Energy 40 (2012) 196e209

Contents lists available at SciVerse ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Sizing, techno-economic and generation management analysis of a stand alone

photovoltaic power unit including storage devices
Rihab Jallouli, Lotfi Krichen*
National Engineering School of Sfax, Advanced Control and Energy Management (ACEM) BP 1173, 3038 Sfax, Tunisia

a r t i c l e i n f o a b s t r a c t

Article history: Due to the mismatch between the load demand and the intermittent solar energy, a stand-alone
Received 10 November 2011 photovoltaic-hydrogen system and an optimal control scheme are designed to maintain the high
Received in revised form system efficiency.
12 January 2012
Based on meteorological and the load demand data, a system sizing technique is proposed to establish
Accepted 1 February 2012
the minimum capacity of the system components, which are a photovoltaic (PV) panel, a proton
Available online 7 March 2012
exchange membrane fuel cell (PEMFC), a battery bank and an alkaline electrolyzer (Elz).
An accurate energy management scheme that is utilized during power transfer is proposed to meet the
Keywords:
Photovoltaic
economic requirements. Case studies are used to verify the efficiency of the energy management strategy
Hydrogen system and system sizing technique. Simulation results illustrate a simple solution to the design and processing
Battery of stand-alone PV-hydrogen (PV-H2) systems.
Energy management Ó 2012 Elsevier Ltd. All rights reserved.
Sizing
Techno-economic analysis

1. Introduction - What is a suitable size for these components?

- At a given excess/deficient power, what is the best economical
Renewable energy sources (RESs) are becoming widely used to combination of energy storage devices to absorb/exude the
overcome the lack of fossil energies and to satisfy the residential energy?
electrical energy demand. Photovoltaic generators are considered
among the most promising RESs for their availability and inex- This study aims to answer these questions for the PVeH2
haustibility. However, the weaknesses of this type of energy remain system.
the intermittence of solar irradiations depending on locations and Ulleberg [2] reported the simulation results of a PVeH2 system
seasons and the time varying load demand. To overcome this consisting of a battery, an electrolyzer, a PV generator and a fuel
problem, the PV panel is always associated with other storage cell. The author presents a control scheme based on the on/off
systems. These systems absorb the excessive energy, store and switching of the fuel cell and the electrolyzer that uses the battery
deliver it at an adequate time and a suitable rate. The most common state of charge (SOC). This work presents the simulation results
design is the H2 system, which comprises an electrolyzer as without presenting the models and sizing of the mathematical
a hydrogen producer and a PEMFC as a hydrogen consumer to components.
produce electrical energy. This H2 system acts as long-term energy Maclay et al. [3] developed an empirical model consisting of a PV
storage. Another device can also be used in such a combination as panel, a reversible fuel cell, a battery bank and an ultracapacitor.
fast energy storage. For example, a buffer may be a rechargeable The authors present the power evolution of multiple combinations
lead acid battery due to its high efficiency, quick charging/dis- without presenting battery and ultracapacitor SOC evolution,
charging ability and capacity for smoothing source fluctuations [1]. which are critical to control.
The use of such a system leads to certain critical questions: Li et al. [4] developed a dynamic model of a stand-alone
photovoltaic power system using an H2 device and a battery bank
as energy storage devices. This work presents a dynamic model of
all system components, makes a cost evaluation of the system
* Corresponding author. Tel.: þ216 74 274 418; fax: þ216 74 275 595. based on size and describes an energy management strategy based
E-mail address: lotfi.krichen@enis.rnu.tn (L. Krichen). on the hydrogen tank and battery SOCs. The load and the

0360-5442/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.energy.2012.02.004
 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209 197

photovoltaic powers said to be used in the energy management are batteries or produce H2 in the electrolyzer using a process called
not described in the paper, and the authors are content with pre- charging. To choose where to send the spare energy, we must
senting the annual battery SOC evolution. evaluate the energy cost of each device (the device used to charge
Zhou et al. [5] build on the work of Ulleberg with the battery the batteries and the device used to form H2 in the electrolyzer).
SOC as the control key for the electrolyzer and fuel cell on/off Thus, the system will choose the method that uses the least cost.
switching. This study gives the sizing equations for different system If, on the contrary, the consumer demand is more than the
components basing on load consumption and irradiation curves photovoltaic supply, then the lack of power can be satisfied in two
over a year. Suggestions for the optimum size of the hybrid system ways: switching on the fuel cell or the battery using a process called
are given for two different locations. This work lacks an economic discharge. Thus, when choosing a certain device (PEMFC or
analysis. battery), we have to estimate the cost of power generation in each
The use of RES and H2 systems for residential applications has system and choose the less costly device.
increased significantly [6]. This work expands on the combination The proposed system can be used as a stand-alone power
of energy storage devices for residential PV systems. Then, we focus system in remote areas where there is no access to the grid and as
on a sizing method depending on PV and residential consumption a support power medium to cover electricity deficiency during
curves. The current study contributes to strategy design for optimal special situations, such as natural disasters.
energy management based on economic challenges and measur-
able SOCs of different storage devices. Mathematical models of
2. System size
PVeH2 system components are built using MATLAB/Simulink, and
the second scale evolution of different source powers is detailed.
Sizing an autonomous PVeH2 system is a complex task because
A scheme of the stand-alone PVeH2 system presented in the
different resources are employed. A properly sized system is
current work is shown in Fig. 1. This model contains the following
a system that responds to different energy requirements to reach
components: a PV generator with maximum power point tracker
a techno-economic autonomous hybrid system [7]
(MPPT), residential power demand, a battery bank with a DC/DC
Based on the energy management strategy described above,
converter for fast energy transitions, a PEMFC with a DC/DC
a system sizing method is developed in the following section.
converter to supply electricity, an alkaline electrolyzer with the DC/
DC converter to produce H2, a compressor, a pressurized tank for H2
conservation and a DC/AC converter for use at the end. 2.1. Weather data and user load demand
A PV panel collects solar irradiations and converts it into usable
electricity. The DC/DC converter behind the PV generator is Precise predictions for long-term performance of PVeH2 system
controlled by an MPPT module to allow more solar energy require consistent data about both solar radiation and residential
extraction. load demand, which comprises TV, lighting, heating and other
The energy produced by the RES must first be used to satisfy the domestic appliances as input data. Enough measurements have to
residential load. If the PV energy is greater than the demanded be accessible to supply the simulation model with the required
energy, the excess power can be used in two ways: to charge the weather and end-use load information.

(L f ,R f )
DC DC
Cdc
DC AC
Filter Line
PV panel Converter1 Inverter Residential
load

DC

DC
Battery Converter 2

DC
H2
DC
PEMFC Converter 3

DC

DC

Electrolyzer Converter 4

Fig. 1. Scheme of a stand-alone PVeH2 system.

198 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209

a 1000 b 400

Monthly averaged solar irradiation (W)

GSYavg
900 GSmax
Daily solar irradiation (Wh/m²) 350
800
300
700
250
600

500 200

400
150
300
100
200
50
100

0 0
2 4 6 8 10 12 14 16 18 20 22 24 1 2 3 4 5 6 7 8 9 10 11 12
Time (Hours) Time (Months)

Fig. 2. Solar irradiation fluctuations.

Photovoltaic power is an intermittent energy that varies Nevertheless, to find the optimum PV panel size, we also have to
frequently according to seasons, temperatures and geographical consider the load required by the user, the power that consumes
locations. Fig. 2 shows daily intermittent and averaged annual solar the system itself and the energy mismatch between PPV and PL.
profiles where GSY avg is the yearly average solar irradiation, and Based on Fig. 4 and according to the control scheme, during sunny
Gsmax is the maximum solar irradiation over a year. periods (t1,t2) and (t3,t4), the excess photovoltaic power will be sent
The daily electrical load profiles differ with time and depend on either to charge the battery or to the electrolyzer to produce
the activity patterns and the types of appliances used. Fig. 3 shows hydrogen. However, during the interval (t2,t3), when the load
the fluctuating daily and annual load demand curves, where PLY avg power exceeds the PV supply, the lack of power will be generated
is the annual average load power and PLMmax is the maximum load either by the battery or by the PEMFC.
power over a year. Thus, the total PV panel area is calculated as follows:
Weather and load parameters are given in Table 1. MPV
APV ¼
ð1 þ hbat Þ=2
2.2. PV panel sizing 0 1
Zt3
B 1
The important required data for sizing the PV panel is the solar B ð1  hC4 $hElz $hFC $hC3 ÞðPini  PL =hIN ÞdtC
C
B t4  t1 C
radiation [W/m2] and the end user demand [W]. These data have to B t2 C
 BAini þ C
be supplied to the simulation model on an hourly basis. B Gsavg $hPV $hMPPT $hC1 $hC2 $hElz $hFC C
B C
The PV panel size is calculated assuming that all the load power, @ A
PL, is provided directly from the photovoltaic power PPV. The initial
area Aini of the PV panel is, as follows [5]: (3)
PLYavg
 2.3. Battery sizing
Aini ¼ m2 (1)
GSYavg $hPV $hMPPT $hIN
In renewable applications, the battery capacity and lifetime are
The photovoltaic power generated by this area is denoted Pini important; thus, accurate battery sizing is necessary for precise
and is described as follows: system optimization. Moreover, the battery state of charge (SOCmax
and SOCmin) is also important to take into account. The battery
Pini ¼ GS $Aini $hPV $hMPPT (2) energy is obtained by the following equation:

a 9000 b PLYavg
7000
Monthly averaged load power (W)

8000 PLMmax
6000
7000
Daily load power (W)

6000 5000

5000 4000
4000
3000
3000
2000
2000
1000
1000

0 0
1 5 10 15 20 24 1 2 3 4 5 6 7 8 9 10 11 12
Time (Hours) Time (Months)

Fig. 3. Load power curves.

 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209 199

Table 1 2500
Load demand and solar irradiation.

Item Value 2000

Solar Maximum average solar irradiation over 309.5

irradiation a year GSmax (W/m2) 1500
Averaged annual irradiation, GSY avg (W/m2) 293.5
Load power Maximum average load power over a year, 5340 1000

Ppv-PL (W)
PLM max (W)
Averaged annual load power, PLY avg (W) 3500
500

0
PLMmax $24h
Ebat ¼ M (4)
ðSOCmax  SOCmin Þ bat -500

-1000
2.4. Electrolyzer size

-1500
The rated power of the electrolyzer is equal to the maximum 1 2 3 4 5 6 7 8 9 10 11 12
excess in the PV generator power over the minimum load power; Time (Months)
thus, it can be calculated using the following equation:
Fig. 5. Difference between monthly photovoltaic production and load demand.
PElz ¼ MElz $Gsmax $APV $hPV $hMPPT $hC1 (5)

which is expressed in m3. For safety reasons, the operating pressure

2.5. Fuel cell size in a tank is limited to20% to 90% of the whole tank pressure [8].

Assuming that the PV generator is unable to produce electricity

and that the battery is fully discharged, the fuel cell has to meet all 3. System modeling
the load demand. For accurate fuel cell sizing, the maximum load
power is considered to be as follow [5]: 3.1. Photovoltaic panel modeling

PFC ¼ MFC $PLMmax =hIN (6) Photovoltaic panels have been studied for more than 20 years
[9]. The most common empirical model is the one with a diode, as
presented in Fig. 6.
2.6. Storage tank size
The empirical output current of a PV panel is described by the
following equation:
The size of the hybrid system is based on the average value of
solar radiation energy over a year and the load demand. Therefore,    
U þ IRS U þI R
the hydrogen tank sizing procedure uses the photovoltaic power IPV ¼ IL  ID  Ish ¼ IL  I0 exp  1  PV PV S (7)
a Rsh
and the load over a month.
The hydrogen amount formed by the electrolyzer during each
month is calculated by assuming that the produced hydrogen is 3.2. Battery model
stored completely before it supplies power to the residential load.
Fig. 5 shows the monthly average difference between photovoltaic Currently, battery technology is being considered in research
production and power used. The amount of hydrogen produced over and development. Thus, different battery models, such as lithium-
a year is then the total excess PV power transformed into volume, ion and lead-acid, show promising results for use in power appli-
cations. In this study, lead-acid battery technology based on the
work verified by Vasebi et al. [10] is utilized. The model considers
an empirical equivalent circuit that comprises a bulk capacitor
Cbulk, a capacitor to model the surface capacitance Csurface,
a terminal resistance Rt, a surface resistance Rs and an end resis-
tance Re, as illustrated in Fig. 7.
Vcb and Vcs are the voltages across the bulk and surface capac-
itors, respectively.

RS IPV

IL ID

RSh UPVU Rcharge

Fig. 4. Profiles of annualload and photovoltaic powers. Fig. 6. Equivalent model of a photovoltaic cell.
200 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209

Table 2
Ibat Initial parameters of a battery cell.

Is Ib parameter Value
Rt
R e ( U) 0.00448
Rs Re Rs (U) 0.00448
V0 R t ( U) 0.00336
Cbulk (F) 66,092
Csurface VCs Cbulk VCb Csurface (F) 66,092
a0 1.9
a1 0.3

Fig. 7. Equivalent model of a lead-acid battery.

where
Rs Re
3.2.1. Description of battery model state variables Að3; 1Þ ¼  þ
Cbulk ðRe þ Rs Þ2 Csurface ðRe þ Rs Þ2
To evaluate the battery voltage V0, Kirchhoff laws were applied
to the circuit presented above as follows: R2s Rs
 þ (13)
 Cbulk Re ðRe þ Rs Þ2 Csurf ace ðRe þ Rs Þ2
V0 ¼ Ibat Rt þ Is Rs þ VCs
(8)
V0 ¼ Ibat Rt þ Ib Re þ VCb and
Rs 1
Að3; 3Þ ¼  (14)
Cbulk Re ðRe þ Rs Þ Csurface ðRe þ Rs Þ
The equality between the two equations in (8) gives:

Ib Re ¼ Is Rs þ VCs  VCb (9) 3.2.2. Determination of battery state of charge (SOC)

The SOC of the battery depends on its open bulk capacitor
The rate of change of the bulk and surface capacitor voltages is voltage Vcb according to the following equation:
given by the following equation:
Vcb ðtÞ ¼ a1 SOCðtÞ þ a0 (15)
2 3 Table 2 provides a summary of the battery parameters.
2 3 1
dVcb  
6 dt 7 6 C ðRe þ Rs ÞCbulk ðRe þ Rs Þ 7 3.3. Fuel cell dynamic modeling
4 6
5 ¼ 4 bulk 7 Vcb
dVcs 1 1 5 Vcs
 A fuel cell is an electrochemical device that transforms the
dt Csurface ðRe þ Rs Þ Csurface ðRe þ Rs Þ
2 3 energy of reactions directly into electricity.
Rs The PEMFC output voltage depends on three terms, as described
6 C 7
þ6 bulk ðRe þ Rs Þ 7I (10) below:
4 Re 5 bat
VFC ¼ ENernst  Vact  Vohm  Vconc (16)
Csurface ðRe þ Rs Þ
Each term in (16) can be expressed by the equations discussed
Because the output voltage V0 can also be expressed as a func- below.
tion of output current Ibat, Eq. (8) is rearranged as follows:
3.3.1. Reversible voltage
     The reversible voltage of cell ENernst is the potential of one cell
Rs Re Vcb Re Rs
V0 ¼ þ Rt þ Ibat (11) obtained in open-circuit thermodynamic conditions.
Re þRs Re þ Rs Vcs R e þ Rs

ENernst ¼ E0th ¼ 1:48  0:9:103 $TFC

The complete state variable equation of battery voltages is
qffiffiffiffiffiffiffi (17)
þ 4:3085:105 $TFC $log PH2 P02
expressed by:

2 3
2 3 1 1 2 3
dVcb =dt 6 C 0 7 Vcb
6 bulk e þ Rs Þ
ðR Cbulk ðRe þ Rs Þ 7
4 dVcs =dt 5 ¼ 6 1 1 7 4 Vcs 5
6 0 7
dV0 =dt 4C 5 V0
surface ðR e þ Rs Þ Csurface ðRe þ Rs Þ
Að3; 1Þ 0 Að3; 3Þ
2 3
Rs
6 Cbulk ðRe þ Rs Þ 7
6 7
6 Re 7
6 7
þ6 Csurface ðRe þ Rs Þ 7 (12)
6 7
6 Re Rs Rt Rt Re Rs 7
4  þ þ 5
Csurface ðRe þ Rs Þ2 C R ðR
bulk e e þ R s Þ C ðR
surface e þ R s Þ Csurface ðRe þ Rs Þ 2
 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209 201

Fig. 8. Energy flow chart.

3.3.2. Activation overpotential dðVconc þ Vact Þ I Vconc þ Vact

¼ FC  (21)
The activation overpotential Vact, including both anode and dt C CRa
cathode, can be written as follows:
To obtain the total output voltage, we multiply the individual

  output cell voltage by the number of cells.
Vact ¼  x1 þ x2 TFC þ x3 TFC log CO2 þ x4 TFC logðIFC Þ (18)
All of the parameters used in the fuel cell model have been
reported previously [11].
3.3.3. Ohmic overpotential
Considering the resistance of the fuel cell membrane and the
3.4. Electrolyzer model
resistance resulting from the electrons conduction through plates
and electrodes, the ohmic overpotential Vohm is calculated using
The electrolyzer is a reversible fuel cell. It transforms electricity
the following equation:
into hydrogen and water.

0 1
2 2:5
Vohm ¼ IFC @181:6ð1 þ 0:03ðIFC =AFC ÞÞ þ 0:062ðTFC =303Þ ðIFC =AFC Þ
TFC 303
þ Rc A (19)
4:18
ðj  0:634  3ðIFC =AFC ÞÞ$e TFC

The output voltage of an alkaline electrolyzer cell can be

3.3.4. Concentration overpotential
expressed as follows:
The voltage drop due to the mass transport, denoted Vconc,, can
be determined as follows: r1 þ r2 TElz

UElz ¼ Urev þ IElz þ S1 þ S2 TElz
AElz
Vconv ¼ B logð1  J=Jmax Þ (20) " #
 t þ t2 =TElz þ t3 =T2Elz
þ S3 T2Elz log 1 IElz þ 1 (22)
3.3.5. Charge double layer AElz
To consider the FC dynamics, we should take into
account the charge double layer phenomenon, which is described All of the parameters can be obtained by measurements and are
by: defined elsewhere [12]. Similar to battery and fuel cell voltages, the
202 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209

Table 3
Different system component sizes, costs and efficiencies.

Cost of charge cycle (€)

Battery
Item Size Initial acquisition Round trip
cost (V) efficiency
(%) [5]
Photovoltaic panel 169,75 [15]
Initial area Aini 143.5 m2 10
Total area APV 203.5 m2
Electrolyzer
Battery capacity Ebat 360 kAh 36,800 [15] 90
SOCmin ¼ 0.35
SOCmax ¼ 0.85
Electrolyzer power PElz 22 kW 57,500 [16] 85
Fuel cell power PFC 12 kW 25,000 [17] 55
Plim-ch P(kW) Storage tank volume VH2 12 m3
(240 bar)
Fig. 9. Determination of the charge power limit. Psto.min ¼ 46 bar
Psto.max ¼ 216 bar

entire output voltage U generated by an electrolyzer is calculated as

follows:
tZ
0 þdt

U ¼ NElz $UElz (23) nElz ðt0 þ dtÞ ¼ nElz ðt0 Þ þ _ Elz ðsÞdðsÞ
m (27)
where NElz is the number of electrolyzer cells. t0
To establish the complete set of phenomena taking place in an To guarantee safe and environmentally friendly conditions, the
electrolyzer, two chemical aspects have to be considered: hydrogen should be compressed and stored in a pressurized tank.

3.4.1. Electro-chemical component 3.5. Compressor modeling

According to Faraday’s law, the hydrogen production rate can be
written as follows: The compressor uses a polytropic compression method. The
relationship between the compressor power Pcomp and the
NElz IElz hydrogen molar flow m_ Elz rate is calculated as follows:
mel ¼ hF (24)
pro
hF
_ Elz
Wcomp  m
where F is Faraday’s constant, n is the number of moles of trans- Pcomp ¼ (28)
acomp
ferred electrons per mole of water and hF is the faraday efficiency.
This parameter, which depends on current density, can be where acomp is the compressor efficiency, and Wcomp is the
expressed by the following empirical equation: compressor polytropic work defined by:
" #
a2 þ a3 TElz þ a4 T2Elz a5 þ a6 TElz þ a7 T2Elz KRTElz h K1
i
hF ¼ a1 exp þ (25) Wcomp ¼ ðPsto =PElz Þ K 1 (29)
IElz =AElz K1
ðIElz =AElz Þ2
where K is the polytropic coefficient, and Psto corresponding
where ai are obtained by empirical measurements. describes the hydrogen pressure in the tank.

3.4.2. Thermo-dynamic component 3.6. Hydrogen tank modeling

According to the perfect gas law, an electrolyzer pressure PElz is
described as follows: The number of H2 moles in the storage volume depends on three
_ Elzout ,
pertinent terms: the outlet flow rate from the electrolyzer m
PElz ¼ ðRTElz =VElz Þ  nElz (26)
where nElz is the amount of hydrogen produced and stored in the 900
cathode. This parameter depends on the hydrogen production rate
m_ Elz and the initial stored hydrogen amount nElz (t0) using the 800
irradiation profile (W/m²)

following equation:
700

600
Cost of providingenergy (€)

Battery 500

400

300

200
Fuel cell
100

0
0 20 40 60 80 100 120 140 160

Plim-disch P (kW) Time (s)

Fig. 10. Determination of the discharge power limit. Fig. 11. Irradiation profile.
 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209 203

4 4
x 10 x 10
2.2
2
2

1.8 1.5

1.6 1

1.4

Ppv-PL (W)
0.5
Ppv(W)

1.2
0
1

0.8 -0.5

0.6
-1
0.4
-1.5
0.2

0 -2
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Time (s) Time (s)

Fig. 12. Photovoltaic power. Fig. 14. Difference of powers.

residential load PL. If the PV panels generate more energy than is

_ fcin and the initial stored
the input flow rate to the fuel cell m demanded, the excess power PPV  PL can be used to charge the
quantity nsto (to): batteries or to produce H2 through the electrolyzer, which is called
tZ
0 þdt tZ
1 þdt
charge. Whether the surplus energy is used to charge the batteries
nsto ðt0 þdtÞ ¼ nsto ðto Þþ _ Elz out ðsÞdðsÞ
m _ FC in ðsÞ
m (30) or to produce H2 depends on the value of the limit charge power
t0 t1 (Plim.ch), as explained in the following subsections.
However, if PV modules generate less energy than is demanded,
The tank pressure Psto is deduced according to the perfect gas the lack of power PL  PPV can be supplied either by battery
law: discharge or by the fuel cell. This process is called discharge. To give
the priority to one device, the limit discharge power (Plim.dis) is
Psto ¼ ðR Tsto =Vsto Þ  nsto (31)
evaluated.
The control process described in Fig. 8 is detailed below.
4. Hybrid system control strategy
- If PPV > PL and PPV  PL < Plim.ch, then SOCbat < SOCmax; thus,
The purpose of the optimization process is to minimize the total the spare power is used to charge batteries up to their
net present cost (NPC) of the system presented in Fig. 1. The NPC is maximum state of charge for economic reasons. Therefore,
the purchase cost plus the investment cost during the lifetime of PPV  PL ¼ Pbat.ch1.
the system. To estimate the NPC of the system, we simulate the - If PPV > PL and PPV  PL < Plim.ch, then SOCbat ¼ SOC max; here,
system during its lifetime, based on assumed component lifetimes, batteries have reached their maximum state of charge, and it is
the energy produced and consumed by the components and the inconvenient to charge them. Thus, the tank pressure is
operating and maintenance costs. checked, if Psto < Psto.max, then PPV  PL ¼ PElz.ch2, otherwise
As a fundamental control rule, the power produced by renew- spare energy will be used for subsequent applications like
able source PPV should be preferentially used to supply the pumping.

0.38
12000

0.37
10000

0.36
8000
PL(W)

SOC min
SOC

6000 0.35

4000 0.34

2000 0.33

0 0.32
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160

Time (s) Time (s)

Fig. 13. Load profile. Fig. 15. Battery state of charge.

204 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209

240 hydrogen tank is undercharged, the PEMFC is unable to trigger

Psto.max the fuel combustion; thus, no possible solution is available, and
220
the system has to be stopped.
200
- If PPV < PL and PL  PPV > Plim.disch, then Psto > Psto.min. The
180 PEMFC will provide the energy needed PL  PPV ¼ PFC.disch1.
160 - If PPV < PL and PL  PPV > Plim.disch, thenPsto < Psto.min. Although
Psto (Bar)

the energy strategy opts for fuel cell use, for economic reasons,
140
batteries are employed to satisfy energy demand if and only if
120
SOCbat > SOCmin; thus, PL  PPV ¼ Pbat.disch2. If the batteries are
100 undercharged, the solution is to stop the system completely.
80
Thus, the amount of discharged battery energy is the sum of two
60 Psto.min
components: Pbat.disch ¼ Pbat.disch1 þ Pbat.disch2. Likewise, the deliv-
40 ered power from the fuel cell can be written as
20 PFC ¼ PFC.disch1 þ PFC.disch2.
0 20 40 60 80 100 120 140 160
(Note that indices 1 and 2 indicate the ways chosen in the flow
chart).
Time (s)

Fig. 16. Storage tank pressure.

5. Charge/discharge process

- If PPV > PL and PPV  PL > Plim.ch, then Psto < Psto.max and the
5.1. Charge process
electrolyzer is switched on; thus PPV  PL ¼ PElz.ch1.
- If PPV > PL and PPV  PL > Plim.ch, then Psto ¼ Psto.max; here,
When more photovoltaic power is available than what is
running the electrolyzer and storing hydrogen in the tank is
demanded by the end user load, the spare energy will be used to
impractical. Thus, the battery state of charge will be checked . If
charge batteries and/or to produce hydrogen using the electrolyzer.
SOCbat < SOCmax, spare energy is used to charge batteries even
The selected charge mode depends on which of the two has the
though it is not an economical alternative and
lowest cycling energy cost for a certain power.
PPV  PL ¼ Pbat.ch2.If not (i.e. SOCbat ¼ SOCmax) PPV  PL will be
used for other applications, such as pumping.
5.1.1. Cost of charge cycle through the batteries
For a certain power P (kW), the cost of the cycling energy
In conclusion, the charge power of the battery has two
through the batteries Ccycle.bat (V) is evaluated asfollows [13]:
components Pbat.ch ¼ Pbat.ch1. þ Pbat.ch2 Similarly the electrolyzer
power is PElz ¼ PElz.ch1 þ PElz.ch2 P$1000$Cbat
Ccycle$bat ¼ (32)
eq $hbat
We now focus in the opposite case (i.e., the renewable energy CN $Nbat p $UDC $Ncycles
produces less energy than demanded by user load). We define the
following states: where Cbat (V) is the battrey bank cost, CN (Ah) is the nominal
capacity of the battery bank, Nbat_p is the number of batteries
- If PPV < PL, then PL  PPV< P lim.disch, and SOCbat > SOCmin; thus, placed in parallel, UDC (V) is the DC bus voltage, Ncycles_eq is the
the lack of energy is satisfied by the battery bank and battery lifetime in equivalent cycles and hbat is the battery
PL  PPV ¼ Pbat.disch1. Batteries will be discharged to their efficiency.
minimum state of charge.
- If PPV < PL and PL  PPV < Plim.disch; then SOCbat ¼ SOCmin; here, 5.1.2. Cost of producing hydrogen in the electrolyzer
batteries have reached their discharge threshold and are unable To produce a certain amount of gaseous hydrogen from a certain
to provide the demanded energy. The hydrogen tank pressure is power P (kW) sent to the electrolyzer then stored in the pressurized
checked. If Psto > Psto.min, the remaining energy needed will be tank for later use by the PEMFC, the cost Ccycle$ElzFC ðVÞ is calcu-
provided by the fuel cell; thus PL  PPV ¼ PFC.disch2. If not, i.e., the lated as follows:

a 2.2
x 10
4 b 2.2 x 10 4 c 2.2
x 10
4

2 2 2
1.8 1.8 1.8
1.6 1.6 1.6
Pbat.ch 2 (W)
Pbat.ch1 (W)

1.4 1.4 1.4

Pbat.ch (W)

1.2 1.2 1.2

1 1 1
0.8 0.8 0.8
0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Time (s) Time (s) Time (s)

Fig. 17. Battery power of charge.

 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209 205

a 2.2
x 10 4 b 2.2
x 10 4 c 2.2 x 10 4

2 2 2
1.8 1.8 1.8
1.6 1.6 1.6
P Elz.ch 1 (W)

P Elz.ch2 (W)
1.4 1.4 1.4

P Elz (W)
1.2 1.2 1.2
1 1 1
0.8 0.8 0.8
0.6 0.6 0.6
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Time (s) Time (s) Time (s)

Fig. 18. Electrolyzer powers.

ðCElz =LifeElz Þ þ CO&MElz þ ðCFC =LifeFC þ CO&MFC Þ 5.2. Discharge process

Ccycle$ElzFC ¼
hElz $hFC
(33) Similarly, in the case of deficient photovoltaic power, the
required energy will be provided either by batteries or by the fuel
where CElz and CFC are the electrolyzer and fuel cell purchase prices cell, depending on which of the two systems has the lowest cycling
(V), respectively, and LifeElz and LifeFC are the electrolyzer and fuel energy cost for a certain power.
cell lifetimes (h), respectively. CO&MElz and CO&MFC correspond to
electrolyzer and fuel cell maintenance and operating costs. 5.2.1. Cost of using batteries to supply energy
We denote the quantity of output energy produced by one kg of The cost of a battery bank providing a certain power P (kW) can
H2 by hFC (kWhoutput/kg of H2). The assumed value of hFC is be written as follows [14]:
explained in more detail elsewhere [13].
P$1000$Cbat
pffiffiffiffiffiffiffiffiffi
5.1.3. Charge power limit CN $Nbat p $UDC $Ncycles eq $ hbat
The charge power limit is the intersection point between the Cp:bat ¼ (35)
hinv
curves representing the costs of the power generated by the battery
and the electrolyzer as depicted in Fig. 9. where hinv is the inverter efficiency.

CElz C 5.2.2. Cost of using a fuel cell to supply energy

þ CO&M Elz þ FC þ CO&M FC
LifeElz LifeFC Assuming that the fuel cell described in this work uses the
Plimch ¼ (34)
1000$Cbat $hElz $hFC hydrogen already produced by the electrolyzer, the cost of
CN $Nbat p $UDC $Ncycles eq $hbat providing a certain power can be written as follows [14]:

CP$FC ¼ CFC =LifeFC þ CO&MFC ¼ CfixedFC (36)

We can see that for low powers (under a certain power Plimch)
it is more appropriate to use batteries to absorb the extra power up
to their maximum state of charge and if more power remains, it 5.2.3. Discharge power limit
should be sent to the electrolyzer to produce hydrogen. However, As revealed in Fig. 10, the discharge power limit is the point of
with high additional power (up to Plimch), it is cheaper to use the intersection between the two curves showing the cost of using
electrolyzer as the main energy absorber. batteries and the fuel cell to supply energy. It can be expressed by

a 0 b 0 c 0

-2000 -2000 -2000

P bat.dis 2 (W)
P bat.dis 1 (W)

-4000 -4000 -4000

P bat.dis (W)

-6000 -6000 -6000

-8000 -8000 -8000

-10000 -10000 -10000

-12000 -12000 -12000

0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Time (s) Temps (s) Temps (s)

Fig. 19. Battery powers of discharge.

206 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209

a 12000
b 12000
c 12000

10000 10000 10000

PFC.dis 1 (W)

PFC.dis 2(W)
8000 8000 8000

PFC (W)
6000 6000 6000

4000 4000 4000

2000 2000 2000

0 0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Time(s) Time (s) Time (s)

Fig. 20. Fuel cell powers of discharge.

hinv $CfixedFC necessary. Therefore, an instantaneous power flow distribution is

Plim$disch ¼ (37)
1000$Cbat considered for the load demand supply.
pffiffiffiffiffiffiffiffiffi
CN $Nbat p $UDC $Ncycles eq $ hbat Simulation results were obtained using MatlabeSimulink
software.
Obviously, for powers less than Plim.disch, it is preferable to use
batteries to compensate for the lack of electrical energy. Never-
6.1. Empty battery mode
theless, for powers up this threshold, it is better to use the fuel cell
to compensate for the lack of electrical energy.
Many cases can be evaluated with the proposed power
The cost study is based on the expected capital expenses of
balancing strategies. All of the transitions are summarized in the
currently available technologies (i.e., actual acquisition cost). The
following photovoltaic and load profiles, which start in the empty
size of different devices and the investment cost values used in this
battery mode.
work are presented in Table 3. Note that these values do not
To show evidence that the control strategy ensures total
represent the estimated potential cost of those technologies that
management of the hybrid system, we applied the radiation profile
are permanently improving rapidly.
represented in Fig. 11 to a period of 160 s. The measured electrical
According to the cost analysis and power flow parameters, the
power production of the PV array is depicted in Fig. 12. The fore-
reference powers used in sizing are Plimch ¼ 15.7 kW and
casted electrical power demand of the residence is given in Fig. 13.
Plimdisch ¼ 8.5 kW.
The first test is to evaluate the power difference between the
one given (photovoltaic power) and the one extracted (electrical
6. Simulation results and discussion
load power), as shown in Fig. 14, to determine whether charge or
discharge is needed.
Most studies include a simulation based on one-day data or
hourly data over the course of a year. However, because of the
 Between 0 s and 20 s, PPV  PL ¼ 5000 W < Plim.ch; thus, the
dynamics of the FC systems and batteries, the data should be
control strategy requires that the battery is charged to absorb
evaluated at the minutes or seconds scale to better estimate the
the excess power provided. Moreover, the initial battery state
energy management system efficiency. Meteorological data are
of charge is in the maximum state of charge
available at an hourly scale; thus, a more precise analysis is
(SOCbat ¼ 0.355 < SOCmax ¼ 0.85), as depicted in Fig. 15; thus,
PPV  PL ¼ Pbat.ch1, as shown in Fig. 17a.
4
x 10
2.2
12000
2

1.8 10000

1.6

1.4 8000
Ppv(W)

1.2
PL(W)

6000
1

0.8
4000
0.6

0.4
2000
0.2

0 0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160
Time (s) Time (s)
Fig. 21. Photovoltaic power. Fig. 22. Load power.
 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209 207

0.9 4
x 10
2

1.8

SOC max 1.6

0.85
1.4
SOC

1.2

P Elz (W)
1
0.8
0.8

0.6

0.4
0.75
0 20 40 60 80 100 120 140 160
0.2
Time (s)
0
Fig. 23. Battery state of charge. 0 20 40 60 80 100 120 140 160

Temps (s)
 Between 20 s and 40 s, PPV  PL ¼ 19,000 W > Plim.ch; thus, the Fig. 25. Electrolyzer power.
excess energy is sent to the electrolyzer to produce hydrogen
then it is stored in the tank up to the maximum storage pres-
sure Psto.max (Fig. 16). Therefore, PPV  PL ¼ PElz.ch1 as described
in Fig. 18.a; the same procedure is used between 100 s and  During the next 20 s, PPV  PL ¼ 6500W < Plim.disch; thus,
120 s. SOCbat ¼ 0.365 > SOCmin and the deficient energy will be
 Between 40 s and 50 s, excess power remains above Plim.ch, but satisfied by the battery bank: -PPV  PL ¼ Pbat.disch1.
instead of releasing the electrolyzer, batteries are charged  During the final seconds, PPV  PL < Plim.disch; thus,
because the hydrogen tank has already reached the maximum SOCbat ¼ 0.355 > SOCmin and the shortage of energy is supplied
storage pressure Psto.max. Thus, PPV  PL ¼ Pbat.ch2, as shown in by batteries until 130 s when SOCbat ¼ 0.35 ¼ SOCmin. Then, the
Fig. 17.b. fuel cell will satisfy the load demand (Fig. 20.b).
 Between 50 s and 80 s, the power difference
PPV  PL ¼ 11 kW < 0. The generated photovoltaic power is
insufficient to ensure the load autonomy. In addition, 6.2. Full battery mode
because (PPV  PL)>Plim.disch, Psto>P sto.minas shown in
Fig. 16.Thus (PPV  PL) ¼ PFC.ch1 is shown in Fig. 20.a. It should be noted that to obtain greater detail, the following
 It is noted that at 71 s, the tank reaches Psto.min; thus, the fuel curves deal with other photovoltaic and load profiles (Figs. 21and
cell cannot continue extracting hydrogen from the tank to 22) with the higher initial battery state of charge shown in Fig. 23.
continue supplying energy. According to the energy manage- We note that the pressure in the tank never exceeds Psto.max
ment system, the batteries will provide the energy needed, as
shown in Fig. 19.b, although it is not an economical substitute.
Thus, PPV PL ¼ Pbat.disch2 12000

240
10000
220
Psto max
200
8000
180
PFC (W)

160
Psto (Bar)

6000
140

120

100 4000

80

60 2000

40

20
0
0 20 40 60 80 100 120 140 160 0 20 40 60 80 100 120 140 160

Time (s) Time (s)

Fig. 24. Storage tank pressure. Fig. 26. Fuel cell power.
208 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209

4 [6] Banos R, Manzano-Agugliaro F, Montoya FG, Gil C, Alcayde A, Gómez J. Opti-

x 10
mization methods applied to renewable and sustainable energy: a review.
Renewable and Sustainable Energy Reviews 2011;15:1753e66.
1.5 [7] Kaabeche A, Belhamel M, Ibtiouen R. Sizing optimization of grid-independent
hybrid photovoltaic/wind power generation system. Energy 2011;36:1214e22.
[8] Ipsakis Dimitris, Voutetakis Spyros, Seferlis Panos, Stergiopoulos Fotis,
Papadopoulo Simira, Elmasides Costas. The effect of the hysteresis band on
1 power management strategies in a stand-alone power system. Energy 2008;
33:1537e50.
[9] Maouedj R, Deliou A, Benyoucef B. Modélisation et simulation des perfor-
mances d’une cellule photovoltaïque. CERE 2006. Hammamet, Tunisia: CD
P bat (W)

0.5 ROM; 2006.

[10] Vasebi A, Bathaee SMT, Partovibakhsh M. Predicting state of charge of lead-
acid batteries for hybrid electric vehicles by extended Kalman filter. Energy
Conversion and Management 2008;49:75e82.
0 [11] Jallouli R, Krichen L, François B, Ouali A. Modelling and control of an optimized
PV array with hydrogen system comprising a PEMFC and an electrolyzer.
International Journal of Electrical and Power Engineering 2007;1:264e73.
[12] Zhou Tao, Francois Bruno, el hadi Lebbal Mohamed, Lecoeuche Stéphane.
-0.5 Modeling and control design of hydrogen production process by using
a causal ordering graph for wind energy conversion system. IEEE International
Symposium on Industrial Electronics. Vigo, Spain: CD ROM; 2007.
[13] Bernal-Agustın Jose L, Dufo-Lopez Rodolfo. Multi-objective design and control
-1 of hybrid systems minimizing costs and unmet load. Electric Power Systems
Research 2009;79:170e80.
0 20 40 60 80 100 120 140 160 [14] Dufo-Lopez Rodolfo, Bernal-Agustın Jose L, Contreras Javier. Optimization of
control strategies for stand-alone renewable energy systems with hydrogen
Temps (s) storage. Renewable Energy 2007;32:1102e26.
[15] Tzamalis G, Zoulias EI, Stamatakis E, Varkaraki E, Lois E, Zannikos F. Techno-
Fig. 27. Battery power. economic analysis of an autonomous power system integrating hydrogen
technology as energy storage medium. Renewable Energy 2011;36:118e24.
[16] Beccali M, Brunone S, Cellura M, Franzitta V. Energy, economic and environ-
(Fig. 24), and the secondary energy systems are never activated mental analysis on RET-hydrogen systems in residential buildings. Renewable
Energy 2008;33:366e82.
together, as shown in Figs. 25e27. Moreover, when the battery [17] Zoulias EI, Lymberopoulos N. Techno-economic analysis of the integration of
reached its maximum state of charge during the last 20 s, it was hydrogen energy technologies in renewable energy-based stand-alone power
switched off, and the fuel cell provided the required energy. systems. Renewable Energy 2007;32:680e96.

The proposed energy management could be employed to stand-

alone renewable H2 systems. Glossary

a0: Output battery voltage when SOC ¼ 0%

AElz: Electrolyzer active area (m2)
7. Conclusion AFC: FC active area (m2)
ai: Parametric coefficients of an electrolyzer i ¼ {1.7}
Aini: Initial PV panel area (m2)
According to the optimal energy management strategy that APV: PV panel area (m2)
takes into account the difference between the intermittent solar B: Parametric coefficients used in concentration losses
irradiation and the varying load demand, a system sizing method is C: Equivalent electric capacitance of an FC (F)
Cbat: Battery bank acquisition cost (V)
developed to design the following system components: the PV Cbulck: Bulk capacitor (F)
panel, battery bank, PEMFC, electrolyzer and H2 storage tank. In CElz: Electrolyzer acquisition cost (V)
terms of device cost, the optimal system provides a simple method CFC: FC acquisition cost (V)
CN: Nominal capacity of the battery bank (Ah)
to evaluate the economic feasibility of a stand-alone photovoltaic CO&M: Operation and maintenance cost(V)
system. Csurface: Capacitor surface (F)
Energy management requires monitoring and controlling the Ebat: Battery bank energy (Ah)
GS: Solar irradiation (W/m2)
power storage levels by making decisions based on the system GSmax: Maximum solar irradiation (W/m2)
operating mode. As a result, it leads to the energy sustainability of GSYavg: Solar irradiation over a year (W/m2)
the power system. The thresholds of the storage levels in the Ibat: Battery current (A)
IElz: Electrolyzer current (A)
batteries and the H2 storage tank are variable but should be adapted IFC: FCcurrent (A)
based on the available equipment and the user’s needs. The results IL: Radiation current (A)
prove that the strategy gives an optimal solution to the general IPV: PV module current (A)
J: FC current density (A/m2)
system design.
Jmax: FC maximum current density (A/m2)
K: Polytropic coefficient
LifeFC: Lifetime of an FC (h)
References LifeElz: Lifetime of an Elz (h)
Mbat: Margin coefficient of a battery
MElz: Margin coefficient of an electrolyzer
[1] Rajkumar RK, Ramachandaramurthy VK, Yong BL, Chia DB. Techno-econom-
MFC: Margin coefficient of an FC
ical optimization of hybrid pv/wind/battery system using neuro-fuzzy. Energy
MPV: Margin coefficient of a PV generator
2011;36:5148e53.
Ncycle-eq: Battery lifetime in equivalent cycles
[2] Ulleberg Ø. The importance of control strategies in PV/hydrogen systems. Sol
Nbat-p: Number of batteries in parallel in a battery bank
Energy 2004;76:323e9.
NElz: Number of cells in electrolyzer stack
[3] Maclay James D, Brouwer Jacob, Samuelsen G Scott. Dynamic modeling of
nElz: Amount of hydrogen produced in the cathode electrolyzer (moles)
hybrid energy storage systems coupled to photovoltaic generation in resi-
nsto: Amount of hydrogen stored in a tank (moles)
dential applications. Journal of Power Sources 2007;163:916e25.
Pbat: Battery power (W)
[4] Li Chun-Hua, Zhu Xin-Jian, Cao Guang-Yi, Sui Sheng, Hu Ming-Ruo. Dynamic
PElz: Electrolyzer power (W)
modeling and sizing optimization of stand-alone photovoltaic power systems
PFC: Fuel cell power (W)
using hybrid energy storage technology. Renewable Energy 2009;34:815e26.
P H2 : Hydrogen pressure (Bar)
[5] Zhou Keliang, Ferreir JA, de Haan SWH. Optimal energy management strategy
PL: Load demand power (W)
and system sizing method for stand-alone photovoltaic-hydrogen systems.
Plim.dis: Limit discharge power (W)
International Journal of Hydrogen Energy 2008;33:477e89.
 R. Jallouli, L. Krichen / Energy 40 (2012) 196e209 209

Plim.ch: Limit charge power (W) UDC: DC bus voltage (V)

PLMmax: Maximum load power over a year (W) UElz: Electrolyzer voltage (V)
PLYavg: Average load power over a year (W) UPV: The PV module voltage (V)
P O2 : Oxygen pressure (Bar) V0: Battery voltage (V)
PPV: Photovoltaic power (W) VCb: Battery bulk voltage (V)
Psto: Storage pressure (bar) VCS: Battery surface voltage (V)
Psto.max: Maximum storage pressure (bar) VFC: FC voltage (V)
Psto.min: Minimum storage pressure (bar) hbat: Battery bank efficiency
R: Universal gas constant (l,atm/mol,K) hC1: DC/DC converter 1 efficiency
Ra: Equivalent electric resistance of an FC (U) hC2: DC/DC converter 2 efficiency
Rc: Membrane equivalent resistance of an FC (U) hC3: DC/DC converter 3 efficiency
Re: End resistance of a battery (U) hC4: DC/DC converter 4 efficiency
RS: Series resistance in a PV module (U) hC5: DC/DC converter 5 efficiency
Rsh: Shunt resistance in a PV module (U) hElz: Electrolyzer efficiency
Rt: Terminal resistance in a battery model (U) hFC: FC efficiency
SOCmax: Maximum state of charge of a battery hIN: DC/AC inverter efficiency
SOCmin: Minimum state of charge of a battery hMPPT: MPPT efficiency
TFC: Fuel cell temperature (K) hPV: PV panel efficiency
TElz: Electrolyzer temperature (K) xi: FC coefficients i ¼ {1,2,3,4]
Tsto: Storage temperature (K) r: Membrane specific resistivity (U/cm2)