaded solar photovoltaic arrays. Sol Energy Mater
Sol Cell 2010;94:1441–7.
[26] Kennedy, J . and Eberhart, R.C., (1995), “Particle SwarmOptimization”, Proc. IEEE Int. Conf. on N.N., pp. 1942-1948.
[27] Eberhart, R. C., Simpson, P. K., and Dobbins, R. W., (1996) “Computational Intelligence PC tools”, 1st ed. ed. Boston, MA:
Academic press professional.
[28] Eberhart, R. C., and Shi, Y. (2000). "Comparing inertia weights and constriction factors in particle swarmoptimization". Proc.
Cong. onEvolutionary Computation, pp 84-88.
[29] El-Abd, M., (2008), "Cooperative Models of Particle SwarmOptimizers", A thesis presented to the University of Waterloo in
fulllment of the thesis requirement for the degree of Doctor of Philosophy in Electrical and Computer Eng.
[30] Zhao B., (2006)," An Improved Particle SwarmOptimization Algorithmfor Global Numerical Optimization", Int. Conf. on
Comput. Science N6, Reading, (Royaume-uni), Vol. 3994, pp. 657-664.
[31] Sun, J ., Feng, B. and Xu, W., (2004), “Particle swarmoptimization with particles having quantumbehavior”, Proc. IEEE/CEC,
Vol. 1, pp. 325 – 331.
[32] Omkar, S. N., Mudigere, D., Narayana Naik, G., and Gopalakrishnan, S., (2008), “Vector evaluated particle swarm
optimization (VEPSO) for multi-objective design optimization of composite structures”, J . Computers and Structures, Vol. 86,
No.1-2. pp.1-14.
[33] M. Clerc, J . Kennedy, “The particle swarm- explosion, stability, and convergence in a multidimensional complex space”,
IEEE Transactions on Evolutionary Computation, vol. 6, pp. 58-73, Feb. 2002.
[34] R. L. Welch, G. K. Venayagamoorthy, “Comparison of two optimal control strategies for a grid independent photovoltaic
system”, 41st Industry Application Conferece, pp. 1120-1127, October 2006.
[35] A. Ratnaweera, S.K. Halgamuge, H.C. Watson, “Self-Organizing Hierarchical Particle SwarmOptimizer with Time-Varying
Acceleration Coefficients”, IEEE Trans. on Evolu. Comput., Vol. 8, pp. 240-55, 2004.
[36] P. Boonyaritdachochai, C. Boonchuay, W. Ongsakul, “Optimal Congestion Management in an Electricity Market Using
Particle SwarmOptimization with Time-Varying Acceleration Coefficients”, Computer and Mathematics Applications, Vol.
60, pp. 1068-77, 2010.