Energy Conversion and Management 216 (2020) 112940

Contents lists available at ScienceDirect

Energy Conversion and Management

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

Coupling electrodialysis desalination with photovoltaic and wind energy T

systems for energy storage: Dynamic simulations and control strategy
Antonino Campionea, Andrea Cipollinaa, , Francesco Caliseb, Alessandro Tamburinia,
⁎

Mosè Galluzzoa, Giorgio Micalea

a
Dipartimento di Ingegneria (DI), Università degli Studi di Palermo, viale delle scienze ed. 6, 90128 Palermo, Italy
b
Department of Industrial Engineering, University of Naples Federico II, Naples, Italy- Piazzale Tecchio 80, 80125 Napoli, Italy

ARTICLE INFO ABSTRACT

Keywords: The presence of desalination systems in polygeneration facilities is usually limited by important difficulties in
Solar energy operating under non-stationary regimes typical of renewable energy sources. Reverse osmosis, namely the most
Time dependent common desalination technology, is characterised by slow dynamics that rarely adapts to the power fluctuations
Renewable energy of renewables. Therefore, the possibility of using electrodialysis coupled with a hybrid photovoltaic/wind energy
Cogeneration
source was investigated in this work. In particular, the combination of photovoltaic and wind energy is very
Desalination
attractive in order to achieve a more stable energy production, while electrodialysis is claimed to be a more
Process control
flexible process compared to reverse osmosis. For this reason, the aim of this work was to analyse the technical
advantages of using electrodialysis in the aforementioned scenarios, Suitable transitory simulation models are
implemented for modelling electrodialysis units, photovoltaic panels and wind turbines. Dynamic scenarios were
analysed, looking at two different time scales. Quasi steady-state simulations were used to study the yearly
operation of 4 electrodialysis units operating in parallel, demonstrating process flexibility over a wide range of
produced flowrates (from 920 to 230 m3/d) and power inputs (5–45 kW) when producing drinking water at a
constant NaCl outlet concentration of 0.25 g/l. Dynamic simulations were adopted to study the daily time scale,
where the desalination unit control system, purposely designed and tuned, was able to maintain a relatively
stable target value in presence of disturbances in power availability, i.e. with a fluctuation of the outlet con-
centration lower than ± 10%, in between 0,23 and 0,27 g/l . Simulation results show how the electrodialysis
process is particularly suitable for the integration within polygeneration systems as energy-buffer.

1. Introduction
Desalination units can generally be part of polygeneration facilities
(i.e. integrated systems with multiple energy outputs such as such as
heating, cooling and electricity and by-products such as desalinated
water, hydrogen, glycerine, etc [1]). Waste heat from thermal power
plants has been used to successfully run thermal desalination plants [2]
or to preheat salt water feeds [3]. Desalination systems are particularly
attractive for integration in polygeneration systems [4], especially
when coupled with renewable energy sources [5]. A number of dif-
ferent desalination technologies are available for the integration in
renewable polygeneration systems. Desalination units may be driven
both by thermal and electrical energy. Thermally-driven desalination
technologies are especially suitable for large-scale district level poly-
generation systems whereas electrically-driven desalination units are
much more flexible and they can be used both for small and large scale
systems [6]. Large-scale desalination systems are very attractive for
their capability to achieve a better economic profitability and higher
efficiencies. On the other hand, large-scale projects also involve several
risks which must be accurately taken into account for the evaluation of
the feasibility of the project [7]. Thermally-driven desalination systems
are also often used in combination with different renewable energy
sources [8], such as geothermal [9] and solar [10]. Usually, poly-
generation systems based on thermally-driven desalination technolo-
gies are very complex and they must include very complex control and
management technologies in order to optimize the operation of the
system itself. Therefore, such systems are often featured by extremely
large capacities and they are very attractive for the production of en-
ergy (electrical, thermal and cooling) and fresh water at district level
[11]. In this context, polygeneration systems based on electric renew-
able sources such as photovoltaic (PV) [12] or wind turbines represent a
very promising scenario. In fact, compared to the systems powered by

⁎
Corresponding author.
E-mail address: andrea.cipollina@unipa.it (A. Cipollina).

https://doi.org/10.1016/j.enconman.2020.112940

Available online 25 May 2020

0196-8904/ © 2020 Elsevier Ltd. All rights reserved.
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940

Nomenclature x Coordinate in the direction of the main flow

z Altitude (km)
Symbols w Parameter of Von Karman correlation
Empirical PV curve-fitting parameter
A Membrane area (m2) Channel thickness (m)
a Axial induction factor Error (g/l)
Ar Rotor area (m2) G Material bang gap energy
B Constant of WT model (K/km) Damping coefficient
b Membrane width (m) Non-Ohmic voltage drop (V)
C Concentration (mol/m3) µ Isc Temperature coefficient of Isc at ref. condition (A/K)
cp Wind turbine power coefficient µVoc Temperature coefficient of Voc at ref. condition (V/K)
GT Insolation (W/m2) Wind speed (m/s)
I Current (A) Air density (kg/m3)
IL Photocurrent (A) D Derivative control time constant (s)
Io Diode reverse saturation current (A) I Integral control time constant (s)
i Current density (A/m2)
Jtot Total molar salt flux (mol/m2s) Abbreviations
K Process gain (g s/m3 l)
k Boltzmann constant (J/K) AEM Anion-exchange membrane
KC Controller gain (m3 l/g s) CEM Cation-exchange membrane
L Stack length (m) ED Electrodialysis
Ncp Number of cell pairs IMC Internal model control
NM Number of PV module connected in series PV Photovoltaic
NS Number of ED stacks RO Reverse osmosis
p Pressure (Pa)
Q Volumetric flow rate (m3/s) Subscripts and superscripts
q Electron charge
qw Total water volumetric flux (m3/m2s) D Dilute
R Universal gas constant elev Height above the ground (elevation)
Rblank Blank specific electrical resistance (Ωm2) in Inlet
Rs PV module series resistance (Ω) mp Point of maximum power
Rtot Cell pair specific electrical resistance (Ωm2) Ns Number of stacks
s Independent variable in the Laplace domain oc Open-circuit
T Temperature (K) ref Reference condition
t Time (s) sc Short-circuit
V Voltage (V) SOL Solution (either concentrate or diluate)
Vcp Voltage drop over one cell pair (V) ss Steady state
Vtot Overall voltage drop (V) tot Total

renewable thermal energy, polygeneration systems including electrical
renewables and desalination exhibit: i) higher flexibility in terms of
capacities; ii) simpler system layout; iii)easier management [13]. In
these cases, desalination technologies can be particularly attractive as
means of using the excess energy that renewable sources produce
during peak periods, producing drinking water instead of using energy
storage devices such as batteries [14].
One of the main issues related to the use of renewable poly-
generation systems consists in the unpredictability of the energy source.
The unavoidable fluctuations of many renewable energy sources (wind,
solar, etc.) make extremely difficult to suitably couple these systems
with the cooling, heating and electrical demands of the users [1]. In this
framework, many researchers investigated several storage technologies
(both thermal and electrical) in order to achieve a more stable power
supply profile [15]. However, such storage technologies are still too
expensive for a good economic profitability. Therefore, some re-
searchers are focusing on innovative solutions that couple renewables
such as solar systems [4], hybrid solar-biomass systems [16] and solar-
geothermal systems [9] with desalination units. These systems are
particularly interesting for small islands [17] (for which economic as-
sessment were even conducted [18]) and isolated communities [19].
Many authors are focusing on specific combinations of renewable en-
ergy sources in order to mitigate the fluctuations typical of using in-
dividual renewable energy sources. For example, the combination of
solar and wind energy is extremely promising. In fact, solar energy is
mainly available during the summer and in the central hours of the day.
Conversely, according to the weather data, wind velocity increases in
winter and during the night. Therefore, combining photovoltaic solar
energy and wind energy the stability of power production profile can be
increased. It is worth noting that, besides solar and wind energy sys-
tems, there are some other innovative processes that can be used in
polygeneration systems. Among them, membrane bioreactors are of
particular interest [20].
The possibility to couple desalination systems with renewable en-
ergy sources has been already demonstrated in the literature. Given the
wider commercial diffusion of reverse osmosis (RO) among desalination
technologies, a lot of works focus on this process. In particular, process
feasibility has been studied and proven for PV-RO [21], wind-RO [22]
and combined PV and wind-RO systems [23]. A number of these works
highlight how the economic feasibility of such processes can be en-
hanced using battery-less power sources, as batteries are associated to a
number of disadvantages such as increased capital costs, limited life-
time and increased maintenance [24]. Nevertheless, when desalination
units are coupled with battery-less systems they are directly subjected
to the fluctuation of power generation, requiring real-time adjusting.
Thomson et al. [25] demonstrated the possibility to apply such systems
to RO from both modelling and experimental points of view. However,
the authors (as well as Manolakos et al. in another work [26]) high-
lighted that long term reliability of the system is not guaranteed due to
the continuous high pressure fluctuation experienced by membranes.

2
A. Campione, et al.
An alternative to compensate for the power fluctuation, by modifying
the production capacity without significantly affect the pressure, is the
use a number of parallel RO plants that are switched on or off based on
the amount of energy available [27]. Nevertheless, this generally im-
plies a system oversizing and an increase in capital costs. Besides the
aforementioned issues, RO systems also suffer from relatively slow and
critical start-ups and shutdowns where a gradual increase/decrease in
system pressure is required before reaching the steady-state [27]. Del-
gado Torres et al. [28] also investigated the integration of reverse os-
mosis with PV and tidal range systems. They performed an energy
analysis aiming at evaluating the phase shift between tidal and solar
energy for a plant at Broome, Australia. They concluded that the use of
the tidal range power plant allows one to significantly increase the
operating time of the desalination plant. The combination of PV and
wind technologies (also including electrical batteries) driving a RO unit
was recently analysed by Atallah et al. [29] for a case study in Egypt.
Their work is based on a simulation developed in Homer. For the
analysed case study, authors concluded that the optimal configuration
consists of 160 kW photo-voltaic (PV) panels, 19 strings of lead-acid
batteries and a 50 kW diesel generator system. A similar configuration
was also analysed by Mito et al. , investigating the hybrid PV-Wind-Ro
configuration for large-scale plants. They also point out that the stabi-
lity of renewable power production is a critical issue for a safe and
reliable operation of RO units. Therefore, short-term energy storage
devices are always mandatory required in order to mitigate such fluc-
tuations. An other possibility may consist in the integration of PV and
with a diesel generator which is used to stabilize the system power
production. They also reviewed several literature studies investigating
the possibility of direct coupling between PV and/or wind and RO units.
They concluded that this configuration is extremely difficult and it re-
quires complex control strategies to be managed. A recent study [30]
also combined PV system with Humidification Dehumidification Desa-
lination (HDD) systems. Authors presented a novel hybrid solar desa-
lination system, implementing a numerical model. In particular, they
focused on nanofluids as solar operating fluid, investigating different
oil-based nanofluids. They found that Cu/oil nanofluid exhibits the
lowest efficiency and they can be suggested only to achieve high
freshwater amounts, at low gain output ratios.
Thus, the above presented literature review clearly shows that hy-
brid PV-Wind-RO systems dramatically suffer for the fluctuations of the
Fig. 1. Scheme of the electrodialysis process, showing the alternated arrangement of anion and cation exchange membranes, channels in which the water flows and
electrode compartments.
Adapted from [49]
3
Energy Conversion and Management 216 (2020) 112940
power produced, requiring the use of some electrical energy storage
device which may significantly affect the overall energy and economic
performance of the system. These issues may be solved using alternative
desalination units, more flexible in terms of fluctuations of power. In
this framework, ElectroDialysis units (Fig. 1) are extremely attractive,
due to their capability to operate under unsteady input energy. The
large interest in the ED process is well demonstrated by the amount of
studies that has been performed in the last years. A comprehensive
review regarding the utilization of ED technology for water desalination
(i.e. the most important application of ED) was recently published [31].
Here, the authors present a detailed review of the technology, model-
ling approaches and ED-related processes. They also introduce the most
innovative configurations under development. Another recent review
[32] focuses on both water and wastewater treatment plants. In this
paper, an overview of the possible sustainability approaches to be used
in the design of ED systems was also provided. The reader may refer to
these review papers for further details regarding ED technology. Ex-
perimental investigation on ED systems working with renewable energy
sources is ongoing since few decades ago [33], including the assessment
of remote areas scenarios [34]. In particular, ED has proven its ad-
vantages to serve off-grid locations with moderately concentrated
brackish waters [35].For this reason, rural areas have been recentrly
chosen to conduct these kind of field tests [36].
ED is an electromembrane process that makes use of ion exchange
membranes and can work in two main operational modes, namely batch
[37] and continuous [38]. In some cases, especially for seawater de-
salination, a multistage configuration can be adopted [39].
As mentioned before, ED can overcome most of the drawbacks that
result from dynamic RO operations. A change in the available power
input can be easily adjusted by changing the applied voltage and by
changing the feed flowrate without the need for particular adjustments,
as low pressures (usually < 1 barg) are involved in the process.
Another advantage is that transient phases during start-ups and shut-
downs are much shorter. Conversely, the main issue that has been
highlighted refers to an increased energy consumption due to the pre-
sence of harmonic disturbances that can be mitigated through filters
[27]. For these reasons, ED is particularly suitable for battery-less sys-
tems. In this context, Malek et al. [40] showed a successful experi-
mental coupling of batch ED with direct wind energy, demonstrating
how the system is insignificantly affected by moderate wind
A. Campione, et al.
Fig. 2. Scheme of the hierarchies of the electrodialysis process model re-
presenting: a) the cell pair model, b) the stack model and c) the overall plant
model, including the energy source, the 4 desalination units in parallel and the
control system (not depicted in the picture).
fluctuations. Similarly, Ortiz et al. [41] proved the operation of a batch
PV-ED system through experiments [42] and simulations [43] in the
scale of hours. Moreover, the dynamics of a PV-ED in continuous mode
was studied and simulated by Troncoso et al. [44]. However, in that
work the presence of a battery was taken into account, so that the
criticalities of battery-less systems were not highlighted. An on-field
demonstration was recently presented by He et al.[36]. Authors pre-
sented an experimental validation of a model of a system including a PV
field and ED reversal desalination system. The study presents the fab-
rication process of the pilot plant tested in India. They proved the ef-
ficiency and the flexibility of their system. In addition, the implemented
methodology allowed to reduce lifetime and capital cost by 35% and
34%, respectively. Solar powered ED has been also used to enhance
nutrient recovery from waste streams [45], showing excellent results.
Xu et al. [46] also investigated the integration of PV and ED technol-
ogies. They directly coupled a PV solar field with ED. Authors devel-
oped a suitable simulation code in MATLAB environment in order to
investigate the influence of voltage, flow rate and ED size. As expected,
they found that the salinity of the outlet stream decreases in case of
higher voltages or larger membrane area.
Besides the technical aspects, ED proved its convenience in different
aspects. From cost estimations, PV-ED resulted more convenient than
PV-RO in the presence of low salinity feeds [47]. A sustainability as-
sessment was also conducted on a PV-ED system, predicting a potential
decrease of one order of magnitude in CO2 emissions compared to grid
mix energy supply [48].
1.1. Aim of the work
The literature review presented in this section clearly shows that the
4
Energy Conversion and Management 216 (2020) 112940
majority of the works available in literature focus on RO desalination
units in different configurations. All these studies pointed out that RO
dramatically suffer for the unpredictability of the renewable power
production. Conversely, ED systems are much more flexible that RO in
terms of fluctuations of input power. However, only a few studies are
available about ED powered by renewables since ED is an emerging
technology compared with RO, which is a much more mature and
commercially available technology. Among THE few works about re-
newable ED systems, most of them focus on the experimental proof of
concept, on coupled process design or on economic analysis. In fact,
attention to the detailed process dynamics is mostly paid only in the
case of batch ED, where the effect of power fluctuations on the product
quality are strongly dampened by the presence of recirculation tanks.
The analysis of the dynamics of the ED systems is a crucial point for the
analysis of the feasibility of renewable ED desalination units.
Unfortunately, as clearly shown by the reported literature review, this
point has not been sufficiently addressed by the scientific community.
Therefore, the present work aims at covering this lack of knowledge,
studying the dynamics of renewable ED systems. In particular, the
present work focuses on a single pass ED unit powered by a hybrid PV/
wind power source in order to prove the flexibility of ED in maintaining
drinking water specifications while changing process conditions, as-
sessing the suitability for integration within polygeneration systems as
energy-buffer. In particular, the analysis was performed by means of
process simulations, focusing on two different time-scales. The first one
is the yearly operational time scale, where quasi-stationary operation of
ED allows for a step by step re-adaptation of working parameters in
order to generate a stable output target with the available energy input.
The other one is the short time scale of transient regimes of the desa-
lination unit, where the intrinsically dynamic behaviour of the unit was
modelled to predict the response to fast disturbances in energy avail-
ability. At this scale, a control system was also designed and im-
plemented.
2. Modelling
The overall simulation model was implemented by a hybrid ap-
proach combining a well-known dynamic simulation tool, TRNSYS [50]
, and a user-developed model. In particular, the model for the solar/
wind system is developed in TRNSYS by using weather data from
Pantelleria, and conventional components included in TRNSYS library
(PV panels, wind turbine, inverter, controllers, etc). The overall elec-
trical production calculated by TRNSYS is subsequently provided as an
input data to the user-developed model, simulating the ED subsystem
implemented into gPROMS Modelbuilder. In the followings, the main
algorithms of the models are briefly presented. For sake of brevity, PV
and wind models are summarized in Appendix, whereas this section
briefly reports ED model.
2.1. Electrodialysis process
The studies presented in this work are based on a steady-state
hierarchical ED model (scheme is reported in Fig. 2) [49] that has been
modified to account for the dynamic behaviour of the system.
The lowest hierarchy of the model (Fig. 2 a), i.e. the cell pair, is the
repeating unit of the system, including an anion-exchange membrane
(AEM), a diluate channel, a cation-exchange membrane (CEM) and a
concentrate channel. At this scale, material balances and transport
phenomena are described. In particular, the dynamic, 1D mass balances
in a generic (concentrate or diluate) channel can be written as:
b
CSOL (x , t )
+
QSOL (x , t ) CSOL (x , t )
= ± bJtot (x , t )
SOL
t x (1)
dQSOL (x , t )
= ± bqw (x , t )
dx (2)
A. Campione, et al.
where A is the membrane active area, SOL is the channel thickess,
CSOL CSOL (x, t) is the salt concentration in solution, b is the channel
width, Jtot Jtot (x, t) is the total salt flux through the membranes (i.e. the
sum of conductive and diffusive salt flux [49]), QSOL QSOL (x, t) re-
presents the local volumetric flow rate of a single cell pair, and qw qw (x)
the local overall volumetric water flux through the membranes (i.e. the
sum of osmotic and electroosmotic fluxes [49]). t and x indicates that
the variables are function of time and space (i.e. the axial direction of
the solution flow through the channels).
Electrical variables of the cell pair are also computed. In particular,
the cell pair voltage drop (Vcp Vcp (t) ) can be written as:
Vcp (t ) = (x , t ) + Rtot (x , t ) i (x , t ) (3)
where is the non-Ohmic voltage drop associated to the back
electromotive force where concentration polarisation is also taken into
account making use of computationally determined Sherwood numbers
[49,51–53], i i(x, t) is the local current density and Rtot is the total areal
Ohmic resistance of cell pair which can be calculated as the sum of
membrane and solution compartment resistances [49].
A number of cell pairs are grouped together in the Stack higher
hierarchy (Fig. 2 b), where the external applied voltage (Vtot Vtot (t) ) is
related to the internal electric variables:
Ncp
Rblank I
Vtot (t ) = + Vcpi (t )
A i=1 (4)
where, Rblank is the blank resistance, accounting for electric voltage
drops in the electrode compartments, I is the total current, calculated as
the integral of the current density over the active area, and Ncp is the
number of cell pairs inside the stack.
The stack model also computes performance parameters such as
current efficiency and specific energy consumption [49].
Finally, the stack can be implemented in the highest hierarchy of the
plant (Fig. 2 c). In the case of this work, the stack model was coupled
with a variable power source and a control unit (Fig. 3). The details of
the controller equation, as well as its design and tuning are discussed in
the control design section.
3. Long time scale simulations
The aim of this work was to simulate an ED unit powered by a
battery-less hybrid PV/Wind energy source. In particular, the energy
system includes a 20 kW wind turbine and a PV array with a peak of
power production of about 25 kW, for an overall peak power produc-
tion of about 45 kW. The simulations were performed using the weather
data of Pantelleria, taken from Meteonorm database. Here, tempera-
ture, radiation, wind velocity and many other weather data are pro-
vided on a hourly basis for a representative year. Therefore, the de-
veloped tool allows one to simulate the operation of the system for
whatever time basis: it can simulate the transient operation of the
system for shot periods (seconds or minutes) up the complete yearly
simulation. Transient simulations are very useful in order to investigate
the off-design performance of the components. Conversely, yearly si-
mulations are very useful in order to calculate the overall energy and
economic performance of the system. A flow chart representing the
simulation procedure is shown in Fig. 4.
The ED plant includes 4 equal stacks operating in parallel that have
to desalinate a feed stream of 5 g/l NaCl concentration down to 0.25 g/l
with 67% recovery (i.e. the diluate to concentrate feed flowrate ratio is
always kept constant to 2:1) . Stacks’ specifications are listed in Table 1.
At nominal conditions, the plant works with ~ 80% of the peak
power (34 kW), producing 800 m3/d (200 per stack) of drinking water.
However, the actual diluate flowrate changes according to the available
power, in order to keep a constant outlet concentration. Based on fea-
sibility conditions, upper and lower limits were set in each stack for the
diluate flowrate. In particular, the maximum diluate feed flowrate was
5
Energy Conversion and Management 216 (2020) 112940
set according to the maximum allowable pressure drop that has been set
at 1.2 bar, corresponding to 920 m3/d (230 m3/d per stack) and 43 kW
of absorbed power. Conversely, the minimum allowable flowrate was
set at ~ 230 m3/d (58 m3/d per stack) and 5 kW of absorbed power,
where the current required to reach 0.25 g/l is already 90% of the es-
timated limiting current [54]. Further reduction in the diluate feed
flowrate would cause limiting current issues and thus an impossibility
to reach the target concentration.
According to the aforementioned limits, the ED system was simu-
lated assuming that the applied voltage changes as a function of the
power produced by the energy system over the entire year. The
boundaries were taken into account, so that if the available power is
higher than the upper limit, the plant does not use all of it. Conversely,
when the power is lower than the minimum the system is switched off.
Given the long time scale, a quasi-steady state approach was adopted
for these simulations.
Fig. 5 a, shows how much the overall power can be different during
the year (according to weather data from Pantelleria). In particular,
peaks are clearly more frequent in summer months, when the PV system
has more solar radiation available. Produced power goes slightly down
in the other periods of the year, even though the wind turbine partially
compensates the reduction of solar radiation.
Fig. 5 b shows the diluate feed flowrate of the ED plant (QDNs ). As
expected, the plant can adapt the feed flowrate to the produced power,
and thus generating different amounts of drinking water. Interestingly,
the set maximum and minimum flowrate boundaries do not excessively
affect the power utilisation on the yearly time scale. In fact, the plant
never goes off for long periods except for few phases during winter
months, while only very few power peaks are not entirely used during
the summer. Fig. 5 also shows cumulative curves (black continuous
lines) that gives an estimation of the total amount of hours in which
power or flowrate were maintained above a certain value. From these
curves it can be seen how the plant stays on for more than 5500 (non-
continuous) hours during the entire year. In addition, it is interesting to
note that the specific energy consumption decreases while reducing the
feed flowrate. This is due to the operating conditions that are closer to
reversibility when the applied voltage is small.
Such findings indicate a large potential for application, especially in
remote or isolated sites, where small electrical grids require effective
buffering strategies when the utilisaiton of renewable energies
Fig. 3. Detailed block diagram of the highest hierarchy of the plant model re-
presenting an electrodialysis unit powered by the Photovoltaic/wind power
delivery system. The control system of the product concentration, acting on the
feed flowrate is also shown.
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940

Fig. 4. Flow chart of the integrated system simulation strategy. The model of the control system has been adopted only for short-time scale simulations.

Table 1
List of the main characteristics of each simulated electrodialysis stack.
L (cm) b (cm) Ncp SOL (μm) Type of membranes Rblank (Ωcm2)

50 50 500 270 FUJIFILM Type 10 3

increases.

4. Control system design and tuning

As the ED unit is powered by a variable power source (i.e. the

coupled PV/Wind plant) the actual energy that is available for the de-
salination process and, thus, the external applied voltage will change
over time. The desalination unit has to maintain the outlet concentra-
tion of the drinking water at the target value. Therefore, a control
system is required in order to continuously adapt the feed flowrate to
ensure that the outlet specifications are met. For this reason, the system
dynamics was first studied with the aim of defining transfer functions,
which link the inlet variables (i.e. feed flow rate and applied voltage) to
the outlet variables (i.e. the product concentration) and can be assumed
as simplified mathematical descriptions of the nonlinear process. The
results of this analysis was then used to define the control strategies and
tune the controller. For the sake of simplicity, the control loop takes
into account only process dynamics, neglecting other contributions
such as the regulation valve.

4.1. Uncontrolled process dynamics

In order to study the process dynamics, it is necessary to char-

acterise the behaviour of the process under transient regime when the
manipulation variable (i.e. the flowrate) shifts from the stationary
value. Given the nature of the process under study, it is not possible to
define a unique reference stationary value, as this will depend on the
available power. Therefore, three reference scenarios were chosen, so
that it was also possible to assess how the dynamics changes with the Fig. 5. a) Simulated yearly power production for the coupled Photovoltaic/
starting condition. In particular, the maximum and the minimum al- Wind energy source, b) Simulated flowrate of the drinking water produced in
lowable flowrates (as discussed in the long time scale simulation sec- the 4 electrodialysis units operating in parallel. Vertical bars represent the
tion) as well as the average flowrate have been taken as a reference. In hourly variation, while the black lines are the cumulative curves for the two
each scenario, the flowrate has been either increased or decreased by variables. electrodialysis units specifications: 50x50 cm2, 500 cell pairs, 270 μm
20% through a step change. spacers and FUJIFILM Type 10 membranes.
Fig. 6a shows time profiles of the diluate outlet concentration after
the step change. At each steady state flowrate, the voltage is set in order

6
A. Campione, et al.
Fig. 6. a) Outlet diluate concentration vs. time in response to a 20% step change in the feed flowrate at different diluate flowrates, b) Variation of the salt
concentration along a diluate at different times for a 20% flowrate step increase starting from an initial flowrate of 144 m3/d. Results refer to a 50x50 cm2
electrodialysis unit equipped with 500 cell pairs, 270 μm spacers and FUJIFILM Type 10 membranes.
to reach 0.25 g/l. Therefore, when the flowrate increases or decreases,
the concentration will reach a new steady state with a higher or lower
concentration respectively, requiring a certain time that depends on the
flowrate itself. When the flowrate is suddenly changed, the spatial
concentration profile in each channel will have to adapt to the new
situation (see Fig. 6b), requiring a mutual interaction between each
adjacent discretisation interval, through which a certain volume of
solution is flowing. Therefore, the resulting dynamic is given by a
number of “virtual” processes that occur in sequence and results in a
high-order behaviour. However, for the sake of simplicity, each of the
curves of Fig. 6a was fitted with a second-order transfer function with a
zero that, in the Laplace domain is expressed as:
K 0s + K
G (s ) =
s+2 s+1 (5)
where K is the process gain, 0 is the time constant of the transfer
function numerator, is the process time constant, is the damping
coefficient and s is the independent variable in the Laplace domain.
Each fitting curve from Fig. 6a was characterised by its own parameters
according to eq. .
4.2. Feedback control design
The aim of the Feed-Back (FB) control is to keep the target con-
centration by adjusting the feed flowrate, according to the following
equation:
1 t d 4.3. Feed Forward control design
QDin, tot (t ) = K C + dt + D + QDin,,sstot (t )
I 0 dt (6)
where K C is the control gain, is the error (i.e. the difference be-
tween the concentration set point and the actual concentration), QDin, tot
and QDin,,sstot are the actual and the steady state diluate flowrate entering
inside a stack and I and D are the integral and derivative time con-
stants of the controller. The controller equation written in this form
includes proportional, integral and derivative actions.
In order to design an effective and robust controller for the process
under study, it is required to identify the best set of parameters (i.e. K C ,
I and D ). Among the different design methods, the internal model
control (IMC) was used to estimate controller parameters [55], as it is
one of the most common method for first estimation. With the IMC
7
Energy Conversion and Management 216 (2020) 112940
method, it was possible to estimate a first set of control parameters for
each reference scenario from the fitted process transfer functions pre-
viously identified (Fig. 6a). These parameters were then averaged and
finely tuned with a trial and error procedure by simulating the response
of the controlled system to step voltage changes on gPROMS Mod-
elbuilder. The final values are reported in Table 2.
Fig. 7 shows simulation results for the controlled ED stack for each
reference flowrate. At the maximum flowrate, the system is subjected to
a 20% step decrease of the applied voltage (Fig. 7 a), the opposite trend
is detected for the minimum flowrate (Fig. 7 d), while at 144 m3/d both
an increase and a decrease of the applied voltage were analysed (Fig. 7
b and c). As it can be observed, going from the highest to the lowest
flowrate (and thus decreasing the channel velocity), the oscillations of
the outlet concentration around the stationary value become more
persistent and take more time to settle, even though they are still at an
acceptable value and any deviation from the set-point disappear after a
time of 300–400 s. This can be explained by the transient behaviour of
the system that significantly changes from high to low flowrates, ac-
cording to the dynamic responses already shown in Fig. 6 a. Besides, a
slower response is expected at lower residence times as the process it-
self needs more time to adapt to the disturbance. It is worth noting that
the derivative term was included in the controller in order to dampen
the oscillations of the outlet concentration generated by the propor-
tional-integral action of the controller. In particular, excessive decrease
in the concentration may cause limiting current issues.
Given the measurable nature of the main disturbance to the process,
namely the available power/voltage at the ED unit, also a Feed Forward
(FF) and a hybrid FB-FF controller were designed and tested via si-
mulations. In order to adapt the controller to the strong non-linear
behaviour of the process, average gain and time constants were adopted
Table 2
Final values of control parameters.
K C (m3 l/ g s) I (s) D (s)
0.01 30 15
A. Campione, et al.
Fig. 7. Dynamic response of the controlled (i.e. the outlet diluate concentration) and the manipulated (i.e. the feed flowrate) variables (PID feedback controller) for a
voltage step change at different initial diluate feed flowrates. a) 230 m3/d, 20% voltage decrease, b) 144 m3/d, 20% voltage increase, c) 144 m3/d, 20% voltage
decrease, d) 58 m3/d, 20% voltage increase. ED unit specifications: 50x50 cm2, 500 cell pairs, 270 μm spacers and FUJIFILM Type 10 membranes.
for the design of the controller, leading to a unified law for the FF
controller.
The same scenarios adopted for the analysis of the FB controller
performance were chosen in this case and results are reported in Fig. 8
for the stand-alone FF controller and in Fig. 9 for the hybrid FB-FF one.
The implementation of the stand-alone FF controller helps in
minimizing the off-set restoring a concentration value close to the
target. However, the control is not ideal and some transient deviation
from the steady-state value of product concentration is observed.
Looking at the mostly common case of hybrid controller, the re-
sponse of the system shows several small improvements compared to
the simple case of FB controller, both in terms of system stability and
amplitude of oscillations.
However, such improvements may have a limited interest in the real
cases analysed hereafter due to the smaller and slower disturbances
occurring in real operation of solar/wind-powered ED systems.
Therefore, the simple FB controller has been adopted for the simula-
tions of short-time-scale scenarios presented in Section 5.
5. Short time scale simulations
The dynamic ED model was used to simulate the operation of the
controlled ED plant in 4 typical days (from 00:00 to 23:59), that were
chosen as representative of the four seasons. At this scale, it is possible
to observe the dynamic effect of the control system. For this reason, the
voltage applied to the ED units was changed with a 3 min step, in order
8
Energy Conversion and Management 216 (2020) 112940
to observe discrete power changes that significantly challenge the sta-
bility of the control system.
Simulation results for the daily operations are shown in Fig. 10. For
each day, the available power, the voltage applied to each ED unit, the
plant diluate flowrate and the outlet diluate concentration are reported.
During the sample winter day (Fig. 10 a), power production is very low
in the very first hours (i.e. before the sunset and with almost no wind).
Then, some small power peaks are observed during light hours while
the power grows up to 20 kW (i.e. the maximum capacity of the wind
turbine) during the night due to an increased wind speed. This means
that for the first hours of the days the unit is not able to operate as the
power is below the power limit, while it keeps running for the rest of
the day, except for an additional short period, reaching its maximum
daily production in the last quarter of the day.
On the other hand, spring and autumn days (Fig. 10 b and d) present
a more unstable behaviour, characterised by a similar non-operational
period of about 6 h in between the first 2 quarters of the day and a
power peak in the third quarter. The main differences between the two
days is that the spring day shows a higher peak power production and
higher peaks in the controlled concentration.
The summer reference day (Fig. 10 c) is the only one in which plant
operation is never interrupted, as the power produced by the PV/Wind
hybrid system is always above the minimum threshold. Despite this, the
day is characterised by a high number of steep variations, especially
during light hours when solar irradiation keeps changing according to
weather conditions. For this reason, the highest concentration peaks are
A. Campione, et al.
Fig. 8. Dynamic electrodialysis response of the controlled (i.e. the outlet diluate concentration) and the manipulated (i.e. the feed flowrate) variables in the case of a
stand-alone Feed Forward controller for a voltage step change at different initial diluate feed flowrates. a) 230 m3/d, 20% voltage decrease, b) 144 m3/d, 20% voltage
increase, c) 144 m3/d, 20% voltage decrease, d) 58 m3/d, 20% voltage increase. electrodialysis unit specifications: 50x50 cm2, 500 cell pairs, 270 μm spacers and
FUJIFILM Type 10 membranes.
observed during this day.
Notably, in all cases the quality of the desalinated water product is
maintained very close to the target value of 0.25 g/lt, while the max-
imum product flow rate ranges from about 400 to 800 m3/day. Such
amount offresh water is suitable for a population of several thousands
persons.
Depending on the site location and energy availability, the size of
the plant can be easily increased up to 10 times thanks to the tech-
nology flexibility and plant modularity, thus reaching a relevant size
also for large communities and small cities.
In addition to the overall daily data, some dynamic details of the
control system action are also shown in Fig. 10. For each day, a re-
sponse to a positive and a negative step disturbance as well as to a ramp
are highlighted. Relatively small peaks and fast responses are found for
step disturbances (as discussed in the controller tuning section). Even in
the summer day, characterised by the highest peaks, the outlet con-
centration stays below ± 10% of the set point value. In the same way,
the oscillatory responses to ramp changes shows very limited oscilla-
tions around the set point.
It might seem that the presence of the aforementioned concentra-
tion peaks is an inconsistent result compared to what was found by
Ortiz et al. [42], where a smooth concentration curve vs time is found.
However, while the present work refers to a continuous ED system, that
is supposed to always produce drinking water at the target concentra-
tion, the work from Ortiz et al. takes into account a batch operation,
where the goal is to reach the target concentration only at the end of the
operation. Therefore, during a batch operation the change in available
9
Energy Conversion and Management 216 (2020) 112940
power only affects the time required to reach the desired concentration
and not the quantity or the quality of the final product.
The behaviour of the outlet concentration can also be compared
with the one found by Thomson et al. [25] in a battery-less PV-reverse
osmosis (RO) system. In that work, the system was able to avoid large
changes in the product flowrate at the cost of larger variations in the
concentration (0.2–0.6 g/l). In addition, the control of the system re-
sulted much more complex in the case of RO due to the fact that it was
necessary to control the rotation of two different pumps in order to
control water recovery ration and maximize the system efficiency.
6. Conclusions
This work presents an analysis of the behaviour of the single pass ED
process when powered by a hybrid PV/Wind energy source via dynamic
simulations. Simulations were performed through an ED process model
in both a yearly and daily time scale.
In the yearly time scale, a quasi-steady state approach was adopted.
4 ED units working in parallel were simulated, in order to demonstrate
how the process can work over the year within a wide range of flow-
rates following the PV or wind power availability. In particular, in order
to achieve the target product water quality (i.e. a concentration of total
dissolve solids of 0.25 g/l) the total diluate flowrate was ranged from
230 to 920 m3/d, corresponding to an absorbed power from 5 kW to
43 kW respectively (with this latter being 25% above the nominal op-
erating condition), using the power produced by the variable power
source. The plant is able to operate for most of the year, even though it
A. Campione, et al.
Fig. 9. Dynamic electrodialysis response of the controlled (i.e. the outlet diluate concentration) and the manipulated (i.e. the feed flowrate)variables in the case of a
hybrid Feedback-FeedForward controller for a voltage step change at different initial diluate feed flowrates. a) 230 m3/d, 20% voltage decrease, b) 144 m3/d, 20%
voltage increase, c) 144 m3/d, 20% voltage decrease, d) 58 m3/d, 20% voltage increase. electrodialysis unit specifications: 50x50 cm2, 500 cell pairs, 270 μm spacers
and FUJIFILM Type 10 membranes.
switches off for short periods (especially during the winter) due to the
excessively low power, reaching a cumulative operational period of
more than 5500 h.
On the other hand, daily simulations aimed to study detailed pro-
cess dynamics in the same configuration studied in the previous case
(i.e. 4 ED units in parallel). A control system based on Feedback and
Feedforward controllers has been designed in order to maintain the
desired product concentration when the available power changes by
acting on the feed flowrate. In particular, the two control systems were
designed based on the study of dynamic behaviour of the ED units to
step changes of the applied voltage (i.e. the disturbance generated by
the change in the available PV/WIND power) and of the feed flowrate
(i.e. the variable that is manipulated in order to keep a constant outlet
concentration). Simulating the power variation over 4 different re-
ference days (one for each season), it was shown that the controlled
system works well throughout each day, always ensuring a stable op-
eration and relatively short settling times of the outlet concentration,
with maximum fluctuations lower than ± 10% of the set point (i.e. in
between 0.23 and 0.27 g/l, as the target concentration was fixed to
0.25 g/l). The advantages of adding a Feedforward controller to the
traditional PID feedback controller resulted to be small and not so re-
levant for the operation in real operating scenarios.
Results show that, given the high flexibility and the fast and con-
trollable process dynamics, ED proved its suitability for coupling with
10
Energy Conversion and Management 216 (2020) 112940
highly non-constant power sources. The operational ranges, in terms of
product water flow rates, can go from few tens of m3/d for very small
applications, up to several thousands of m3/d for large industrial plant,
thus potentially covering the needs for fresh water supply in remote
villages up to small cities. Within such scenarios, the process could be
successfully implemented in battery-less polygeneration systems for
fresh water production, also acting as an energy buffer device.
CRediT authorship contribution statement
Antonino Campione: Conceptualization, Methodology, Software.
Andrea Cipollina: Conceptualization, Supervision, Funding acquisi-
tion. Francesco Calise: Conceptualization, Methodology, Software.
Alessandro Tamburini: Methodology, Supervision. Mosè Galluzzo:
Conceptualization, Methodology, Software. Giorgio Micale:
Supervision, Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940

Fig. 10. Results for daily simulations of the 4 feedback controlled electrodialysis stacks powered by a Photovoltaic/Wind system whose power output changes every
3 min throughout the day. Graphs show available power, applied voltage, inlet flowrate and diluate outlet concentration (for the latter the overall daily trend as well
as detailed dynamics of few minutes is shown) for 4 days representative of a) Winter, b) Spring, c) Summer, d) Autumn. Units’ specifications: 50x50 cm2, 500 cell
pairs, 270 μm spacers and FUJIFILM Type 10 membranes.

11
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940

Acknowledgments ElectroDialysis systems). The REvivED water project has received

funding from the European Union’s Horizon 2020 research and in-
This work has been performed within the REvivED water project novation programme under Grant Agreement no. 685579 (www.revi-
(Low energy solutions for drinking water production by a REvival of vedwater.eu).

Appendix:. Photovoltaic and wind turbines models

A.1 Photovoltaic panels

In order to simulate the PV panels, the four parameter model was used. It assumes that the slope of the IV curve is zero at the short-circuit
condition:
dI
=0
dV v=0 (7)
The four parameters included in the model are: i) IL,ref (module photocurrent at reference conditions), ii) Io,ref (diode reverse saturation current at
reference conditions), iii) γ (empirical PV curve-fitting parameter), iv) Rs (module series resistance). The software uses parameters values from
manufacturers’ data in order to generate an IV curve at each time step.
The current–voltage equation of the circuit is:

GT q
I = IL, ref Io exp (V + IRs ) 1
GT , ref kTc (8)
where GT and GT,ref are the insolation and the reference insolation respectively, q is the electron charge, k is the Boltzmann constant and Tc is the
temperature of the cells.
The diode reverse saturation current Io is a temperature dependent function, such as:
3
Io Tc
=
Io, ref Tc, ref (9)
where Io,ref is one of the four parameters defined above.
Once Io is obtained, the Newton’s method is employed to calculate the PV current, whereas an iterative search routine finds the current (Imp) and
voltage (Vmp) at the point of maximum power along the IV curve. To solve the four equivalent circuit characteristics, current and voltage at open-
circuit, short circuit, and maximum power conditions are substituted into eq. (8), yielding, after some rearrangement, to the following three eqs.
(10)-(12) related to IL,ref, Io,ref, γ:
IL, ref Isc, ref (10)

q (Vmp, ref Voc, ref + Imp, ref Rs )

=
kTc, ref ln 1 ( Imp, ref
Isc, ref ) (11)
Isc, ref
Io, ref =
exp ( qVoc, ref
kTc, ref ) (12)
A fourth equation, eq. (13), derived by taking the analytical derivative of voltage with respect to temperature at the reference open-circuit
condition, is needed in order to determine the last unknown parameter:

Voc k Isc, ref Tc µisc qG

= µ voc = ln + 3+
Tc q Io, ref Isc , ref kTc, ref
NM (13)
This analytical value is matched to the open circuit temperature coefficient (manufactures’ specification). Finally, an iterative search routine is
followed to calculate the equivalent circuit characteristics.

A.2 Wind turbine

The model calculates the power output P of the WT through the power coefficient of WT, cp, multiplied by the area of the rotor and the wind
power, as reported in eq. (14):
P = cp Ar v 3 = 4a (1 a) 2Ar v 3 (14)
2 3
here, v is the wind speed (m/s), Ar is the rotor area (m ) and ρ is the air density (kg/m ). The cp is a function of the axial induction factor, a, and
its maximum value of 59.3%, obtained for a = 1/3, was first derived by Betz in 1919 (known as Betz's limit).
The WT power calculation is based on a power versus wind speed characteristic (Fig. A.1), provided by the manufacturer.
This model takes into account the air density changes and wind speed increases with height above the ground (elevation). Air density at a certain
elevation is a function of the combined effects of pressure and temperature, according to the ideal gas law, and it is calculated as:
pelev
=
elev
RT (15)
Regarding the variability of the temperature as a function of the elevation, the temperature “lapse rate” is also considered as shown in the

12
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940

following eq.16 :
T (z ) = T0 B z (16)
where B = 6.5 K/km of altitude and T0 = 288 K.
The change in wind speed per change in height above the ground is based on the theoretical work of Von Karman [56]. Here, the relation
between the elevation and the wind speed is formulated as follows:
w
v1 z
= 1
v2 z2 (17)
A single parameter, αw, determines the rate of wind speed increase as a function of height. Under ideal boundary layer conditions, the value of αw
is 1/7 (0.14). However, under actual conditions, the value of αw constantly varies, and depends on a plurality of factors, affecting vertical turbulence
intensity (surface roughness, mountains, buildings, atmospheric stability, etc.).
The main design parameters of wind turbine and PV panels adopted in the models are reported in Table A.2.

Fig. A1. Power (left) and cp (right) versus wind speed characteristic curve.

Table A2
Main design parameters of wind turbine and PV panels.
Parameter Description Value Unit

PV PANELS Amodule,PV PV module area 1.609 m2

Pmodule,PV Unit peak power 260 W
Isc,ref Module short-circuit current at reference conditions 6.50 A
Voc, ref Module open-circuit voltage at reference conditions 21.6 V
Tc, ref Reference temperature 298 K
GT,ref Reference insolation 1000 W/m2
Vmp,ref Module voltage at max power point and reference conditions 17 V
Imp,ref Module current at max power point and reference conditions 5.9 A
μIsc Temperature coefficient of Isc at (ref. condition) 0.02 A/K
μvoc Temperature coefficient of Voc (ref. condition) −0.079 V/K
Tc, NOCT Module temperature at NOCT 313 K
Tc, ref Ambient temperature at NOCT 293 K

WIND TURBINE z Site elevation 205 m

H Data collection Height 18 m
Hhub Hub height 10.2 m
NWT Number of turbines 1 –
PWT Wind turbine rated power 10 kW
vrated Wind turbine rated speed 6.5 m/s
vcut,in Cut-in speed 2.6 m/s
vcut,off Cut-off speed 16 m/s

13
A. Campione, et al.
References
[1] Calise F. de Notaristefani di Vastogirardi G, Dentice d’Accadia M, Vicidomini M.
Simulation of polygeneration systems. Energy 2018;163:290–337. https://doi.org/
10.1016/J.ENERGY.2018.08.052.
[2] Cardona E, Culotta S, Mangiaracina L, Terranova R. Multi-flash desalination and
cogeneration to solve the water supply problems in southeastern Sicily. Desalination
1996;106:121–30. https://doi.org/10.1016/S0011-9164(96)00100-2.
[3] Kamal I. Cogeneration desalination with reverse osmosis: A means of augmenting
water supplies in Southern California. Desalination 1992;88:355–69. https://doi.
org/10.1016/0011-9164(92)80127-U.
[4] Calise F, Dentice d’Accadia M, Vanoli R, Vicidomini M. Transient analysis of solar
polygeneration systems including seawater desalination: A comparison between
linear Fresnel and evacuated solar collectors. Energy 2019;172:647–60. https://doi.
org/10.1016/J.ENERGY.2019.02.001.
[5] Mollahosseini A, Abdelrasoul A, Sheibany S, Amini M, Salestan SK. Renewable
energy-driven desalination opportunities – A case study. J Environ Manage
2019;239:187–97. https://doi.org/10.1016/j.jenvman.2019.03.044.
[6] Wang Z, Lin X, Tong N, Li Z, Sun S, Liu C. Optimal planning of a 100% renewable
energy island supply system based on the integration of a concentrating solar power
plant and desalination units. Int J Electr Power Energy Syst
2020;117:105707https://doi.org/10.1016/j.ijepes.2019.105707.
[7] Zhang Y, Wang R, Huang P, Wang X, Wang S. Risk evaluation of large-scale sea-
water desalination projects based on an integrated fuzzy comprehensive evaluation
and analytic hierarchy process method. Desalination 2020;478:114286https://doi.
org/10.1016/j.desal.2019.114286.
[8] Calise F, d’Accadia MD, Quiriti E, Vicidomini M, Piacentino A. Trigeneration and
Polygeneration Configurations for Desalination and Other Beneficial Processes.
Sustain. Desalin. Handb. Plant Sel. Des. Implement., Elsevier 2018:99–199. https://
doi.org/10.1016/B978-0-12-809240-8.00004-6.
[9] Calise F, Dentice d’Accadia M, Macaluso A, Vanoli L, Piacentino A. A novel solar-
geothermal trigeneration system integrating water desalination: Design, dynamic
simulation and economic assessment. Energy 2016;115:1533–47.
[10] Calise F, Dentice d’Accadia M, Piacentino A. A novel solar trigeneration system
integrating PVT (photovoltaic/thermal collectors) and SW (seawater) desalination:
Dynamic simulation and economic assessment. Energy 2014;67:129–48. https://
doi.org/10.1016/j.energy.2013.12.060.
[11] Colmenar-Santos A, Palomo-Torrejón E, Mur-Pérez F, Rosales-Asensio E. Thermal
desalination potential with parabolic trough collectors and geothermal energy in
the Spanish southeast. Appl Energy 2020;262:114433https://doi.org/10.1016/j.
apenergy.2019.114433.
[12] Calise F, Cappiello FL, Vanoli R, Vicidomini M. Economic assessment of renewable
energy systems integrating photovoltaic panels, seawater desalination and water
storage. Appl Energy 2019;253:113575https://doi.org/10.1016/j.apenergy.2019.
113575.
[13] Khan EU, Martin AR. Optimization of hybrid renewable energy polygeneration
system with membrane distillation for rural households in Bangladesh. Energy
2015;93:1116–27. https://doi.org/10.1016/J.ENERGY.2015.09.109.
[14] Usón S, Uche J, Martínez A, del Amo A, Acevedo L, Bayod Á. Exergy assessment and
exergy cost analysis of a renewable-based and hybrid trigeneration scheme for
domestic water and energy supply. Energy 2019;168:662–83. https://doi.org/10.
1016/J.ENERGY.2018.11.124.
[15] Jana K, Ray A, Majoumerd MM, Assadi M, De S. Polygeneration as a future sus-
tainable energy solution – A comprehensive review. Appl Energy 2017;202:88–111.
https://doi.org/10.1016/J.APENERGY.2017.05.129.
[16] Sahoo U, Kumar R, Pant PC, Chaudhary R. Development of an innovative poly-
generation process in hybrid solar-biomass system for combined power, cooling and
desalination. Appl Therm Eng 2017;120:560–7. https://doi.org/10.1016/J.
APPLTHERMALENG.2017.04.034.
[17] Calise F, Macaluso A, Piacentino A, Vanoli L. A novel hybrid polygeneration system
supplying energy and desalinated water by renewable sources in Pantelleria Island.
Energy 2017;137:1086–106. https://doi.org/10.1016/J.ENERGY.2017.03.165.
[18] Calise F, Cipollina A, Dentice d’Accadia M, Piacentino A. A novel renewable poly-
generation system for a small Mediterranean volcanic island for the combined
production of energy and water: Dynamic simulation and economic assessment.
Appl Energy 2014;135:675–93. https://doi.org/10.1016/J.APENERGY.2014.03.
064.
[19] Calise F, Dentice d’Accadia M, Piacentino A. Exergetic and exergoeconomic analysis
of a renewable polygeneration system and viability study for small isolated com-
munities. Energy 2015;92:290–307. https://doi.org/10.1016/J.ENERGY.2015.03.
056.
[20] Sepehri A, Sarrafzadeh MH. Effect of nitrifiers community on fouling mitigation and
nitrification efficiency in a membrane bioreactor. Chem Eng Process - Process
Intensif 2018;128:10–8. https://doi.org/10.1016/j.cep.2018.04.006.
[21] Garg MC, Joshi H. A Review on PV-RO Process: Solution to Drinking Water Scarcity
due to High Salinity in Non-Electrified Rural Areas. Sep Sci Technol
2015;50:1270–83. https://doi.org/10.1080/01496395.2014.951725.
[22] García-Rodríguez L. Desalination by Wind Power. Wind Eng 2005;28:453–63.
https://doi.org/10.1260/0309524042886405.
[23] Caldera U, Bogdanov D, Breyer C. Local cost of seawater RO desalination based on
solar PV and wind energy: A global estimate. Desalination 2016;385:207–16.
https://doi.org/10.1016/J.DESAL.2016.02.004.
[24] Shawky HA, Abdel Fatah AA, Abo ElFadl MMS, El-Aassar AHM. Design of a small
mobile PV-driven RO water desalination plant to be deployed at the northwest coast
of Egypt. Desalin Water Treat 2015;55:3755–66. https://doi.org/10.1080/
14
Energy Conversion and Management 216 (2020) 112940
19443994.2015.1080447.
[25] Thomson M, Infield D. A photovoltaic-powered seawater reverse-osmosis system
without batteries. Desalination 2003;153:1–8. https://doi.org/10.1016/S0011-
9164(03)80004-8.
[26] Manolakos D, Mohamed ES, Karagiannis I, Papadakis G. Technical and economic
comparison between PV-RO system and RO-Solar Rankine system. Case study:
Thirasia island. Desalination 2008;221:37–46. https://doi.org/10.1016/J.DESAL.
2007.01.066.
[27] Carta JA, González J, Subiela V. The SDAWES project: An ambitious R and D
prototype for wind-powered desalination. Desalination 2004;161:33–48. https://
doi.org/10.1016/S0011-9164(04)90038-0.
[28] Delgado-Torres AM, García-Rodríguez L, del Moral MJ. Preliminary assessment of
innovative seawater reverse osmosis (SWRO) desalination powered by a hybrid
solar photovoltaic (PV) - Tidal range energy system. Desalination
2020;477:114247https://doi.org/10.1016/j.desal.2019.114247.
[29] Atallah MO, Farahat MA, Lotfy ME, Senjyu T. Operation of conventional and un-
conventional energy sources to drive a reverse osmosis desalination plant in Sinai
Peninsula. Egypt. Renew Energy 2020;145:141–52. https://doi.org/10.1016/j.
renene.2019.05.138.
[30] Rafiei A, Loni R, Mahadzir SB, Najafi G, Pavlovic S, Solar Bellos E. Desalination
System with a Focal Point Concentrator using Different Nanofluids. Appl Therm Eng
2020::115058https://doi.org/10.1016/j.applthermaleng.2020.115058.
[31] Campione A, Gurreri L, Ciofalo M, Micale G, Tamburini A, Cipollina A.
Electrodialysis for water desalination: A critical assessment of recent developments
on process fundamentals, models and applications. Desalination 2018;434:121–60.
https://doi.org/10.1016/j.desal.2017.12.044.
[32] Al-Amshawee S, Yunus MYBM, Azoddein AAM, Hassell DG, Dakhil IH, Hasan HA.
Electrodialysis desalination for water and wastewater: A review. Chem Eng J
2020;380:122231https://doi.org/10.1016/j.cej.2019.122231.
[33] Lundstrom JE. Water desalting by solar powered electrodialysis. Desalination
1979;31:469–88. https://doi.org/10.1016/S0011-9164(00)88551-3.
[34] Ishimaru N. Solar photovoltaic desalination of brackish water in remote areas by
electrodialysis. Desalination 1994;98:485–93. https://doi.org/10.1016/0011-
9164(94)00175-8.
[35] Wright NC, Winter AG. Justification for community-scale photovoltaic-powered
electrodialysis desalination systems for inland rural villages in India. Desalination
2014;352:82–91. https://doi.org/10.1016/j.desal.2014.07.035.
[36] He W, Amrose S, Wright NC, Buonassisi T, Peters IM, Winter AG. Field demon-
stration of a cost-optimized solar powered electrodialysis reversal desalination
system in rural India. Desalination 2020;476:114217https://doi.org/10.1016/j.
desal.2019.114217.
[37] Galama AH, Saakes M, Bruning H, Rijnaarts HHM, Post JW. Seawater pre-
desalination with electrodialysis. Desalination 2014;342:61–9. https://doi.org/10.
1016/j.desal.2013.07.012.
[38] Ion-exchange membrane separation processes 2004;vol:9.
[39] Campione A, Cipollina A, Toet E, Gurreri L, Bogle IDL, Micale G. Water desalination
by capacitive electrodialysis: Experiments and modelling. Desalination
2020;473:114–50. https://doi.org/10.1016/j.desal.2019.114150.
[40] Malek P, Ortiz JM, Schulte-Herbrüggen HMA. Decentralized desalination of
brackish water using an electrodialysis system directly powered by wind energy.
Desalination 2016;377:54–64. https://doi.org/10.1016/j.desal.2015.08.023.
[41] Ortiz JM, Expósito E, Gallud F, García-García V, Montiel V, Aldaz A. Photovoltaic
electrodialysis system for brackish water desalination: Modeling of global process. J
Memb Sci 2006;274:138–49. https://doi.org/10.1016/j.memsci.2005.08.006.
[42] Ortiz JM, Expósito E, Gallud F, García-García V, Montiel V, Aldaz A. Electrodialysis
of brackish water powered by photovoltaic energy without batteries: direct con-
nection behaviour. Desalination 2007;208:89–100. https://doi.org/10.1016/J.
DESAL.2006.05.026.
[43] Ortiz JM, Expósito E, Gallud F, García-García V, Montiel V, Aldaz VA. Desalination
of underground brackish waters using an electrodialysis system powered directly by
photovoltaic energy. Sol Energy Mater Sol Cells 2008;92:1677–88. https://doi.org/
10.1016/j.solmat.2008.07.020.
[44] Troncoso P, Mantz R, Battaiotto P. Electrodialysis Processes Assisted by
Photovoltaic Panels. Concentration Control. IEEE Lat Am Trans 2014;12:864–70.
https://doi.org/10.1109/TLA.2014.6872897.
[45] Pan Y, Zhu T, He Z. Minimizing effects of chloride and calcium towards enhanced
nutrient recovery from sidestream centrate in a decoupled electrodialysis driven by
solar energy. J Clean Prod 2020::121419https://doi.org/10.1016/j.jclepro.2020.
121419.
[46] Xu H, Ji X, Wang L, Huang J, Han J, Wang Y. Performance study on a small-scale
photovoltaic electrodialysis system for desalination. Renew Energy
2020;154:1008–13. https://doi.org/10.1016/j.renene.2020.03.066.
[47] Karimi L, Abkar L, Aghajani M, Ghassemi A. Technical feasibility comparison of off-
grid PV-EDR and PV-RO desalination systems via their energy consumption. Sep
Purif Technol 2015;151:82–94. https://doi.org/10.1016/j.seppur.2015.07.023.
[48] Fernandez-Gonzalez C, Dominguez-Ramos A, Ibañez R, Irabien A. Sustainability
assessment of electrodialysis powered by photovoltaic solar energy for freshwater
production. Renew Sustain Energy Rev 2015;47:604–15. https://doi.org/10.1016/
j.rser.2015.03.018.
[49] Campione A, Cipollina A, Bogle IDL, Gurreri L, Tamburini A, Tedesco M, et al. A
hierarchical model for novel schemes of electrodialysis desalination. Desalination
2019;465:79–93. https://doi.org/10.1016/j.desal.2019.04.020.
[50] Sa Klein. Beckman WA, Mitchell JW, Duffie JA, Duffie NA, Freeman TL. Solar en-
ergy laboratory: TRNSYS. A Transient Syst Simul Program Univ Wisconsin,
Madison; 2006.
[51] Gurreri L, Tamburini A, Cipollina A, Micale G, Ciofalo M. CFD prediction of
A. Campione, et al.
concentration polarization phenomena in spacer-filled channels for reverse elec-
trodialysis. J Memb Sci 2014;468:133–48. https://doi.org/10.1016/j.memsci.2014.
05.058.
[52] Gurreri L, Tamburini A, Cipollina A, Micale G, Ciofalo M. Flow and mass transfer in
spacer-filled channels for reverse electrodialysis: a CFD parametrical study. J Memb
Sci 2016;497:300–17. https://doi.org/10.1016/j.memsci.2015.09.006.
[53] Gurreri L, Ciofalo M, Cipollina A, Tamburini A, Van Baak W, Micale G. CFD mod-
elling of profiled-membrane channels for reverse electrodialysis. Desalin Water
Treat 2015;55:1–20. https://doi.org/10.1080/19443994.2014.940651.
15
Energy Conversion and Management 216 (2020) 112940
[54] La Cerva M, Gurreri L, Tedesco M, Cipollina A, Ciofalo M, Tamburini A, et al.
Determination of limiting current density and current efficiency in electrodialysis
units. Desalination 2018;445:138–48. https://doi.org/10.1016/J.DESAL.2018.07.
028.
[55] Seborg DE, Edgar TF, Mellichamp DA, Doyle FJ. Process dynamics and control.
Wiley; 2016.
[56] Koeppl GW. Putnam’s power from the wind. New York, NY: Van Nostrand Reinhold
Company; 1982.