Solar Energy 139 (2016) 666–675

Contents lists available at ScienceDirect

Solar Energy

journal homepage: www.elsevier .com/locate /solener

Design of a cost-effective wind/photovoltaic/hydrogen energy system forsupplying a desalination unit by a heuristic approach

http://dx.doi.org/10.1016/j.solener.2016.09.0280038-092X/� 2016 Elsevier Ltd. All rights reserved.

⇑ Corresponding authors.E-mail addresses: akbar.maleki20@yahoo.com, a_maleki@ut.ac.ir (A. Maleki),

pourfayaz@ut.ac.ir (F. Pourfayaz).

Akbar Maleki ⇑, Fathollah Pourfayaz ⇑, Mohammad Hossein AhmadiDepartment of Renewable Energies, Faculty of New Science & Technologies, University of Tehran, Tehran, Iran

a r t i c l e i n f o

Article history:Received 4 June 2016Received in revised form 2 August 2016Accepted 22 September 2016Available online 4 November 2016

Keywords:Hybrid renewable energy systemFuel cell/electrolyser storage systemDesalination unitOptimization

a b s t r a c t

The techno-economic assessment, optimization, and sizing of grid independent renewable energy sys-tems affects not only the likelihood of deployment but also their reliability to supply potable waterand electricity where needed. Despite many investigations on hybrid renewable energy systems (photo-voltaic/wind), the reverse-osmosis desalination unit powered by solar and wind electricity productionsystems with hydrogen energy storage, and effects of integrating water desalination alongside meetingload demand, is rarely found. In this paper, a hybrid photovoltaic/wind/hydrogen/reverse osmosis desali-nation system is modeled and designed for increasing the fresh water availability and to meet the loaddemand to a stand-alone region in Iran. The configuration of the proposed hybrid system is optimallydetermined with respect to two optimization criteria, the life cycle cost of the economic evaluationand the loss of power supply probability concept for the reliability. For this aim, an efficient metaheuristictechnique based on artificial bee swarm optimization is used. From the results it is seen that at a max-imum loss of power supply probability set to 0–10% the photovoltaic/hydrogen/reverse osmosis desali-nation is the most cost-effective energy system and photovoltaic/wind/hydrogen/reverse osmosisdesalination and wind/hydrogen/reverse osmosis desalination systems are in the other ranks.

� 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Water and energy are necessary for the existence and the devel-opment of modern societies. However, most water, approximately97%, is saltwater in the oceans and the remaining 3% is freshwater,and about 25% of the world’s population does not have access toadequate quantities of fresh water (Koutroulis and Kolokotsa,2010). Additionally, freshwater scarcity is becoming an increas-ingly significant problem in many areas around the world (Heet al., 2015). Water desalination offers a promising and viable tech-nology for providing drinking water (Shannon et al., 2008), but itswidespread use is impeded by its high economic cost, especiallydue to high energy consumption (Fritzmann et al., 2007; Heet al., 2015). Furthermore, the current use of traditional fossil fuelsas the main power source for desalination is increasing concernsabout climate change and the need to reduce greenhouse gas emis-sions (He et al., 2015; Subramani et al., 2011).

Nowadays the method of reverse osmosis (RO) dominates glob-ally because of its ability to desalinate water with relatively lowenergy requirements and costs, it requires only electricity, and

can operate using renewable energy technologies such as wind tur-bines and photovoltaics (Garcí, 2003; Koutroulis and Kolokotsa,2010; Spyrou and Anagnostopoulos, 2010). The use of renewableenergy sources for driving RO desalination units is particularlyfavored in remote areas (Bourouni et al., 2011; Koroneos et al.,2007). So, desalination systems powered by hybrid renewableenergy systems (HRESs) offer a promising option for many remotesmall villages and cities in mainland regions and many small citiesand villages in coastal areas.

If desalination systems powered by HRESs are optimized, theycan be more cost-effective and reliable. The main advantage of adesalination system powered by HRESs is the ability to developsmall scale desalination plants. The electricity from photovoltaic(PV) collectors and wind turbines (WTs) can be used to drivehigh-pressure pumps in reverse osmosis plants. The fuel cell actsas a backup power supply for periods when the demand is high.The system includes an electrolyser to produce hydrogen fromexcess electrical energy generated by the renewable energysources (solar and wind).

RO desalination units having various energy supply combina-tions have been reported, including PV/RO (Joyce et al., 2001),WT/battery/RO (Tzen and Morris, 2003), PV/battery/RO (Massonet al., 2005), WT/RO (de la Nuez Pestana et al., 2004; Koklas andPapathanassiou, 2006), PV/diesel/RO (Helal et al., 2008), PV/WT/

A. Maleki et al. / Solar Energy 139 (2016) 666–675 667

diesel/RO (Setiawan et al., 2009; Voivontas et al., 2001), PV/WT/-battery/RO (Bourouni et al., 2011; MOkheimer et al., 2013), andPV/FC/battery/RO (Clarke et al., 2015). Al Malki et al. (1998) inte-grated renewable energy sources (solar and wind) to power anRO system for desalinating brackish water in Oman and demon-strated, for the first time in Oman, that solar power can be usedto run an RO desalination plant to produce fresh water but thatit needs a backup energy source for continuous operation. de laNuez Pestana et al. (2004) studied an RO plant connected directlyto a wind system without energy storage. Spyrou andAnagnostopoulos (2010) developed a computer algorithm to simu-late and economically evaluate a stand-alone desalination systempowered by renewable energy sources and a pumped storage unit.Stochastic optimization software based on evolutionary algorithmsis implemented to optimize the plant design for various objectives,like minimization of fresh water production cost and the maxi-mization of water needs satisfaction. Koutroulis and Kolokotsa(2010) presented a methodology for the optimal sizing of desalina-tion systems, driven by PV modules and WTs. The purpose of theproposed methodology is to derive, among a list of commerciallyavailable system devices, the optimal number and type of unitsso that the 20-year total system cost is minimized consumer waterdemands are completely met. The total cost function minimizationis implemented using genetic algorithms and it was found that thetotal cost of the desalination system is highly affected by the oper-ational characteristics of the devices comprising the system, whichaffect the degree of exploitation of the available solar and windenergy potentials. Cherif and Belhadj (2011) estimated and evalu-ated over a large time period energy and water production from aPV/WT hybrid system coupled to a RO desalination unit in south-ern Tunisia. Bourouni et al. (2011) presented a new model basedon genetic algorithms for coupling a small RO unit to renewableenergy sources (PV/WT/batteries/RO) for Ksar Ghilène village insouthern Tunisia. Hossam-Eldin et al. (2012) used a mathematicalmodel and a related computer program for sizing hybrid renewableenergy systems combined with RO desalination. MOkheimer et al.(2013) proposed a model for the optimization of a hybrid wind–solar-powered RO water desalination system in Saudi Arabia. Heet al. (2015) proposed a novel RO seawater desalination plant pow-ered by photovoltaics and pressure-retarded osmosis, and investi-gated the feasibility of two stand-alone schemes: salinity-solarpowered RO operation and salinity powered RO operation.Although studies on various aspects of RO desalination-basedhybrid systems have been reported in the literature, informativemodels and efficient optimization tools for optimal sizing andtechno-economic analysis are seldom found.

Nevertheless, some optimization techniques for sizing hybridsystems have been reported (Ekren and Ekren, 2010; Maleki andAskarzadeh, 2014b; Merei et al., 2013). Nonlinear programming(Roy, 1997) and HOMER (Hybrid Optimization of Multiple EnergyResources) (Fazelpour et al., 2014; Sen and Bhattacharyya, 2014)are two common algorithms for the optimal design of hybrid sys-tems. HOMER energy modeling software is a particularly powerfultool for designing this type of system, but when the number of pos-sible design points is very high; this method can require excessivecalculation time. Another design tool, HOGA (Hybrid Optimisationby Genetic Algorithms) developed by Bernal-Agustín and Dufo-López (2009), uses an evolutionary algorithm for the design ofhybrid systems. HOGA is capable of applying an enumerative algo-rithm and is useful for validating the results reached by means ofthe evolutionary algorithm (Ekren and Ekren, 2010). Therefore,we propose here using heuristic methods, such as heuristic algo-rithms, to solve these kinds of optimization problems, based in parton the usefulness of heuristic algorithms in solving such time-consuming optimizations. Heuristic algorithms such as harmonysearch (Maleki and Askarzadeh, 2014b), genetic algorithm (Mellit

et al., 2010; Merei et al., 2013), simulated annealing (Ekren andEkren, 2010; El-Naggar et al., 2012; Garlík and Krivan, 2013), par-ticle swarm optimization (Maleki et al., 2016a), and others (Malekiet al., 2015; Sinha and Chandel, 2015) are techniques for sizinghybrid systems.

In this article, we consider the continuous variables (photo-voltaic area and the swept area of wind turbines) and the integervariable (number of hydrogen tanks) in the optimization modelfor a hybrid PV/WT/hydrogen/RO system. We propose an efficientversion of heuristic algorithms for optimally designing stand-alone desalination systems driven by renewable energy sources(PV and WT) in the remote area of Davarzan, Khorasan, Iran. Forachieving this objective, an artificial bee swarm optimization(ABSO) algorithm is proposed for determining the optimum sizingof the PV/WT/hydrogen/RO system.

2. Mathematical modeling

Fig. 1 presents the structure of the hybrid energy system whichforms the focus of this study. This structure consists of wind tur-bine (WT), photovoltaic (PV) panel, a fuel cell (FC) coupled to anelectrolyser, DC/DC converters and an inverter which is used tointerface the DC voltage to the consumer load AC requirements.The WT module and the PV work together to meet the loaddemand. If the total electric energy generated by the renewableenergy sources is greater than the load, the excess electric poweris used to operate the electrolyser and produce hydrogen. Whenpower shortage takes place, stored hydrogen converts into electri-cal energy by using FC which is considered as a secondary powersource. The optimization of a hybrid renewable energy system(HRES) is important and leads to a good balance between perfor-mance and cost. A water tank is used to store the surplusdesalinated-water produced, which is not used directly by thedesalination system consumer.

2.1. Basic models

2.1.1. Modeling of PV powerThe output power of a PV panel can be obtained as follows:

PPV ðtÞ ¼ gPVRtAPV ð1Þwhere PPV is the power generated by PV panels (kW), Rt is the solarinsolation (kW/m2), APV is the total area occupied by the set of PVpanels (m2) and gPV represents the efficiency of the PV panels,which is given by Bakelli et al. (2011):

gPV ¼ grgpc 1� b Tair þ NOCT � 20800

� �Rt

� �� Tref

� �� �ð2Þ

where gpc is the power conditioning efficiency (equal to 1 if a per-fect maximum power tracker is used), gr is the reference moduleefficiency, Tref is the cell temperature at the reference conditions(usually set to 25 �C), b is the photovoltaic panel efficiency temper-ature coefficient (�3.7 � 10�3 �C�1 for mono and polycrystalline sil-icon (Ismail et al., 2013)), Tair is the ambient air temperature (in �C),R is the solar radiation, and NOCT (normal operating cell tempera-ture) is the nominal cell operating temperature (43 �C here), typi-cally stipulated by the manufacturer.

2.1.2. Modeling of wind powerThe output power of a wind turbine is calculated as follows

(Caballero et al., 2013):

PWTðtÞ ¼0; vðtÞ 6 Vci or vðtÞ P Vco

Pr �v3ðtÞðV3

r �V3ci� Pr �V3

ci

V3r �V3

cið Þ ; Vci < vðtÞ < Vr

Pr ; Vr 6 vðtÞ < Vco

8>><>>:

ð3Þ

Electrolyser

Fuel Cell

Wind Turbine

PV Panel

Load

Power Management Unit

RO UnitsWaterStorage

HydrogenStorage

Fig. 1. Schematic of the hybrid renewable energy system.

668 A. Maleki et al. / Solar Energy 139 (2016) 666–675

where Vci is cut-in speed, v is the wind speed, Vr is the rated windspeed, Vco is cut-off speeds of the wind turbine (in m/s), and Pr isthe rated power of the wind turbine (in kW) which is defined onthe area swept by the wind turbine blades AWT, the power coeffi-cient CP, the reducer efficiency gr, the air density qa, and themechanical efficiency of wind turbine gWT:

Pr ¼ 12AWTCPgrqagWTV

3r ð4Þ

2.1.3. Modeling of fuel cell/electrolyserA simplified fuel cell (FC) model is used in this study. We

assume that the FC works at a fixed operation point, so that thechoice of appropriate size of the hydrogen tanks requires an inclu-sive analysis of the electrolyser (Ele) and FC requirements. Thetotal power generated (PG) by the PVs andWTs at hour t is obtainedby:

PG ¼ PPV þ PWT ð5ÞAt a specific time, depending on the load demand (PL), the total

power generated may or may not be sufficient to supply therequired power. A one hour time step is used.

The storage system works as follows:If PGðtÞ P PLðtÞ

gInv, then the electrolyser is used to fill the hydrogen

tanks. The amount of hydrogen stored in the hydrogen tanks(HST) can be obtained as follows:

HSTðtÞ ¼ HSTðt � 1Þ þ EGðtÞ � ELðtÞgInv

� �gEle ð6Þ

When PGðtÞ < PLðtÞgInv

, the FC is used to supply the load. In this case, the

amount of hydrogen in the tanks at hour t is obtained as follows:

HSTðtÞ ¼ HSTðt � 1Þ � ELðtÞgInv

� EGðtÞ� �

=gFC ð7Þ

where HST(t) and HST(t � 1) are the energy stored in the hydrogentanks at hours t and t � 1, respectively, gEle is the efficiency of theelectrolyser, EG is the generated energies, gInv is the efficiency ofthe inverter, EL is the load demand (in kW h), and gFC is the effi-ciency of the FC.

2.1.4. Modeling of reverse-osmosis desalination (ROD) unitThe correlation between the electric power required for desali-

nation and the hourly water demand is obtained assuming an aver-age specific energy consumption for desalination SDC and an hourlywater demand HWD (in m3). Today, reverse-osmosis desalination(ROD) units require 2–5 kW h to produce 1 m3 of fresh water(MOkheimer et al., 2013; Spyrou and Anagnostopoulos, 2010),

and a value of 4 kW h/m3 is assumed in the present study to berepresentative of the entire desalination process and equipment(membranes, pumps and energy recovery systems). Thus, the cor-relation between the desalination electric power (PDEM) and thehourly water demand is obtained as follows:

PDEM ¼ HWDSDC ð8ÞThe desalination unit installed power (PDI) is considered as a

system free variable, so the desalination water production capabil-ity per day (DWC) can be expressed as (Spyrou andAnagnostopoulos, 2010)

DWC ¼ 24PDI

SDC

� �ð9Þ

The ROD unit can operate between its a minimum (PMD) andnominal loads, because of the membrane’s characteristic curve(Manolakos et al., 2009; Spyrou and Anagnostopoulos, 2010):

PMD 6 PDES 6 PDI ð10Þwhere PDI is the installed power (or nominal load) of the ROD unit,PDES is the instantaneous power consumption by the ROD unit andPMD is the a minimum load of the ROD unit. The lower operationlimit corresponds to the minimum required pressure to overcomethe osmotic pressure and to design the desalination unit. Workingwith small loads may affect the quality of the produced fresh water,but this is not a subject of the current study (Spyrou andAnagnostopoulos, 2010). In this study, the technical minimum istaken at 25% of the nominal power.

The ROD system includes a fresh water tank, the capacity ofwhich is proportional to the total daily fresh water demand, DWD

(in m3). A period of two days is considered as the desirable auton-omy of the desalination system. Therefore, the capacity of the freshwater tank, VWTa, is obtained as:

VWTa ¼ 2DWD ð11Þ

2.2. Life cycle cost (LCC)

The Life Cycle Cost (LCC) approach is used in this study for thecost analyses of the hybrid PV/WT/hydrogen/RO, PV/hydrogen/ROand WT/hydrogen/RO systems. LCC analysis involves finding thenet present value of all costs that are incurred over the life spanof the hybrid system. The LCC is formulated as follows:

LCC ¼ CCnpv þMC þ TCMR þ TCCH ð12Þwhere CC is the annual capital cost, MC is the annual maintenancecost, TCMR is the annual membrane replacement cost and TCCH is

A. Maleki et al. / Solar Energy 139 (2016) 666–675 669

the annual cost of chemicals of the RO desalination unit, all in 2014US dollars. The subscript of ‘npv’ denotes the net present value ofeach factor.

2.2.1. LCC of PV arrayBy considering the area APV for the PV panels of the hybrid sys-

tem, the initial capital cost of the PV panels can be determined as

CCnpv ;PV ¼ APVCPVirð1þ irÞnp

ð1þ irÞnp � 1

� �ð13Þ

where APV is the area of solar panels; and CPV is the unit cost of PVpanels, which is defined as the sum of the panel price and panelinstallation fee, np denotes the life span and ir is the interest rateof the system.

The annual maintenance cost can be evaluated as follows:

MCPV ¼ CMnt�PVAPV ð14Þwhere CMnt-PV is the annual maintenance cost of each PV panel.

2.2.2. LCC of WTsAs for the PV system, the capital cost of the WTs is obtained by

considering the area swept by the wind turbine blades (AWT) for thehybrid system:

CCnpv ;WT ¼ AWTCWT � irð1þ irÞnpð1þ irÞnp � 1

� �ð15Þ

where CWT is the unit cost of theWTs, which is defined as the sum ofWT price and WT installation fee.

The annual maintenance cost is expressible as

MCWT ¼ CMnt�WTAWT ð16Þwhere CMnt-WT is the annual maintenance cost of each WT.

2.2.3. LCC of FCWe now consider the fuel cell (FC) for the storage system. The

capital cost of the FC is obtained as:

CCnpv ;FC ¼ PWFC � irð1þ irÞnpð1þ irÞnp � 1

� �ð17Þ

Some components of PV/WT/hydrogen/RO system need to bereplaced several times over the project’s lifetime. Here, the lifetimeof the fuel cell is assumed to be five years (Khan and Iqbal, 2005).Using the single payment present worth factor (PWFC), we have

PWFC ¼ CFC

Xk¼0;5;10;15

1

ð1þ irÞkð18Þ

where CFC is the FC cost.The FC annual maintenance cost can be expressed as

MCFC ¼ CMnt-FC ð19Þwhere CMnt-FC is the annual maintenance cost of each fuel cell.

2.2.4. LCC of electrolyserThe LCC for the electrolyser is obtained in the same way as

explained for the fuel cell in Section 2.2.3, except with the super-script ‘FC’ replaced by ‘Ele’.

2.2.5. LCC of hydrogen tank (H2)The capital cost of NH2 hydrogen tanks for the storage system

can be determined as follows:

CCnpv ;HT ¼ NH2 � CH2 � irð1þ irÞnpð1þ irÞnp � 1

� �ð20Þ

where CH2 is unit cost of one hydrogen storage tank, and NH2 is thenumber of storage tanks.

2.2.6. LCC of ROD unitSimilarly, the capital cost of the investment for ROD for the

hybrid system is obtained as

CCnpv;ROD ¼ CRODCaWD � irð1þ irÞnpð1þ irÞnp � 1

� �ð21Þ

where CaWD is the capacity of the ROD system (in m3/day) and CROD

is the cost of ROD system per m3 of the capacity of the desalinatedwater of the ROD system per day.

The annual maintenance cost is expressible as

MCROD ¼ CMnt�ROD � DWD ð22Þwhere CMnt-ROD is the annual maintenance cost of the ROD unit.

The capital cost of the fresh water tank of the ROD unit isexpressed as:

CCnpv;ROD ¼ CWTaVWTa � irð1þ irÞnpð1þ irÞnp � 1

� �ð23Þ

where CWTa is the water tank cost, which is defined as the sum ofwater tank price and water tank installation fee.

The annual membrane replacement cost is expressible as

TCMR ¼ CMR � CaWD � NMe ð24Þwhere CMR is the membrane replacement cost and NMe is the num-ber of membrane replacements per year.

The annual cost of chemicals of the ROD unit is expressible as

TCCH ¼ CCHDWD ð25Þwhere CCH is the cost of chemicals.

2.3. The proposed constraint objective function

The optimization goal is to minimize the life cycle cost byadjusting three decision variables subject to some constraints.The objective function is defined as follows:

Minimize LCCðAWT ;APV ;NH2Þ ¼X

m2WT;PV ; HT;FC;Ele; ROD

CCnpv;m

þMCm þ TCMR þ TCCH ð26Þand the subsequent constraints are satisfied:

AWT P 0APV P 0 ð27ÞNH2 P 0LPSP� P LPSP

In this problem, the three decision variables (NH2, AWT, and APV) areoptimally modified, where NH2 is an integer decision variable andAWT and APV are continuous decision variables.

Since the maximum energy stored in the reservoir tanks cannotexceed the maximum HST (HSTmax), the following relation isapplied during optimization:

HSTðtÞ 6 HSTmax ð28ÞIn order to supply the load, the HST must satisfy the minimum

HST (HSTmin). So, during use of the FC, the following constraintmust be satisfied:

HSTðtÞ P HSTmin ð29ÞDuring use of the FC, if HST(t) < HSTmin, a percentage of the load

is not satisfied. In this case, HST(t) is set to HSTmin (HST(t) = HSTmin)and the loss of power supply (LPS) is calculated by:

LPSðtÞ ¼ ELðtÞgInv

� EGðtÞ � ½HSTðt � 1Þ � HSTmin�gFC ð30Þ

670 A. Maleki et al. / Solar Energy 139 (2016) 666–675

In hybrid renewable energy systems, good reliability requiresconsideration of the loss of power supply probability (LPSP), whereLPSP is a number between 1 and 0. A LPSP of 0 means that the loadwill always be satisfied, while a LPSP of 1 means that the load willnever be satisfied. For a specified period T, the LPSP can be evalu-ated as follows:

LPSP ¼PT

t¼1LPSðtÞPTt¼1ELoadðtÞ

ð31Þ

where LPSP⁄ is the maximum allowable LPSP. Fig. 2 shows the com-puter program used in analyzing the performance of the HRES.

Historical data(Wind speed, Solar radiation,

Temperature, Water demand, and Electricity demand)

Send po

Sizing of the hybrid energy system by ABSO

Optimal siz

Hydrogen pro

Simulation of the hybrid energy system during each hour

Fuel cell on to met the load demand

End

Over production?

Init

ReseYes

Reservoir tanks energy > 0

No

Y

Overall dema

No

Yes

No

Start

Fig. 2. Flowchart for the computer program used

3. Artificial bee swarm optimization

Inspired by the intelligent behaviors of honey bees (collectionand processing of nectar), a group of optimization algorithmsbased on bee swarm behaviors has been developed. The maindifference between a bee swarm and other population-basedalgorithms such as particle swarm optimization (PSO) andgenetic algorithm (GA) is that in a bee swarm, different kindsof bees are employed which use different types of trajectoriesto amend their positions. In order to implement the ABSO(Artificial Bee Swarm Optimization), the steps shown in Fig. 3should be followed. More explanation about the ABSO can be

wer to dump load

ing successful

duction with electrolyser

LPS (t)

ial conditions

rvoir tanks full?

es

No

nd satisfied ? Fitness function satisfied ?Yes

No

Yes

in analyzing the performance of the HRES.

Step 1: A swarm of bees are randomly initialized in the search space

Step 3: The bees are ranked based on their objective function values

Step 4: Onlooker and scout bees are specified

Step 5: The position of the onlooker and scout bees is updated according to their patterns

Step 6: If a bee exceeds the search space, it is replaced with the previous position

Step 7: Steps 2 to 6 are repeated until itermax is met

Step 8: The best achievement of the bee swarm is selected as the optimal solution

Step 2: The value of the objective function for each bee is computed

Steps of ABSO implementation:Steps of ABSO implementation:

Fig. 3. The steps of ABSO implementation.

Fig. 4. (a) Irradiance and temperature monthly data for one year. (b) Wind speed at10 m height and electricity demand during one year for Davarzan.

A. Maleki et al. / Solar Energy 139 (2016) 666–675 671

found in (Maleki and Askarzadeh, 2014a; Maleki and Pourfayaz,2015).

Fig. 5. Typical hourly variation of water consumption during a day.

4. Results and discussion

The results are presented of the optimal sizing and operationsimulation of the examined HRES, based on actual data. Firstly, ahybrid wind/photovoltaic/hydrogen/reverse osmosis desalinationsystem having a proper design is examined and the results are ana-lyzed. Then, the hybrid photovoltaic/hydrogen/reverse osmosisdesalination and wind/hydrogen/reverse osmosis desalination sys-tems are examined, and general conclusions are deduced.

The ambient temperature, solar radiation, and wind speed ofthe location are measured by the solar radiation and meteorologi-cal station located in the Davarzan, Iran, which has the followinggeographical coordinates: latitude = 36.35� (36�2100600N), longi-tude = 56.83� (56�5204200E), and altitude = 912 m (approx.) abovesea level. The data are recorded each ten minutes per day. Themonthly mean values of the solar irradiation on a horizontal plane,air temperature and wind speed are calculated and illustrated inFig. 4. The monthly load profile during a year is depicted inFig. 4b. Variation curves for typical fresh water demand is depictedin Fig. 5. Finally, parameters related to the components are pre-sented in Table 1 (Maleki et al., 2016b, MOkheimer et al., 2013;WTP, 2014; WSS, 2014).

In order to implement the proposed methodology, the HRESoptimization problem is coded in the software of MATLAB. Param-eters related to the heuristic algorithm are given in Table 2 (Malekiand Pourfayaz, 2015).

The heuristic algorithm seek the optimum design for the PVpanels, wind turbines and hydrogen tanks in different HRESs so

as to minimize the life cycle cost and satisfy the constraints regard-ing LPSP⁄ (maximum loss of power supply probability). The algo-rithm considers the hybrid PV/WT/hydrogen/RO system, thehybrid PV/hydrogen/RO system, and the hybrid WT/hydrogen/ROsystem. This problem includes three decision variables (NH2, AWT,and APV) where NH2 is an integer decision variable and AWT andAPV are continuous decision variables. Because of the stochasticnature of ABSO algorithm, different runs lead to different results.For this aim, ABSO algorithm is run 30 times and the best solutionof the algorithm (the minimum fitness function value found by thealgorithm over the runs) is reported. In ABSO, swarm size and max-imum number of iterations are set to 30 and 100, respectively.

4.1. Hybrid photovoltaic/wind/hydrogen/reverse osmosis desalinationsystem

In this section, the RO desalination station powered by therenewable energy system is examined to assess its potential for

Table 3The optimal characteristics of the hybrid wind/photovoltaic/hydrogen/reverse osmosis desalination system by a heuristic approach over 30 independent runs.

LPSP⁄ (%) APV (m2) FC units AWT (m2) Electrolyser units NH2 RO units LCC ($) LPSP (%)

0 275.52 1 72.22 1 50,163 1 8327623.9 01 260.76 1 75.36 1 45,416 1 7541882.6 0.99992 359.16 1 6.28 1 36,972 1 6141994.3 1.99995 177.12 1 109.9 1 30,997 1 5156018.9 4.999910 180.4 1 87.92 1 22,344 1 3722457.6 9.9999

Table 1The parameters used in the studied system.

Parameters Value Parameters Value

Interest rate (ir) 5% FC efficiency (gFC ) 50%Project lifetime (np) 20 years FC lifetime 5 yearsNominal wind turbine power (Pr-WT ) 1 kW FC cost (CFC ) $20,000Cut-in wind speed (Vci) 3 m/s Replacement cost of FC $1400Cut-out wind speed (Vco) 20 m/s Nominal electrolyser power 3 kWNominal wind speed (Vr) 9 m/s Electrolyser efficiency (gEle) 74%Wind turbine price (CWT ) $1804 Electrolyser lifetime 5 yearsArea swept by wind turbine blades (AWT ) 3.14 m2 Electrolyser cost (CEle) $20,000Annual maintenance costs of wind turbine (CMnt-WT ) 100 $/year Replacement cost of electrolyser $1400Power coefficient (CP) 0.59 Reservoir tanks cost (CHT ) $2000Air density (qa) 1.225 kg/m3 Nominal capacity of hydrogen tank 0.3 kW hWind turbine efficiency (gWT ) 85% Hydrogen tank lifetime 20 yearsWind turbine lifetime 20 years Annual maintenance costs of hydrogen tank 5 $/yearNominal PV power (Pr-PV ) 260 W Capacity of ROD system 10 m3/dayPV panel cost (CPV ) $468 Cost of RO system per m3 of capacity of the desalinated

water of RO system/day (CROD)532 $/m3/day

Annual maintenance costs of PV (CMnt-PV ) 10 $/year Maintenance cost of RO system (CMnt-ROD) 0.2 $/m3

PV area (APV ) 1.64 m2 Membrane replacement cost (CMR) 0.06 $/m3

PV efficiency (gPV ) 15.8% Number of membrane replacements per year (NMe) 2PV lifetime 20 years Cost of chemicals (CCH) 0.06 $/m3

Nominal FC power 3 kW Water tank cost (CWTa) 255.4 $/m3

Nominal converter/inverter power 3 kW Converter/inverter efficiency (gInv ) 95%Converter/inverter price PConv=Inv $1583 Converter/inverter lifetime 10 yearsAnnual maintenance costs of converter/inverter (CMnt-Inv) 15 $/year

Table 2The parameters used for ABSO algorithm.

Parameter Colony size ne wbmax wemax wbmin wemin itermax smax smin

Value 30 5 2.5 2.5 1.25 1.25 100 0.2 0.2

Fig. 6. The loss of power supply probability and life cycle cost of hybridphotovoltaic/wind/hydrogen/reverse osmosis desalination system for differentLPSP⁄.

672 A. Maleki et al. / Solar Energy 139 (2016) 666–675

increasing the fresh water availability and to meet the loaddemand for a small region. Table 3 shows the size optimizationof the PV/WT/hydrogen/RO system. As can be seen, in this table,the optimal system size has been obtained with respect to differentvalues of LPSP⁄ (0–10%). By increase of loss of power supply prob-ability, the reliability of the system decreases. At LPSP⁄ = 0%, theoptimal values of the decision variables found by ABSO areAPV = 275.52 m2, AWT = 72.22 m2 and NH2 = 50,163. In this case,the values of LCC and LPSP are $8327623.9 and 0% respectively.At LPSP⁄ = 2% which is a desired LPSP value in power systems tosatisfy the system reliability, the optimal values of the decisionvariables found by ABSO are APV = 359.16 m2, AWT = 6.28 m2, andNH2 = 36,972. By increase of LPSP⁄, as it was predictable, the valueof NH2 and LCC decreases so that at LPSP⁄ = 10%, we haveAPV = 180.4 m2, AWT = 87.92 m2, NH2 = 22,344, LCC = 3722457.6 $,and LPSP = 9.9999%.

The life cycle cost of hybrid systems for different LPSP is calcu-lated by the proposed optimal sizing method. The results for ahybrid PV/WT/hydrogen/RO system are demonstrated in Fig. 6. Itis seen that higher power reliable systems are more expensive thanlower requirement systems. Selection of an optimal system config-uration according to system power reliability requirements canhelp avoid blind capital spending and save investments.

Because the reliable supply of the load demand at any timedepends strongly on the amount of the stored energy, Fig. 7 illus-trates the hourly expected amount of stored energy in the hydro-gen tanks during a year for the different LPSP⁄ values and hybrid

Fig. 7. Hourly expected amount of stored energy in the hydrogen tanks during ayear for the desired LPSP.

Fig. 8. System configurations for the desired LPSP.

A. Maleki et al. / Solar Energy 139 (2016) 666–675 673

PV/WT/hydrogen/RO system. It is obvious that loss of load ismostly probable at the hours that stored mass of hydrogen reachesto its minimum allowable limit. Therefore, using the resultingoptimally sized hybrid PV/WT/hydrogen/RO system enables bothreliable power-supply to the RO desalination unit and full exploita-tion of the energy storage capability, contributing towards theminimization of the life cycle cost.

Fig. 9. The loss of power supply probability and life cycle cost of hybrid wind/hydrogen/reverse osmosis desalination system for different LPSP⁄.

4.2. Hybrid wind/hydrogen/reverse osmosis desalination system

The results obtained using the considered algorithms for theHRES are listed in Table 4. In this study, the hybrid WT/hydro-gen/RO system is considered for fresh water supply and to meetthe load demand for a remote area. It is seen that the optimal val-ues of AWT and NH2 at LPSP⁄ = 0% are determined to be 436.46 m2

and 88,744, respectively. In comparison with the PV/WT/hydro-gen/RO system, LCC increases to $14732390.3. At LPSP⁄ = 10%, theoptimal values of the decision variables found by ABSO areAWT = 357.96 m2 and NH2 = 51,273. In this case, the value of LCCand LPSP are $8525376.4 and 9.7439%, respectively. As can be seen,in comparison with LPSP⁄ = 0%, the number of storage tanks andtotal swept area of the rotating turbine blades decreases signifi-cantly (Fig. 8).

The loss of power supply probability and life cycle costs ofhybrid WT/hydrogen/RO system for different LPSP⁄ are demon-strated in Fig. 9.

4.3. Hybrid photovoltaic/hydrogen/reverse osmosis desalinationsystem

In this subsection, the hybrid photovoltaic/hydrogen/reverseosmosis desalination system is considered for fresh water supplyand to meet the load demand for a remote area. In this system,photovoltaic panels produce power to supply the load and the fuelcell-electrolyser with hydrogen tanks is used to store excessenergy and deliver it in deficit conditions. The optimal design ofthe hybrid photovoltaic/hydrogen/reverse osmosis desalination

Table 4The optimal characteristics of the hybrid wind/hydrogen/reverse osmosis desalination sys

LPSP⁄ (%) FC units AWT (m2) Electrolyser units

0 1 436.46 11 1 423.9 12 1 408.2 15 1 379.94 110 1 357.96 1

system corresponding to the Best solution obtained with the ABSOalgorithm is provided in Table 5.

Fig. 10 illustrates the convergence process of ABSO algorithmfor finding the optimum size of the hybrid photovoltaic/hydro-gen/reverse osmosis desalination system with respect to differentvalues of LPSP⁄. Convergence process shows the best value of theobjective function at each iteration of the ABSO algorithm. This fig-ure is corresponding to the best performance of the ABSO algo-rithm over 30 runs in each case.

At LPSP⁄ = 0%, the optimal value of NH2 is determined to be46,629. In comparison with the PV/WT/hydrogen/RO system, LCCdecreases to $7740153.8, and APV increases to 393.6 m2. AtLPSP⁄ = 2%, the optimal values of the decision variables found byABSO are APV = 372.28 m2 and NH2 = 36,136. In this case, the valuesof LCC and LPSP are $6003099.6 and 1.9892% respectively. Toenhance understanding, the loss of power supply probability andlife cycle costs of hybrid photovoltaic/hydrogen/reverse osmosisdesalination system for different LPSP⁄ are demonstrated in Fig. 11.

The relationships between system reliabilities or LPSP and sys-tem configurations are studied. Fig. 12 shows the results of therelationship between LPSP⁄ values and system configurations. It

tem by a heuristic approach over 30 independent runs.

NH2 RO units LCC ($) LPSP (%)

88,744 1 14732390.3 081,094 1 13465449.7 0.970572,936 1 12118665.9 1.964059,342 1 9862389.6 4.399551,273 1 8525376.4 9.7439

Fig. 13. Life cycle cost for the optimized hybrid systems for the desired LPSP.Fig. 10. Convergence process of ABSO for finding the optimum size of the hybridphotovoltaic/hydrogen/reverse osmosis desalination system for the desired LPSP.

Fig. 11. The loss of power supply probability and life cycle cost of hybridphotovoltaic/hydrogen/reverse osmosis desalination system for different LPSP⁄.

Table 5The optimal characteristics of the hybrid photovoltaic/hydrogen/reverse osmosis desalination system by a heuristic approach over 30 independent runs.

LPSP⁄ (%) APV (m2) FC units Electrolyser units NH2 RO units LCC ($) LPSP (%)

0 393.6 1 1 46,629 1 7740153.8 01 382.12 1 1 41,343 1 6865066.5 0.99992 372.28 1 1 36,136 1 6003099.6 1.98925 337.84 1 1 19,821 1 3302210.4 4.999910 309.96 1 1 13,204 1 2206386.6 9.9999

Fig. 12. System configurations for the desired LPSP.

674 A. Maleki et al. / Solar Energy 139 (2016) 666–675

is shows that when the system reliability is higher; the systemconfiguration (number of storage tanks and total area occupiedby the set of photovoltaic panels) is higher too for the same capac-ity of water and load.

When we set LPSP⁄ to 0–10%, it is seen that economically thephotovoltaic/hydrogen/reverse osmosis desalination system is abetter choice since its life cycle cost are $7740153.8, $6865066.5,$6003099.6, $3302210.4, and $2206386.6 which is lower than thatof the other hybrid systems (PV/WT/hydrogen/RO and WT/hydro-gen/RO).

Fig. 13 graphically compares the life cycle cost of differenthybrid systems for the desired LPSP⁄ (0–10%). The cost analysisof three hybrid systems namely: PV/WT/hydrogen/RO, PV/hydro-gen/RO, and WT/hydrogen/RO are scrutinized in this figure. It canbe seen that the PV/hydrogen/RO is the most cost-effective hybridsystem and PV/WT/hydrogen/RO and WT/hydrogen/RO systemsare in the other ranks, and with increase of LPSP⁄, the life cycle costof the system decreases.

5. Conclusions

This article presents a framework for size optimization of a photovoltaic/wind/hydrogen/reverse osmosis desalination hybridenergy system in which three decision variables of the numberof hydrogen tanks, photovoltaic area, and swept area of wind tur-bines are optimally determined for having a cost-effective and reli-able energy system. The configuration of the proposed hybridsystem is optimally determined with respect to two optimizationcriteria, the life cycle cost of the economic evaluation and the lossof power supply probability concept for the reliability. The opti-mization problem is solved by consideration of three different sys-tems and effects of integrating water desalination alongsidemeeting load demand using artificial bee swarm optimization(ABSO). From the obtained results it can be drawn that the combi-nation of photovoltaic, hydrogen, and desalination systems (photovoltaic/hydrogen/reverse osmosis desalination) leads to the mostcost-effective system for different maximum loss of power supplyprobability (0–10%). Moreover, ABSO can be efficiently used tosolve the optimum sizing problems. Also, the metaheuristic algo-rithm could be used to evaluate larger systems, with more thanone of each type of energy source. The present research providesa useful reference for the design of hybrid renewable energy

A. Maleki et al. / Solar Energy 139 (2016) 666–675 675

sources with a reverse osmosis desalination unit. Finally, it isworthwhile to mention that the results obtained in this study arevalid for the considered case study since the results of hybridrenewable energy systems are directly affected by the environ-mental conditions (i.e. solar radiation, temperature, and windspeed) and the components type. Designing a hybrid system inanother location with other types of the components than the onesused in this study may lead to different results.

References

Al Malki, A., Al Amri, M., Al, Jabri H., 1998. Experimental study of using renewableenergy in the rural areas of Oman. Renew. Energy 14, 319–324.

Bakelli, Y., Arab, A. Hadj, Azoui, B., 2011. Optimal sizing of photovoltaic pumpingsystem with water tank storage using LPSP concept. Sol. Energy 85, 288–294.

Bernal-Agustín, J.L., Dufo-López, R., 2009. Efficient design of hybrid renewableenergy systems using evolutionary algorithms. Energy Convers. Manage. 50,479–489.

Bourouni, K., M’Barek, T.B., Al, Taee A., 2011. Design and optimization ofdesalination reverse osmosis plants driven by renewable energies usinggenetic algorithms. Renew. Energy 36, 936–950.

Caballero, F., Sauma, E., Yanine, F., 2013. Business optimal design of a grid-connected hybrid PV (photovoltaic)-wind energy system without energystorage for an Easter Island’s block. Energy 61, 248–261.

Cherif, H., Belhadj, J., 2011. Large-scale time evaluation for energy estimation ofstand-alone hybrid photovoltaic–wind system feeding a reverse osmosisdesalination unit. Energy 36, 6058–6067.

Clarke, D.P., Al-Abdeli, Y.M., Kothapalli, G., 2015. Multi-objective optimisation ofrenewable hybrid energy systems with desalination. Energy 88, 457–468.

de la Nuez Pestana, I., Latorre, F.J.G., Espinoza, C.A., Gotor, A.G., 2004. Optimizationof RO desalination systems powered by renewable energies. Part I: wind energy.Desalination 160, 293–299.

Ekren, O., Ekren, B.Y., 2010. Size optimization of a PV/wind hybrid energyconversion system with battery storage using simulated annealing. Appl.Energy 87, 592–598.

El-Naggar, K., AlRashidi, M., AlHajri, M., Al-Othman, A., 2012. Simulated annealingalgorithm for photovoltaic parameters identification. Sol. Energy 86, 266–274.

Fazelpour, F., Soltani, N., Rosen, M.A., 2014. Feasibility of satisfying electrical energyneeds with hybrid systems for a medium-size hotel on Kish Island, Iran. Energy73, 856–865.

Fritzmann, C., Löwenberg, J., Wintgens, T., Melin, T., 2007. State-of-the-art of reverseosmosis desalination. Desalination 216, 1–76.

Garcí, L., 2003. Renewable energy applications in desalination: state of the art. Sol.Energy 75, 381–393.

Garlík, B., Krivan, M., 2013. Renewable energy unit commitment, with differentacceptance of balanced power, solved by simulated annealing. Energy Build. 67,392–402.

He, W., Wang, Y., Shaheed, M.H., 2015. Stand-alone seawater RO (reverse osmosis)desalination powered by PV (photovoltaic) and PRO (pressure retardedosmosis). Energy 86, 423–435.

Helal, A., Al-Malek, S., Al-Katheeri, E., 2008. Economic feasibility of alternativedesigns of a PV-RO desalination unit for remote areas in the United ArabEmirates. Desalination 221, 1–16.

Hossam-Eldin, A., El-Nashar, A., Ismaiel, A., 2012. Investigation into economicaldesalination using optimized hybrid renewable energy system. Int. J. Electr.Power Energy Syst. 43, 1393–1400.

Ismail, M., Moghavvemi, M., Mahlia, T., 2013. Techno-economic analysis of anoptimized photovoltaic and diesel generator hybrid power system for remotehouses in a tropical climate. Energy Convers. Manage. 69, 163–173.

Joyce, A., Loureiro, D., Rodrigues, C., Castro, S., 2001. Small reverse osmosis unitsusing PV systems for water purification in rural places. Desalination 137, 39–44.

Khan, M., Iqbal, M., 2005. Pre-feasibility study of stand-alone hybrid energy systemsfor applications in Newfoundland. Renew. Energy 30, 835–854.

Koklas, P.A., Papathanassiou, S.A., 2006. Component sizing for an autonomous wind-driven desalination plant. Renew. Energy 31, 2122–2139.

Koroneos, C., Dompros, A., Roumbas, G., 2007. Renewable energy drivendesalination systems modelling. J. Clean. Prod. 15, 449–464.

Koutroulis, E., Kolokotsa, D., 2010. Design optimization of desalination systemspower-supplied by PV and W/G energy sources. Desalination 258, 171–181.

Maleki, A., Askarzadeh, A., 2014a. Artificial bee swarm optimization for optimumsizing of a stand-alone PV/WT/FC hybrid system considering LPSP concept. Sol.Energy 107, 227–235.

Maleki, A., Askarzadeh, A., 2014b. Optimal sizing of a PV/wind/diesel system withbattery storage for electrification to an off-grid remote region: a case study ofRafsanjan, Iran. Sustain. Energy Technol. Assess. 7, 147–153.

Maleki, A., Pourfayaz, F., 2015. Optimal sizing of autonomous hybrid photovoltaic/wind/battery power system with LPSP technology by using evolutionaryalgorithms. Sol. Energy 115, 471–483.

Maleki, A., Ameri, M., Keynia, F., 2015. Scrutiny of multifarious particle swarmoptimization for finding the optimal size of a PV/wind/battery hybrid system.Renew. Energy 80, 552–563.

Maleki, A., Khajeh, M.G., Ameri, M., 2016a. Optimal sizing of a grid independenthybrid renewable energy system incorporating resource uncertainty, and loaduncertainty. Int. J. Electr. Power Energy Syst. 83, 514–524.

Maleki, A., Pourfayaz, F., Rosen, M.A., 2016b. A novel framework for optimal designof hybrid renewable energy-based autonomous energy systems: a case studyfor Namin, Iran. Energy 98, 168–180.

Manolakos, D., Kosmadakis, G., Kyritsis, S., Papadakis, G., 2009. Identification ofbehaviour and evaluation of performance of small scale, low-temperatureOrganic Rankine Cycle system coupled with a RO desalination unit. Energy 34,767–774.

Masson, L., Richards, B.S., Schäfer, A.I., 2005. System design and performance testingof a hybrid membrane—photovltaic desalination system. Desalination 179, 51–59.

Mellit, A., Kalogirou, S.A., Drif, M., 2010. Application of neural networks and geneticalgorithms for sizing of photovoltaic systems. Renew. Energy 35, 2881–2893.

Merei, G., Berger, C., Sauer, D.U., 2013. Optimization of an off-grid hybrid PV–Wind–Diesel system with different battery technologies using genetic algorithm. Sol.Energy 97, 460–473.

MOkheimer, E.M., Sahin, A.Z., Al-Sharafi, A., Ali, AI., 2013. Modeling andoptimization of hybrid wind–solar-powered reverse osmosis waterdesalination system in Saudi Arabia. Energy Convers. Manage. 75, 86–97.

Roy, S., 1997. Optimal planning of wind energy conversion systems over an energyscenario. IEEE Trans. Energy Convers. 12, 248–254.

Sen, R., Bhattacharyya, S.C., 2014. Off-grid electricity generation with renewableenergy technologies in India: an application of HOMER. Renew. Energy 62, 388–398.

Setiawan, A.A., Zhao, Y., Nayar, C.V., 2009. Design, economic analysis andenvironmental considerations of mini-grid hybrid power system with reverseosmosis desalination plant for remote areas. Renew. Energy 34, 374–383.

Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G., Marinas, B.J., Mayes, A.M.,2008. Science and technology for water purification in the coming decades.Nature 452, 301–310.

Sinha, S., Chandel, S., 2015. Review of recent trends in optimization techniques forsolar photovoltaic–wind based hybrid energy systems. Renew. Sustain. EnergyRev. 50, 755–769.

Spyrou, I.D., Anagnostopoulos, J.S., 2010. Design study of a stand-alone desalinationsystem powered by renewable energy sources and a pumped storage unit.Desalination 257, 137–149.

Subramani, A., Badruzzaman, M., Oppenheimer, J., Jacangelo, J.G., 2011. Energyminimization strategies and renewable energy utilization for desalination: areview. Water Res. 45, 1907–1920.

Tzen, E., Morris, R., 2003. Renewable energy sources for desalination. Sol. Energy 75,375–379.

Voivontas, D., Misirlis, K., Manoli, E., Arampatzis, G., Assimacopoulos, D., 2001. Atool for the design of desalination plants powered by renewable energies.Desalination 133, 175–198.

WSS, 2014. <http://www.wholesalesolar.com/> (Date Accessed: 2014/26/04).WTP, 2014. <http://www.alibaba.com/products/water_tanks_prices.html> (Date

Accessed: 2014/26/04).

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具