See discussions, stats, and author profiles for this publication at: [616842]

See discussions, stats, and author profiles for this publication at:
https://www.researchgate.net/publication/245107654
Modeling, control and simulation of an
autonomous wind turbine/photovoltaic/fuel
cell/ultra-capacitor…
Article

in

Journal of Power Sources
· December 2008
DOI: 10.1016/j.jpowsour.2008.08.083
CITATIONS
114
READS
794
3 authors
, including:
Mehmet Uzunoglu
Yildiz Technical University
151

PUBLICATIONS

2,073

CITATIONS

SEE PROFILE
Mohammad S. Alam
Texas A&M University – Kingsville
479

PUBLICATIONS

4,475

CITATIONS

SEE PROFILE
All content following this page was uploaded by
Mehmet Uzunoglu
on 09 April 2014.
The user has requested enhancement of the downloaded file.
All in-text references
underlined in blue
are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.

Journal of Power Sources 185 (2008) 1273–1283
Contents lists available at ScienceDirect
Journal of Power Sources
journal homepage: www.elsevier.com/locate/jpowsour
Modeling, control and simulation of an autonomous wind
turbine/photovoltaic/fuel cell/ultra-capacitor hybrid power system
O.C. Onara,b, M. Uzunoglua,b,∗, M.S. Alama
aDepartment of Electrical and Computer Engineering, University of South Alabama, Mobile, AL 36688, USA
bYildiz Technical University, Istanbul 34349, Turkey
article info
Article history:
Received 17 June 2008Received in revised form 25 August 2008Accepted 26 August 2008Available online 3 September 2008
Keywords:Dynamic modelFuel cellHybrid power generationPhotovoltaicUltra-capacitorWind powerabstract
This paper focuses on the combination of wind turbine (WT), photovoltaic (PV), fuel cell (FC) and ultra-
capacitor (UC) systems for grid-independent applications. The dynamic behavior of the proposed hybridsystem is tested under various wind speed, solar radiation and load demand conditions. The developedmodel and its control strategy exhibit excellent performance for the simulation of a complete day. In thesimulation, the solar radiation and power demand data are based on real world measurements, while thewind speed data are quasi-real because it is simulated based on special wind speed generation algorithms.
© 2008 Elsevier B.V. All rights reserved.
1. Introduction
Renewable energy sources and conversion units such as PVs,
WTs, and FCs have been found to be promising energy sourcestoward building a sustainable and environment friendly energyeconomy in the next decade. However, these renewable energysources suffer from some deficiencies when used as stand-aloneenergy sources. The power generated by WT and PV systems ishighly dependent on weather conditions. Natural variations inwind speed and sunlight causes power fluctuations in WT and PVsystems, respectively. In addition, it is difficult to store the powergenerated by a PV or WT system for future use. To alleviate theseproblems, WT and PV sources can be integrated with other alter-native systems using hybrid topologies. A hybrid power systemconsists of a combination of two or more energy sources, convert-ers and/or storage devices. Thus, higher efficiency can be obtainedby making the best use of their features while overcoming theirlimitations [1–4] .
Alternate energy conversion systems such as PV panels and wind
turbines can be combined with FC power plants to satisfy sustained
∗Corresponding author at: Department of Electrical Engineering, Yildiz Technical
University, Istanbul 34349, Turkey. Tel.: +90 212 383 2441; fax: +90 212 259 4869.
E-mail addresses: conar@yildiz.edu.tr (O.C. Onar), uzunoglu@yildiz.edu.tr ,
mehmet uzunoglu@yahoo.com (M. Uzunoglu), malam@usouthal.edu (M.S. Alam).load demands. An FC power plant uses hydrogen and oxygen to con-
vert chemical energy into electrical energy. In this study, among thevarious types of FC systems, a Proton Exchange Membrane (PEM) FCpower plant is used because PEMFC power plants have been foundto be especially suitable for hybrid energy systems [5] since they
have low operating temperature (80–1000
◦C) and fast start-up.
Currently, a 5 kW commercial PEMFC power plant is operational inthe authors’ lab. Moreover, PEMFC is suitable for using the hydrogenproduced from the electrolysis of water since hydrogen is free fromany carbon monoxide [1]. However, assisting an FC hybrid power
plant with a parallel UC bank makes economic sense when satisfy-ing the peak power demands or transient conditions since the UCbank supplies the extra power, thereby reducing the size and cost ofthe FC system. Ultra-capacitors are electrical energy storage devices(a few Farads to several thousand Farads per cell) with high powerdensities when compared to batteries [6]. Ultra-capacitor banks
can be used for short-term energy storage due to their high cyclingefficiency, convenience for charging/discharging, and additionallyto meet the instantaneous load ripples/spikes. Besides, overloadingfuel cell systems may cause gas starvation thus decreasing its per-formance and lifetime. As another feature, the load tracking delaysand mismatches of the FCs should be compensated by an auxiliarysystem such as a battery or ultra-capacitor bank.
In this paper, the dynamic model, design and simulation of a new
WT/PV/FC/UC autonomous hybrid system is proposed. To achieveautonomous operation and energy management, appropriate
0378-7753/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi: 10.1016/j.jpowsour.2008.08.083

1274 O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283
Fig. 1. Overall system configuration.
components and a control system is used to achieve adequate per-
formance for the overall system. Here, the control system has beendesigned to draw the maximum available energy from the WT andPV sources. The excess power generated by renewable sources isused for UC charging and for the electrolyzer to produce hydrogen.When the energy generated from the WT and PV systems becomesinsufficient with respect to the load demand, the stored hydrogen isfed to the FC system. The tracking mismatches of the FC system andthe load demand exceeding the power generated by WT/PV/FC, issupplied by the UC bank. Using the developed model, the purpose oftesting the overall performance of the complete system is achieved.Besides, the dynamic models of each component are integratedusing the proposed topology with good performance. The modelallows to design and size the hybrid system for a variety of load-ing and meteorological conditions. The developed model providesthe integration of renewable energy systems with energy man-agement strategy to overcome the deficiencies of each renewablesystem. Modeling and simulations are performed using MATLAB
®,
Simulink®and SimPowerSystems®[7–9] software packages and
test results of the autonomous operation and energy managementof the system show that this technique performs successfully.
2. System description and methodology
The overall configuration of the hybrid power generation sys-
tem is shown in Fig. 1 . The PV/FC/UC hybrid power system model
reported in Ref. [10] and the dynamic PV system model reported in
Ref. [1 1] are enhanced and integrated with the wind turbine, elec-
trolyzer, storage, and load models. The primary energy sources ofthe system consist of a 10 kW WT and a 3 kW (peak) PV array. Theexcess energy with respect to the load requirement is used for UCcharging and hydrogen production. The excess power demand ofthe user is supplied by the FC system up to 5 kW. When FC systemis in operation, not only the excess demand over 5 kW but also thetracking mismatches and delays of the FC system are compensatedby the UC bank.
The dc/dc converters depicted in Fig. 1 include two cascade con-
nected converters. The bus side converters are used for bus voltageregulation and the source side converters are used for referencepower tracking.2.1. Wind speed, wind turbine and generator models
The Wind Turbine Blockset, developed in Ref. [12] ,i su s e di n
this study to realize and employ the wind speed, wind turbine andgenerator models. The model of the wind speed profile is based onKaimal spectra [12] . This model takes into account the rotational
turbulences and the tower shadow. The built-in band-limited whitenoise generator is used from Simulink and the wind speed is calcu-lated as an averaging of the fixed-point wind speed over the wholerotor. The model of the wind turbine is based on the torque coef-ficient C
Qusing a look-up table as a function of /NAKand /DC2for the
variable pitch wind turbine. The torque coefficient CQis used to
directly determine the aerodynamic turbine torque, given by
Tωt=0.5/EM/SUBR/GS2CQ. (1)
The parameters used in the wind turbine model are as follows:
CQ torque coefficient
R blade radius
/NAK tip speed ratio
/SUB air density
/DC2 pitch angle
/GS wind speed
An active stall controller is also employed to determine the opti-
mal pitch angle and to limit the electrical output power to thenominal mechanical power. The cut-in and cut-off wind speed ofthe turbine is set to be 3 m s
−1and 15 m s−1, respectively. The out-
put torque of the wind turbine is used to drive the squirrel cageinduction generator shaft. In the induction generator model, thesteady-state voltage equations are used. Using this approach, theinduction machine can be completely characterized in terms ofsteady-state values of the stator and rotor currents, phase angle,electromagnetic torque and electrical power. The electrical powerwhich is supplied to the dc bus through the ac/dc and dc/dc convert-ers is obtained from the stator windings of the induction generator.All detailed information on the components of WT system suchas wind model, active stall controller, aerodynamic turbine model,and induction generator models can be found in Ref. [12] . The WT
system configuration is given in Fig. 2 .

O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283 1275
Fig. 2. WT system configuration.
2.2. PV system characteristics and model
A PV system consists of many cells connected in series and paral-
lel to provide the desired output terminal voltage and current, andexhibits a nonlinear I–Vcharacteristic [1 1,13] . Masoum et al. [1 1]
introduced a model for Silicon solar PV panel. The PV cell equivalentmodel, which represents the dynamic nonlinear I–Vcharacteristics
of the PV system shown in Fig. 3 , is used within the hybrid sys-
tem as a controlled voltage source by utilizing it in the Simulinkand SimPowerSystems toolboxes of MATLAB. The subsystems of thecomplete PV model are described in details in Ref. [10] ,w h i c hi sa
previous study of the authors, hence; those are not repeated in thisstudy.
The parameters used in the mathematical modeling of the PV
system are as follows:
a ideality or completion factor
I
0 PV cell reverse saturation current [A]
IPV PV cell output current [A]
Isc short-circuit cell current [A]
k Boltzmann’s constant [J K−1]
Np the number of parallel strings
Ns the number of series cells per string
q electron charge [C]
Rs series resistance of PV cell [ /DEL]
T PV cell temperature [K]
VPV terminal voltage for PV cell [V]The output voltage characteristic of the PV system can be
expressed as in Ref. [1 1] and [13] ,
VPV=NsakT
qln/bracketleftbigg
Isc−IPV+Np
NpI0/bracketrightbigg
−Ns
NpRsIPV. (2)
The manufacturer’s datasheet provides necessary information
for most of the parameters of Eq. (2) [14] . The current–voltage char-
acteristics of the PV array (for Np= 30 and Ns= 1 12) are obtained
according to the different short-circuit current values using Eq. (2)
as shown in Fig. 4 .
Using the current–voltage curves, the current–power curves can
be obtained as shown in Fig. 5 to determine the maximum power
to be drawn from the PV array under various short-circuit currentvalues.
The maximum power output of the PV array varies accord-
ing to solar radiation or load current. Therefore, a control systemis needed to exploit the solar array more effectively as an elec-tric power source by building a maximum power point tracker(MPPT) [15,16] . Different maximum power point tracker systems
can be implemented. In this case, a computational approach is usedinvolving the model of the nonlinear I–Vcharacteristics of the solar
panel. Based on the I–Vcharacteristics, the corresponding maxi-
mum power points are computed for different load conditions asa function of cell short-circuit currents which varies as a functionof the solar radiation. To determine the relationship between thesolar radiation and short-circuit current, a curve fitting based com-putational method is employed. Then, an MPPT system, where the
Fig. 3. PV system with MPPT, controllers and converters.

1276 O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283
Fig. 4. Current–voltage curves for different short-circuit levels.
hardware part is based on a dc/dc converter is built to draw all the
maximum available power from the PV generator.
The voltage based PV model exhibits the I–Vcharacteristics of
the PV system using a built-in dc/dc converter as shown in Fig. 3
(PV system model). The pulse width of the switching signal is deter-mined using reference and measured voltages to represent thecharacteristic given in Eq. (2). Then, the pulse width command and
the carrier signal are compared to obtain the switching PWM sig-nals. The MPPT system calculates the maximum available powerfrom the PV system, and then compares the calculated power withthe actual drawn power for generating the switching signals of thedc/dc converter of the MPPT system. Another dc/dc converter cas-caded after the MPPT converter which helps to keep the outputvoltage equal to the dc bus voltage.
The values of the PV system parameters are listed in Table 1 .
2.3. PEMFC model
The PEMFC model described in Ref. [17] is modified using the
relationship between output voltage and partial pressure of hydro-gen, oxygen and water. The FC system model parameters used inthis model are as follows:
Fig. 5. Current–power curves for different short-circuit current values.Table 1
PV system model parameters [1 1]
Specification Value
The number of series cells per string ( Ns)3 0
The number of parallel cells per strings ( Np)1 1 2
Ideality or completion factor ( a)1 . 9
Boltzmann’s constant ( k) 1.3805e −23 [J K−1]
PV cell temperature ( T) 298 [K]
Electron charge ( q) 1.6e −19 C
Short-circuit cell current (representing insulation level) ( Isc) 2.926 [A]
PV cell reverse saturation current ( I0) 0.00005 [A]
Series resistance of PV cell ( Rs) 0.0277 [ /DEL]
B,C constants to simulate the activation over voltage in PEMFC
system and [V]
CV conversion factor [kmol of hydrogen per kmol of
methane]
E Nernst instantaneous voltage [V]
E0 standard no load voltage [V]
F Faraday’s constant [C (kmol)−1]
IFC FC system feedback current [A]
k1 PI gain
Kan anode valve constant [/radicalbig
kmol kg(atm s)−1]
KH2hydrogen valve molar constant [/radicalbig
kmol kg(atm s)−1]
KH2O water valve molar constant [/radicalbig
kmol kg(atm s)−1]
KO2oxygen valve molar constant [/radicalbig
kmol kg(atm s)−1]
Kr modeling constant [kmol (s A)−1]
MH2molar mass of hydrogen (kg kmol−1)
N0 number of series fuel cells in the stack
pH2hydrogen partial pressure [atm]
pH2O water partial pressure [atm]
pO2oxygen partial pressure [atm]
qH2molar flow of hydrogen [kmol s−1]
qO2input molar flow of oxygen [kmol s−1]
qmethane methane flow rate [kmol s−1]
qin
H2hydrogen input flow [kmol s−1]
qout
H2hydrogen output flow [kmol s−1]
qr
H2hydrogen flow that reacts [kmol s−1]
qreq
H2amount of hydrogen flow required to meet the load
change [kmol s−1]
R universal gas constant [(1 atm) (kmol K)−1]
RintFC internal resistance [ /DEL]
rH–O the hydrogen–oxygen flow ratio
T absolute temperature [K]
U utilization rate
Van volume of the anode [m3]
Vcell dc output voltage of FC system [V]
/FSH2hydrogen time constant [s]
/FSO2oxygen time constant [s]
/FSH2O water time constant [s]
/DC1act activation over voltage [V]
/DC1ohmic ohmic over voltage [V]
The relationship between the molar flow of any gas (hydrogen)
through the valve and its partial pressure inside the channel can beexpressed as [18]
qH
2
pH2=Kan/radicalbig
MH2=KH2. (3)
For hydrogen molar flow, there are three significant factors:
hydrogen input flow, hydrogen output flow, and hydrogen flow dur-ing the reaction [18] . The relationship among these factors can be

O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283 1277
expressed as
d
dtpH2=RT
Van(qin
H2−qout
H2−qr
H2). (4)
According to the basic electrochemical relationship between the
hydrogen flow and the FC system current, the flow rate of reactedhydrogen is given by [18]
q
r
H2=N0NsIFC
2F=2KrIFC. (5)
Using Eqs. (3) and (5) , and applying Laplace transform, the hydrogen
partial pressure can be obtained in the sdomain as [19]
pH2=1/K H2
1+/FSH2s(qin
H2−2KrIFC), (6)
where
/FSH2=Van
KH2RT. (7)
Similarly, water partial pressure and oxygen partial pressure can
be obtained. The polarization curve for the PEMFC is obtained fromthe sum of the Nernst’s voltage, the activation over voltage, andthe ohmic over voltage. Assuming certain temperature and oxygenconcentration, the FC output voltage may be expressed as [17,20]
V
cell=E+/DC1act+/DC1ohmic , (8)
where/DC1
act=−Bln(CIFC) (9)
and/DC1
ohmic =−RintIFC. (10)
Now, the Nernst’s instantaneous voltage may be expressed as [17]
E=N0/bracketleftBigg
E0+RT
2Flog/bracketleftBigg
pH2/radicalbig
PO2
pH2O/bracketrightBigg/bracketrightBigg
. (1 1)
The FC system consumes hydrogen according to the power
demand. The hydrogen is obtained from the high-pressure hydro-gen tanks. During operational conditions, to control hydrogen flowrate according to the output power of the FC system, a feedbackcontrol strategy is utilized. To achieve this feedback control, FC cur-rent from the output is taken back to the input while converting thehydrogen into molar form [17,21] . The amount of hydrogen available
from the hydrogen tank is given by
q
req
H2=N0NsIFC
2FU. (12)
Depending on the FC system configuration, and the flow of
hydrogen and oxygen, the FC system produces the dc outputvoltage. The hydrogen–oxygen flow ratio r
H–O in the FC system
determines the oxygen flow rate [17] .
The values of the FC system parameters are listed in Table 2 .
2.4. Ultra-capacitor model
The parameters used in the mathematical modeling of the UC
are as follows:
C capacitance [F]
CUC-total the total UC system capacitance [F]
EPR equivalent parallel resistance [ /DEL]
ESR, R equivalent series internal resistance [ /DEL]
EUC the amount of energy drawn from the UC bank [Ws]
ns the number of capacitors connected in series
np the number of series strings in parallelTable 2
FC system model parameters
Specification Value
Activation voltage constant ( B) 0.04777 [A−1]
Activation voltage constant ( C) 0.0136 [V]
Conversion factor (CV) 2Faraday’s constant (F) 96484600 [C kmol
−1]
Hydrogen time constant ( /FSH2) 3.37 [s]
Hydrogen valve constant ( KH2) 4.22 ×10−5[kmol (s atm)−1]
Hydrogen–oxygen flow ratio (r H–O )1 . 1 6 8
Krconstant = No/4F 2.2802 ×10−7[kmol (s A)−1]
Methane reference signal ( Qmethref ) 0.00001 5 [kmol s−1)
No load voltage (Eo) 0.8 [V]Number of cells (No) 88Number of stacks ( N
S)1
Oxygen time constant ( /FSO2) 6.74 [s]
Oxygen valve constant ( kO2)2 . 1 1 ×10−5[kmol (s atm)−1]
FC system internal resistance ( Rint)N o ×0.00303 [ /DEL]
FC absolute temperature ( T) 343 [K]
Universal gas constant ( R) 8314.47 [J (kmol K)−1]
Utilization factor ( U) 0.8
Water time constant ( /FSH2O) 18.418 [s]
Water valve constant ( KH2O)7 . 7 1 6 ×10−6[kmol (s atm)−1]
RUC-total the total UC system resistance [ /DEL]
T time [s]
V(t) the voltage at time t[V]
Vi the initial voltage before discharging starts [V]
Vf the final voltage after discharging ends [V]
Fig. 6 shows the classical equivalent circuit of the UC unit. The
model consists of a capacitance ( C), an equivalent series resis-
tance (ESR, R) representing the charging and discharging resistance,
and an equivalent parallel resistance (EPR) representing the self-discharging losses [21,22] . The EPR models the leakage effects and
only impacts long-term energy storage performance of the UC[23,24] . The voltage state of an UC (i.e., RC circuit) with a capac-
itance Cdraining charge into a resistance Rmay be expressed as
[25]
V(t)=V
iexp/parenleftBig
−t
RC/parenrightBig
. (13)
The RC time constant determines the effective period of the charg-
ing and discharging processes for some initial voltage on thecapacitor [25] . The state of charge (SOC) of an UC bank is the avail-
able energy capacity expressed as a percentage of its rated energycapacity which is related to its voltage. The amount of energy drawnfrom the UC bank is directly proportional to the capacitance and thechange in terminal voltage [23,6] , which is expressed as
E
UC=1
2C(V2
i−V2
f). (14)
The effective specific energy for a prescribed load can be sup-
plied by various UC bank configurations. In practical applications,the required amount of terminal voltage and energy or the capaci-tance of UC storage system can be built using multiple UCs in series
Fig. 6. Classical equivalent model for the UC unit.

1278 O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283
Table 3
Maxwell Boostcap BMOD2600-48EA ulta-capacitor specifications
Specification Value
Capacitance 144 [F]Voltage 48.6 [V]ESR, dc 1 1 [m /DEL]
ESR @ 1 kHz 6.8 [m /DEL]
Thermal resistance 0.27 [
◦CW−1]
Short-circuit current 4800 [A]Energy density (max) 3.40 [W h kg
−1]
Power density 2600 [W kg−1]
Volume 13.4 [L]Mass 13.5 [kg]
and parallel. The terminal voltage determines the number of capaci-
tors which must be connected in series to form a bank and the totalcapacitance determines the number of capacitors which must beconnected in parallel in the bank. The total system resistance andthe total system capacitance of the UC bank can be calculated as[21,23]
R
UC-total =nsESR
np, (15)
CUC-total =npC
ns. (16)
In this research, Maxwell Technologies Boostcap®BMOD2600-
48EA type ultra-capacitor bank is used which is currentlyoperational in the authors’ lab. The reasons for selecting thismodule are the ultra low internal resistance, low RC time con-stant and high power performance. In this study, 4 parallel UCmodules are used, where each module includes 18 UC units con-nected in series. The specifications of the UC bank [26] is given in
Table 3 .2.5. Electrolyzer and storage models
Water can be decomposed into its elementary components by
passing electric current between two electrodes separated by anaqueous electrolyte [27–29] . The electrochemical reaction of water
electrolysis is given by
H
2O(liquid) +electrical energy ⇒H2+1
2O2. (17)
The Electrolyzer model involves the following parameters:
F Faraday constant [C kmol−1]
ie electrolyzer current [A]
nc the number of electrolyzer cells in series
/DC1F Faraday efficiency
nH2produced hydrogen moles per second [mol s−1]
According to Faraday’s law, hydrogen production rate of an elec-
trolyzer cell is directly proportional to the electrical current in theequivalent electrolyzer circuit [27] , which can be expressed as
n
H2=/DC1FnCie
2F. (18)
The ratio between the actual and the theoretical maximum amount
of hydrogen produced in the electrolyzer is known as Faraday effi-ciency. Assuming that the working temperature of the electrolyzeris 40
◦C, Faraday efficiency is expressed by Refs. [27,29]
/DC1F=96.5e(0.09/ie−75.5/i2
e). (19)
According to Eqs. (18) and (19) , a simple electrolyzer model is devel-
oped using Simulink.
The amount of hydrogen required by the PEMFC is sent directly
from the electrolyzer system according to the relationship betweenthe output power and the hydrogen requirement of the PEMFCsystem. The remaining amount of hydrogen, i.e., the difference
Fig. 7. The general decision algorithm used in the main controller.

O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283 1279
Fig. 8. Power demand of the micro-grid neighborhood consisting of 3 residential
units.
between the produced and consumed hydrogen is sent to the stor-
age tank.
One of the hydrogen storage techniques is physical hydrogen
storage, which involves using tanks to store either compressedhydrogen gas or liquid hydrogen. The hydrogen storage model isbased on Eq. (20) and it directly calculates the tank pressure using
the ratio of hydrogen flow to the tank. The produced hydrogen isstored in the tank, whose system dynamics can be expressed as [30]
P
b∧
−Pbi=zNH2RT b
MH2Vb. (20)
The parameters used in the hydrogen storage system are listed
below:
MH2Molar mass of hydrogen [kg kmol−1]
NH2hydrogen moles per second delivered to the storage tank[kmol s
−1]
Pb pressure of tank [Pa]
Pbi Initial pressure of the storage tank [Pa]
R universal (Rydberg) gas constant [J (kmol K)−1]
Tb operating temperature [K]
Vb volume of the tank [m3]
z compressibility factor as a function of pressure
Fig. 9. Wind speed profile.
Fig. 10. Power produced by the wind generator.
The compression dynamics and the compression energy
requirements are not considered in our calculations. All auxil-iary power requirements such as pumps, valves and compressionmotors are ignored in the dynamic model since their investigationis out of the paper’s scope.
2.6. Energy management and control strategy
A main controller is employed for the entire energy measure-
ment and management purposes. The main controller providesthe autonomous operation with the required measurements, deci-sions and controls by collecting data through the sensors andproducing the commands for the power converters connected tothe inputs/outputs of the components used in the hybrid system.The operating strategies used in the main controller are as fol-lows:
– The use of power generated by the PV and WT systems has priority
in satisfying power demand over that provided by the FC systemor by the UC bank.
– If the total power generated by the PV and WT systems is higher
than the demand, the additional power will be used to charge theUC bank.
– After charging the UC bank, the remaining power is used for the
electrolyzer through a power converter to generate hydrogen.
Fig. 1 1. Global hourly solar radiation.

1280 O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283
Fig. 12. Short-circuit current versus power variation.
– The electrolyzer is kept in operation as long as available power
exists and the tank pressure does not exceed the prescribed valueof 15,000 Pa. The electrolyzer maximum power is also set to10 kW. The additional power generated by the WT and PV systemscan be transferred to the electrolyzer up to the 10 kW limit.
– Dumping power is necessary if the UC bank is fully charged and
the electrolyzer power is at 10 kW limit.
– If the total electric power generated by the PV and WT systems
is less than the demand, power will be supplied from the FC sys-tem up to the maximum 5 kW limit. If the load demand exceedsthe power generated by WT/PV/FC combination, the difference issupplied by the UC bank.
– The UC bank not only provides the excess power demand over
5 kW, but also compensates the tracking mismatches and delaysof the FC system which has relatively slow response time.
– The UC bank maximum voltage is set to be its nominal volt-
age of 48.6 V and its minimum voltage is set to be 25% offull state of charge (SOC) to avoid overcharging and deepdischarging.
To implement the aforementioned power management strategy,
appropriate power control systems are used at relevant points andthe total system components are integrated. The block diagram ofthe main controller is illustrated in Fig. 7 .
Fig. 13. Hourly power drawn from the solar system.
Fig. 14. The difference between the load power and generated power (PV + WT).
3. Results and discussions
In the simulation process, the aim is to observe the proposed
system’s behavior over a long period of time including day and nightcases as well as low or no wind conditions. The load profile, windspeed variation and solar radiation profiles are all used to test theperformance of the proposed hybrid system for a typical day of theneighborhood.
The 5 kW PEMFC system currently operated in our laboratory
f e e d sa5 0 0 f t
2house. Energy consumption varies from house to
house, but to simulate a real-world scenario, load profiles are alsoobtained from a neighborhood comprising of three houses, whereeach house is approximately 2500 ft
2and includes all household
electrical devices. The load profile of the micro-grid neighborhoodis illustrated in Fig. 8 .
The wind speed profile is generated by the wind speed model
described in Section 2.1 and the hourly wind speed profile is
depicted in Fig. 9 , which yields the power generated by WT as
shown in Fig. 10 .
The hourly averaged global solar radiation on a horizontal sur-
face (W m
−2) is taken from Ref. [31] , where the record coordinates
are: Latitude 41◦10/primeNorth and Longitude 40◦20/primeEast. The exper-
imental solar radiation data which is recorded by the regionalmeteorological station in the month of June is given in Fig. 1 1 .
Fig. 15. Power produced by the FC system.

O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283 1281
Fig. 16. Tracking mismatches of the FC system.
Using the nonlinear curve fitting based computation, the rela-
tionship between the solar radiation and the short-circuit currentis obtained. Then using the maximum power points for differentshort-circuit values given in Fig. 5 , the short-circuit current and the
maximum power relationship are obtained as shown in Fig. 12 .I t
is evident from Fig. 12 that the short-circuit current varies almost
linearly with the solar radiation.
The maximum power values are used in the MPPT dc/dc con-
verter as power commands and the maximum power drawn fromthe PV system to the dc bus is obtained as illustrated in Fig. 13 .F r o m
Fig. 13 , we observe that in the early morning before 4:08 and in
the evening after 19:38, the solar power is unavailable due to non-existence of solar radiation. The maximum power from PV systemcan be transferred to the dc bus at about 12.02 in the noon.
As mentioned earlier, the use of power generated by the PV and
WT systems has priority in satisfying load demand. Thus, the totalpower generated by the WT and PV systems is transferred to the dcbus directly. The difference between the power generated by WTand PV systems and the user power demand is demonstrated inFig. 14 . The positive regions of Fig. 14 are the additional power gen-
erated by PV and WT sources which is used for the UC charging andhydrogen production within the electrolyzer system. Furthermore,the negative regions are the exceeded power demands which aremet by the FC system and UC bank discharge.
Fig. 17. Amount of hydrogen required by the FC system.
Fig. 18. Amount of hydrogen produced by the electrolyzer.
The FC system power command is obtained according to the
negative regions of Fig. 14 (exceeded user demands) up to the 5 kW
limit. The power produced by FC system is given in Fig. 15 , which
slightly differs from the power command due to the time constantsand relatively slow response characteristic of the FC system. How-ever, since the required hydrogen is not processed by a reformerand provided directly from the hydrogen tank, the reformation timedelays are eliminated.
The tracking mismatches of the FC system are given in Fig. 16
which is supposed to be supplied by the UC bank. The power pro-duced by the FC system shown in Fig. 15 corresponds to the amount
of hydrogen flow to the FC system as depicted in Fig. 17 . The amount
of power transferred to the electrolyzer corresponds to the hydro-gen flow from the electrolyzer to the storage tank as illustrated inFig. 18 .
The hydrogen produced by the electrolyzer and consumed by the
FC system causes the pressure of the storage tank to vary as showninFig. 19 . The pressure of the storage tank increases when the elec-
trolyzer produces hydrogen, and the tank pressure decreases whenthe FC system uses the stored hydrogen.
The power supplied by the UC bank is shown in Fig. 20 which
includes the exceeded load demand and the tracking mismatches(Fig. 16 ) of the FC system. Due to the scaling differences of
Figs. 16 and 20 , the low mismatches cannot be seen in Fig. 20 . These
Fig. 19. The pressure variation of the hydrogen storage tank.

1282 O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283
Fig. 20. Power supplied by UC bank including tracking delays and all mismatches.
Fig. 21. UC bank charge and discharge power.
mismatches can be observed by zooming into the areas of interest
ofFig. 20 .
The charge and discharge power of the UC bank is recorded as
shown in Fig. 21 , where positive power region represents the power
released by the UC bank and negative power region represents thepower captured by the UC bank through the bi-directional dc/dcconverter.
Fig. 22. UC bank terminal voltage.The charging and discharging power of the UC bank corresponds
to the terminal voltage variation as demonstrated in Fig. 22 .F r o m
Fig. 22 , it is obvious that the UC bank terminal voltage varies with
the charge and discharge state of the UC bank within the limitedvalues.
4. Conclusions
A novel WT/PV/FC/UC hybrid power system is designed and
modeled for grid-independent applications. A detailed simulationmodel has been developed which allows designing and analyzingany WT/PV/FC/UC hybrid system with various power levels andparameters. The dynamic behavior of the hybrid system is testedunder various wind speed, solar radiation and load demand con-ditions. The solar radiation and power demand data are based onreal world records and the wind speed is artificially generated inthe computer environment using a suitable algorithm.
The main contribution of this work is the hybridization of the
renewable energy sources with FC systems using long and short-term storage strategies with appropriate power controllers andcontrol strategy to build an autonomous system. The developedsystem and its control strategy exhibit excellent performance forthe simulation of a complete day or longer periods of time. Theproposed system can be used for non-interconnected remote areasor isolated cogeneration power applications as a reliable and tech-nically feasible system.
Acknowledgement
This work was supported in part by the U.S. Department of
Energy under Grant DE-FG02-05CH1 1295.
References
[1] K. Rajashekara, IEEE Trans. Ind. Appl. 41 (3) (2005) 682–689.
[2] K. Agbossou, M. Kolhe, J. Hamelin, T.K. Bose, IEEE Trans. Energy Convers. 19 (3)
(2004) 633–640.
[3] A. Bilodeau, K. Agbossou, J. Power Sources 162 (2) (2006) 757–764.[4] O.C. Onar, M. Uzunoglu, M.S. Alam, J. Power Sources 161 (1) (2006) 707–722.[5] L. Gao, Z. Jiang, R.A. Daugal, J. Power Sources 130 (1–2) (2004) 202–207.[6] A. Burke, J. Power Sources 91 (1) (2000) 37–50.[7] M. Uzunoglu, A. Kizil, O.C. Onar, Her Yonu ile MATLAB (Proficiency in MATLAB,
2nd Edition), Turkmen Kitabevi, Istanbul, 2003.
[8] MATLAB Simulink Simulation and Model Based Design User’s Guide, Version
6.3 [Online]. Available: http://www.mathworks.com/access/helpdesk/help/
pdf
doc/simulink/sl using.pdf .
[9] MATLAB SimPowerSystems for Use with Simulink User’s Guide, Ver-
sion 4.1.1 [Online]. Available: http://www.mathworks.com/access/helpdesk/
help/pdf doc/physmod/powersys/powersys.pdf .
[10] M. Uzunoglu, O.C. Onar, M.S. Alam, Modeling, control and simulation of
a PV/FC/UC based hybrid power generation system for stand-alone appli-cations, Renewable Energy, accepted for publication, DOI information:10.1016/j.renene.2008.06.009.
[1 1] M.A.S. Masoum, H. Dehbonei, E.F. Fuchs, IEEE Trans. Energy Convers. 17 (4)
(2002) 514–522.
[12] F. Iov, A.D. Hansen, P. Sørensen, F. Blaabjerg, Wind Turbine Blockset in MAT-
LAB/Simulink (A Simulation Platform to Model, Optimize and Design WindTurbines), Project Participants: Aalborg University Institute of Technology,RISØ National Laboratory, Support from Danish Energy Agency through Grant1363/01-0013, 2004–2007.
[13] M. Veerachary, T. Senjyu, K. Uezato, IEEE Trans. Aerospace Electron. Syst. 38 (1)
(2002) 262–270.
[14] Y. Sukamongkol, S. Chungpaibulpatana, W. Ongsakul, Renew. Energy 27 (2)
(2002) 237–258.
[15] R. Chedid, S. Rahman, IEEE Trans. Energy Convers. 13 (1) (1998) 76–83.
[16] K. Kobayashi, H. Matsuo, Y. Sekine, IEEE Trans. Ind. Electron. 53 (1) (2006)
281–286.
[17] M.Y. El-Shark, A. Rahman, M.S. Alam, P.C. Byrne, A.A. Sakla, T. Thomas, J. Power
Sources 138 (1–2) (2004) 199–204.
[18] J. Padulles, G.W. Ault, J.R. McDonald, J. Power Sources 86 (1–2) (2000) 495–500.
[19] L. Gao, R.A. Dougal, S. Liu, IEEE Trans. Power Electron. 20 (1) (2005)
236–243.
[20] J. Hamelin, K. Agbossou, A. Laperriere, F. Laurencelle, T.K. Bose, Int. J. Hydrogen
Energy 26 (6) (2001) 625–629.

O.C. Onar et al. / Journal of Power Sources 185 (2008) 1273–1283 1283
[21] K.-H. Hauer, Analysis tool for fuel cell vehicle hardware and software (controls)
with an application to fuel economy comparisons of alternative system designs,Ph.D. Dissertation, Dept. Transportation Technology and Policy, University ofCalifornia, Davis, 2001.
[22] R.M. Nelms, D.R. Cahela, B.J. Tatarchuk, IEEE Trans. Aerospace Electron. Syst. 39
(2) (2003) 430–438.
[23] R.L. Spyker, R.M. Nelms, IEEE Trans. Aerospace Electron. Syst. 36 (4) (2000)
1439–1443.
[24] R.L. Spyker, R.M. Nelms, IEEE Trans. Aerospace Electron. Syst. 36 (3) (2000)
829–836.
[25] B.E. Conway, Electrochemical Supercapacitors-Scientific Fundamentals and
Technological Applications, Kluwer Academic/Plenum Publishers, New York,1999, pp. 497–547.[26] Maxwell Technologies, BOOSTCAP
®Energy Storage Module BMOD2600-48EA
(104567), User’s Manual.
[27] M.J. Khan, M.T. Iqbal, Renew. Energy 30 (3) (2005) 421–439.
[28] O. Ulleberg, Stand-alone Power Systems for the Future: Optimal Design.
Operation and Control of Solar-Hydrogen Energy Systems, Ph.D. Dissertation,Norwegian University of Science and Technology, 1998.
[29] K. Sapru, N.T. Stetson, S.R. Ovshinsky, Development of a Small Scale Hydrogen
Production Storage System for Hydrogen Applications, presented at the EnergyConversion Engineering Conference, in Proc. The 32nd Intersociety, Honolulu,HI, USA, 1997, pp. 1947–1952.
[30] H. Gorgun, Int. J. Hydrogen Energy 31 (1) (2006) 29–38.
[31] K. Kaygusuz, T. Ayhan, Energy Convers. Manage. 40 (5) (1999) 545–556.