Design of a Hybrid Wind-PV-Fuel Cell System for Powering a Desalination Plant

Riad Chedid and Hiba El Khoury Department of Electrical & Computer Engineering

American University of Beirut, 3 Dag Hammarskjold Plaza, 8-th floor

New York, NY 10017-2303, USA

Abstract- This paper presents a methodology for the design of an autonomous renewable energy system with fuel cell as the back-up generator to supply electricity to a reverse osmosis desalination plant. The sizing of system’s units is based on an operational algorithm that considers the operational features of the system. Since the algorithm generates several workable designs, of importance is only the optimal solution that satisfies all the objectives or provides the best compromise among them. For this purpose, the trade-off method is used to generate multiple plans/solutions under different futures and obtain the trade off curves. Best solutions are selected at the knee of the curve and the optimal solution is the one with the least distance from the origin to the tangent to the curve. A case study is presented to illustrate the usefulness of the proposed methodology.1 Keywords: Renewable energy, fuel cells, desalination, unit sizing.

NONMENCLATURE FC: Fuel cell LP: Linear programming LPSP: Loss of power supply probability Nm3: Normal cubic meter NWTmin: Minimum number of wind turbines NWTmax: Maximum number of wind turbines NPVmin: Minimum number of PV panels NPVmax: Maximum number of PV panels NFmin: Minimum number of Fuel cells NFmax: Maximum number of Fuel cells NWT: Number of wind turbines NPV: Number of PV panels NF: Number of fuel cells Pl(t) : Load power at hour t (kW) PWT : Electric power of the WT PPV : Electric power of a PV panel RE: Renewable energy RO: Reverse osmosis SPC: Specific power consumption Vhmax: Maximum volume of hydrogen that can be stored Vhmin: Minimum volume of hydrogen that has to remain Vh(t): Volume of hydrogen at hour t WT: Wind turbine

1

I. INTRODUCTION Until recently, desalination has only been used in

extreme circumstances due to the high energy consumption of the process. With the continuous increase in global energy concerns, renewable energy (RE) technologies offer very attractive economic solutions for deserts, remote areas, and islands that usually have abundant wind and solar resources and frequently face problems in freshwater supply. Extensive literature on unit sizing of wind-solar power plants exists. In [1], a linear programming (LP) model to determine the optimal design of autonomous or grid linked-applications is presented. The proposed methodology uses LP to minimize electricity production cost and energy related emissions while meeting load requirements reliably. Sometimes the objectives in unit-sizing problems are conflicting and the improvement of one objective leads to the deterioration of others. In [2], multi-objective LP is used to design a hybrid solar-wind system with battery storage and diesel generators. The design objectives are the minimization of capital, operation, and maintenance costs and greenhouse gas emission and the maximization of system reliability. Heuristic algorithms may be developed to get over the complexities of LP and may yield effective and feasible solutions though the optimality is not guaranteed. Such an approach is used in [3] where sizing of an autonomous wind-solar system is presented; the authors set constraints on the number of photovoltaic (PV) panels and battery charge and discharge rates, thus producing several plans from which the one satisfying a minimum set reliability criterion with minimum electricity cost is chosen to be the optimal one. In [4], the trade-off method is adopted to find the optimal wind-PV-battery configuration by combining hourly meteorological and load data. The annual total loss of power supply probability (LPSP) of different PV and battery bank combinations are calculated, and the trade-off curve between battery bank and PV array capacity is drawn for the given LPSP value. The optimal configuration is selected as the tangent to the trade-off curve with slope representing the relationship between PV panel and battery costs. Also in [5], a decision support technique is presented to help decision makers study the influencing factors in the design of a hybrid solar-wind power system (HSWPS) for grid-linked applications. Here both the Analytic Hierarchy Process and the trade-off/risk method are applied to reach an optimal solution. Finally, a software tool for the design of desalination energy plants powered by RE systems is discussed in [5]. Designers using this tool are provided with different RE technology combinations with the possibility of

1-4244-1298-6/07/$25.00 ©2007 IEEE.

comparing alternative options on the basis of different economic indicators to choose the optimal solution for a specific case.

The aim of this paper is to provide a design methodology

for an autonomous wind-solar energy system with a fuel cell (FC) back-up generator operating on hydrogen produced by the RE sources through an electrolyzer. Such a system will be used to power a reverse osmosis (RO) desalination plant. The objectives of the design are to minimize energy cost, thus decrease water cost, and maximize reliability while taking into consideration the energy available from the RE sources. An operational algorithm is proposed to generate all possible design combinations, and the trade-off method is then applied to choose the best solution.

II. PROBLEM DEFINITION The energy cost of a desalination plant represents an economic burden which makes freshwater price very high. Therefore, it is of utmost importance that the design of the autonomous RE system be economic while guaranteeing reliable power supply under variable weather conditions. The RE system used in this paper consists of wind turbines (WTs) and PV modules complemented by a hydrogen operating FC backup as shown in Fig. 1.The design objective is to find the optimal size of each RE unit, the electrolyzer and the fuel cells while keeping at a minimum the electricity production cost ($/kWh) and the expected energy not supplied (EENS)

Fig 1. System configuration

The electric power generated by the PV modules (Ppv) and the WTs (Pwt) have priority over FC (Pf) in satisfying the load. If the total electric power generated by the RE units is greater than the load, the additional electric power is used to operate the electrolyzer and produce hydrogen. If additional electric energy occurs, it is dumped through a resistor. When RE units cannot satisfy the load, the FC uses the hydrogen stored to produce electric energy. If the demand is still not satisfied, then load shedding will result.

III.SYSTEM DESCRIPTION

A. Wind Turbines

The design variable is the total number of wind turbines of pre-selected characteristics. Following are the equations used to mathematically model the output power of a wind turbine Pw:

⎪⎪

⎩

⎪⎪

⎨

⎧

==

−=

=

0

03

w

rw

rw

w

P

PP

bPaVP

P

co

cor

rci

ci

VV

VVV

VVV

VV

><<<<

<

(1)

Where:

33cir

r

VV

Pa

−= and

33

3

cir

ci

VV

Vb

−=

Pr is the rated power, and Vci, Vr, and Vco are the cut-in, rated, and cut-out wind speeds respectively

The electric power PWT of the WT is calculated as follows:

Wwwt PP η= (2)

Where Pw is the power of the wind, ηW is the combined efficiency of the gearbox, generator and associated electronics.

B. Solar Panels

PVs, are environmentally friendly RE technologies. The major obstacles for the widespread of PV systems are their initial investment cost and their relatively low conversion efficiency. In this paper, the design variable is the total number of PV panels of pre-selected characteristics. The output power PPV of a PV panel with an area A (m2) subject to a solar irradiance r (kW/m2) is given by:

PVrAPPV η= (3)

Where ηPV is the combined efficiency of the PV panel and its corresponding converter.

C. Backup Energy System

Solar and wind energy are intermittent power resources which cannot meet the energy requirements of the RO plant at all times. Normally, power systems that power RO units adopt different operation strategies such as ON/OFF unit switching if no energy storage systems exist. Yet, these approaches are not recommended since they deteriorate the performance of RO units and shorten their lifetime. In this paper, the FC system is incorporated with the isolated RE system to guarantee reliability. The hydrogen storage system consists of an electrolyzer and hydrogen storage tanks. Hydrogen is produced by water electrolysis using the surplus electric energy generated by the RE units, and is stored in tanks to be later used by the FC in periods when the energy produced by the RE units is not sufficient to meet the RO load. The storage tank has a limit on the upper and lower levels of hydrogen storage and a minimum amount of hydrogen should remain in the tank to maintain a security limit. Electrolyzers and FC are DC-current devices, and DC/AC conversion is required to connect the FC directly to an AC load or to operate the electrolyzer in parallel with the RE system. The relations between the power and hydrogen flow rate both in the electrolyzer and the FC are simplified by a linear approximation that takes into account stack and power conversion losses [6]:

)()( tVSPCtP Heee = (4)

)()( tVSPCtP Hfff = (5)

Where Pe(t) and Pf(t) are the powers consumed by the electrolyzer and produced by the FC respectively at time t. SPCe and SPCf are the specific power consumption of electrolyzer and FC respectively, both in kW/Nm3. VHe(t)and VHf(t) are the volumes of hydrogen produced by the electrolyzer and consumed by the FC respectively in Nm3.

D. Reliability Calculation

The EENS is used to model system reliability. At every hour of the year, if the total energy provided by the RE units with or without the backup system is greater than the required load; the Energy Not Supplied (ENS) for this hour is 0. Otherwise the ENS at this hour is the difference between energy required by the load and energy supplied by both the RE and the backup system. Finally, EENS is calculated by summing all ENS for every hour of the year divided by the total amount of energy required by the load Eload over the year.

load

t

E

ENS

EENS∑==

8760

1 (6)

E. Economic Analysis

The cost of a kWh produced by the RE and the FC backup system takes into consideration the capital cost, salvage value, and operation and maintenance costs. The cost is generated by dividing the summation of the present worth of all salvage values (Sk) of the equipment, the yearly operation and maintenance costs (OMk) and the initial capital costs (Ik) by Eload and the lifetime of the project, N (years). Hence, the cost is expressed as follows [1]:

NEOMSI

loadpkpk

tk

14

1⎟⎟

⎠

⎞

⎜⎜

⎝

⎛+−∑

=

(7)

Where k is an index used to account for the 4 energy system components (PV, wind, electrolyzer, and FC). Full details about the implementation of (7) can be found in [1,3].

IV. THE RO DESALINATION PLANT The load under consideration is an RO desalination plant mainly consisting of five units each with a run water transfer pump, an acid dosing system, an anti-scalant dosing system, a degasification system, a cartridge filter, two high pressure pumps, RO modules and associated feed, reject and product water piping, and an evaporation pond. The power requirement for the equipment is given in Table 1 [7].

Groundwater is lifted to a storage tank and treated. It is then divided into five streams of 0.72 m3/h each served by separate high-pressure pumps and the RO units. The final product comprises 65±70% of the original feed volume and the remaining brine passes to an evaporation pond. The high-pressure pump continuously applies the pressure required to overcome the osmotic pressure of the water and the system pressure drop, and the feed water is pumped into the RO system. The feed stream is divided into a permeate stream low in dissolved salt, and a reject stream which then becomes

Table 1 Daily energy requirements for one RO unit

Item Model and rating

Hours Wh

Pre-feed pump Grundfos CR 2-20/10.8A, 230V

7 1094.8

Anti-scalant dosing pump

LMI A7 0.2A, 230V

7 273.7

High-pressure pump Procon Model 2607XZ 3.6A, 230V

3 2111.4

High-pressure pump Procon Model 2607XZ 3.6A, 230V

7 4926.6

Stirrer operated for 10 min, 3 times/ day

Wingert 240V,37W

0.5 18.5

Total 8425 the feed to the second stage, where it is again divided into four streams of reject and product. The second stage reject is fed to the third stage as feed and the final reject from the third stage is fed into the evaporation pond. The permeate water from all the three stages is taken into the product water tank. Fig 2 illustrates the hourly load of the plant over a day. The peak load of the plant during a day is 6.5604 kW occurring between hours 9:00 and 11:00. At 8:00, the load is 0.798 kW. Between hours 12:00 and 15:00, the load is 3.5972 kW. For the rest of the day, the load is zero. The total amount of energy required by the load over 1 year is 12727 kWh. The water output of 20m3/day during 5 hours of operation per day is achieved with a plant lifetime of 10 years.

V. SOLUTION METHODOLOGY

A. Operational Algorithm

The operational algorithm suggested is inspired from the work in [3], but modified to respond to the proposed design and requirements set in section II above. A flowchart for the algorithm is shown in Fig. 3. The input data are the number of system’s units: NWTmin, NWTmax, NPVmin, NPVmax, NFmin, NFmax (see Nomenclature), hourly wind speed (m/s) and solar irradiance (kWh/m2), the units generated power Pf, SPCe, SPCf, the load and electrolyzer power Pl(t) and Pe, and the limits on the hydrogen storage Vhmax, and Vhmin. For a given number of FCs and PVs, NWT is increased until it reaches NWTmax. It is then reset to NWTmin and NPV is increased by 1. Again, NWT is increased until it reaches NWTmax. This loop continues until NPV becomes NPVmax, it is then reset to minimum and NF is increased by 1.

Fig 2. Hourly load over 1 day

NF=NFmin

NF>NFmax

END

NPV=NPVmin

Y

N

NPV>NPVmax

NF=NF+1

NWT=NWTmin

NWT>NWTmax

Y

N

NPV=NPV+1

Y

t=1

N

NwtPwt(t)+NpvPpv(t)-Pl(t)>=0

Pf=0

Pe=NWTPWT(t)+NPVPPV(t)-Pl(t)

Y

Pe<Pemin

Pe=0

Pe>Pemax

Y

N

Pe=Pemax N

Vh(t+1)=Vh(t)+Pe/SPCe

Vh(t+1)>Vhmax

Vh(t+1)=Vhmax

Pe=(Vhmax-Vh(t))SPCe

EENS(t)=0

Y

N

t=t+1

t>TN

Pe=0

Pf=Pl(t)-NWTPWT(t)-NPVPPV(t)

N

Pf>Pfmax

Pf=Pfmax

Y

Vh(t+1)=Vh(t)-Pf/SPCf

N

Vh(t+1)<Vhmin

Vh(t+1)=Vhmin

Y

Pf=(Vh(t)-Vh(t+1))SPCf

Calculate EENS(t)

N

Y Calculate electricity cost

NWT=NWT+1

Y

Fig 3. Algorithm flowchart

The process is repeated until the NFmax is attained. The total number of combinations generated is:

maxmaxmax FPVWT NNN ×× (8)

At each iteration, the hourly power available from the RE system is calculated as follows:

)()( tPNtPN PVPVWTWT + (9)

The power calculated in (9) is compared to load power

Pl(t). If it exceeds Pl(t), EENS(t) is zero and the excess energy Pe is used to charge the electrolyzer. If Pe is less than Pemin, the electrolyzer is not charged and Pe is dumped through a resistor, else it is compared to Pemax. If Pe is greater that Pemax, Pe is set to Pemax. Vh produced at t is added to existing Vh. If Vh(t)+Vh(t-1)>Vhmax, Vh(t+1) is set to Vhmax and thus Pe(t)=(Vhmax-Vh(t))SPCe.

Conversely, if the power calculated in (9) is less than

Pl(t), the FC should supply the deficient power to cover the load requirements. The power supplied by FC should not

exceed Pfmax, thus if Pf>Pfmax, Pf is set to Pfmax. A minimum volume of hydrogen Vhmin should remain in the tank. If this constraint is violated, and Vh(t+1)<Vhmin, then Vh(t)=Vh(t+1)-Vhmin, and Pf is modified accordingly. Finally, the EENS(t) is calculated for every hour by:

)()()(()()( tPNtPNtPNtPtEENS fFPVPVWTWTl ++−= (10)

This procedure is repeated for every hour t of the year

for a total of T=8760 hours for all combinations generated. After the EENS(t) is calculated for every hour of the year the sum of all EENS(t) is divided by the total yearly energy required by the load to obtain the overall EENS. The algorithm was programmed using MATLAB.

B. Trade-off solution

In many studies, optimal sizing of stand alone systems has been based on optimization techniques that minimize a cost function. In such approaches, average values for varying renewable resources as well as the load are

adopted. In this paper, the sizing of system’s units is based on a simulation that considers operational features of the system. Therefore, since several designs may be considered workable solutions, of importance is only the optimal solution that satisfies all the objectives or provides the best compromise among them. For this purpose, the trade-off method is used to generate multiple plans under different futures and obtain the trade off curves. In this respect, a future is understood as a set of outcomes or realizations of all the uncertainties, and a plan is considered as a set of specified options [5]. Hence, all the combinations generated by the algorithm discussed above are plotted on a 2-D graph to obtain the trade-off curves [8]. Then, all combinations having the EENS more than0.06 are considered inferior designs and hence are eliminated. The reliability and the cost objectives are set to have equal importance and are assigned a weight of 50% each. Best solutions are selected at the knee of the curve and the optimal solution is found to be the one with the least distance from the origin to the tangent to the curve.

VI. RESULTS AND DISCUSSION Seven futures and 60 plans are developed to study the

robustness of the optimal plan [8]. Table 2 summarizes the futures that are considered in the tradeoff analysis. For all futures, the inflation rate is considered to be 3%, and a decrease in the capital costs of wind turbines, PVs, and fuel cells is projected. The capital cost is set to decrease to $750/kW for wind turbines, to $3000/kW for PV, and to either $5400/kW or $5100/kW for the fuel cells. In addition the interest rate (IR) is made to vary between 5%, 10%, and 15%. Table 3 shows the range and variation step of the design variables.

Table 2 Description of futures

Future #

WT cost $/kW

PV cost $/kW

FC cost $/kW

IR Electricity cost $/kWh

1 1015 4401 6000 10% 0.5407 2 750 4401 6000 10% 0.5350 3 1015 3000 6000 10% 0.3974 4 1015 4401 5400 10% 0.5361 5 1015 4401 5100 10% 0.5338 6 1015 4401 6000 5% 0.5098 7 1015 4401 6000 15% 0.5592

Table 3

Range and variation step of variables Number of

wind turbines Number of PV panels

Number of fuel cells

Min 1 1 1 Max 2 10 3 Step 1 1 1

Fig.4 illustrates the trade-off curves, obtained through

the algorithm shown in Fig.3, for future 1 whose characteristics are given in table 2. The minimum EENS is 0.0087 for an electricity cost of $2.3413/kWh and the minimum electricity cost is $0.3071/kWh for an EENS of 0.0999. According to the design criterion, the EENS should not exceed 0.06, and it is observed that only 38 plans out of a total of 60 plans generated satisfy this criterion. From the acceptable plans, the least distance concept is applied and the optimal plan is determined. The latter is found to have two 5 kW wind turbines, two 1 kW PV panels, and one 5kW fuel cell. The electricity cost was

$0.5407 /kWh for an EENS of 0.0571. The share of each component in the annual cost is illustrated in Fig.5. It can be seen that the PV, for example, has the maximum share of 83.44% with a contribution of only 11.32% to the total energy production. A decrease in the PV capital cost will decrease the cost of generated electricity significantly. WTs have the minimum cost share of 3.17% but with a maximum annual energy contribution of 82.06 %. Similarly, the cost of fuel cells contributes to 8.82% of total cost but their contribution to energy production is limited to only 6.62%.

Fig. 4. Trade-off curve

Fig 5. Share of system components in the total electricity cost.

The autonomous system operation is tested during different months. Fig.6 shows the simulation results of a day in April with a deficiency in RE existing between 9:00 and 11:00 a.m. Therefore the FC has to supply the deficient power. Fig. 7 shows the simulated renewable energy power versus load power during April. The volume of hydrogen in the tank during April is shown in Fig. 8, and is used to supply the fuel cell whose generated power in April is shown in Fig. 9.

Fig 6. Hourly renewable, load, and FC power during a day in April

Fig 7. Renewable power versus load power during April

Fig 8. Volume of hydrogen in the tank during April

Fig 9. Power supplied by the fuel cell in April

VII. CONCLUSIONS

This paper has presented a methodology for designing a hybrid wind-PV-Fuel cell system to power a desalination plant taking into consideration the operational features of the system. An operational algorithm has been proposed for that purpose, and so it calculates the electric power generated by the PV modules and the WTs and uses it to satisfy the load. If the total electric power generated by the RE system is bigger than the load, the additional electric power is used to operate an electrolyzer and produce hydrogen. When RE sources cannot satisfy the load, the FC uses the hydrogen stored to produce energy. The problem was constrained by a minimum cost and maximum reliability criteria, and hence the algorithm was used to generate a number of design combinations and then the desired solution was obtained by using the tradeoff method. The design does that not claim to be optimal yet it allows the identification of plans representing a reasonable compromise among the set objectives. A case study based on a hypothetical site was

presented to illustrate the usefulness of the proposed methodology.

VIII. REFERENCES 1. R.Chedid and S.Rahman, ‘Unit Sizing and Control of hybrid

Wind-Solar Power Systems’, IEEE Trans. on Energy Conversion, Vol. 12, No.1,1997, pp.79-85.

2. R. Chedid, S. Karaki and A. Rifai. A multi-Objective Design Methodology for Hybrid Renewable Energy Systems. Proceedings of the IEEE PowerTech Conference, Saint Petersburg, Russia, June 29, 2005.

3. Z.M.Salameh and B.S.Borowy, ‘Methodology for Optimally Sizing the Combination of a Battery Bank and PV Array in a Wind/PV Hybrid System’, IEEE Trans. on Energy Conversion, Vol. 11, No. 2, 1996, pp.367-375.

4. B.Ai, H.Yang, H. Shen, and X. Liao, ‘Computer-aided Design of PV/Wind Hybrid System’, Renewable Energy, Vol. 28, No. 10, 2003, pp. 1491-1512.

5. Chedid R, Akiki H, Rahman S. ‘A decision support technique for the design of hybrid solar-wind power systems’. IEEE Trans. Energy Conversion, 13 (1), 1998, pp 76-83

6. Manolakos D., Papadakis G., Papantonis D., and Kryitis S., ‘A simulation-optimisation programme for designing hybrid energy systems for supplying electricity and fresh water through desalination to remote areas Case study: the Merssini village, Dounoussa island, Aegean Sea, Greece’, Energy, Vol. 26, No.7, 2001, pp 679-704.

7. N. Korpas, ‘Distributed Energy systems with Wind Power and Energy Storage’, PhD thesis, Norwegian University of Science and Technology, Trondheim, March 2004.

8. Al Suleimani Z. and Nair V.R., ‘Desalination by solar-powered reverse osmosis in a remote area of the Sultanate of Oman’, Applied Energy, Vol .65, No.1, 2000, pp 367-380.

BIBLIOGRAPHIES Riad B. Chedid received his Ph.D. in Electrical Engineering from the University of London, and the DIC from Imperial College of Science Technology and Medicine, UK. At present, Dr. Chedid is full Prof. at the Department of Electrical and Computer Engineering, American University of Beirut. His research interests include alternative energy systems and energy policy and planning. Hiba El Khoury received her B.E. and M.S. degree from the American University of Beirut in 2003 and 2005 respectively. Her research interests include renewable energy systems and power system deregulation.

Appendix

Characteristics of future 1: SPCe: 4.9 kWh/Nm3, SPCf: 1.5 kWh/Nm3 Inflation rate: 3%, Interest rate:10%. Project life span: 10 years Wind turbine life span: 20 years, PV panel life span: 25 years Fuel Cell life span: 5 years, Eelectrolyzer life span: 15 years Wind turbine capital cost: 1015 $/kW, PV panel capital cost: 4401 $/kW Fuel cell capital cost: 6000 $/kW Electrolyzer capital cost: 3000 $/kW Wind turbine OM costs: 25$/kW, PV OM costs: 88.01 $/kW Fuel Cell OM costs: 240 $/kW, Electrolyzer OM costs: 120 $/kW WT salvage value: 203$/kW, PV salvage value: 880 $/kW Fuel cell salvage value: 300 $/kW, Electrolyzer salvage value: 150 $/kW NWTmin: 1; NWTmax: 2; NPVmin: 1; NPVmax: 10; NFmin: 1; NFmax: 3; Vhmax: 20 Nm3; Vhmin: 5 Nm3. Vci: 2.8 m/s, Vr: 10.5 m/s, Vco: 14 m/s Rated wind turbine power: 5 kW, Rater PV power: 1kW Rated electrolyzer power: 5kW, Rated fuel cell power: 5kW