SWRO Desalination unit

The variable load operation of SWRO units results in lower specific energy consumption. However, variable operation of a SWRO unit (variable feed flow rate and pressure) affects the flow rate and the electrical conductivity (salinity) of the produced fresh water. Particularly, when the high-pressure pump of a SWRO unit is supplied by lower power than the nominal one, then the membrane inlet pressure is decreased and as a result the product flow rate is decreased while the salinity of the desalinated water is increased [15,16]. In order to determine the specific energy consumption and the product conductivity as functions of membrane inlet pressure, a SWRO desalination unit was initially studied individually identifying its performance at variable load operation.

The Sea water Reverse Osmosis desalination system (Figure 1) includes a high-pressure pump equipped with two types of Danfoss pumps, coupled with a DC electrical motor and a DC driver for the control speed of the motor responsible for the control speed of the rotational speed of the motor pump assembly. Additionally, the SWRO system is equipped with a feed water pump, a pretreatment system, four membrane modules, a mixing tank and measurement instruments (flow meters, pressure, and electrical conductivity transmitters). The SWRO desalination system has been discussed in detail in [16]. The main technical specifications of the SWRO desalination unit, are shown in Table 1.

Figure 1 Reverse Osmosis desalination unit.

Table 1 Technical specifications of the SWRO system.

Plant material, growth conditions and treatments

Regarding irrigated fresh vegetables, lettuce (Lactuca sativa) is categorized as being moderately salt tolerant [23]; on account of this, an experimental study of lettuce irrigation was executed to study the growth rate, the quality characteristics, and the yield under six different salinity levels of irrigation water as would be the case if water is produced by a desalination plant that operates at variable load. Hence, the objective of the experimental study is to identify the impact of water salinity on lettuce yield/quality, in order to determine the SWRO unit operation limits for the production of adequate water quality, while operating under variable load conditions with the least energy consumption for irrigation purposes.

The experimental study of lettuce irrigation was executed to assess the response of lettuce under different salinity levels of irrigation water by evaluating the growth rate, the quality characteristics, and the yield. The water was stored in six different water tanks according to each water salinity level (300, 800, 1200, 1700, 2200 and 3000 μS/cm). The crop irrigation experiment was conducted in spring 2020 in a glasshouse at the Agricultural University of Athens (N 37°59′10″, E 23°42′29″, altitude 24 m). On February 24^th, 2020, lettuce (Lactuca sativa) seedlings at the stage of three true leaves were transplanted in 4 L pots filled with sandy soil as a substrate. The characteristics of soil are presented in Table 2.

Table 2 Soil characteristics.

Twelve plants were accommodated in an irrigation channel with the dimensions of 5.3 m in length, 0.42 m in width, and 0.07 m in height and supporting on scaffoldings with 4% inclination. The total irrigation channels were six as the different salinities of the irrigation water produced by the SWRO system. The plants into the irrigation channels are presented graphically in Figure 2a. For the identification of the plants level alphabetic letters were used in each salinity level. Three from twelve plants in each salinity level were used for the experimental measurements (height, diameter and dry biomass) and they are depicted in Figure 2a with green color.

The plants treated with a conventional basic nutrient solution (N:12%, P₂O₅:8%, K₂O:16%). The optimum fertilization rate was set to 12 grams of each pot as reference [24]. The experimental layout is shown graphically in Figure 2a. Each salinity level was divided in twelve samples. Alphabetic letters were used for the identifications of the samples in each salinity level. The samples with green color (sub-samples) were used for the experimental measurements. Irrigation flow was set at 32 mL/min*pot and irrigation time was set to 3 min/2 times per day (morning and afternoon). Lettuce plants were grown under natural light conditions. Inside the glasshouse, the daily air temperature was always maintained below 25 °C, while the night temperature was always higher than 12 °C. A perspective of crop irrigation experiment in the glasshouse is presented in Figure 2b.

Figure 2 (a) The experimental layout of lettuce cultivation; (b) Lettuce irrigation experiment in the glasshouse.

Fuzzy Cognitive Maps

The Fuzzy Cognitive Maps (FCM) constitutes a modeling methodology based on human knowledge and belongs to the category of computational intelligence methods. The Fuzzy Cognitive Maps have been introduced by Kosko [25] and have been widely used in complex nonlinear systems containing uncertainties [26]. FCMs consist of nodes (C_i), which can be inputs, outputs, rules or intermediate states and each concept is characterized by a value (A_i) in the space [0, 1]. Each node can interact with other nodes. The relationship between the nodes is expressed by weights which take values in a range between [−1, 1]. The definition of the model nodes and the degree and type of the interaction between them can be determined by the knowledge of the experts. The FCM theory has been discussed in detail in [17, 24].

A typical FCM is depicted in Figure 3 where C_i represents the nodes-concepts and the weights (w_ij) express their relationships.

Figure 3 Structure of a typical FCM.

The weights of an FCM can be presented in a matrix form (W_i,j), as follows:

The values of the FCM nodes are calculated according to Eq. 1.

Particle Swarm Optimization (PSO)

Optimization methods aims to define the optimal solution of an objective function, considering the constraints and restrictions of the optimization problem. Several optimization algorithms have been proposed for engineering problems in the literature [27]. Indicatively, heuristic techniques such as Genetic Algorithms, Simulated Annealing, Particle Swarm Optimization (PSO) and Ant Colony Optimization, are used to solve complex multi-dimensional optimization problems [28]. In the present paper, the PSO method was used to find the optimum sizing of the components of the PV-SWRO system. The PSO method is a stochastic optimization, evolutionary and simulating algorithm and its purpose is to find the optimum solution of an objective function by performing a stochastic search based on a population. PSO has been used because of the good results it has produced in energy systems optimizations [19,29,30]. The used PSO settings are presented in Table 3 [31].

Table 3 PSO settings.

Introduction

The proposed tool is a computational tool for optimal design of desalination systems powered by PVs for crop irrigation. The computational tool targets at the optimization of both desalination unit and RES, together with the coordination of the control system. In order to find the most cost-effective system that meets the demand for irrigation water at the lowest possible cost, the computational tool carries out technical and economic optimization of the following variables:

• Number of membranes connected in series and in parallel.

• Number of membrane cases connected in series and in parallel.

• Technical characteristics of the energy recovery device in the RO unit.

• Choice of control strategy (full load or variable load).

• Number of PV panels.

• Number of batteries connected in series and in parallel.

• Size of the water tank.

The design after the optimization will lead to systems that provide lower water cost and improved operational efficiency.

Database of SWRO desalination plants

Initially, a database was created, which includes SWRO desalination systems for production capacity of 50 m³ per day to 300 m³ per day. An important element for the optimal desalination system, according to the desired water production, is the composition of the feed solution (salt water). Thus, the optimal desalination system is determined by the quality of the feed water (total amount of solid particles, in mg per L of water - Total Dissolved Solids (TDS)). A desalination system, for a specific water production, can be designed by selecting the type and the number of membranes, their arrangement, the number of membrane cases and the number of desalination stages. The optimal desalination system will be the system with the lowest energy requirements per unit of amount of desalinated water produced (kWh/m³). For the creation of the database the software package SQL Server Management Studio was used [32].

Weather forecast algorithms

Weather is an important factor in agricultural production, as it has a significant impact on crop growth, their water needs, their fertilizer requirements as well as on their yield. Weather changes can cause damages to crop, produce soil erosion, and affect crop quality. In addition, weather conditions are the main factor in predicting the energy production from PV panels. Therefore, a weather forecasting algorithm was developed, based on machine learning, with the use of time series and specifically the ARIMA model [33]. The main variables that must be forecasted are solar radiation, relative humidity, wind speed, duration of rainfall and ambient temperature. A short-term (3 days) weather forecast algorithm was developed using the Python programming language [34].

Energy management and control system

An energy management and control system for the desalination unit was developed to determine the optimal operating point of the desalination unit, based on techniques of computational intelligence, particularly on Fuzzy Cognitive Maps. This system was integrated into the computational tool, providing more flexibility for the designer to choose between two options:

(a) the system where the desalination plant operates at its rated load and (b) the system where the desalination plant operates at variable load [29,30,35].

The logical diagram of the energy management and control system is presented in Figure 4. The inputs for the energy management and control system are the water tank level (W_level(t)), the state of charge the batteries (SOC) and the energy produced by the photovoltaic panels (PV_power). If the water tank level is equal to the maximum value (W_max) or the SOC of the batteries is less or equal to 20% (minimum SOC for the batteries) then the operating point of the desalination unit (C4) is set to zero. In all other cases, the energy management and control system use a fuzzy cognitive map (FCM) to decide the operating point of the desalination plant (C4), taking as concepts the stored water in the water tank (C1), the SOC of batteries (C2), and the energy production by PVs (C3). All concepts are fuzzified in the space [0, 1]. Using the centroid defuzzification method, the concept of the operation point of the desalination unit (C4), takes a numerical value. This value is a percentage of the nominal power of the desalination unit. The weights W_ij express the relationship between the concepts and their values in space [0, 1]. The design and the programming of the energy management and control system were implemented using the MATLAB software package.

Figure 4 Logical diagram of the energy management and control system.

Optimization parameters

The optimization aims to minimize the net present cost (NPC) of the PV-SWRO system for a 20-year investment. Hence, a cost function was formulated, and it is presented in Eq. 2.

where n_PV is the number of the installed PV modules, P_PV is the rated power of the PV modules (W), Cost_PV is the price of the PV modules (€/W), E_bat is the battery bank’s rated energy capacity, Cost_bat is the price of the battery bank (€/kWh), C_des is the capacity of the desalination unit (m³/h), Cost_des is the cost of the desalination unit (€), Cost_ele is the cost of electronics (€), R is the annuity factor equal to 8%, C_ins,PV is the insurance cost of the PVs (€), C_main is the maintenance costs (€), C_chem,des is the chemicals costs (€/m³), j is the annual increase of prices at 1%, P_i is the annual fresh water production (m³), and i is the interest rate was equal to 5%.

The overall system cost analysis of the 20-year investment is presented in Table 4. The prices of the equipment are in accordance with the market prices [19].

Table 4 Costs and lifetime of the PV-SWRO components.

Modelling and Simulation framework of the computational tool

The environment of the computational tool was modelled and implemented using three existing software packages. The main software used is TRNSYS. TRNSYS is a flexible software environment used to simulate the behavior of transient systems and is a dynamic simulation software for energy systems. In addition to TRNSYS library, which includes various subroutines for photovoltaics, and batteries, the user can include tailored made subroutines (e.g., desalination system, controllers, etc.) in FORTRAN. Based on its modular structure, it can solve large groups of equations described by the various subroutines. GenOpt is open-source optimization software that allows multidimensional optimization of an objective function calculated by a simulation program. It is a standalone platform written in java that can minimize a cost function evaluated by an external simulation program such as TRNSYS and can be successfully connected with TRNSYS. GenOpt has been developed for optimization problems where the cost function is computationally accurate, and its derivatives are not available or may be not yet exist. TRNOPT software was used to connect GenOpt to TRNSYS. The TRNOPT transmits data from TRNSYS to GenOpt and vice versa without any interaction from the user. Finally, Matlab software was used for the development of the energy management system and control system. MATLAB is a high-level language and an interactive environment for arithmetic calculations and programming. The structure of the modelling and simulation framework of the computational tool is shown in Figure 5.

Figure 5 Schematic of the computational tool.

SWRO desalination unit experimental results

The SWRO desalination unit was initially studied to identify its performance at variable load conditions (variable pressure and variable feed water supply) by varying the power input, in order to determine its optimum operational window with the lowest specific energy consumption. The specific energy consumption and the electrical conductivity of the produced water as a function of the membrane inlet pressure of the desalination unit, are presented in Figure 6. The specific energy consumption (kWh/m³) is calculated as the ratio of the energy consumed by the SWRO motor (kWh) to the freshwater production (m³). The experimental results showed that the SWRO desalination unit can operate in a range of conductivity within the optimum operational window in which the SEC is around 5 kWh/m³.

Figure 6 Specific energy consumption and electrical conductivity as a function of the membrane inlet pressure.

Statistical analysis of variance

The height and diameter of canopy were measured every three days during the growing period, in order to identify the impact of irrigation water salinity on lettuce yield/quality. The experimental results with regard to lettuce height and diameter, under different salinity levels are presented in Figures 7 and 8, respectively. The experimental measurements showed that the maximum height was noticed for the salinity of 300 μS/cm at 25 cm whereas the minimum height was for the salinity of 2200 μS/cm at 19 cm (Figure 6). The same situation was observed in the experimental results of diameter with the maximum rate at 27 cm for the salinity of 1700 μS/cm and the minimum at 25.3cm for the salinity of 2200 μS/cm (Figure 7).

The experimental measurements showed that, the lettuce height under the salinity levels of 800–3000 μS/cm was lower than the lettuce height under the reference salinity point of 273 μS/cm and the best height was noticed in the salinities 800, 1200 and 1700 μS/cm (Figure 7). The opposite situation was observed in the experimental results of diameter (Figure 8) where the lettuce leaves at the reference salinity point of 273 μS/cm had lower dimensions than the lettuce leaves at the salinity levels of 800–3000 μS/cm while the best dimeter was noticed in the salinities 800, 1200 and 1700 μS/cm. Additionally, the dry biomass of lettuce leaves was also measured, the results of which, are presented in Figure 9. Specifically, Figure 9 shows the weight of dry biomass of lettuce canopy in different salinity levels. It was observed that the maximum weight of dry biomass of lettuce canopy was for the salinity of 300 μS/cm and the rest salinities had approximately the same weight of dry biomass with the reference salinity point (273 μS/cm). However, as it can be noticed in Figures 6–8 there aren’t any significant difference during the lettuce growth between the 6 salinity levels. This was validated through the analysis of variance (one way-ANOVA) which was caried out using the software package Statistica for Windows 9.0 (Tulsa, OK, USA). Hence, the best salinities level for lettuce irrigation purposes and simultaneously with the least energy consumption for the desalinated water by a PV-SWRO system are 300, 800 and 1200 μS/cm with specific energy consumption of around 5 kWh/m³.

Figure 7 The growing rate of lettuce height in different salinity levels.

Figure 8 The growing rate of lettuce diameter in different salinity levels.

Figure 9 The weight of dry biomass of lettuce canopy in different salinity.

Simulation and optimization results

Case study scenario

In order to evaluate the computational tool, a case study was conducted that includes a SWRO desalination system powered by PVs for lettuce irrigation. The PV-SWRO desalination system was considered to be installed in an Aegean Island, namely Naxos, at Cyclades complex, to irrigate a 0.1 ha lettuce farm. Given that the Aegean islands are characterized by high solar and wind potential, the combination of RES with SWRO desalination units is the most appropriate way to address water issues for crop irrigation on these islands. Specifically, lettuce is a cool-season crop and the optimal temperatures for its growth are 23 ^oC during day and 7 ^oC at night [36]. The solar radiation and the temperature per month for the selected area is presented in Figure 10.

The amount of irrigation water depends on biotic (fungi etc.) and abiotic factors (temperature, type of soil, its season year and soil moisture). The water requirements of a lettuce crop are usually 340 to 750 m³ per ha and the growing season is from September to April [36,37]. This high range of irrigation rate, together with the irrigation experience in Naxos Island (private communication with farmers) were used to identify the average consumption presented in Figure 11 for the 0.1 ha land under consideration. In this case study scenario, the water requirement is presented on a monthly basis in Figure 11 [38,39] in order to prevent water stress. In addition, it was considered that a drip irrigation system was installed for lettuce irrigation as drip irrigation transfers small amounts of water to the periphery of the roots of plants and as a result it can save water and increase crop production [40,41]. The rest of the period, the PV-SWRO unit can be used to produce water for other purposes.

Figure 10 The solar radiation and temperature per month

Figure 11 Daily water consumption for irrigation per month.

Simulation results

The results of the application of the computational tool are presented in Table 5. It can be observed that a PV powered SWRO desalination system that operates at variable load and combined with a small capacity battery bank and an energy management and control system based on fuzzy cognitive maps, is the optimum system since this system can ensure the required water production for irrigation needs at lowest cost compared with the system that computed with the conventional way of design, covering in parallel all the technical set constraints.

Table 5 Computational tool results compared with a conventional design.

In addition, the operation of the PV-SWRO desalination unit under variable load can further reduce the NPC. Specifically, the PV-SWRO desalination system, operating only at nominal load, designed by the computational tool presented lower NPC by almost 25% compared to the conventional way of designing, and the PV-SWRO desalination system that operated at variable load presented lower NPC by almost 39% compared to the conventional method of designing.

Ethics Statement

Not applicable.

Consent for Publication

Not applicable.

Availability of Data and Material

Not applicable.

Funding

This experimental work is co-financed by Greece and the European Union (European Social FundESF) through the Operational Program «Human Resources Development, Education and Lifelong Learning 2014-2020» and the Program encoded EDBM103, titled “Support for researchers with an emphasis on young researchers-cycle B’”, in the context of the project “Development and validation of a computational tool for designing RO-PV desalination systems for irrigation” (MIS 5048478).

Competing Interests

George Papadakis is Editor-in-Chief of the journal Green Energy and Sustainability. All the authors have declared that no competing interests exist. The authors were not involved in the journal’s review or decisions related to this manuscript.