Professional Documents
Culture Documents
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 systems [6]. Large-scale desalination systems are very attractive for
their capability to achieve a better economic profitability and higher
Desalination units can generally be part of polygeneration facilities efficiencies. On the other hand, large-scale projects also involve several
(i.e. integrated systems with multiple energy outputs such as such as risks which must be accurately taken into account for the evaluation of
heating, cooling and electricity and by-products such as desalinated the feasibility of the project [7]. Thermally-driven desalination systems
water, hydrogen, glycerine, etc [1]). Waste heat from thermal power are also often used in combination with different renewable energy
plants has been used to successfully run thermal desalination plants [2] sources [8], such as geothermal [9] and solar [10]. Usually, poly-
or to preheat salt water feeds [3]. Desalination systems are particularly generation systems based on thermally-driven desalination technolo-
attractive for integration in polygeneration systems [4], especially gies are very complex and they must include very complex control and
when coupled with renewable energy sources [5]. A number of dif- management technologies in order to optimize the operation of the
ferent desalination technologies are available for the integration in system itself. Therefore, such systems are often featured by extremely
renewable polygeneration systems. Desalination units may be driven large capacities and they are very attractive for the production of en-
both by thermal and electrical energy. Thermally-driven desalination ergy (electrical, thermal and cooling) and fresh water at district level
technologies are especially suitable for large-scale district level poly- [11]. In this context, polygeneration systems based on electric renew-
generation systems whereas electrically-driven desalination units are able sources such as photovoltaic (PV) [12] or wind turbines represent a
much more flexible and they can be used both for small and large scale 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
renewable thermal energy, polygeneration systems including electrical mainly available during the summer and in the central hours of the day.
renewables and desalination exhibit: i) higher flexibility in terms of Conversely, according to the weather data, wind velocity increases in
capacities; ii) simpler system layout; iii)easier management [13]. In winter and during the night. Therefore, combining photovoltaic solar
these cases, desalination technologies can be particularly attractive as energy and wind energy the stability of power production profile can be
means of using the excess energy that renewable sources produce increased. It is worth noting that, besides solar and wind energy sys-
during peak periods, producing drinking water instead of using energy tems, there are some other innovative processes that can be used in
storage devices such as batteries [14]. polygeneration systems. Among them, membrane bioreactors are of
One of the main issues related to the use of renewable poly- particular interest [20].
generation systems consists in the unpredictability of the energy source. The possibility to couple desalination systems with renewable en-
The unavoidable fluctuations of many renewable energy sources (wind, ergy sources has been already demonstrated in the literature. Given the
solar, etc.) make extremely difficult to suitably couple these systems wider commercial diffusion of reverse osmosis (RO) among desalination
with the cooling, heating and electrical demands of the users [1]. In this technologies, a lot of works focus on this process. In particular, process
framework, many researchers investigated several storage technologies feasibility has been studied and proven for PV-RO [21], wind-RO [22]
(both thermal and electrical) in order to achieve a more stable power and combined PV and wind-RO systems [23]. A number of these works
supply profile [15]. However, such storage technologies are still too highlight how the economic feasibility of such processes can be en-
expensive for a good economic profitability. Therefore, some re- hanced using battery-less power sources, as batteries are associated to a
searchers are focusing on innovative solutions that couple renewables number of disadvantages such as increased capital costs, limited life-
such as solar systems [4], hybrid solar-biomass systems [16] and solar- time and increased maintenance [24]. Nevertheless, when desalination
geothermal systems [9] with desalination units. These systems are units are coupled with battery-less systems they are directly subjected
particularly interesting for small islands [17] (for which economic as- to the fluctuation of power generation, requiring real-time adjusting.
sessment were even conducted [18]) and isolated communities [19]. Thomson et al. [25] demonstrated the possibility to apply such systems
Many authors are focusing on specific combinations of renewable en- to RO from both modelling and experimental points of view. However,
ergy sources in order to mitigate the fluctuations typical of using in- the authors (as well as Manolakos et al. in another work [26]) high-
dividual renewable energy sources. For example, the combination of lighted that long term reliability of the system is not guaranteed due to
solar and wind energy is extremely promising. In fact, solar energy is the continuous high pressure fluctuation experienced by membranes.
2
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
An alternative to compensate for the power fluctuation, by modifying power produced, requiring the use of some electrical energy storage
the production capacity without significantly affect the pressure, is the device which may significantly affect the overall energy and economic
use a number of parallel RO plants that are switched on or off based on performance of the system. These issues may be solved using alternative
the amount of energy available [27]. Nevertheless, this generally im- desalination units, more flexible in terms of fluctuations of power. In
plies a system oversizing and an increase in capital costs. Besides the this framework, ElectroDialysis units (Fig. 1) are extremely attractive,
aforementioned issues, RO systems also suffer from relatively slow and due to their capability to operate under unsteady input energy. The
critical start-ups and shutdowns where a gradual increase/decrease in large interest in the ED process is well demonstrated by the amount of
system pressure is required before reaching the steady-state [27]. Del- studies that has been performed in the last years. A comprehensive
gado Torres et al. [28] also investigated the integration of reverse os- review regarding the utilization of ED technology for water desalination
mosis with PV and tidal range systems. They performed an energy (i.e. the most important application of ED) was recently published [31].
analysis aiming at evaluating the phase shift between tidal and solar Here, the authors present a detailed review of the technology, model-
energy for a plant at Broome, Australia. They concluded that the use of ling approaches and ED-related processes. They also introduce the most
the tidal range power plant allows one to significantly increase the innovative configurations under development. Another recent review
operating time of the desalination plant. The combination of PV and [32] focuses on both water and wastewater treatment plants. In this
wind technologies (also including electrical batteries) driving a RO unit paper, an overview of the possible sustainability approaches to be used
was recently analysed by Atallah et al. [29] for a case study in Egypt. in the design of ED systems was also provided. The reader may refer to
Their work is based on a simulation developed in Homer. For the these review papers for further details regarding ED technology. Ex-
analysed case study, authors concluded that the optimal configuration perimental investigation on ED systems working with renewable energy
consists of 160 kW photo-voltaic (PV) panels, 19 strings of lead-acid sources is ongoing since few decades ago [33], including the assessment
batteries and a 50 kW diesel generator system. A similar configuration of remote areas scenarios [34]. In particular, ED has proven its ad-
was also analysed by Mito et al. , investigating the hybrid PV-Wind-Ro vantages to serve off-grid locations with moderately concentrated
configuration for large-scale plants. They also point out that the stabi- brackish waters [35].For this reason, rural areas have been recentrly
lity of renewable power production is a critical issue for a safe and chosen to conduct these kind of field tests [36].
reliable operation of RO units. Therefore, short-term energy storage ED is an electromembrane process that makes use of ion exchange
devices are always mandatory required in order to mitigate such fluc- membranes and can work in two main operational modes, namely batch
tuations. An other possibility may consist in the integration of PV and [37] and continuous [38]. In some cases, especially for seawater de-
with a diesel generator which is used to stabilize the system power salination, a multistage configuration can be adopted [39].
production. They also reviewed several literature studies investigating As mentioned before, ED can overcome most of the drawbacks that
the possibility of direct coupling between PV and/or wind and RO units. result from dynamic RO operations. A change in the available power
They concluded that this configuration is extremely difficult and it re- input can be easily adjusted by changing the applied voltage and by
quires complex control strategies to be managed. A recent study [30] changing the feed flowrate without the need for particular adjustments,
also combined PV system with Humidification Dehumidification Desa- as low pressures (usually < 1 barg) are involved in the process.
lination (HDD) systems. Authors presented a novel hybrid solar desa- Another advantage is that transient phases during start-ups and shut-
lination system, implementing a numerical model. In particular, they downs are much shorter. Conversely, the main issue that has been
focused on nanofluids as solar operating fluid, investigating different highlighted refers to an increased energy consumption due to the pre-
oil-based nanofluids. They found that Cu/oil nanofluid exhibits the sence of harmonic disturbances that can be mitigated through filters
lowest efficiency and they can be suggested only to achieve high [27]. For these reasons, ED is particularly suitable for battery-less sys-
freshwater amounts, at low gain output ratios. tems. In this context, Malek et al. [40] showed a successful experi-
Thus, the above presented literature review clearly shows that hy- mental coupling of batch ED with direct wind energy, demonstrating
brid PV-Wind-RO systems dramatically suffer for the fluctuations of the how the system is insignificantly affected by moderate wind
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
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
fluctuations. Similarly, Ortiz et al. [41] proved the operation of a batch The overall simulation model was implemented by a hybrid ap-
PV-ED system through experiments [42] and simulations [43] in the proach combining a well-known dynamic simulation tool, TRNSYS [50]
scale of hours. Moreover, the dynamics of a PV-ED in continuous mode , and a user-developed model. In particular, the model for the solar/
was studied and simulated by Troncoso et al. [44]. However, in that wind system is developed in TRNSYS by using weather data from
work the presence of a battery was taken into account, so that the Pantelleria, and conventional components included in TRNSYS library
criticalities of battery-less systems were not highlighted. An on-field (PV panels, wind turbine, inverter, controllers, etc). The overall elec-
demonstration was recently presented by He et al.[36]. Authors pre- trical production calculated by TRNSYS is subsequently provided as an
sented an experimental validation of a model of a system including a PV input data to the user-developed model, simulating the ED subsystem
field and ED reversal desalination system. The study presents the fab- implemented into gPROMS Modelbuilder. In the followings, the main
rication process of the pilot plant tested in India. They proved the ef- algorithms of the models are briefly presented. For sake of brevity, PV
ficiency and the flexibility of their system. In addition, the implemented and wind models are summarized in Appendix, whereas this section
methodology allowed to reduce lifetime and capital cost by 35% and briefly reports ED model.
34%, respectively. Solar powered ED has been also used to enhance
nutrient recovery from waste streams [45], showing excellent results.
2.1. Electrodialysis process
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-
The studies presented in this work are based on a steady-state
oped a suitable simulation code in MATLAB environment in order to
hierarchical ED model (scheme is reported in Fig. 2) [49] that has been
investigate the influence of voltage, flow rate and ED size. As expected,
modified to account for the dynamic behaviour of the system.
they found that the salinity of the outlet stream decreases in case of
The lowest hierarchy of the model (Fig. 2 a), i.e. the cell pair, is the
higher voltages or larger membrane area.
repeating unit of the system, including an anion-exchange membrane
Besides the technical aspects, ED proved its convenience in different
(AEM), a diluate channel, a cation-exchange membrane (CEM) and a
aspects. From cost estimations, PV-ED resulted more convenient than
concentrate channel. At this scale, material balances and transport
PV-RO in the presence of low salinity feeds [47]. A sustainability as-
phenomena are described. In particular, the dynamic, 1D mass balances
sessment was also conducted on a PV-ED system, predicting a potential
in a generic (concentrate or diluate) channel can be written as:
decrease of one order of magnitude in CO2 emissions compared to grid
mix energy supply [48]. b
CSOL (x , t )
+
QSOL (x , t ) CSOL (x , t )
= ± bJtot (x , t )
SOL
t x (1)
1.1. Aim of the work
dQSOL (x , t )
= ± bqw (x , t )
dx (2)
The literature review presented in this section clearly shows that the
4
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
where A is the membrane active area, SOL is the channel thickess, set according to the maximum allowable pressure drop that has been set
CSOL CSOL (x, t) is the salt concentration in solution, b is the channel at 1.2 bar, corresponding to 920 m3/d (230 m3/d per stack) and 43 kW
width, Jtot Jtot (x, t) is the total salt flux through the membranes (i.e. the of absorbed power. Conversely, the minimum allowable flowrate was
sum of conductive and diffusive salt flux [49]), QSOL QSOL (x, t) re- set at ~ 230 m3/d (58 m3/d per stack) and 5 kW of absorbed power,
presents the local volumetric flow rate of a single cell pair, and qw qw (x) where the current required to reach 0.25 g/l is already 90% of the es-
the local overall volumetric water flux through the membranes (i.e. the timated limiting current [54]. Further reduction in the diluate feed
sum of osmotic and electroosmotic fluxes [49]). t and x indicates that flowrate would cause limiting current issues and thus an impossibility
the variables are function of time and space (i.e. the axial direction of to reach the target concentration.
the solution flow through the channels). According to the aforementioned limits, the ED system was simu-
Electrical variables of the cell pair are also computed. In particular, lated assuming that the applied voltage changes as a function of the
the cell pair voltage drop (Vcp Vcp (t) ) can be written as: power produced by the energy system over the entire year. The
boundaries were taken into account, so that if the available power is
Vcp (t ) = (x , t ) + Rtot (x , t ) i (x , t ) (3)
higher than the upper limit, the plant does not use all of it. Conversely,
where is the non-Ohmic voltage drop associated to the back when the power is lower than the minimum the system is switched off.
electromotive force where concentration polarisation is also taken into Given the long time scale, a quasi-steady state approach was adopted
account making use of computationally determined Sherwood numbers for these simulations.
[49,51–53], i i(x, t) is the local current density and Rtot is the total areal Fig. 5 a, shows how much the overall power can be different during
Ohmic resistance of cell pair which can be calculated as the sum of the year (according to weather data from Pantelleria). In particular,
membrane and solution compartment resistances [49]. peaks are clearly more frequent in summer months, when the PV system
A number of cell pairs are grouped together in the Stack higher has more solar radiation available. Produced power goes slightly down
hierarchy (Fig. 2 b), where the external applied voltage (Vtot Vtot (t) ) is in the other periods of the year, even though the wind turbine partially
related to the internal electric variables: compensates the reduction of solar radiation.
Fig. 5 b shows the diluate feed flowrate of the ED plant (QDNs ). As
Ncp
Rblank I expected, the plant can adapt the feed flowrate to the produced power,
Vtot (t ) = + Vcpi (t )
A i=1 (4) and thus generating different amounts of drinking water. Interestingly,
the set maximum and minimum flowrate boundaries do not excessively
where, Rblank is the blank resistance, accounting for electric voltage affect the power utilisation on the yearly time scale. In fact, the plant
drops in the electrode compartments, I is the total current, calculated as never goes off for long periods except for few phases during winter
the integral of the current density over the active area, and Ncp is the months, while only very few power peaks are not entirely used during
number of cell pairs inside the stack. the summer. Fig. 5 also shows cumulative curves (black continuous
The stack model also computes performance parameters such as lines) that gives an estimation of the total amount of hours in which
current efficiency and specific energy consumption [49]. power or flowrate were maintained above a certain value. From these
Finally, the stack can be implemented in the highest hierarchy of the curves it can be seen how the plant stays on for more than 5500 (non-
plant (Fig. 2 c). In the case of this work, the stack model was coupled continuous) hours during the entire year. In addition, it is interesting to
with a variable power source and a control unit (Fig. 3). The details of note that the specific energy consumption decreases while reducing the
the controller equation, as well as its design and tuning are discussed in feed flowrate. This is due to the operating conditions that are closer to
the control design section. reversibility when the applied voltage is small.
Such findings indicate a large potential for application, especially in
3. Long time scale simulations remote or isolated sites, where small electrical grids require effective
buffering strategies when the utilisaiton of renewable energies
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 Fig. 3. Detailed block diagram of the highest hierarchy of the plant model re-
power, in order to keep a constant outlet concentration. Based on fea- presenting an electrodialysis unit powered by the Photovoltaic/wind power
sibility conditions, upper and lower limits were set in each stack for the delivery system. The control system of the product concentration, acting on the
diluate flowrate. In particular, the maximum diluate feed flowrate was feed flowrate is also shown.
5
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)
increases.
6
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
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, method, it was possible to estimate a first set of control parameters for
the concentration will reach a new steady state with a higher or lower each reference scenario from the fitted process transfer functions pre-
concentration respectively, requiring a certain time that depends on the viously identified (Fig. 6a). These parameters were then averaged and
flowrate itself. When the flowrate is suddenly changed, the spatial finely tuned with a trial and error procedure by simulating the response
concentration profile in each channel will have to adapt to the new of the controlled system to step voltage changes on gPROMS Mod-
situation (see Fig. 6b), requiring a mutual interaction between each elbuilder. The final values are reported in Table 2.
adjacent discretisation interval, through which a certain volume of Fig. 7 shows simulation results for the controlled ED stack for each
solution is flowing. Therefore, the resulting dynamic is given by a reference flowrate. At the maximum flowrate, the system is subjected to
number of “virtual” processes that occur in sequence and results in a a 20% step decrease of the applied voltage (Fig. 7 a), the opposite trend
high-order behaviour. However, for the sake of simplicity, each of the is detected for the minimum flowrate (Fig. 7 d), while at 144 m3/d both
curves of Fig. 6a was fitted with a second-order transfer function with a an increase and a decrease of the applied voltage were analysed (Fig. 7
zero that, in the Laplace domain is expressed as: 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
K 0s + K
G (s ) = the outlet concentration around the stationary value become more
s+2 s+1 (5)
persistent and take more time to settle, even though they are still at an
where K is the process gain, 0 is the time constant of the transfer acceptable value and any deviation from the set-point disappear after a
function numerator, is the process time constant, is the damping time of 300–400 s. This can be explained by the transient behaviour of
coefficient and s is the independent variable in the Laplace domain. the system that significantly changes from high to low flowrates, ac-
Each fitting curve from Fig. 6a was characterised by its own parameters cording to the dynamic responses already shown in Fig. 6 a. Besides, a
according to eq. . 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
4.2. Feedback control design the derivative term was included in the controller in order to dampen
the oscillations of the outlet concentration generated by the propor-
The aim of the Feed-Back (FB) control is to keep the target con- tional-integral action of the controller. In particular, excessive decrease
centration by adjusting the feed flowrate, according to the following in the concentration may cause limiting current issues.
equation:
7
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
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 to observe discrete power changes that significantly challenge the sta-
controller. bility of the control system.
The same scenarios adopted for the analysis of the FB controller Simulation results for the daily operations are shown in Fig. 10. For
performance were chosen in this case and results are reported in Fig. 8 each day, the available power, the voltage applied to each ED unit, the
for the stand-alone FF controller and in Fig. 9 for the hybrid FB-FF one. plant diluate flowrate and the outlet diluate concentration are reported.
The implementation of the stand-alone FF controller helps in During the sample winter day (Fig. 10 a), power production is very low
minimizing the off-set restoring a concentration value close to the in the very first hours (i.e. before the sunset and with almost no wind).
target. However, the control is not ideal and some transient deviation Then, some small power peaks are observed during light hours while
from the steady-state value of product concentration is observed. the power grows up to 20 kW (i.e. the maximum capacity of the wind
Looking at the mostly common case of hybrid controller, the re- turbine) during the night due to an increased wind speed. This means
sponse of the system shows several small improvements compared to that for the first hours of the days the unit is not able to operate as the
the simple case of FB controller, both in terms of system stability and power is below the power limit, while it keeps running for the rest of
amplitude of oscillations. the day, except for an additional short period, reaching its maximum
However, such improvements may have a limited interest in the real daily production in the last quarter of the day.
cases analysed hereafter due to the smaller and slower disturbances On the other hand, spring and autumn days (Fig. 10 b and d) present
occurring in real operation of solar/wind-powered ED systems. a more unstable behaviour, characterised by a similar non-operational
Therefore, the simple FB controller has been adopted for the simula- period of about 6 h in between the first 2 quarters of the day and a
tions of short-time-scale scenarios presented in Section 5. 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.
5. Short time scale simulations 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
The dynamic ED model was used to simulate the operation of the hybrid system is always above the minimum threshold. Despite this, the
controlled ED plant in 4 typical days (from 00:00 to 23:59), that were day is characterised by a high number of steep variations, especially
chosen as representative of the four seasons. At this scale, it is possible during light hours when solar irradiation keeps changing according to
to observe the dynamic effect of the control system. For this reason, the weather conditions. For this reason, the highest concentration peaks are
voltage applied to the ED units was changed with a 3 min step, in order
8
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
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. power only affects the time required to reach the desired concentration
Notably, in all cases the quality of the desalinated water product is and not the quantity or the quality of the final product.
maintained very close to the target value of 0.25 g/lt, while the max- The behaviour of the outlet concentration can also be compared
imum product flow rate ranges from about 400 to 800 m3/day. Such with the one found by Thomson et al. [25] in a battery-less PV-reverse
amount offresh water is suitable for a population of several thousands osmosis (RO) system. In that work, the system was able to avoid large
persons. changes in the product flowrate at the cost of larger variations in the
Depending on the site location and energy availability, the size of concentration (0.2–0.6 g/l). In addition, the control of the system re-
the plant can be easily increased up to 10 times thanks to the tech- sulted much more complex in the case of RO due to the fact that it was
nology flexibility and plant modularity, thus reaching a relevant size necessary to control the rotation of two different pumps in order to
also for large communities and small cities. control water recovery ration and maximize the system efficiency.
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 6. Conclusions
are highlighted. Relatively small peaks and fast responses are found for
step disturbances (as discussed in the controller tuning section). Even in This work presents an analysis of the behaviour of the single pass ED
the summer day, characterised by the highest peaks, the outlet con- process when powered by a hybrid PV/Wind energy source via dynamic
centration stays below ± 10% of the set point value. In the same way, simulations. Simulations were performed through an ED process model
the oscillatory responses to ramp changes shows very limited oscilla- in both a yearly and daily time scale.
tions around the set point. In the yearly time scale, a quasi-steady state approach was adopted.
It might seem that the presence of the aforementioned concentra- 4 ED units working in parallel were simulated, in order to demonstrate
tion peaks is an inconsistent result compared to what was found by how the process can work over the year within a wide range of flow-
Ortiz et al. [42], where a smooth concentration curve vs time is found. rates following the PV or wind power availability. In particular, in order
However, while the present work refers to a continuous ED system, that to achieve the target product water quality (i.e. a concentration of total
is supposed to always produce drinking water at the target concentra- dissolve solids of 0.25 g/l) the total diluate flowrate was ranged from
tion, the work from Ortiz et al. takes into account a batch operation, 230 to 920 m3/d, corresponding to an absorbed power from 5 kW to
where the goal is to reach the target concentration only at the end of the 43 kW respectively (with this latter being 25% above the nominal op-
operation. Therefore, during a batch operation the change in available 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
9
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
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 highly non-constant power sources. The operational ranges, in terms of
excessively low power, reaching a cumulative operational period of product water flow rates, can go from few tens of m3/d for very small
more than 5500 h. applications, up to several thousands of m3/d for large industrial plant,
On the other hand, daily simulations aimed to study detailed pro- thus potentially covering the needs for fresh water supply in remote
cess dynamics in the same configuration studied in the previous case villages up to small cities. Within such scenarios, the process could be
(i.e. 4 ED units in parallel). A control system based on Feedback and successfully implemented in battery-less polygeneration systems for
Feedforward controllers has been designed in order to maintain the fresh water production, also acting as an energy buffer device.
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 CRediT authorship contribution statement
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 Antonino Campione: Conceptualization, Methodology, Software.
(i.e. the variable that is manipulated in order to keep a constant outlet Andrea Cipollina: Conceptualization, Supervision, Funding acquisi-
concentration). Simulating the power variation over 4 different re- tion. Francesco Calise: Conceptualization, Methodology, Software.
ference days (one for each season), it was shown that the controlled Alessandro Tamburini: Methodology, Supervision. Mosè Galluzzo:
system works well throughout each day, always ensuring a stable op- Conceptualization, Methodology, Software. Giorgio Micale:
eration and relatively short settling times of the outlet concentration, Supervision, Funding acquisition.
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 Declaration of Competing Interest
traditional PID feedback controller resulted to be small and not so re-
levant for the operation in real operating scenarios. The authors declare that they have no known competing financial
Results show that, given the high flexibility and the fast and con- interests or personal relationships that could have appeared to influ-
trollable process dynamics, ED proved its suitability for coupling with ence the work reported in this paper.
10
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
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)
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
13
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
References 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-
[1] Calise F. de Notaristefani di Vastogirardi G, Dentice d’Accadia M, Vicidomini M. 9164(03)80004-8.
Simulation of polygeneration systems. Energy 2018;163:290–337. https://doi.org/ [26] Manolakos D, Mohamed ES, Karagiannis I, Papadakis G. Technical and economic
10.1016/J.ENERGY.2018.08.052. comparison between PV-RO system and RO-Solar Rankine system. Case study:
[2] Cardona E, Culotta S, Mangiaracina L, Terranova R. Multi-flash desalination and Thirasia island. Desalination 2008;221:37–46. https://doi.org/10.1016/J.DESAL.
cogeneration to solve the water supply problems in southeastern Sicily. Desalination 2007.01.066.
1996;106:121–30. https://doi.org/10.1016/S0011-9164(96)00100-2. [27] Carta JA, González J, Subiela V. The SDAWES project: An ambitious R and D
[3] Kamal I. Cogeneration desalination with reverse osmosis: A means of augmenting prototype for wind-powered desalination. Desalination 2004;161:33–48. https://
water supplies in Southern California. Desalination 1992;88:355–69. https://doi. doi.org/10.1016/S0011-9164(04)90038-0.
org/10.1016/0011-9164(92)80127-U. [28] Delgado-Torres AM, García-Rodríguez L, del Moral MJ. Preliminary assessment of
[4] Calise F, Dentice d’Accadia M, Vanoli R, Vicidomini M. Transient analysis of solar innovative seawater reverse osmosis (SWRO) desalination powered by a hybrid
polygeneration systems including seawater desalination: A comparison between solar photovoltaic (PV) - Tidal range energy system. Desalination
linear Fresnel and evacuated solar collectors. Energy 2019;172:647–60. https://doi. 2020;477:114247https://doi.org/10.1016/j.desal.2019.114247.
org/10.1016/J.ENERGY.2019.02.001. [29] Atallah MO, Farahat MA, Lotfy ME, Senjyu T. Operation of conventional and un-
[5] Mollahosseini A, Abdelrasoul A, Sheibany S, Amini M, Salestan SK. Renewable conventional energy sources to drive a reverse osmosis desalination plant in Sinai
energy-driven desalination opportunities – A case study. J Environ Manage Peninsula. Egypt. Renew Energy 2020;145:141–52. https://doi.org/10.1016/j.
2019;239:187–97. https://doi.org/10.1016/j.jenvman.2019.03.044. renene.2019.05.138.
[6] Wang Z, Lin X, Tong N, Li Z, Sun S, Liu C. Optimal planning of a 100% renewable [30] Rafiei A, Loni R, Mahadzir SB, Najafi G, Pavlovic S, Solar Bellos E. Desalination
energy island supply system based on the integration of a concentrating solar power System with a Focal Point Concentrator using Different Nanofluids. Appl Therm Eng
plant and desalination units. Int J Electr Power Energy Syst 2020::115058https://doi.org/10.1016/j.applthermaleng.2020.115058.
2020;117:105707https://doi.org/10.1016/j.ijepes.2019.105707. [31] Campione A, Gurreri L, Ciofalo M, Micale G, Tamburini A, Cipollina A.
[7] Zhang Y, Wang R, Huang P, Wang X, Wang S. Risk evaluation of large-scale sea- Electrodialysis for water desalination: A critical assessment of recent developments
water desalination projects based on an integrated fuzzy comprehensive evaluation on process fundamentals, models and applications. Desalination 2018;434:121–60.
and analytic hierarchy process method. Desalination 2020;478:114286https://doi. https://doi.org/10.1016/j.desal.2017.12.044.
org/10.1016/j.desal.2019.114286. [32] Al-Amshawee S, Yunus MYBM, Azoddein AAM, Hassell DG, Dakhil IH, Hasan HA.
[8] Calise F, d’Accadia MD, Quiriti E, Vicidomini M, Piacentino A. Trigeneration and Electrodialysis desalination for water and wastewater: A review. Chem Eng J
Polygeneration Configurations for Desalination and Other Beneficial Processes. 2020;380:122231https://doi.org/10.1016/j.cej.2019.122231.
Sustain. Desalin. Handb. Plant Sel. Des. Implement., Elsevier 2018:99–199. https:// [33] Lundstrom JE. Water desalting by solar powered electrodialysis. Desalination
doi.org/10.1016/B978-0-12-809240-8.00004-6. 1979;31:469–88. https://doi.org/10.1016/S0011-9164(00)88551-3.
[9] Calise F, Dentice d’Accadia M, Macaluso A, Vanoli L, Piacentino A. A novel solar- [34] Ishimaru N. Solar photovoltaic desalination of brackish water in remote areas by
geothermal trigeneration system integrating water desalination: Design, dynamic electrodialysis. Desalination 1994;98:485–93. https://doi.org/10.1016/0011-
simulation and economic assessment. Energy 2016;115:1533–47. 9164(94)00175-8.
[10] Calise F, Dentice d’Accadia M, Piacentino A. A novel solar trigeneration system [35] Wright NC, Winter AG. Justification for community-scale photovoltaic-powered
integrating PVT (photovoltaic/thermal collectors) and SW (seawater) desalination: electrodialysis desalination systems for inland rural villages in India. Desalination
Dynamic simulation and economic assessment. Energy 2014;67:129–48. https:// 2014;352:82–91. https://doi.org/10.1016/j.desal.2014.07.035.
doi.org/10.1016/j.energy.2013.12.060. [36] He W, Amrose S, Wright NC, Buonassisi T, Peters IM, Winter AG. Field demon-
[11] Colmenar-Santos A, Palomo-Torrejón E, Mur-Pérez F, Rosales-Asensio E. Thermal stration of a cost-optimized solar powered electrodialysis reversal desalination
desalination potential with parabolic trough collectors and geothermal energy in system in rural India. Desalination 2020;476:114217https://doi.org/10.1016/j.
the Spanish southeast. Appl Energy 2020;262:114433https://doi.org/10.1016/j. desal.2019.114217.
apenergy.2019.114433. [37] Galama AH, Saakes M, Bruning H, Rijnaarts HHM, Post JW. Seawater pre-
[12] Calise F, Cappiello FL, Vanoli R, Vicidomini M. Economic assessment of renewable desalination with electrodialysis. Desalination 2014;342:61–9. https://doi.org/10.
energy systems integrating photovoltaic panels, seawater desalination and water 1016/j.desal.2013.07.012.
storage. Appl Energy 2019;253:113575https://doi.org/10.1016/j.apenergy.2019. [38] Ion-exchange membrane separation processes 2004;vol:9.
113575. [39] Campione A, Cipollina A, Toet E, Gurreri L, Bogle IDL, Micale G. Water desalination
[13] Khan EU, Martin AR. Optimization of hybrid renewable energy polygeneration by capacitive electrodialysis: Experiments and modelling. Desalination
system with membrane distillation for rural households in Bangladesh. Energy 2020;473:114–50. https://doi.org/10.1016/j.desal.2019.114150.
2015;93:1116–27. https://doi.org/10.1016/J.ENERGY.2015.09.109. [40] Malek P, Ortiz JM, Schulte-Herbrüggen HMA. Decentralized desalination of
[14] Usón S, Uche J, Martínez A, del Amo A, Acevedo L, Bayod Á. Exergy assessment and brackish water using an electrodialysis system directly powered by wind energy.
exergy cost analysis of a renewable-based and hybrid trigeneration scheme for Desalination 2016;377:54–64. https://doi.org/10.1016/j.desal.2015.08.023.
domestic water and energy supply. Energy 2019;168:662–83. https://doi.org/10. [41] Ortiz JM, Expósito E, Gallud F, García-García V, Montiel V, Aldaz A. Photovoltaic
1016/J.ENERGY.2018.11.124. electrodialysis system for brackish water desalination: Modeling of global process. J
[15] Jana K, Ray A, Majoumerd MM, Assadi M, De S. Polygeneration as a future sus- Memb Sci 2006;274:138–49. https://doi.org/10.1016/j.memsci.2005.08.006.
tainable energy solution – A comprehensive review. Appl Energy 2017;202:88–111. [42] Ortiz JM, Expósito E, Gallud F, García-García V, Montiel V, Aldaz A. Electrodialysis
https://doi.org/10.1016/J.APENERGY.2017.05.129. of brackish water powered by photovoltaic energy without batteries: direct con-
[16] Sahoo U, Kumar R, Pant PC, Chaudhary R. Development of an innovative poly- nection behaviour. Desalination 2007;208:89–100. https://doi.org/10.1016/J.
generation process in hybrid solar-biomass system for combined power, cooling and DESAL.2006.05.026.
desalination. Appl Therm Eng 2017;120:560–7. https://doi.org/10.1016/J. [43] Ortiz JM, Expósito E, Gallud F, García-García V, Montiel V, Aldaz VA. Desalination
APPLTHERMALENG.2017.04.034. of underground brackish waters using an electrodialysis system powered directly by
[17] Calise F, Macaluso A, Piacentino A, Vanoli L. A novel hybrid polygeneration system photovoltaic energy. Sol Energy Mater Sol Cells 2008;92:1677–88. https://doi.org/
supplying energy and desalinated water by renewable sources in Pantelleria Island. 10.1016/j.solmat.2008.07.020.
Energy 2017;137:1086–106. https://doi.org/10.1016/J.ENERGY.2017.03.165. [44] Troncoso P, Mantz R, Battaiotto P. Electrodialysis Processes Assisted by
[18] Calise F, Cipollina A, Dentice d’Accadia M, Piacentino A. A novel renewable poly- Photovoltaic Panels. Concentration Control. IEEE Lat Am Trans 2014;12:864–70.
generation system for a small Mediterranean volcanic island for the combined https://doi.org/10.1109/TLA.2014.6872897.
production of energy and water: Dynamic simulation and economic assessment. [45] Pan Y, Zhu T, He Z. Minimizing effects of chloride and calcium towards enhanced
Appl Energy 2014;135:675–93. https://doi.org/10.1016/J.APENERGY.2014.03. nutrient recovery from sidestream centrate in a decoupled electrodialysis driven by
064. solar energy. J Clean Prod 2020::121419https://doi.org/10.1016/j.jclepro.2020.
[19] Calise F, Dentice d’Accadia M, Piacentino A. Exergetic and exergoeconomic analysis 121419.
of a renewable polygeneration system and viability study for small isolated com- [46] Xu H, Ji X, Wang L, Huang J, Han J, Wang Y. Performance study on a small-scale
munities. Energy 2015;92:290–307. https://doi.org/10.1016/J.ENERGY.2015.03. photovoltaic electrodialysis system for desalination. Renew Energy
056. 2020;154:1008–13. https://doi.org/10.1016/j.renene.2020.03.066.
[20] Sepehri A, Sarrafzadeh MH. Effect of nitrifiers community on fouling mitigation and [47] Karimi L, Abkar L, Aghajani M, Ghassemi A. Technical feasibility comparison of off-
nitrification efficiency in a membrane bioreactor. Chem Eng Process - Process grid PV-EDR and PV-RO desalination systems via their energy consumption. Sep
Intensif 2018;128:10–8. https://doi.org/10.1016/j.cep.2018.04.006. Purif Technol 2015;151:82–94. https://doi.org/10.1016/j.seppur.2015.07.023.
[21] Garg MC, Joshi H. A Review on PV-RO Process: Solution to Drinking Water Scarcity [48] Fernandez-Gonzalez C, Dominguez-Ramos A, Ibañez R, Irabien A. Sustainability
due to High Salinity in Non-Electrified Rural Areas. Sep Sci Technol assessment of electrodialysis powered by photovoltaic solar energy for freshwater
2015;50:1270–83. https://doi.org/10.1080/01496395.2014.951725. production. Renew Sustain Energy Rev 2015;47:604–15. https://doi.org/10.1016/
[22] García-Rodríguez L. Desalination by Wind Power. Wind Eng 2005;28:453–63. j.rser.2015.03.018.
https://doi.org/10.1260/0309524042886405. [49] Campione A, Cipollina A, Bogle IDL, Gurreri L, Tamburini A, Tedesco M, et al. A
[23] Caldera U, Bogdanov D, Breyer C. Local cost of seawater RO desalination based on hierarchical model for novel schemes of electrodialysis desalination. Desalination
solar PV and wind energy: A global estimate. Desalination 2016;385:207–16. 2019;465:79–93. https://doi.org/10.1016/j.desal.2019.04.020.
https://doi.org/10.1016/J.DESAL.2016.02.004. [50] Sa Klein. Beckman WA, Mitchell JW, Duffie JA, Duffie NA, Freeman TL. Solar en-
[24] Shawky HA, Abdel Fatah AA, Abo ElFadl MMS, El-Aassar AHM. Design of a small ergy laboratory: TRNSYS. A Transient Syst Simul Program Univ Wisconsin,
mobile PV-driven RO water desalination plant to be deployed at the northwest coast Madison; 2006.
of Egypt. Desalin Water Treat 2015;55:3755–66. https://doi.org/10.1080/ [51] Gurreri L, Tamburini A, Cipollina A, Micale G, Ciofalo M. CFD prediction of
14
A. Campione, et al. Energy Conversion and Management 216 (2020) 112940
concentration polarization phenomena in spacer-filled channels for reverse elec- [54] La Cerva M, Gurreri L, Tedesco M, Cipollina A, Ciofalo M, Tamburini A, et al.
trodialysis. J Memb Sci 2014;468:133–48. https://doi.org/10.1016/j.memsci.2014. Determination of limiting current density and current efficiency in electrodialysis
05.058. units. Desalination 2018;445:138–48. https://doi.org/10.1016/J.DESAL.2018.07.
[52] Gurreri L, Tamburini A, Cipollina A, Micale G, Ciofalo M. Flow and mass transfer in 028.
spacer-filled channels for reverse electrodialysis: a CFD parametrical study. J Memb [55] Seborg DE, Edgar TF, Mellichamp DA, Doyle FJ. Process dynamics and control.
Sci 2016;497:300–17. https://doi.org/10.1016/j.memsci.2015.09.006. Wiley; 2016.
[53] Gurreri L, Ciofalo M, Cipollina A, Tamburini A, Van Baak W, Micale G. CFD mod- [56] Koeppl GW. Putnam’s power from the wind. New York, NY: Van Nostrand Reinhold
elling of profiled-membrane channels for reverse electrodialysis. Desalin Water Company; 1982.
Treat 2015;55:1–20. https://doi.org/10.1080/19443994.2014.940651.
15