Desalination 257 (2010) 137149

Contents lists available at ScienceDirect

Desalination
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / d e s a l

Design study of a stand-alone desalination system powered by renewable energy

sources and a pumped storage unit
Ioannis D. Spyrou, John S. Anagnostopoulos
School of Mechanical Engineering/Fluids Section, National Technical University of Athens, 9 Heroon Polytechniou ave., Zografou, 15780 Athens, Greece

a r t i c l e

i n f o

Article history:
Received 12 December 2009
Received in revised form 20 February 2010
Accepted 23 February 2010
Available online 29 March 2010
Keywords:
Stand-alone desalination
Wind power
Photovoltaics
Pumped storage
Plant operation simulation
Design optimization

a b s t r a c t
The aim of this work is to investigate in detail the optimum design and operation strategy of a stand-alone
hybrid desalination scheme, capable to fulll the fresh water demand of an island or other remote coastal
regions. The scheme consists of a reverse-osmosis desalination unit powered by wind and solar electricity
production systems and by a pumped storage unit.
A specic computer algorithm is developed to simulate in detail the entire plant operation and also to
perform economic evaluation of the investment. A stochastic optimization software based on evolutionary
algorithms is implemented to accomplish design optimization studies of the plant for various objectives, like
the minimization of fresh water production cost or the maximization of water needs satisfaction.
Miscellaneous parametric studies are also conducted in order to analyze the effects of various critical
parameters, as population, water pricing, water demand satisfaction rate and photovoltaics cost are.
The results demonstrate not only the performance, the role and the contribution of each subsystem but also
the production and economic results of the whole plant. An optimally designed scheme is found to be
economically viable investment, although energy rejections are signicant and there is a clear need for better
exploitation of renewable energy production surplus.
2010 Elsevier B.V. All rights reserved.

1. Introduction
Nowadays it is observed, globally, an extensive phenomenon of
drought. Especially in Greece, many isolated areas, such as Aegean
islands, suffer from drought [1]. The problem becomes worse in
summer when the water demand increases up to 45 times compared
to winter because of tourism [2]. In most islands the existing water
stocks cannot satisfy such increasing demand; thus the problem that
comes up must be solved with permanent and viable solutions. At
present, this water demand is being satised by tank transportation
with the considerably high cost of about 58 /m3 for Cyclades and
Dodecanese complex [1].
Seawater desalination can play an important role towards a
permanent confrontation of the problem [2]. The installation of
desalination units is a common solution throughout the world, in
areas with drought. In the last decades the number of desalination
applications has greatly increased, while desalination is the subject of
several research works. As a result, new desalination methods have
been developed, experience has been gained, system operation has
been amended and the equipment production has become massive
[3]. Thus, the two most important performance characteristics of such

Corresponding author. Tel.: + 30 2107721080.

E-mail address: anagno@uid.mech.ntua.gr (J.S. Anagnostopoulos).
0011-9164/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.desal.2010.02.033

applications, which are the quality of produced water and the water
production cost are continuously being improved.
A critical technical parameter of desalination applications is the
way the system is powered. This decision is taken according to the
selected method of desalination and the characteristics of the
candidate area [4]. Nowadays the method of reverse osmosis
dominates globally; it requires only electricity, has a quite low
specic energy demand, and can cooperate with technologies of
renewable energy sources (RES) such as wind turbines and photovoltaics [47]. Concerning Aegean islands that suffer from drought,
most of them are isolated and the electricity is provided by local
conventional power stations operating at very high production costs.
In addition, the power demand of large desalination units may not be
satised by the existing power stations. On the other hand, Aegean
islands feature an abundance of RES like solar and wind energy.
Consequently, a desalination system powered by hybrid renewable
energy technologies would be a very promising solution for those
regions.
Several simulation studies of desalination units cooperating with
RES and conventional thermal units have already taken place [811].
Stand-alone reverse-osmosis desalination units powered by wind
turbines and/or photovoltaics and supported by batteries have been
the research topic of several works that have shown that such systems
could be a viable solution at present conditions [1214]. Energy
storage is an important aspect of such autonomous systems, although
some systems without storage have been simulated and tested

138

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

[15,16]. Because of the stochastic and intermittent nature of RES, a

storage system is usually required in order to avoid excessive
rejection of the energy production, as well as to guarantee the
desalination unit operation during unfavorable weather conditions.
The most common system for energy storage are batteries which are
included in most studies. The use of batteries has the disadvantages of
short life cycle, high-cost maintenance and environmentally unfriendly content [17,18]. As a result, batteries have been proposed
only for small-scale plants.
The present work aims to study an alternative means of energy
storage, a pumped storage subsystem, in stand-alone desalination
plants, and to investigate its role on the operation of the whole
scheme. Pumped storage in hybrid wind-hydro power production
plants has been studied applying numerical design optimization
methodologies in some previous works [19,20]. The optimum sizing of
all desalination system components that maximize its energy and/or
economic results for small to large islands constitutes an objective of
the present work, as well.

2. System description
The considered system is schematically shown in Fig. 1. The standalone desalination unit is powered by a hybrid RES system (windphotovoltaics), and includes a fresh water tank to provide autonomy
to the area for a determinate time period, in case that the system is out
of order for some reason (e.g. maintenance, failure).
Due to the intermittent nature of those RES and the difculty in
predicting the energy production rate, a means of energy storage is
required to operate the desalination unit even during unfavorable
weather conditions. A pumped storage subsystem is considered, as an

alternative to batteries. A typical pumped storage unit consists of a

pumping and a turbining station, two water reservoirs at different
altitude, and the necessary pipelines (Fig. 1). A number of pumps are
usually installed in parallel operation, equipped with variable speed
motors in order to be able to absorb the uctuating production of RES
with no power gaps [19]. On the other hand, the type of hydroturbine(s)
depends mainly on the available head between the reservoirs.
During periods of excessive RES production, the power surplus is
used to operate the pumps and store hydraulic (dynamic) energy in
the upper reservoir (Fig. 1). On the other hand, when the primary
energy production cannot satisfy the desalination demands, then the
hydroturbine re-transforms the stored energy into electricity, which
powers the desalination unit. Details on the energy transfer rules and
constrictions are given in Section 3.
3. Algorithm description
In order to simulate the entire system operation and its
subsystems interaction, a specic computer algorithm is developed.
The software makes also an economic evaluation of the system, based
on empirical cost relations for all its main components (pumps, hydro
and wind turbines, photovoltaics, pipelines, etc.) [19,20]. The
algorithm is divided into three sections: data input, application of
system's logic operation, and techno-economic evaluation. These
parts are described in more detail in the present section.
3.1. Data input
Data needed for the simulation of system's operation are either xed
or free variables, the value or the range of which is predetermined by the

Fig. 1. Sketch of the examined desalination system set-up.

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

user. In the present study xed quantities include the hydraulic head
(here 400 m), the pipeline dimensions, the number of pumps and
turbines, and the time-variation curves of wind speed and solar
radiation. On the other hand, the size of all main system components
(desalination unit, wind and photovoltaic parks, pumping and turbining
systems, reservoirs and fresh water tank), constitutes the free design
variables. Finally, some critical parameters like the resident population,
the water demand satisfaction rate and the photovoltaics (PV) cost are
also introduced and examined in the present work. Detailed description
of the main system parameters, along with their technical and
operational constrictions, is given below.
3.1.1. Water demand
Calculation of the hourly water demand of a candidate area is
based on an ideal island of constant resident population, RP. The
seasonal population, SP, of the island is assumed here to follow a
monthly variation curve like the one drawn in Fig. 2, which is typical
for a Greek island at the Aegean Sea: population in summer months
may increase due to tourism up to 3 or 4 times [1]. The resident
population is a major parameter of the system, and usually depends
on the island size.
Assuming a seasonally varying specic daily consumption, SWC for
seasonal population and RWC for resident population (Table 1), the
total daily fresh water demand, DWD (m3), is obtained as:
DWD = RP SWC + SP RWC :

Then, based on typical fresh water demand variation curves [21],

as shown in Fig. 3, a percentage of daily water demand, WDP,
corresponds to every hour i of the day. As a result, the hourly water
demand, HWD (m3), of the entire population is calculated as:
HWD = WDP DWD :

3.1.2. Desalination unit

The correlation between hourly water demand, and the
corresponding power for desalination, PDEM (or the hourly energy
demand, EDEM) is given by Eq. (3), assuming an average specic
energy consumption for desalination, SDC. Today, reverse-osmosis
desalination units require 24 kW h to produce 1 m3 of fresh water
[5], and a value of 3 kW h/m3 is taken in the present study to stand for
the entire desalination process and apparatus (pumps, membranes
and energy recovery systems).
PDEM = EDEM = HWD SDC :

Fig. 2. Monthly variation of seasonal, in respect to resident population.

139

Table 1
Typical specic water consumption data.
Daily water consumption (l/day/person)
Population type
Resident, RWC
Seasonal, SWC

Winter
150
200

Summer
250
300

The desalination unit installed power, PD,I, is considered as a

system free variable. Thus, desalination water production capability
per day, DFC, can be computed from the relation:
DFC = 24PD = SDC :

The reverse osmosis unit can operate between its nominal and a
minimum load, PMD, because of membrane's characteristic curve [22]:
PMD PDES PD;I

where PDES is the instant power consumed by desalination unit. The

lower operation limit corresponds to the minimum required pressure to
overcome the osmotic pressure and to set desalination unit up. Working
in partial loads may affect the quality of the produced fresh water, but
this is not a subject of the present work. In the present simulation, the
technical minimum is taken at 25% of the nominal power.
The desalination system includes a fresh water tank, the capacity
of which is proportional to the island size (or to the resident
population). A period of two summer days (August) was considered
as the desirable autonomy of the system. Thus, the capacity of the
fresh water tank, VTC, is:
VTC = 2DWD :

In order to conserve the above autonomy throughout the year, the

hydroturbine is put in operation when the content of the tank drops
below 90%, providing that there exist available storage in the upper
reservoir.
3.1.3. RES features and production
Hourly energy production by wind turbines and photovoltaics is
being calculated using the corresponding meteorological data for
wind speed, V, and solar radiation, G (Fig. 4). Specic data time series
are used for all cases examined here, assuming that they are
representative of the Aegean islands complex [1].
Concerning the calculation of wind farm production, the power
curve of a typical wind turbine of specic nominal power, PMWT, is
used. The installed wind power, PW,I, is a free variable of the system.
For every hour the given wind speed corresponds to a specic power

Fig. 3. Typical hourly variation of water consumption during a day.

140

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

given by the following relation, where HP is the pumping head and P

is the total efciency of the pumps:

QP =

PPS P
3600
gHP

PMP PPS PP;I :

12

13

Hydroturbines can operate at variable power load, PHT, in order to

supplement the RES production. However, operation at partial loads
may affect efciency. In this work the latter follows a typical
characteristic curve of Pelton turbines, the technical minimum of
which is taken 20% of the nominal power load, PMT. Therefore, the
hourly water ow, QT (m3), through the hydroturbine is computed as:

QT =
Fig. 4. Indicative meteo data time series (typical week of August).

PHT
3600
T gHT

PMT PHT PT;I

production, PNWT, and the total wind farm production, PWT (or EWT in
kW h), is calculated by the following equations, where WT is the wind
farm's total efciency:
PWT = EWT = WT kPNWT

14

15

where HT is the net head and T is the turbine efciency. The upper
and lower reservoirs have equal useful capacity, which is a system's
design parameter.

7
3.2. System operation strategy

PW;I
:
k=
PMWT

On the other hand, the power production by the PV array, PPV (or
EPV in kW h), is computed by Eq. (9), assuming that there are NPV
identical PV panels of a specic nominal power PNPV each, average
efciency PV, and panel's area A.
PPV = EPV = PV NPV AG:

The total installed photovoltaic power is considered as a free

variable of the system and is given by:
PPV;I = NPV PNPV :

10

In every time-step of the plant operation simulation, the total RES

production, PRES, is the sum of wind turbines and PV production:
PRES = ERES = PPV + PWT :

11

The time-step used in the present study was always 1 h, hence the
values of the produced energy ERES (in kW h), and power PRES (in kW)
are exactly the same, allowing the treatment of even the energy
quantities in terms of power. This practice is followed throughout this
work.
3.1.4. Pumped storage subsystem
The pumping station is consisted by a number of variable speed
pumps in parallel operation, with total installed power PP,I. The pumps
operate only when the available RES power exceeds a technical
minimum, which is taken 15% of the installed power. This corresponds
to a number of at least 4 pumps, with the capability of reduction of
their rotation speed by about 15%. For any available wind power, PPS,
the hourly water ow, QP, from the lower to the upper reservoir is

The basic power/energy balance relations for every time-step (1 h)

are given below. The rst stands for the RES produced power, which
can be spent for desalination, PAD, or for pumping, PPS, while the
remaining (if any), PREJ, is rejected
PRES = PAD + PPS + PREJ :

16

The desalination unit can be powered by RES, PAD, or hydroturbine

power, PHT, or both:
PDES = PAD + PHT :

17

Finally, if the existing water in the tank, along with the additional
production from RES and hydroturbine, cannot satisfy the water
demand, then there is a desalination power shortage, PN:
PN = PDEM PTANK + PAD + PHT ; PTANK = VT SDC

18

where PTANK is the desalination power equivalent (in kW) of the fresh
water tank content VT (in m3).
The fresh water demand is always being fed through the tank.
Desalination unit operates in order to replace the consumed water
and to retain a 90% minimum fulllment of this tank. The unit, as also
the pumps and hydroturbines does not operate if the available power
is below its corresponding technical minima. In addition, desalination
stops when the water tank is full, and the same is done for the pumps
if the upper reservoir is full.
Following the above constrictions, the operation strategy of the
entire system is presented in the ow chart of Fig. 5. The diagram
consists of two parts, explaining the management of RES and of
hydroturbine production, respectively.
In the rst part of Fig. 5, the power produced by the primary
generators (wind turbines and photovoltaics) can be used either for
desalination or for pumping storage. Priority is given to desalination

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

Fig. 5. Flow chart of the plant operation algorithm.

141

142

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

unit, provided that there is an empty space in the fresh water tank,
namely PTANK,F N 0, where:
PTANK;F = VTC VT SDC :

21

The latter condition always happens rst, hence it controls the

second ow chart diagram of Fig. 5. Finally, the hydropower stops
when the upper reservoir reaches its lowest acceptable level, here
PUPR = 0, or VR = 0.
3.3. Techno-economic evaluation
At the end of the simulated period of plant operation (one year) all
technical and economic evaluation indicators are computed. At this
section, the most important and those that will be presented and
discussed in the results are dened.
3.3.1. Technical evaluation
System's rst priority is to satisfy the island's water needs at a
predetermined percentage limit. Consequently, in terms of desalination power, we dene an indicator of demand satisfaction rate, FDS

FDS = 1

8760
j = 1 PN
8760
j = 1 PDEM

22

where the summation is over the hours of a year (24 365). In order
to assess the hydroturbine contribution to the desalination power
feed, the next indicator gives the ratio of the hydropower production
divided by the total desalination energy absorbed during the year:

FHT =

8760
j = 1 PHT
8760
j = 1 PDES

23

Also, an important energy indicator is dened to express the

portion of RES production that cannot be exploited and it is nally
rejected.
FREJ =

8760
j = 1 PREJ
8760
j = 1 PRES

CFW =

CFT =

20

and VR is the current water volume content, then the RES energy
surplus is rejected, as also when PRES exceeds the installed pumping
power.
On the other hand, the hydroturbine is set to operate in case of
desalination power shortage (PN N 0), but also when the fresh water
level in the tank drops below the limit VTG = 0.9 VTC, namely:
PTANK;G = VTG VT SDC N 0:

19

When the upper reservoir is full (PUPR,F = 0 or VR = VRC), where:

PUPR;F = gVRC VR HT T

8760

24

j=1


PWT
PAD + PPS
PRES
PW;I 8760

8760
j = 1 PHT
PT;I T8760

26

27

where PPV,I, PW,I and PT,I are the corresponding installed power,
respectively, and the rest power symbols represent hourly values.
3.3.2. Economic evaluation
The developed algorithm includes empirical equations from the
literature [2325] (some of them updated using recent available data),
in order to calculate the investment cost (purchase and construction) of
each subsystem, as well as the operation and maintenance cost of the
plant. Each cost depends either on a parameter's value or an operation
quantity that comes up from the simulation. In the following relations
power is measured in kW, water volume in m3 and costs in .
Wind turbines
Investment cost: ICW = 1300 PW,I
O&M cost: OMW = 0.02 ICW
Photovoltaics
Investment cost: ICPV = 6000 PPV,I
O&M cost: OMPV = 0.02 ICPV
Desalination (reverse osmosis)
Unit investment cost: ICD = 2270 D0.875
FC
Tank cost: ICT = 1090.8 V0.61
TC
Investment cost: ICDES = ICT + ICD
O&M cost: OMDES = 0.6WAP
Where WAP is the total fresh water production:

WAP =

8760
j = 1 PDES
:
SDC

28

Pumped storage subsystem

Hydroturbine investment cost: ICHT = 18,000 P0.48
T,I
Pumps investment cost: ICP = NP 1700 P0.82
P,I
Reservoirs investment cost: ICR = 2 420 V0.7
RC .
The pipe investment cost, ICPIPE, is considered as the summary of
material, welding, and coating costs, which are computed as function
of the pipe dimensions, whereas the latter is optimally selected from
standardized tables depending on the water ow rate and head.
Other costs (electric and electronic equipment, unpredictables, etc.):
ICOTHER = 0:2ICHT + ICP + ICR + ICPIPE :

Finally, the capacity factors of the three power generation

subsystems (PV, wind turbines, and hydroturbine) are computed by
the following relations and constitute the sizing indicators of this standalone system.
8760

CFP =

j=1


PPV
PAD + PPS
PRES
PPV;I 8760

Pumped storage subsystem total cost: ICPS = ICHT + ICP + ICR + ICPIPE +
ICOTHER
O&M cost: OMPS = 0.02 ICPS.
Total costs of the plant

25

System investment cost: ICSYSTEM = ICW + ICPV + ICDES + ICPS

System O&M cost: OMSYSTEM = OMW + OMPV + OMDES + OMPS.

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

The economic evaluation of the investment is based on the specic

water production cost and the dynamic index of Internal Return Rate
(IRR). The annual water production cost is obtained as the sum of the
annual depreciation, DA, of the total investment costs plus the total
O&M cost. Then, the specic water cost is computed as:

WPC =

DA =

DA + OMSYSTEM
WAP

29

ICSYSTEM r
11 + r n

143

desalination power of 480 kW is needed (3 kW h/m3). Also, a fresh

water tank of 8500 m3 capacity in order to provide 2 days of
autonomy in August is included (Eq. (5)). Then, because of the
stochastic nature of RES, the cumulative installed power of wind
turbines and PV is taken about two times greater than the desalination
unit power, namely 800 kW and 200 kW, respectively.
On the other hand, hydroturbine installed power is taken equal to
the desalination power (480 kW), in order to provide sufcient power

30

where r is the discount rate and n is the plant's life cycle. The specic
water cost will be compared to the current water transportation cost.
IRR index is obtained as the solution of the equation:
n

Bj

j=0

1 + IRR

Cj

j=1

1 + IRRj

= ICSYSTEM +

31

where the annual incomes Bj and expenses Cj are:

Bj = Bj1 e; B0 = WAP FWP

32

Cj = Cj1 e; C1 = OMSYSTEM :

33

In the above relations, e is the average ination rate and FWP the
sell price of the produced fresh water. Every value is converted to
present.
4. Results and discussion
This section is devoted to the presentation of results from the
operation simulation and optimal sizing of the examined hybrid
system. Firstly, a reference system with reasonable design is
examined and the results are analyzed in order to acquire a detailed
view of the plant operation and the subsystems' interaction. Then,
various single and double-objective optimization studies take place,
and some general conclusions are deduced.
4.1. Reference plant simulation
The various design parameters of an indicative desalination system
taken as reference are selected according to the following reasoning:
The system is installed in a medium size island of 5000 residents.
Having in mind the water consumption proles in Section 3.1.1, a
yearly average hourly fresh water demand would be 160 m3. Thus, a

Table 2
Dimensioning of the examined systems.
Design parameter
1 Island resident population
(people)
2 Tank capacity (m3)
3 Desalination installed power (kW)
4 Wind turbines installed power
(kW)
5 Photovoltaics installed power
(kW)
6 Hydroturbine installed power
(kW)
7 Pumps installed power (kW)
8 Storage reservoir capacity (m3)

Symbol Reference Optimum Optimum

system
System A System B
RP

5000

5000

5000

VTC
PD,I
PW,I

8.500
480
800

8500
498
1210

8500
612
1500

PPV,I

200

PHT,I

480

127

326

PP,I
VRC

600
40.000

386
12.120

593
56.900

Fig. 6. Plant monitoring during a winter week: a) energy production, b) water needs
satisfaction, c) energy consumption, d) tank and reservoir content.

144

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

even in the most adverse case of simultaneous apnea, cloudiness and

empty fresh water tank. Pumping station installed power is taken
between hydroturbine and total RES power (wind plus PV), at
600 kW. Concerning the reservoirs, a net capacity of 40,000 m3 was
chosen that corresponds to about 3 days of continuous turbine
production. All the above data are tabulated in Table 2 (Reference
system).

Graphs of plant performance during a typical winter week and a

typical summer week are presented in Figs. 6 and 7, showing the
energy production and exploitation and the water needs satisfaction
by desalination.
In the winter the energy is produced mostly by wind turbines
(Fig. 6a), whereas hydroturbine does not operate at all. Water needs
are small and they are fully satised even in periods of low
desalination production (Fig. 6b), while the fresh water tank remains
almost full (Fig. 6d). As a result, a large percentage of RES production
cannot be either used or stored, and it is rejected (Fig. 6c).
On the other hand, during the summer-week results of Fig. 7 the
water demand is higher, and the hydroturbine is continuously used to
supplement the desalination power (Fig. 7a), as long as there is
available hydraulic energy stored in the upper reservoir (Fig. 7d).
Nevertheless, the water needs are not totally satised, and there exist
certain hours every day when the fresh water tank empties (Fig. 7b
and d). This problem becomes worse towards the end of the week,
due to the reduced wind production (Fig. 7a). The only positive effect
is that energy rejections are now minimal, since the RES production
surplus can be pumped and stored to the upper reservoir (Fig. 7c).
From quantitative point of view, the performance of the reference
system can be evaluated from the results in the corresponding column
of Table 3, which were obtained by the simulation algorithm for a yearly
operation of the system. It is concluded that the examined system is
capable to satisfy the island's water demand at a high degree (90.3%),
but not fully. Also, a considerable part (39.5%) of the RES production
cannot either be consumed for desalination or stored, and the capacity
factors of all three generators are low. Hydroturbine participation in
desalination powering is not so strong (18.1%) but justies its
implementation in the system. Concerning the economic results, the
estimated specic production cost of the water is 2.53 /m3, much lower
than the current transportation cost (58 /m3). Using the former as
initial (rst year's) sell price for the system, the obtained IRR index value
exceeds 14%, for a depreciation period of 20 years.

4.2. System optimal design

The values of the system free design variables are numerically
optimized for various objectives. Towards this purpose a general
optimization software was used, which is developed and brought to
market by the Lab. of Thermal Turbomachinery, NTUA [26]. The
software performs stochastic optimization based on evolutionary
algorithms, and it is very effective for multi-parametric and multiobjective optimization of complex and discontinuous cost functions,
like in the present simulation. The optimization algorithm works with
populations of candidate solutions and in order to create the next
improved generation it mimics the biological evolution of species
generations, using processes like cross-over and mutation [26]. It has

Table 3
Systems energy and nancial results.

1
2
3

Fig. 7. Plant monitoring during a summer week: a) energy production, b) water needs
satisfaction, c) energy consumption, d) tank and reservoir content.

4
5
6
7
8
9
10

Description

Reference
system

Optimum
System A

Optimum
System B

Water demand satisfaction, FDS (%)

Energy rejection, FREJ (%)
Capacity factors (%)
Wind turbines, CFW
Photovoltaics, CFPV
Hydroturbine, CFT
Hydroturbine contribution, FHT (%)
Water production, WAP (m3/year)
Pumped storage unit cost, ICPS (K)
System investment cost, ICSYSTEM (K)
System O&M cost, OMSYSTEM (K)
Water production cost, WPC (/m3)
Internal Return Rate, IRR (%)

90.3
39.5

90.0
57.2

99.5
58.1

21.5
13.2
6.7
18.1
516.070
2.070
7.729
396.6
2.53
14.2

15.7
0
19.1
13.8
514.000
897
6.010
359
2.07
18.9

14.8
0
12.8
21.5
568.500
2.389
8.530
429.4
2.52
14.2

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

been successfully used by one of the authors in various previous

studies [19,20].

4.2.1. Single-objective optimization

During preliminary tests it resulted that the water production cost
reduces increasing the hydroturbine power up to a limit, beyond
which a minor further decrease in production cost is possible, but the
required hydroturbine power exhibits an abrupt and undesirable
further increase. After repeated computations for a range of different
populations and water demand satisfaction limits, a maximum
effective and reasonable size of the hydroturbine in relation to the
desalination unit power is extracted and tabulated, in order to be used
as additional constriction in all subsequent evaluations.
At rst, the desalination system for 5000 residents is optimized so
as to minimize the specic water production cost, WPC, while the
coverage of the annual water needs is kept almost equal to the
reference system performance (90%). The convergence rate of the
optimizer is shown in Fig. 8. Wide variation limits were used for all
design variables, and for this reason a great number of about 15,000
evaluations is required to minimize the cost function value. A single
evaluation is taken after simulating the entire system operation for a
period of one year. The obtained optimal values of the design
parameters, as well as the most important performance and economic
indexes of the so-called System A are concentrated in Tables 2 and 3,
respectively, in comparison to the reference system results.
The optimal system exhibits a considerably reduced water
production cost (2.07 /m3), compared to the reference design
corresponding cost (2.53 /m3, Table 3). This gain comes mainly
from the smaller installed size of the pumped storage system
(Table 2), which is now much better exploited (capacity factor of
hydroturbine, Table 3). On the other hand, the optimum power of the
wind farm is 50% increased, but its capacity factor reduces. Hence,
from the energy point of view the optimum system exhibits less RES
exploitation and increased energy rejections (Table 3).
In a second optimization study the desired water demand
satisfaction limit for the same population is set at a much higher
level, 99.5%, and the results are also given in the last columns of
Tables 2 and 3 (System B). Compared to the previous optimal System
A, the obtained System B conguration has a larger desalination unit
and wind farm power. Moreover, in order to guarantee the above
fresh water sufciency, the optimal size of both hydraulic turbine and
reservoirs are now obtained much larger (Table 2). As a result, the
capacity factors of generators (wind and hydroturbines) are not
satisfactory. In other words, the system is rather oversized, but this is
necessary in order to achieve the high demand satisfaction requirement, especially during summer. However, compared to the initial
non-optimal system, the new design still achieves a slightly less
production cost of fresh water (2.5 /m3), in spite of its higher total

Fig. 8. Indicative convergence history of the optimization algorithm.

145

investment cost (Table 3), thanks to the better energy management

among its subsystems.
On the other hand, a remarkable result is that both optimal
Systems A and B do not include photovoltaics at all, because of the
high investment cost of the latter. Consequently, wind turbines are
the primary power generators of an economically optimized investment, at least for the current cost-effectiveness of PV units.
4.2.2. Parametric studies
Aiming to analyze further the system performance and results, a
parametric study of the effect of island size (resident population) is
carried out. A single-objective optimization is performed for every
different size examined hence the results always correspond to
optimal systems.
The graphs in Fig. 9 depict the minimum achievable water production
cost for two different acceptable limits of water demand satisfaction, as
function of the island's resident population. For 90% satisfaction the
specic cost of produced water ranges from about 2.4 /m3 for a small
island of 1000 residents, to about 1.8 /m3 for a large island (20,000
residents). The production cost becomes, as expected, quite higher for
99% satisfaction requirement (Fig. 9), because the entire system must be
oversized to cover also the peak water demand periods.
This is evident also in the comparative results of Fig. 10a and b. An
almost linear dependence on the population for all subsystems installed
power can be observed in these results. The optimum wind power
remains 2 to 2.5 times greater than the desalination power, whereas the
pumping station power is almost equal to the difference between them,
in order to store the RES production surplus (Fig. 10a,b). Also, the
optimal hydroturbine has only about half of the desalination power,
since the former is used as an auxiliary and supplementary power
source. As it can be seen in Fig. 11, hydroturbine participates in
desalination unit feeding at a percentage in the range of 13 to 16% for
90% demand satisfaction. Its contribution increases for higher covering
limits (18 to 23% for 99% satisfaction, Fig. 11), in order to fulll the
increased needs for guaranteed power. Photovoltaics are again not
included in the optimum systems conguration.
The optimal capacity of the energy storage reservoirs (not shown
in gure), increases also linearly with the population, e.g. from about
2000 m3 for 1000 residents to 65,000 m3 for 20,000 residents and for
90% satisfaction of demands. The above values become 4 times larger
for 99% satisfaction limit (from 8000 to 265,000 m3, respectively).
The capacity factors of the two powering units of the desalination
system are drawn in Fig. 12. Their values are low and exhibit only a
small increase with the population, indicating that they are at certain
degree oversized. As a result, a signicant portion of the primary
energy production that ranges between 55 and 60% is rejected. The

Fig. 9. Water production cost for various island sizes.

146

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

Fig. 12. Capacity factor of main power generators for various populations and water
demand satisfaction limits.

Fig. 10. Optimal subsystems size for various populations, and demand satisfaction limit:
a) 90%, and b) 99%.

In a second parametric study the sell price of the produced fresh

water is considered as a parameter to assess the economic feasibility
of the investment. In this case the objective of system optimizations is
to maximize the IRR value. The obtained results for a small and a large
island and for two demand satisfaction limits are plotted in Fig. 13. It
can be observed that systems designed for larger islands and lower
demand satisfaction are the most attractive for an investor, exhibiting
much higher IRR values for a given water pricing.
Considering a typical IRR threshold value of the order of 10%, the
results in Fig. 13 show that for all cases the investment becomes
economically viable for water present price above 2.5 /m3, which is
much smaller than the current purchase cost (58 /m3). Moreover,
the economic results could be even better under more favorable
nancial strategies (e.g. subsidy), or by further exploitation of the
stored energy.

capacity factor of the primary energy source, wind turbines, does not
change much for the high demand covering (99%). On the contrary, in
that case the hydroturbine capacity factor reduces substantially
(Fig. 12), as the machine becomes larger to cover the peak demand
needs. A possibility to exploit the rejected energy to other consumptions or sell it to an electric grid appears to be the only way to increase
the capacity factors and to reduce energy rejections.

4.2.3. Double-objective optimizations

According to the results of single-optimization studies, the low
water production cost, the small pumped storage unit size, and the
high demand satisfaction rate are competitive objectives among each
other. Consequently, simultaneous optimization of them would be of
technical and economic importance.
The rst such optimization study concerning the minimization of
both hydroturbine power and water production cost, is carried out for
various populations and demand satisfaction limits. In this case the
optimizer converges to a series of optimum solutions distributed on

Fig. 11. Hydroturbine contribution to desalination.

Fig. 13. Effect of water pricing on the IRR value of the investment.

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

147

Fig. 16. Effect of demand satisfaction limit on the water production cost.

Fig. 14. Two-objective optimization results for various populations, and 99.5% demand
satisfaction limit.

curves, called Pareto Fronts [26], which are illustrated in Figs. 14 and
15. As expected due to economy of scale, the water production cost
reduces as population and plant size increase (Fig. 14). However, in all
cases there is an upper limit of normalized hydroturbine size, above
which the production cost reduction becomes negligible. For large
islands and/or for high demand satisfaction rates this limit is in the
range of 0.40.5 of the installed desalination power (Figs. 14 and 15,
respectively), whereas for small populations and/or reduced satisfaction rates it becomes lower: about 0.3 for 1000 residents (Fig. 14), and
about 0.2 for demand satisfaction 90% or lower (Fig. 15).
A remarkable result is that for reduced water needs satisfaction
requirements the energy storage unit may not be included at all,
because its incorporation only slightly decreases the water production
cost (Fig. 15). More detailed computations for the present data
showed that optimal systems obtained for populations larger than
5000 residents always include a pumped storage unit for any demand
satisfaction. However, for smaller islands the energy storage becomes
unnecessary when the water demand satisfaction limit is below 75%.
A more comprehensive picture of the effect of demand satisfaction
limit on the water production cost is given in Fig. 16 that contains
Pareto Fronts obtained from corresponding two-objective optimiza-

Fig. 15. Two-objective optimization results for various demand satisfaction limits and
5000 residents.

tions, for various populations. In all the three curves the production
cost increases almost linearly with the satisfaction limit, and only for
high percentage rates, above 95%, it rises more steeply, due to the
need of larger-sized power production and pumped storage units.
Consequently, the optimum selection of this limit depends on the
local conditions and the adopted strategy in order to cover the total
water demand needs of an island, that may include penalty prices for
insufcient production or shipping transportation cost of supplementary water.
4.2.4. Photovoltaic production
As mentioned before, all previously obtained lowest production
cost systems do not incorporate photovoltaics, due to the high
investment cost of this technology that makes it non-competitive
compared to wind turbines for such stand-alone desalination plants.
However, in the last decade this cost has been considerably decreased
and is expected to keep reducing in the next years. Consequently, we
are approaching to a threshold, below which the solar energy
production will be economically favorable. In order to estimate that
value for the hybrid desalination units examined here, a parametric
study of the inuence of PV investment cost on the optimum unit
conguration is performed.
The results for two different demand satisfaction limits are plotted
in Fig. 17, for an indicative number of 5000 residents, and they are
similar for smaller or larger populations. The photovoltaic production
unit becomes part of the optimal system when its installation cost
reduces below 3000 /kW. As the cost is further reduced, the PV
optimal size increases linearly, and exceeds the wind farm power in
the range between 1500 and 2000 /kW, depending on the demand
satisfaction limit (Fig. 17). This PV cost is quite higher than the

Fig. 17. Effect of investment cost on PV participation in hybrid desalination units.

148

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

A pumped storage subsystem is necessary to guarantee the desired

fresh water production throughout the year. Its contribution to the
desalination power feed varies between 13% and 23%, and its
optimum installed power becomes greater for larger islands or higher
water demand satisfaction requirements.
Photovoltaics are still non-competitive solution for the studied
hybrid system. However, PV production pattern ts well with the daily
water consumption needs, and such units are expected to become basic
component of hybrid desalination systems in the forthcoming years,
when their investment costs will drop below 3000 /kW.
The capacity factor of the primary and secondary power generators
of an optimally designed system (RES and hydroturbine, respectively) is
relatively low in all cases, while a signicant portion of the RES
production is rejected. Consequently, the capability of exploiting the
pumped storage unit to produce electricity for other parallel usages
provides to be the only effective way to improve the capacity factor of all
subsystems. This would also minimize the amount of rejected energy,
and improve the economic results of the entire desalination plant.
Fig. 18. Power generators capacity factor for optimum system with photovoltaics.
List of symbols

corresponding wind generators cost (taken 1300 /kW, Section 3.3.2).

The reason is that the entire solar production is available during the
daytime, namely within the high water demand hours, in contrast to
the wind power, a considerable part of which may be generated
during the night. This advantageous feature reduces the need for
energy storage, resulting in smaller optimum reservoirs size and
pumped storage machinery. The latter compensates the higher
installation cost of photovoltaics compared to wind generators.
On the other hand, the optimum relation between hydroturbine
and desalination installed power, discussed in Section 4.2.3, is not
affected by the PV production, because it is mainly associated with
periods of peak water demand and/or insufcient RES production.
For the optimal systems obtained without photovoltaics, the wind
farm and hydroturbine capacity factors are always below 20%, even
for lower satisfaction limits than those shown in Fig. 12. This picture
changes when the cost of PV system is reduced, allowing it to be part
of the optimum hybrid unit. Assuming an investment cost for
photovoltaics of the order of 1800 /kW, for which the optimum
size of both RES units (wind and solar) are about the same (Fig. 17),
the corresponding capacity factors for 5000 resident population are
illustrated in Fig. 18.
The values for all three generators increase as demand satisfaction
limit decreases, because smaller satisfaction needs can be fullled
with smaller installed power. Compared to Fig. 12, the exploitation of
the wind farm production is remarkably improved by introducing the
PV unit, which reduces the optimum wind farm size. On the other
hand however, hydroturbine capacity factor exhibits a considerable
drop for demand satisfaction rates above 80% (Fig. 18). This happens
because part of its duty during daytime hours is now replaced by PV
production, whereas its installed power remains high in order to be
able to provide the required power for desalination at periods of
insufcient RES production.

5. Conclusions
A numerical algorithm is developed and applied in order to
investigate in detail the operation and performance of a hybrid standalone desalination system. The results of the various parametric and
optimum design studies of the system carried out in this work are
summarized below.
The studied stand-alone hybrid desalination system is capable to
fulll the water demand of areas such as Greek islands, having an
attractive specic water production cost (1.53.0 /m3), which is much
competitive to the present water transportation pricing (58 /m3).

CFP
CFT
CFW
DWD
FDS
FHT
FREJ
G
g
HP
HT
HWT
ICSYSTEM
OMSYSTEM
PAD
PD,I
PDEM
PDES
PHT
PMD
PMP
PMT
PN
PP,I
PPS
PPV
PPV,I
PREJ
PRES
PT,I
PTANK
PWT
PWT,I
QP
QT
RP
RWC
SDC
SP
SWC
VR
VRC
VT
VTC
WDP
WPC
P
T

Capacity factor of pumps,

Capacity factor of hydroturbines,
Capacity factor of wind turbines,
Daily fresh water demand,
Water demand satisfaction rate,
Hydroturbine contribution indicator,
Energy exploitation indicator,
Hourly solar radiation,
Gravity acceleration,
Pumping head,
Hydroturbine net head,
Hourly fresh water demand,
Total investment cost of the hybrid system,
Annual operation and maintenance costs,
RES production spent for desalination
Desalination unit installed power,
Hourly power demand for desalination,
Desalination unit consumption,
Hydroturbine production,
Technical minimum of desalination unit,
Technical minimum of hydroturbines,
Technical minimum of pumping station,
Desalination power shortage,
Pumping station installed power,
Pumping absorbed power,
Photovoltaics production,
Photovoltaics installed power,
Rejected power,
Cumulative RES production,
Hydroturbine installed power,
Power equivalent of water tank content,
Wind turbines production,
Wind turbine installed power,
Pumped water ow rate,
Water ow rate through hydroturbine,
Monthly island resident population,
Water consumption of resident population,
Specic desalination consumption,
Monthly island seasonal population,
Water consumption of seasonal population,
Water volume in the reservoir,
Reservoirs capacity,
Water volume in the tank,
Water tank capacity,
Hourly percentage of daily water demand,
Specic fresh water production cost,
Pumping efciency,
Hydro turbining efciency,

%
%
%
m3
01 or %
01 or %
01 or %
kW h/m2
m/s2
m
m
m3

kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
kW
m3/h
m3/h
people
l/day/human
kW h/m3
people
l/day/human
m3
m3
m3
m3
%
/m3
%
%

References
[1] J.K. Kaldellis, E.M. Kondili, The water shortage problem in the Aegean archipelago
islands: cost-effective desalination prospects, Desalination 216 (2007) 123138.

I.D. Spyrou, J.S. Anagnostopoulos / Desalination 257 (2010) 137149

[2] I.C. Karagiannis, P.G. Soldatos, Current status of water desalination in the Aegean
Islands, Desalination 203 (2006) 5661.
[3] A.D. Khawaji, I.K. Kutubkhanah, J.M. Wie, Advances in seawater desalination
technologies, Desalination 221 (2007) 4769.
[4] S.A. Kalogirou, Seawater desalination using renewable energy sources, Progress in
Energy and Combustion Science 31 (2005) 242281.
[5] C. Fritzmann, J. Lowenberg, T. Wintgens, T. Melin, State-of-the-art of reverse
osmosis desalination, Desalination 216 (2007) 176.
[6] E. Mathioulakis, V. Belesiotis, E. Delyannis, Desalination by using alternative
energy: review and state-of-the-art, Desalination 203 (2007) 346365.
[7] L. Garcia-Rodriguez, Renewable energy applications in desalination: state of the
art, Solar Energy 75 (2003) 381393.
[8] C. Lopez-Fernandez, A. Viedma, R. Herrero, A.S. Kaiser, Seawater integrated
desalination plant without brine discharge and powered by renewable energy
sources, Desalination 235 (2009) 179198.
[9] S.A. Kershman, J. Rheinlander, Th. Neumann, O. Goebel, Hybrid wind/PV and
conventional power for desalination in LibyaGECOL's facility for medium and
small scale research at Ras Ejder, Desalination 183 (2005) 112.
[10] D. Manolakos, E.Sh. Mohamed, I. Karagiannis, G. Papadakis, Technical and
economic comparison between PV-RO system and RO-Solar Rankine system,
Case study: Thirasia island, Desalination 221 (2008) 3746.
[11] M. Wilf, C. Bartels, Optimization of seawater RO systems design, Desalination 173
(2005) 112.
[12] E. Tzen, K. Perrakis, P. Baltas, Design of a stand alone PV-desalination system for
rural areas, Desalination 119 (1998) 327334.
[13] P.A. Koklas, S.A. Papathanasiou, Component sizing for an autonomous winddriven desalination plant, Renewable Energy 31 (2006) 21222139.
[14] E.Sh. Mohamed, G. Papadakis, Design, simulation and economic analysis of a
stand-alone reverse osmosis desalination unit powered by wind turbines and
photovoltaics, Desalination 164 (2004) 8797.
[15] M. Thompson, D. Ineld, Laboratory demonstration of a photovoltaic-powered
seawater reverse-osmosis system without batteries, Desalination 183 (2005)
105111.

149

[16] E.Sh. Mohamed, G. Papadakis, E. Mathioulakis, V.A. Belessiotis, Direct coupled

photovoltaic seawater reverse osmosis desalination system toward battery based
systems a technical and economical experimental comparative study, Desalination
221 (2008) 1722.
[17] M. Patel, Wind and Solar Power Systems, CRC Press, 1999.
[18] J.K. Kaldellis, D. Zarakis, K. Kavadias, Techno-economic of energy storage systems
for island autonomous electrical networks, Renewable & Sustainable energy
reviews 13 (2009) 378392.
[19] J.S. Anagnostopoulos, D.E. Papantonis, Pumping station design for a pumpedstorage wind-hydro power plant, Energy Conversion & Management 48 (2007)
30093017.
[20] J.S. Anagnostopoulos, D.E. Papantonis, Simulation and size optimization of
pumped storage power plant for the recovery of wind-farms rejected energy,
Renewable Energy 33 (2008) 16851694.
[21] D. Manolakos, G. Papadakis, D. Papantonis, S.A. Kyritsis, Simulationoptimisation
programme for designing hybrid energy systems for supplying electricity and
fresh water through desalination to remote areas Case study: the Merssini village,
Donoussa island, Aegean Sea, Greece, Energy 26 (2001) 679704.
[22] D. Manolakos, G. Kosmadakis, G. Kyritsis, G. Papadakis, Identication of behaviour
and evaluation of performance of small scale, low-temperature Organic Rankine
Cycle system coupled with a RO desalination unit, Energy 34 (2009) 767774.
[23] D. Papantonis, Small Hydroelectric Plants, Symeon, Athens, 2008 (in Greeks).
[24] D. Voivontas, G. Arampatzis, E. Manoli, C. Karavitis, D. Assimacopoulos, Water
supply modeling towards sustainable environmental management in small
islands: the case of Paros, Greece, Desalination 156 (2003) 127135.
[25] C. Koroneos, A. Dompros, G. Roumbas, Renewable energy driven desalination
systems modeling, Journal of Cleaner Production 15 (2007) 449464.
[26] M. Karakasis, K.C. Giannakoglou, On the use of metamodel-assisted multiobjective evolutionary algorithms, Engineering Optimization 38 (2006) 941957.