Vous êtes sur la page 1sur 14

ARTICLE IN PRESS

Renewable Energy 33 (2008) 617630


www.elsevier.com/locate/renene

Designing cost-effective seawater reverse osmosis system


under optimal energy options
Asmerom M. Gilau, Mitchell J. Small1
Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA
Received 21 November 2006; accepted 17 March 2007
Available online 11 May 2007

Abstract
Today, three billion people around the world have no access to clean drinking water and about 1.76 billion people live in areas already
facing a high degree of water stress. This paper analyzes the cost-effectiveness of a stand alone small-scale renewable energy-powered
seawater reverse osmosis (SWRO) system for developing countries. In this paper, we have introduced a new methodology; an energy
optimization model which simulates hourly power production from renewable energy sources. Applying the model using the wind and
solar radiation conditions for Eritrea, East Africa, we have computed hourly water production for a two-stage SWRO system with a
capacity of 35 m3/day. According to our results, specic energy consumption is about 2.33 kW h/m3, which is a lower value than that
achieved in most of the previous designs. The use of a booster pump, energy recovery turbine and an appropriate membrane, allows the
specic energy consumption to be decreased by about 70% compared to less efcient design without these features. The energy recovery
turbine results in a reduction in the water cost of about 41%. Our results show that a wind-powered system is the least cost and a PVpowered system the most expensive, with nished water costs of about 0.50 and 1.00$/m3, respectively. By international standards, for
example, in China, these values are considered economically feasible. Detailed simulations of the RO system design, energy options, and
power, water, and life-cycle costs are presented.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Reverse osmosis; Energy recovery; Optimal energy options; Energy storage; Power cost; Water cost

1. Background
Today, about three billion people around the world have
no access to clean drinking water. According to the World
Water Council, by 2020, the world will be about 17% short
of the fresh water needed to sustain the world population.
Moreover, about 1.76 billion people live in areas already
facing a high degree of water stress [1]. Water stress is at
the top of the international agenda of critical problems, at
least as rmly as climate change [2]. As a result, the need
for desalination is increasing, even in regions where water
supply is currently adequate.

Corresponding author. Tel.: +1 412 268 3426; fax: +1 412 268 3757.

E-mail address: asmeromg@andrew.cmu.edu (A.M. Gilau).


H. John Heinz III Professor of Environmental Engineering, Carnegie
Mellon University, Civil & Environmental Engineering, and Engineering
& Public Policy, Pittsburgh, PA, USA.
1

0960-1481/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2007.03.019

As part of one of the most affected arid areas of the


world, Eritrea, a moderate size nation located along
the northeastern coast of Africa, has been the victim of
recurrent droughts and water shortage. The problems of
water supply vary from place to place and one of the most
severe problems exists in the coastal areas and islands [3].
Thus, this case study assesses the use of renewable energy
for seawater reverse osmosis (SWRO) for the coastal
village of Beraesoli, located at the Southern Red Sea,
Eritrea.
Generally, desalination processes can be categorized
into two major types: (i) phase-change/thermal; and
(ii) membrane process separation. Some of the phasechange processes include multi-stage ash, multiple effect
boiling, vapor compression, freezing, humidication/dehumidication and solar stills. Membrane-based processes
include reverse osmosis (RO) and electrodialysis. Kalogirou [4] has provided details on each process and all
processes are available in the market. Preferred options

ARTICLE IN PRESS
618

A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

depend upon the need for, and availability of, energy


sources, as well as local climatic conditions. In Eritrea, for
example, in 1992, Gilau [5] experimented with solar still
desalination in the coastal city of Massawa, Eritrea. That
study reported that, in the month of September, where the
temperature usually reaches about 36 1C, using a glass area
(upper roof area) of 5.27 m2, about 7 L of desalinated water
per day could be collected. Noted above, this study will
focus renewable energy powered SWRO.
Seawater desalination is an energy-intensive process
[6,7]. Most of the available large-scale desalination plants
around the world are powered by fossil fuel. Due to the
concerns regarding global warming and increasing fuel
costs, alternative energy sources have been proposed for
desalination purposes. For example, the International
Atomic Energy Agency (IAEA) has proposed the use
of nuclear power for large-scale desalination plants [7].
In addition, the use of renewable energy sources for
small-scale desalination plants is emerging. Autonomous
wind-powered seawater RO potentials have been studied
[6,8] and pilot projects are in progress [7]. However,
the application of renewable energy for desalination
has not yet reached sufcient maturity to be applied
widely. Moreover, most of the proposed designs are
connected to the conventional power grids [7,9]. Thus,
the main objective of this research is to design a costeffective reverse osmosis system, which considers both
non-renewable and renewable power options, and determines the optimal mix of energy options. This paper
introduces a new methodology for energy optimization for
an SWRO system.
In the analysis, two major models are applied, including
the Reverse Osmosis System Analysis (ROSA) model, a
sophisticated RO design program that predicts the
performance of membranes in user-specied systems [10],
and the hybrid optimization model for electric renewables
(HOMER), which is an optimization model for hybrid as
well as stand-alone power systems [11]. Life cycle analyses
are performed to examine the performance of the system,
determine water costs, and undertake comparative analysis
of different power supply options.

2. Situation analysis
2.1. Water demand
Recent studies conduced by the Swiss Federal Institute
for Environmental Science and Technology [1] indicate
that East African countries have renewable freshwater
resources below the calculated threshold of 1500 m3 per
capita per year. Eritrea is already in a condition of water
decit. According to recent studies, in the coastal parts of
the country, water demand is 116 m3/household/year [12].
Thus, for the Berasoel village of 108 households, we are
assuming an average water demand of 35 m3/day and
13,000 m3/year.

2.2. Energy supply


In Eritrea, there is a strong potential for wind powered
electricity generation [13] and sunshine is abundant. The
Southeast coast of Eritrea has as much as 100200 km of 6
and 7 wind classes. At these sites, wind turbines may
operate at a 4060% capacity factor [14], which implies
that the wind potential ranges from excellent to exceptional
[15], and could be potentially used for commercial and
industrial purposes as well. The average annual wind speed
and solar radiation are about 6.8 m/s and 6.8 kW h/m2,
respectively.
3. Modeling SWRO design and optimal energy options
3.1. SWRO design considerations
3.1.1. Model description
ROSA 6.0.1 software is the latest version, used in the
analysis to determine the performance of a membrane and
energy requirements for desalination. The use of the model
is inuenced by the need to design a technically feasible RO
system. The ROSA model has been used for designing
desalination plants in different parts of the world [1621].
We have extended the application of the model in creating
an operating window for an RO system that could operate
under intermittent power supply. This is done by running
the model multiple times under different water ow rates
and pressures. The main inputs to the model include the
amount of feed water and its chemical characteristics, feed
water ow rate, feed water and concentrate feed pressures,
temperature and pH. Then, a conguration of the number
of membranes, pressure vessels, and type of membrane,
and feed and booster pumps is determined. After performing calculations, the model provides the amount of water
produced and the energy required. The energy required to
produce an intended amount of drinking water with
acceptable water quality is then determined by running
the model multiple times. Booster pumps and an energy
recovery turbine can also be included in the design.
3.1.2. RO system design and energy consumption
Using ROSA, we determined several RO design options
capable of producing 35 m3/day potable water. After
examining several design alternatives, our preferred design
is a two-stage design with three membrane elements in each
stage (Fig. 1). The reason for choosing a two-stage system
is that this enables an increased water productivity by
applying a booster pump that recovers a signicant amount
of energy. The type of the membrane used in the analysis is
FILMTEC SW30HRLE-400. The membrane is widely
used, has a high ow rate capacity (37.2 m3/day), high
boron rejection ability (91%) and resists up to about 83 bar
of pressure. Moreover, the membrane is designed to
properly function under intermittent energy supply, resists
fouling, and enables effective cleaning [22].

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

619

Seawater Reverse Osmosis Desalination System for Beraso'ele Village, Southern Red Sea, Eritrea

Low pressure
pump
Pr1

Feed water
from Red Sea

Pw1

Pretreatment

1st stage
membranes

Q1
Q2 Pr2

High pressure
pump (Booster)

Pw2

2nd stage
membranes

Pr3
Q3

Energy
Recovery
Turbine

Q7

Product
tank

Q6
Q4 P
r4

Pw3

Concentrate

Post
Treatment

Q5

Product/
permeate
water

Fig. 1. Schematic two-stage RO system for the village of Beraesoli, Southeastern Red Sea, Eritrea.

In RO desalination systems, energy is a major consideration. Power consumption by the system includes
power for seawater pumping, high-pressure pumping, i.e. a
booster, and chemical treatment. The total power requirement is calculated using Eq. (1) [23]:
Pwn

Qn  Prn
,
En

(1)

where Pwn (kW) is the power consumed by feed, lowpressure, high-pressure and chemical water treatment
pumps, Qn (m3/s) the rates of feed water (Q1), fresh water
production (Q5+Q6), boosted water (Q3), Prn (kPa) the
feed pressure (Pr1), boosted pressure (Pr3), rejection
pressure (Pr2 and Pr4); and En (net efciency of feed
pump) Ep (pump efciency)  En (motor efciency) for
the high-pressure pump (booster) and energy recovery
turbine.
According to Darwish et al. [23], the low-pressure pump
consumes the highest energy (Pw1), and the rest constitutes
about 20% of the LP pump. The power required for the
systems LP pump, at 10 m3/h feed water ow rate, 45 bar
pressure, and 0.85 pump efciency, is about 14.7 kW. An
additional 2.9 kW is needed for the booster, feed water,
chemical treatment and other pumps, which is about 20%
of the LP power requirement. Thus, the total power
required for the RO system design is about 17.6 kW. Using
ROSA software, we have obtained a similar result,
17.2 kW. The specic work done is about 3.92 kW h/m3
potable water produced. Without a booster pump, the
system requires about 7.87 kW h/m3, and its water quality
deteriorates from 270 to 800 ppm total dissolved solids.
Moreover, if the booster pump is not applied, water
production in the second stage could decrease by about
33% per hour, i.e. from 2.1 to 1.4 m3/h. Thus, in terms of
water production, water quality, and energy recovery, a
two-stage RO design is preferable. The system design has
an average conversion factor of 55% (relation between
product water ow and feed water ow), producing about
4.4 m3/h of potable water.

3.1.3. Performance prediction


In designing the SWRO system that uses intermittent
energy sources, it is very important to design an RO system
that can operate under broad operational windows. The
main thresholds of the operational window include the
maximum feed pressure (determined by the membrane
mechanical resistance); maximum brine ow rate (should
not be exceeded to avoid membrane deterioration);
minimum brine ow rate (should be maintained to avoid
precipitation and consequent membrane fouling); and
maximum product concentration (if the applied pressure
is less than a determined value, the permeate concentration
will be too high) [8].
Using chemical characteristics of water of the study area
[24], and varying the values of the operational window
thresholds, we have run the model several times. According
to the results of the analysis, at 25 1C, the maximum
allowable pressure, maximum brine ow rate, minimum feed
ow rate, and minimum pressure of our design are about
50 bar, 16 m3/h, 7 m3/h, and 30 bar, respectively (Fig. 2a).
This design allows the production of potable water with
an average water quality of about 500 ppm total dissolved
solids or less, which is the World Health Organization
(WHO)s water quality acceptable standard. Likewise, in
order to operate under this operational window, the energy
supply or pumping power should not be less than 7 kW and
not exceed 26 kW (Fig. 2b). This means that, without
interrupting the operation of the system, it can operate at
as low as 7 kW and as high as 26 kW power supply, which
is a wide operating window. This exibility is one of the
aspects that could potentially make renewable energy
resources for SWRO systems more attractive.
The average conversion factor of the system, at 25 1C, is
about 55%. Carta et al. [7] indicated that for an increase of
1 1C, water production increases by about 4%. Thus, since
the climatic conditions of the study area are very hot, with
average monthly temperatures varying from 26.5 to 35.5 1C
[25], it is expected that most of the time of the year, the
conversion factor could reach as high as 70%.

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

620

Water quality and flowrate threshholds of the SWRO design


Water Quality (ppm of total TDS)

80
0
70
0
60
0
30 bar

50
0
40
0

35 bar

30
0

40 bar
45 bar
50 bar

20
0
10
0
0
6

10

11

12

13

Feedwater flow rate (m

14

15

16

17

3/h)

Power Thresholds for The SWRO design


30
50 bar

Pumping Power (kW)

25

45 bar
40 bar

20

35 bar
30 bar

15
10
5
6

10

11

12

Feedwater flow rate

13

14

15

16

17

(m3/h)

Fig. 2. (a) Water quality and feed water ow rate and (b) power thresholds of the SWRO system.

3.1.4. Energy recovery


Energy recovery should be considered if brine exits the
system at 300 psig or more, and if system recovery is less
than 80% [26]. Since the brine rejection pressure of our
design is well above 300 psig, with the design pressure
760 psi (52.82 bar), the potential for energy recovery is very
high. For example, using an energy recovery turbine with
an efciency of 0.85, at concentrate pressure of 52.82 bar,
and a concentrate water ow of 5.6 m3/h, about 7 kW
(Eq. (2)) of energy could be recovered. Thus, our design
can potentially reduce power consumption by nearly half,
from 15 kW to about 8 kW.
Pwrn Qn  Prn  E t ,

(2)

where Et is the turbine efciency, Prn (kPa) is the feed


pressure, and Qn (m3/s) is the ow rate of the feed brine.
The type of energy recovery turbine under consideration
is the pressure exchange (PX) turbine, PX45s [27]. Using
this energy recovery device, net energy consumption could
be reduced from 17.23 to 10.25 kW, and specic energy
consumption from 3.92 to 2.33 kW h/m3, assuming 45 bar
feed pressure and 10 m3/s feed ow rate. This is about a
40% energy recovery, and it is a substantial energy

recovery opportunity for a small-scale SWRO system.


Therefore, using a booster pump and energy recovery
turbine, energy consumption can be decreased from
about 7.87 kW h/m3 (at maximum ow rate of 16 m3/s,
i.e. during high wind speed periods) to about 2.33 kW h/m3.
A specic energy consumption of an RO system in Kuwait
and in the Caribbean (Curacao) islands has been reported
to be about 4.52 and 3.15 kW h/m3, respectively, for
5700 m3/day desalination water capacity [23]. Compared
with these results, our design results are low, though
compatible.
Thus, depending on the feed ow rate and pressure
exerted, our operating window of the system can potentially recover anywhere from 4 to 12 kW. The RO system is
expected to operate under a semi-instantaneous base load
of 17 kW. Thus, assuming a feed pressure of 45 bar, a
boosting pressure of 10 bar, and a feed water ow rate of
10 m3/h, the system can recover about 47 kW/h (Fig. 3).
Moreover, increasing the feed water ow rate at low
pressure could substantially increase energy recovery and
water production. In this regard, within the operational
window of the system, any other points of operation could
be selected as an initial point of operation.

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

621

Energy consumption/recovery(kW)

Initial energy consumption, potential energy recovery, and net energy


consumption at 45bar feed pressure of the RO system
30

Energy recovery(kW)
Initial energy consumption(kW)
Net energy consumption(kW)

25
20
15
10
5
0
6

10

11

12

13

14

15

16

17

Feed water (m3/hr)

Fig. 3. Energy recovery potential of the RO system at 45 bar feed pressure.

3.2. Modeling optimal energy options


The main objective of this analysis is to determine
cost-effective and feasible energy options to produce the
required amount of water using the SWRO design
discussed in the previous section. The types of energy
options under consideration include stand alone as well as
hybrid energy sources of wind, photovoltaic (PV) cells, and
diesel. To determine least cost energy options, we used an
optimization model, HOMER, developed by the National
Renewable Energy Laboratory (NREL).
3.2.1. Model description
HOMER evaluates different energy options by simulating hourly energy ows for complex hybrid as well as stand
alone power systems. The model simulates power system
congurations and optimizes for lifecycle costs. It simulates
the operation of a system by making energy balance
calculations for each of the 8760 h of the year. For each
hour, the model compares the electric load in the hour
to the energy that the system can supply in that hour.
In the presence of energy storage devices, the model
determines when to discharge and charge electricity. The
model also estimates the lifecycle cost of the system
based on capital, replacement, operation and maintenance,
and fuel costs of each component including PV cells,
wind turbines, batteries, generators, and inverters. After
simulating the system congurations, the model displays
a list of feasible systems, sorted by lifecycle cost,
based on their net present values (NPVs). Then, based
on the results, one may choose from among the least
cost and feasible systems. However, all the least cost
systems are not necessarily feasible. Thus, reliability and
other issues need to be considered in choosing the optimal
energy option. The HOMER model has been used
for different household electrication feasibility studies
[2833]. However, HOMER has not yet been used for the
design of a power system for SWRO.

3.2.2. Determining SWRO base load


In RO desalination systems, energy is a major consideration. Depending on the capacity of the RO systems,
estimates of energy use range from 2 to 10 kW h/m3 of
water produced [34]. The average base load for our RO
system is 17 kW/h with a specic work done of about
2.33 kW h/m3. In order to make the assumptions of SWRO
energy demands more realistic, 5% and 2% noises are
added in the model for daily and hourly loads, respectively.
The challenge is optimizing to meet the required constant
load using highly variable power supply systems.
In order to solve the challenge, we have designed the
SWRO system to operate anywhere between 7 and 27 kW,
which are the allowable operating power thresholds. Any
power below and above about 7 and 27 kW could
ultimately be considered as excess power. Thus, we are
trying to optimize the regular SWRO power load under an
irregular power supply, which makes the analysis complex,
especially when wind energy is considered. Based on the
base load power requirement, different energy models are
simulated with different base load congurations, with and
without energy storage devices. An energy dispatch
strategy is designed to meet the required base load of the
RO system, depending on the type of energy option
selected.
3.2.3. Wind power
3.2.3.1. Implications of wind speed variations. The average wind speed of the study area is 6.8 m/s [3]. Wind speed
distributions can typically be described in terms of the
Weibull distribution [6]. According to Rosen [35], the
shape parameter for the Southern Red Sea area, particularly Port Assab, is k 2.4. The standard Weibull shape
parameter is k 2.5. Using the Danish Industry Wind
Association power calculator [36], at 25 1C, 10 m above
sea level, 101.21 kPa atmospheric pressure, 1.22 kg/m3
air density and a 2.4 shape parameter, the scale parameter
is c 7.67 m/s. Thus, using Eq. (3), the probability

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

622

Seasonal variation of wind speed


0.25
Season 1(Oct.to April)
Season 2 (May toSeptember)
Annual distribution

Probability Density

0.20
0.15
0.10
0.05
0.00
0

8
10
12
Wind speed (m/s)

14

16

18

20

Fig. 4. Seasonal variations of wind speed Weibull distribution of the study area.

distribution function of the wind speed is computed. The


probability density function, fv(v), is given by
   
k vk1
v k
exp 
f vv
,
(3)
c c
c
where v is the wind speed in m/s, k the shape parameter,
and c the scale parameter (m/s).
The average wind speed is calculated as


1
,
(4)
Ev cG 1
k
where G is the gamma function.
The mean wind speed is thus computed to be 6.8 m/s.
However, wind speeds throughout the seasons are irregular, and the average wind speed exhibits high seasonal
variation. According to the US Navy Climatic Study of the
Red Sea report conducted in 1982 [35], for 7 months
(OctoberApril), mean wind speeds are between 6 and
10 m/s, and the remaining months (MaySeptember), mean
wind speeds are between 3 and 5 m/s. Assuming an average
wind speed of 8 m/s for season 1 (OctoberApril) and 4 m/s
for season 2 (MaySeptember), the scale parameters are 9
and 4.5 m/s, respectively. The Weibull probability density
function of wind speeds for the two seasons are shown in
Fig. 4. A very high wind speed in one season and a very low
wind speed in the other season make the modeling process
more complex. That is, the irregularity of seasonal wind
speed makes the system more complex to optimize without
expecting very high excess electricity during the high wind
speed season, and signicant use of energy storage systems
during the low-wind-speed season. Most studies indicate
that to overcome power shortages during the low-windspeed seasons, a large number of wind turbines should be
installed [37]. This is one of the issues that we address in
this analysis.
3.2.3.2. Model input. In this analysis, we assume that the
RO system will be powered by a stand-alone wind turbine.
In this model, a 17 kW base load is considered in order to
produce potable water at about 2.33 kW h/m3. This

approach allows capturing all wind energy outputs that


are produced whenever the wind is blowing. Taking into
consideration the above-mentioned seasonal wind speed
and energy variations, we have considered monthly average
wind speeds, with an annual average wind speed of 6.8 m/s.
The monthly average wind speeds are then used to simulate
hourly wind speeds for the 8760 h of a year. Using a FL2
30 kW wind turbine, the results of the simulation are shown
in Fig. 5.
The results indicate that the effective annual number of
hours with adequate wind power for the SWRO operation
is more than 4000 h. Thus, the wind turbine seems to be a
good candidate for the analysis. However, in doing so,
excess electricity is expected, especially during the high
wind speed season. To complete the cost and performance
analysis of the FL 30 wind turbine, we assume a lifetime of
20 years [38,39], a capital cost of $130,000, annual
operation and maintenance costs of $3900 (2.5% of capital
cost) [40], and a replacement cost of 85% of the capital
cost.
3.2.3.3. Results of the analysis. According to the optimization results, the least cost feasible wind energy option
consists of 1 FL 30 kW wind turbine, 10 batteries, and 10
converters. The cost of this option is about $0.17/kW h
with a capacity shortage of about 51%. This means that
our RO system will be functional at full capacity for about
4000 h of the year, producing about 30,000 m3/year. Since
most of the high wind speeds are during the day, the system
operates for more than 812 h/day for the whole year to
yield the projected 4000 h/year. In order to deal with water
production increases during high temperature and wind
speed seasons, there might be a need to build a water
storage tank in order to store water to be used during lowwind-speed periods.
It should be noted that the use of a battery as an energy
storage device is not solely meant to increase energy
2
Fuhrlander (FL) 30 kW is a type of 30 kW wind turbine developed by
Fuhrlander AG. http://www.fuhrlaender.de/start.php.

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

623

Fig. 5. Daily wind power outputs and the required load of FL 30 kW wind turbines.
40

Base Load (kW)


Wind Power (kW)
Water Produced(W+B)(m3)
Battery Power Dispatch(kW)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

350

400

-5
-1
-15

Fig. 6. Base load, wind power production, battery power dispatched, and water production simulated for the rst.

production. The primary objective is to keep the power


supply at a semi-instantaneous condition (Fig. 6) so that
the RO system could produce water continuously. When
wind power is starting to decrease, electricity is dispatched
accordingly to meet the demand. This could also potentially minimize the deterioration of membranes. According
to the results of the model, the lowest wind speeds occur
from midnight up to early in the morning. Thus, there is a
possibility of turning off the RO system or adjusting it to
function with battery power during this time interval.
Annual wind power output, base load, and water production scenarios are shown in Fig. 7.
3.2.4. Photovoltaic (PV) power
3.2.4.1. Model input. This analysis assumes simulating a
stand-alone PV power system without applying a sunshine

tracking system. Using a 17 kW base load for about 8 h/day


throughout the year and daily solar radiation at Beraesoli
(131300 N and 421430 E), the performance of a PV-powered
system has been analyzed. In the analysis, we assume a 25year lifetime of the PV cells, $4500/kW capital cost [41],
and $3800 replacement cost.
It is important to realize that the assumptions of RO
base load for the wind powered system, which is
continuous for 24 h (Fig. 7), is different from the PV
powered RO system, which is congured to operate during
the day only. The main reason is, unlike daily predictions
of sunshine, for wind power it is very difcult to predict
when the wind will be blowing. Therefore, in the case of the
wind powered conguration, we have considered a
continuous power demand in order to capture all available
wind energy whenever the wind speed is high. In the case of

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

624

RO Base load, wind power and water production

Power(kW) and Water Production(m3/hr)

40
Wind Power (kW)
Base Load (kW)
Water Produced(W+B) (m3)

35
30

Wind Power

25
20

Base Load

15
10
5

Water Produced
0
0

1000

2000

3000

4000

5000

6000

7000

8000

-5
Time for one year (hr)

Fig. 7. Annual wind power, base load, and water production scenarios.

Baseload of RO, PV Power, Battey Power Storage and Dispatch and Water Production
35
RO Baseload kW
PV Power kW
Water Produced
Battery Power kW

Power out put (kW) &


Water Production (m3/hr)

30
25
20
15
10
5
0

10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260 270 280 290 300 310 320 330 340

-5
-10
-15

Time (hr)

Fig. 8. Simulated RO base load for the rst 2 weeks of the month of January, PV power produced.

a PV powered system, the optimal load is congured to


operate during the day assuming that the energy load is
only required for 810 h of the day. Thus, this conguration assumes operation during the day only, producing the
intended amount of water.
3.2.4.2. Results of the analysis. The analysis considers a
PV power system with and without battery energy storage.
The result of the analysis shows that unmet load is
substantially decreased when battery energy storage is
applied. Using energy storage, unmet capacity shortage
is decreased from 23% to 0%. Thus, in terms of attaining a
steady power supply for about 8 h/day, the PV with battery
energy storage system might be a good candidate compared
with the PV model without battery. Compared with the
cost of electricity using wind power (0.17$/kW h), the cost
of electricity for the PV system with battery energy storage
is more than double, about 0.40$/kW h.
The RO system base load, PV power produced and
battery power dispatched for the rst 2 weeks (330 h) of
January are shown in Fig. 8. According to the simulated
results, most of the time, the PV power is above or equal to
the base load. Since the RO system can operate at up to

27 kW, all the electricity could be used producing as high as


8 m3/h. Under this scenario, on average water production is
about 5 m3/h. Without battery storage, annual water
production is about 11,600 m3 and with battery storage,
about 14,000 m3. Thus, PV power with battery storage
increases water production by about 20%. Fig. 9 shows the
annual load, power and water produced. From this
analysis, it can be concluded that the PV-battery conguration is an appropriate conguration option, enough to
achieve daily water needs without requiring investment in a
large water storage tank.
3.2.5. Diesel power
3.2.5.1. Model input. In this model a 50 kW diesel
generator is considered. The estimated capital and replacement costs are $14,000, and $10,000, respectively, the
O&M cost is $1.5/h, and the diesel cost in Eritrea is
assumed to be $1.00/l (2006 Eritrean diesel price). The
lifetime of the generator is 20,000 h. In Eritrea, the cost of
electricity produced by a diesel generator is at least twice as
expensive as grid electricity [42]. This is mainly due to the
high diesel transporting cost to remote areas, high
maintenance costs, and high fuel consumption. Despite

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

625

RO Base load, PV Power Produced and Water Produced for one year

RO Baseload kW
Water Produced

35
30
Produced(m3/hr)

Base load and PV Power (kW), and Water

PV Power kW

40

PV Power

25
20
15

RO Baseload

10

Water Produced

5
0
0

1000

2000

3000

4000

5000

6000

7000

8000

Time (hr)

Fig. 9. Annual RO base load, PV power, and water production.

Table 1
Summary of power and water costs for different energy options
Energy options

Power cost ($/kW h)


Water cost ($/m3)

W_B (1)

W_PV_B (2)

W_D_B (3)

W_PV_D_B (4)

W_D (5)

W_PV_D (6)

PV (7)

D (8)

PV_B (9)

0.18
0.53

0.18
0.57

0.26
0.74

0.27
0.74

0.29
0.79

0.30
0.82

0.30
0.82

0.40
1.04

0.40
1.05

Note: W_B: wind and battery; W_PV_B: wind, photovoltaic, and battery; W_D_B: wind diesel battery; W_PV_D_B: wind, photovoltaic, diesel, and
battery; W_D: wind and diesel; W_PV_D: wind, photovoltaic, and diesel; PV: photovoltaic; D: diesel only; and PV_B: photovoltaic and battery.

such high costs, users tend to give high value to the power
generated by diesel generator as it is exible to meet the
energy needs of both domestic and income generating
activities.
3.2.5.2. Results of the analysis. According to the results
of the analysis, a 25 kW diesel generator is enough to meet
the intended water demand, operating for about 8 h/day.
The lowest electricity cost for a feasible option using a
diesel generator is about 0.39$/kW h, similar to the result
for the PV system.
3.2.6. Integrated power system
This option simulates the combination of the three
options such as wind, PV and diesel energy sources with a
battery as energy storage. In this scenario, the challenge is
the assumptions made in choosing the base load congurations. In the case of stand-alone power systems, it is
possible to congure the base load according to the
behavior of the energy sources. In an integrated case, we
assume that the base load is the same as the wind energy
conguration, 24 h/day and 8760 h/year. Then, based on
the simulation results, in addition to least cost options, we
determine the most feasible options.
In an integrated approach, whenever wind energy is
available, it is the cheapest option. Under this scenario, the
cost of electricity of the feasible options ranges from 0.175
to 0.183$/kW h, for stand alone wind turbine with battery,

and wind turbine and PV with battery (W_PV_B),


respectively. A 5 kW PV, 30 kW wind turbine, and 10 kW
battery are one of the least cost (0.183$/kW h) choices.
Moreover, it seems that whenever the wind speed is high
usually during the day, PV energy is also high. Thus,
assuming that water demand will increase, this option
could become more attractive. Wind energy with a battery
is the cheapest option ($0.175/kW h), and it is a feasible
choice with the lowest capacity shortage of about 48%.
While either of the options could be feasible, the PVwindbattery system might be more reliable for the SWRO
system.
4. Discussion
4.1. Energy options
Based on the renewable energy potential of the study
area, both wind and PV energy resources are feasible. With
respect to power cost, the optimization model does not
incorporate price uncertainty. All costs are simulated based
on the NPV. According to the optimization model, at
about a 6% interest rate, the PV-powered RO system is
more expensive than the wind powered RO system, with
the costs about 0.399 and 0.175$/kW h, respectively
(Table 1). However, the results of our sensitivity analysis
show that PV could be competitive with wind energy as the
interest rate increases. As a result of changes in interest

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

626

Power Costs of Different Energy Options


0.55
0.50

Power cost($/kWh)

0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0

10

Energy Options
Fig. 10. Power costs with 25% error.

rates, production, and other factors, energy prices are


volatile, especially for diesel. Thus, assuming about 25%
error, we have calculated the maximum and minimum
power costs (Fig. 10). According to the results, the
minimum and maximum wind power costs are about 0.13
and 0.22$/kW h, respectively. Likewise, for PV power, the
minimum and maximum costs are about 0.30 and 0.50$/
kW h, respectively. In either case, wind power is cheaper
than the PV power. In terms of reliability, for this
particular area of study, it is likely that PV will be more
reliable than wind. This is due to the fact that the period of
availability of sunshine is more predictable than the wind
speed. As a result, it is easier to schedule the operation of
PV-powered RO systems than a wind-powered system.
4.2. Water cost
In the last 20 years, the water cost from SWRO has
dramatically decreased from about $2.8/m3 to $1.5/m3.
This has been attributed to the use of energy recovery
devices, and efcient membranes that allow rejection of salt
at low pressure [43]. Water costs are analyzed based on the
overall life cycle of the RO system, which includes the
capital costs of the initial membrane, pressure vessel,
pumps, energy recovery turbine, and operational and
maintenance costs for membrane replacement and electricity. Based on the schematic two-stage RO system design
shown in Fig. 1 which is analyzed in detail using the ROSA
software package, we have extended the Element Value
Analysis (EVA) model developed by FilmTec Corporation
[22] to determine the life cost of the RO system. The RO
system has two pressure vessels and six system elements, in
which the costs of per vessel and per element (FILMTEC
SW30HRLE-400) are $1000 and $750, respectively. The
total cost of the vessels and system elements is about
$6,500. The capital costs of a pressure exchange (PX)
turbine, PX45s pumps and miscellaneous costs are $12,000,
$5000, and $3000, respectively. We have also assumed a

membrane replacement price of about $750, which gives a


replacement cost of $585, assuming a 13% replacement
rate per year per membrane. Based on these assumptions,
the total capital cost of the seawater reverse osmosis system
excluding the cost of the power system is estimated to be
about $27,000. Energy costs are determined using the
HOMER model in Section 3 as shown also in Table 1.
Assuming that the operation period of the system is about
10 years and a 10% interest rate, the net present cost of the
renewable energy powered seawater reverse osmosis system
ranges from $126,000 for wind with energy storage (W_B),
to $250,000 for PV with energy storage (PV).
The power costs (Table 1) are the results of the
optimization model for the selected energy options. With
the assumed 10-year lifetime of the RO system and 10%
interest rate, water costs are computed as shown in Table 1.
According to the results of the analysis, energy expenses
per cubic meter of water produced range from 0.53$/m3 for
wind powered, to 1.05$/m3 for the PV-powered RO system.
PV powered desalination is more expensive than wind
powered by about a factor of two. Recent studies indicate
that today a cost of $1/m3 for seawater desalination would
be feasible and competitive [44]. With an estimated 25%
water cost error, water costs for each option are shown in
Fig. 11.
According to the results, power cost constitutes the
highest proportion of the total costs. Regardless of the type
of energy option chosen, energy expenses are 80% of the
total expenses. The proportion of the expense for the other
costs constitutes less than 20%, including the initial
membrane and pressure vessel (3%), the capital cost of
the energy recovery turbine, pumps and miscellaneous
(8%), and the membrane replacement cost (3%).
4.3. Water productivity
According to the model, wind powered (Option 1) and
PV powered (Option 9) RO systems could produce about

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

627

Water Costs at different Energy Options


1.50

Water costs($/m3)

1.25
1.00
0.75
0.50
0.25
0.00
0

10

Energy Options

Fig. 11. Water costs with 25% error. (Note: the number of the energy options corresponds to the types of energy options shown in Table 1.)

PV and Wind Powered RO Systems' Monthly Water Production

Water Production(m3/month)

5000
4500
4000
PV Powered

3500

Wind Powered

3000
2500
2000
1500
1000
500

r
em

be

be
D

ov
N

ec

em

ob
ct
O

em

er

r
be

st
Se

pt

Au

gu

ly
Ju

ne
Ju

ay
M

ril
Ap

ar

y
au
br
Fe

ar

y
ar
nu
Ja

ch

Months
Fig. 12. Monthly PV- and wind-powered water production.

30,000 and 15,000 m3/year, respectively. PV produces


about the exact demand at high price and wind produces
more than double of the required amount at the lowest
price. Wind powered water production for the months of
November through March is predicted to be about
4000 m3/month, and for the months of April through
October, about 1600 m3/month (Fig. 12). The PV-powered
RO system produces a constant amount of water, 1200 m3/
month. Based on the potential of water productivity and
water cost, the wind-powered RO system is indicated to be
the best option.
4.4. Implications of battery storage
We have tried to compare the implications of energy
storage in wind and PV-powered systems. For wind and PV
systems with energy storage, the average water production

increases by about 6% and 11%, respectively (Fig. 13).


This implies that, using our model, the use of battery
storage in an RO system is more important for the PVpowered than the wind-powered RO system. Moreover, in
addition to increasing the water production at a minimum
incremental cost of about 6% for battery in wind and 2%
for battery in PV, it helps to decrease the intermittency of
the treated water ow.
4.5. Implication of energy recovery
An attempt has also been made to compare the impact of
energy cost with and without the energy recovery system
(Fig. 14). The results show that using a pressure exchange
(PX) 45s energy recovery turbine, the energy expenses for
the wind powered option, for example, could be reduced
by about 41% from 0.69 to 0.41$/m3. However, it should

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

628

Percentage of water production Increase using battery storage system


Water Production Percentage Increase

14%
PV Powered

Wind Powered

12%

10%

8%

6%

4%

2%

r
em
ec
D

em

be

be

er
N

ov

O
ct
ob

be
Se

pt
em

gu

st

ly
Au

Ju

Ju

ne

ay
M

ril
Ap

ua
Fe
b

ra

ar

ry

y
ar
nu
Ja

ch

0%

Months
Fig. 13. Water productivity and energy storage.

$ per m3 of water

Energy Expense with and without the energy recovery(ER) system under different energy options
$1.6
$1.5
$1.4
$1.3
$1.2
$1.1
$1.0
$0.9
$0.8
$0.7
$0.6
$0.5
$0.4
$0.3
$0.2
$0.1
$0.0

Energy expenses withER ($/m3)


Energy expenseswithout ER($/m3)

W_B

W_PV_B

W_D

W_D_B

W_PV_D_B W_PV_D

PV_B

PV

Energy options

Fig. 14. Energy cost with and without energy recovery.

be noted that the PX 45 turbine has some limitations


with respect to its use for the water ow rates considered.
An upgraded PX 45 turbine to cover all ow rates would
be critical and important for successful and wide
application. The capital cost of a PX 45s energy recovery
turbine, which is about $12,000, could be paid back within
a year.
5. Conclusion
This paper has analyzed the performance of a renewable
energy powered small-scale seawater reverse osmosis
(SWRO) system particularly in terms of water productivity
and energy cost. The RO system has been designed with a
exible operating system capable of operating under
intermittent energy supply. Results of our analysis show

that the contribution of the booster pump and energy


recovery is not negligible in increasing water productivity
and decreasing energy consumption per cubic meter
of water produced. Our results show that using the
available technologies, it is possible now to produce
water at about 2.33 kW h/m3, an amount that is below
previous estimates [14,45]. Using a booster pump and
energy recovery turbine, energy consumption can be
decreased from about 7.87 kW h/m3 to about 2.33 kW h/m3.
The results indicate that wind powered water production
(0.50$/m3) is economically feasible. According to [46], in
China, 1.0$/m3 desalinated water is considered as economically feasible. This implies that even a PV powered system
could be competitive, and much of the low latitude world,
with consistent solar radiation, could benet. Generally, in
Eritrea, water production and tariff costs are about $0.30/m3

ARTICLE IN PRESS
A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

and $0.43/m3, respectively. This implies that the windpowered system could be competitive.
Our energy optimization model for RO desalination
system is a rst step toward facilitating its wider application. Moreover, the energy optimization model, which has
not been used before for such a purpose, provides a useful
and relatively straight forward approach, particularly for
simulating hourly renewable energy production and consequently synchronizing this with the energy load in order
to operate under the wide operating system parameters of
the RO plant. However, in order to determine the
robustness of the methodology, we believe that the model
requires further testing, especially for the design of an
actual SWRO system. With current results, we recommend
that, since the availability of energy recovery systems
capable of operating under highly variable conditions is
limited, for higher water productivity and continuous
energy recovery, it is important to operate above 10 m3/h
feed water at low pressures between 30 and 45 bar. The
lower the pressure, the lower the energy required for
producing a limited quantity of water. In contrast, if a
greater rate of water production is needed, then a higher
energy/higher pressure design is required.

References
[1] Vorosmarty CJ, Green P, Salisbury J, Lammers RB. Water stress in
to todays and tomorrows world. Global water resources: vulnerability from climate change and population growth. Water 2001;3(1);
Science 289:2848.
[2] Vaknin S. The emerging water wars. The progress report. Progress.org; 2005. Accessed December 29, 2005.
[3] Government of Eritrea (GoE). Asmara Power Distribution and Rural
Electrication Project, Environmental and Social Assessment (ESA).
Report including Environmental and Social Management and
Monitoring Plan (ESMMP). Ministry of Energy and Mines, Asmara,
Eritrea, January 2004.
[4] Kalogirou A. Seawater desalination using renewable energy sources.
Prog Energy Combust Sci 2005;31(3):24281.
[5] Gilau A. Desalination by solar distillation. Ministry of Energy,
Mining and Water Resources, Asmara, Eritrea; September 1992.
[6] Feron P. The use of wind power in autonomous reverse osmosis sea
water desalination. Wind Energy Group, the Netherlands. Wind Eng
1985;9(3):1985.
[7] Carta J, Gonzalez J, Subiela V. Operational analysis of an innovative
wind powered reverse osmosis system installed in the Canary Islands.
Sol Energy 2003;75:15368.
[8] Miranda MS, Ineld D. A wind-powered seawater reverse-osmosis
system without batteries. Desalination 2003;153(1):916.
[9] Sultan A, Rheinlander J, Gabler H. Seawater reverse osmosis
powered from renewable energy sourceshybrid wind/photovoltaic/grid power supply for small-scale desalination in Libya.
Desalination 2002;153:1723.
[10] DOW Chemical Company. ROSA 6.0.1 Software, A guide to ROSA
6.0; 2005. /http://www.dow.comS. Accessed October 2005.
[11] National Renewable Energy Laboratory (NREL). HOMER, The
Hybrid Optimization Model for Electric Renewables, 2005. /http://
www.nrel.gov/homer/S. Accessed 2005.
[12] Marie A, Pedersen J. Urban households and urban economy in
Eritrea. Analytical report from the urban Eritrean household income
and expenditure survey 1996/97. Statistics and Evaluation Ofce,
Asmara, Eritrea, May 2001.

629

[13] Habtesion S, Tsighe Z, Anebrhan. Sustainable energy in Eritrea. In:


Proceedings of a national policy seminar. Asmara, Eritrea, October
31, 2001.
[14] Habtesion S, Tsighe Z. Current energy utilization and future
options in rural areas. Department of Energy, Eritrea, October
2002.
[15] African Development Bank. Strategic study of wind energy deployment in Africa. Helimax Energie Inc.; March 2004.
[16] Avlonitis S. Optimization of the design and operation of seawater RO
desalination plants separation. Sci Technol 2005;40:266378.
[17] Busch M, Mickols W. Reducing energy consumption in seawater
desalination. Desalination 2004;165:299312.
[18] Chen J, Li G. Marine reverse osmosis desalination planta case
study. Desalination 2005;174:299303.
[19] Moreno F, Pinilla A. Preliminary experimental study of a small
reverse osmosis wind-powered desalination plant. Desalination
2004;171:25765.
[20] Nisan S, Commercon B, Dardour S. A new method for the treatment
of the reverse osmosis process, with preheating of the feed water.
Desalination 2005;182:48395.
[21] Redondo J, Casanas A. Designing seawater RO for clean and fouling
RO feed, desalination experiences with the FilmTec SW30HR-380
and SW30HR-320 elementstechnicaleconomic review. Desalination 2001;134:8392.
[22] DOW Chemical Company. FILMTEC membrane, system design:
membrane system design guidelines. Form No. 609-02054-604;
2005.
[23] Darwish M, Asfour F, Al-Najem N. Energy consumption in
equivalent work by different desalinating methods: case study for
Kuwait. Desalination 2002;152:8392.
[24] Thomson M, Gwillinm J, Draisely I. Batteryless photovoltaic reverse
osmosis desalination. CREST and Dulas Ltd.; 2001.
[25] Van Buskirk R, Garbesi K, Rosen K. South Red Sea wind and solar
radiation maps. /http://www.punchdown.org/rvbS. Accessed July
2005.
[26] Amjad Z. Reverse osmosis: membrane technology, water chemistry,
and industrial application. New York, NY: Van Nostrand Reinhold
Publishing Co.; 1993.
[27] Energy Recovery Inc. Data sheet: performance curves PX-45s
pressure exchanger; 2005. /http://www.energy-recovery.com/pdfS.
Accessed October 2005.
[28] Lew DJ, Barley CD, Flowers LT. Hybrid wind/photovoltaic systems
for households in Inner Mongolia. In: International conference on
village electrication through renewable energy, New Delhi, 35
March 1997.
[29] Farret A, Simoes G. Integration of alternative sources of energy.
New York: Wiley-Interscience; 2006.
[30] Fung C, Rattanongphisat W, Nayar C. A simulation study on the
economic aspects of hybrid energy systems for remote islands in
Thailand. In: IEEE Region 10 conference on computers, communications, control and power engineering, TENCON 02 proceedings,
vol. 3, 2831 October 2002. p. 19669.
[31] Iqbal M. A feasibility study of a zero energy home in newfoundland.
Renewable Energy Int J 2003;29(2):27789.
[32] Jubran B, Al-Hinai H, Zurigat Y, Al-Salti S. Feasibility of using
various photovoltaic systems for window-type air-conditioning units
under hot-arid climates. Renew Energy 2003;28(10):154553.
[33] Manwell F, McGowan G, Blanco G. Wind/hybrid power system
applications for New England Islands. Wind Eng 2003;27(2)
14355.
[34] Hafez A, El-Manharawy S. Economics of seawater RO desalination
in the Red Sea region, Egypt. Part 1. A case study. Desalination
2002;153:33547.
[35] Rosen K. An assessment of the potential for utility-scale wind power
generation in Eritrea. Masters thesis in environmental studies, August
1999.
[36] Danish Industry Wind Association. /http://www.windpower.orgS.
Accessed October 2005.

ARTICLE IN PRESS
630

A.M. Gilau, M.J. Small / Renewable Energy 33 (2008) 617630

[37] DeCarolis J, Keith DW. Is the answer to climate change blowing in


the wind? In: Proceedings of the rst international doctoral
consortium on technology, policy, and management. Delft University, Delfelt; 2002.
[38] Lorax Energy Systems, LLC. Fuhrlander wind turbines. /http://
www.lorax-energy.comS. Accessed 2005.
[39] Manwell J, McGowan J, Rogers A. Wind energy explained: theory,
design and application. USA: University of Massachusetts; 2002.
[40] DuPont H. Wind diesel workshop. Lorax Energy Systems LLC.
Block Island, RI, September 29, 2004.
[41] United Nations Environment Program (UNEP). Africa environment
outlook 2Our environment, our wealth (AEO-2). UNEP, 2006.

[42] Habtesion S, Ghebrehiwet D, Buskirk R, Lebassi B. The potentials of


wind energy applications in Eritrea. Asmara, Eritrea; 2001. /http://
www.punchdown.org/rvbS. Accessed October 2005.
[43] Meyer-Steele S, Gottberg A. New sea water reverse osmosis plants for
the Caribbean. Energy recovery, brine recovery and cost reduction.
Ionics Incorporated, USA, 2005.
[44] Zhou Y, Tol R. Evaluating the costs of desalination and water
transport. Water Resour Res 2005;41.
[45] Grundisch A, Schneider B. Optimizing energy consumption in SWRO
systems with brine concentrators. Desalination 2001;138:2239.
[46] Zhou Y, Tol R. Implications of desalination for water resources in
Chinaan economic perspective. Desalination 2004;164:22540.