1. Introduction
In the present society, the shortage of fresh water resources is an indisputable fact. As a necessity of human life, the lack of fresh water resources have become a constraint on social development in many countries of the world [
1,
2]. With the development of industrial technology and the world population quick increase, the global demand for desalination is increasing [
3,
4]. At the present time, desalination plants produce around 95 million m
3/day, Middle East and North Africa (MENA) regions are responsible for 48% of the global installed desalination capacity [
5,
6]. Generally, the use of multi-stage flash (MSF), multi-effect distillation (MED), electrodialysis (ED), and reverse osmosis (RO) processes has attracted significant attention in attempts to improve the reliability and the performance of freshwater production processes. Among them, MSF desalination is an important technology in the desalination industry due to its high reliability and good distillate quality. However, it is acknowledged that its production cost is high and the system performance is greatly affected by the seawater temperature, fouling factor and etc. [
7,
8,
9,
10,
11]. Therefore, it is of great significance to study how to reduce the operating cost of the MSF desalination system. In order to solve this problem, it is necessary to establish a complete mathematical model.
In past decades, the simulation, performance analysis and optimization of MSF systems were widely performed to reveal the desalination mechanism and to improve the performance with lower cost. Hellal [
12] et al. established a detailed mathematical model of MSF desalination system, and then proposed an efficient and reliable method to solve the nonlinear equations describing the multi-stage flash desalination system. After linearization, the equations are decomposed into some subsets based on their characters, and then the enthalpy balance equations are expressed in the form of triangular matrix. This method has good stability and fast convergence speed, but the heat loss of the flash process is ignored in the model. Therefore, Marina Rosso [
13] et al. established a detailed steady-state mathematical model for the analysis of MSF desalination process. The model takes into account the physical and structural factors, such as the geometry of each flash chamber, the changes in physical properties of water with temperature, non-equilibrium temperature difference and so on. The established mathematical model allows the analysis of the influence of operation and design variables on the system in the process. This information can be used not only for design purposes but also for the development of dynamic models. Based on the above model, El-Dessouky [
14] further considered the factors, including fouling factor, pressure drop through demister, non-equilibrium allowance and so on into the model, making it stricter and more accurate. Considering that the parameters such as the boiling point elevation, heat capacity of brine, latent heat of vaporization are functions of temperature and salinity, and the non-equilibrium loss in the MSF process should not be ignored, Wu [
15] et al. established a mathematical model of the MSF desalination system. However, in order to reduce the non-linearity of the system, the temperature of each flash chamber is assumed to follow an equal temperature drop distribution, which reduced the accuracy the rationality of the modeling process. Therefore, Tanvir and Mujtaba [
16] proposed an effective prediction method of boiling point elevation of brine based on neural network, and then Tanvir and Mujtaba [
17] embedded this method into the MSF desalination process model written in gPROMS platform. Based on the above work, Tanvir and Mujtaba studied the influence of seawater temperature and steam temperature on the performance of the MSF system. Since brine heater fouling can significantly reduce the heat transformation efficiency and cause performance deterioration of the MSF system, Al-Rawajfeh et al. [
18] studied the deposition of calcium carbonate in flash chambers in once-through MSF (MSF-OT) and brine recirculation MSF (MSF-BR) processes by correlating the deposition of calcium carbonate to the released rate of carbon dioxide in a steady state model based on coupling of mass transfer with chemical reaction. Then they extended their work to include deposition of calcium sulphate with calcium carbonate inside the tubes and flash chambers [
19]. Mujtaba considered the brine heater fouling in his MSF process simulation and optimization. Based on actual plant data, a simple linear dynamic fouling factor profile was developed which allows calculation of the fouling factor at different times [
20]. Furthermore, Alsadaie and Mujtaba [
10] presented very detailed fouling model that considered the attached and removal rate of calcium carbonate and magnesium hydroxide and also it considered the effect of temperature, velocity and salinity. The model was applied on the MSF-BR process. In 2019, they presented a dynamic model of fouling to predict the crystallization of calcium carbonate and magnesium hydroxide inside the condenser tubes of the once-through MSF process [
21].
Based on the mechanism analysis and mathematical modeling of MSF desalination, Hu et al. [
22] developed an improved optimization model for the MSF desalination system using the annual average distillate cost as the objective function, and a more detailed optimization design analysis with market price change was carried out for a 3000-ton/day MSF desalination plant. Based on the strict mathematical model of MSF system, Mussati et al. [
23] studied the design optimization model to obtain a better structure of an MSF system, and generalized gradient algorithm under the GAMS platform was used to solve the problem. Said et al. [
24] considered a storage tank between the desalination system and the users, then they built the design optimization and operational optimization problem to minimize the daily operational cost of the MSF desalination system under various conditions. Hawaidi and Mujtaba [
20] considered the dynamic relationship between scaling factor of the brine heater, operation time and seawater temperature change, they also studied the operational optimization problem with the goal of minimum annual operation cost with a fixed freshwater demand. Considering the dynamic change of operational conditions and the aim of better control of a MSF system, dynamic modeling and optimization of the MSF process were also studied in past years. Based on a detailed description of the fundamental elementary phenomena involved in the process, Mazzotti et al. [
25] developed a model for the dynamic simulation of MSF desalination units and analyzed the non-linear dynamic features under different disturbances. Gambier et al. [
26] collected some different dynamic models from the literatures, and analyzed their advantages and drawbacks taking into account simulation and automatic control purposes. Considering that there were not enough detailed models and analyses of the dynamics of the MSF process and the demister, Al-Fulaij [
27] developed lumped parameter dynamic models for the once-through (MSF-OT) and the brine circulation (MSF-BC) processes. He also coded the models with the gPROMS modelling program for analysis. By coupling the dynamic equations of mass, energy and momentum, Bodalal et al. [
28] presented a mathematical model to predict the performance of MSF plant systems under transient conditions. The model describing the dynamic behavior of each stage in terms of some key physical parameters were solved by using the fifth order Runge–Kutta method. Lappalainen et al. [
29] presented a new method for one-dimensional modelling and dynamic simulation of thermal desalination processes. The approach combines the simultaneous mass, momentum, and energy solution, local phase equilibrium by Rachford-Rice equation, and rigorous calculation of the seawater properties as function of temperature, pressure and salinity. Computing results show that it is a competent approach for dynamic simulation of thermal desalination processes. With known dynamics of the MSF process, advanced control can be obtained for performance improvement and cost saving. Alsadaie and Mujtaba [
30] coded a dynamic model of MSF process with gPROMS model builder, and successfully applied a generic model control (GMC) algorithm to provide better performance over a conventional PID (proportional-integral-derivative) controller. Their work has guiding significance for the optimal operation and control of MSF process in the short term to long term, and is very helpful and enlightening to our work.
The modelling and optimization studies mentioned above established the solid foundation for reliable for economic operation of MSF system. But few of them considered the effect of discharged recycle mass flow and pump energy cost for the whole flowsheet. We also found that for the easy use and solution of the MSF process model, ideal assumptions and some heat loss were ignored, and the components of operating expenses are quite different if we consider the power loss of pumps and steam price. Therefore, in order to obtain a more complete and elaborated model, this paper considers the effects of factors such as boiling point elevation, influence of heat loss of condenser tubes, the phenomenon of unequal temperature drop in each flash stage, and at the same time the compensation effect of the reject seawater recycle mass flowrate on the seawater temperature is also accounted for in the proposed model. Since we just consider daily operational optimization of the MSF process, a fouling factor was used in the process model, and the dynamic model of exchanger fouling was not incorporated into the established model. The established model can fully reflect the non-linearity of the system, and can reflect the influence of various operating variables on the system performance well in short term. Based on the model, the operating states of the MSF system were simulated under different parameter conditions. And in addition, considering the changes of seawater temperature and distillate demand on the system performance, the optimal operation problems were carried out for maximizing the GOR (gained output ratio) and minimizing the daily operational cost, respectively.
2. Flowsheet of Multi-Stage Flash (MSF) Process
MSF desalination system has two classic structures, which are one-through multi-stage flash (OT-MSF) seawater desalination system and brine recirculation multi-stage flash (BR-MSF) seawater desalination system [
27,
30,
31,
32,
33]. Among them, the structure of OT-MSF system is relatively simple, and the brine circulation type is more widely used because of its better comprehensive performance. The flowsheet of BR-MSF system is shown in
Figure 1. The simulation and optimization research of this paper are based on this kind of BR-MSF system.
The BR-MSF system is composed of four parts: brine heater, heat rejection section, heat recovery section, splitters and mixers. The heat rejection section aims to reject the remaining heat energy from the system, thereby cooling the distilled product and concentrated brine to the lowest possible temperature. In addition to completing the energy release from the heat rejection section, the splitters and mixers also recycle part of the waste brine back into the system to reuse the energy.
As can be seen from
Figure 1, after preheating of the heat rejection section, the feed seawater is divided into two parts, one is returned to the sea and the other is mixed with the recycled brine. The mixed brine is pumped into the end of the heat recovery section. When it flows through a series of heat exchangers from the right to the left, it is gradually heated, and the flash steam in the flash chamber is condensed. Finally, seawater comes out of the first stage flash chamber of the heat recovery section, flows into the brine heater to be further heated, and then flows into the first stage flash chamber again at the highest temperature. The pressure in the flash chamber is gradually reduced, so that the flash stream enters the flash chamber and evaporates immediately. The generated steam passes through the condensers and drops into the distillate trays after being condensed. The process is repeated until the last stage is reached, then the concentrated brine is rejected, the distillate is extracted, and a part of the brine is reused as recycle brine.
5. Operational Optimization of MSF System
Considering that the seawater temperature, distillate demand and other operational parameters change all the time, the optimal operation of the MSF system for cost saving is quite important. Based on the established model of MSF desalination system, this paper studies the optimal operation problems with two different objectives. The first is to maximize the GOR under different operating conditions, and the second is to minimize daily operational cost in order to obtain the optimal manipulated variables at different times. Since the seawater temperature and freshwater demand are the most important and most frequently changing parameters, the optimization work selects the actual temperature from some seawater desalination plant in Zhejiang province as the background, which can be seen in
Table 4. The hourly distillate demand is as shown in
Table 5.
5.1. Optimal Operation Problem to Maximize Gained Output Ratio (GOR)
To a large extent, the GOR is known as the traditional index to describe the performance of the MSF system. According to the flowsheet of BR-MSF shown in
Figure 1, there are four manipulated variables that can be adjusted to reach the optimal operational point. These manipulated variables are recycle stream mass flowrate (
), rejected seawater mass flowrate, steam temperature and reject seawater recycle mass flowrate (
). With bound constraints, given distillate demand and a well-established process model, the optimal operation problem to maximize GOR was formulated as the following problem called OPT1.
The superscripts
L and
U here denote the lower and upper bounds of the parameter, respectively. The equation
represents the mechanism model of the system. Since the optimization period is a short-term problem, the fouling factor of the brine heater and the exchanger in the flash chambers are assumed to be a constant value of 1.86 × 10
−4 (h·m
2·K)/kcal. At the same time, the seawater temperature and distillate demand in each time period are shown in
Table 4 and
Table 5, respectively. The optimal operation problem was solved with interior point algorithm under the MATLAB.
The solutions including the optimal variables and the optimal objective function values are listed in
Table 6.
Having assigned the steam temperature () of the system, the seawater temperature () and distillate demand () for each time period, for the Opt1 problem that maximizes the GOR, optimal operations of the series of the recycle stream mass flowrate (), mass flowrate to the reject seawater splitter () and reject seawater recycle mass flowrate () were obtained. Here, note that the equipment is not working during maintenance from 00.00 to 02.00 at night.
The overall optimization results also show that lower seawater temperature () can result in higher GOR. During the time period of 22:00–00:00, the seawater temperature shows the lowest value in a day. At the same time, the reject seawater recycle mass flowrate () reaches a maximum value, and the GOR reaches a peak value, which is to compensate for the effect of the low seawater temperature on the system performance. It is also found that, at relatively high temperature and high distillate demand, the value of S is quite small. This means in the conditions that the discharged brine from rejection section should not be reused and mixed with origin feed seawater.
Figure 14 shows the relationship of distillate demand and rejected recycle mass flowrate (
), from which a conclusion can be drawn.
Figure 15 shows the profile of optimal the recycle stream mass flowrate (
) and the reject seawater splitter (
) under the given ambient parameter change to maximize the GOR.
5.2. Optimal Operation Problem to Minimize Daily Operational Cost
The production cost is directly affected by the higher operating cost of the MSF desalination system. The electricity price, seawater temperature etc. change frequently over time, causing the operation control difficulty of MSF system. Therefore, the optimal operation of the MSF system was further studied below. The optimal operation problem to minimize the daily operational cost can be formulated as:
Here
is the total distillate production,
is the given distillate demand, which is shown in
Table 5, and
is expected value of
TBT.
TOC denotes the total operational cost of each hour, which can be calculated as follows:
TOC1 denotes steam cost,
TOC2 denotes pump energy consumption,
TOC3 denotes chemical additive cost,
TOC4 denotes labor cost,
TOC5 denotes system maintenance and servicing costs. The electricity price of each time period is shown in
Figure 16. The steam price at 92–115 °C is 23.72 CNY/Million kJ.
Each part of the operational cost is as follows:
where,
.
With the operational cost equations and the established MSF system model, the optimal operation problem named as OPT2 was studied and solved by an interior point algorithm under the MATLAB platform. The problem takes into account the changes of the seawater temperature (
), distillate demand and electricity price at different hours in a day. The problem was successfully solved and the key results are listed in
Table 7. Since the optimization results show that the optimal reject seawater recycle mass flowrate (
) is 0 at each time interval, it is not included in
Table 7 for convenience.
After assigning the seawater temperature () and distillate demand () for each time period, for the Opt2 problem that minimizes the TOC, the optimal operations of the series of the recycle stream mass flowrate (), mass flowrate to the reject seawater splitter (), Steam temperature () and Steam mass flowrate () were obtained. Here note that the equipment is not working during the maintenance from 0 to 2 o’clock at night.
The overall optimization results also show that from two o’clock in the morning, the seawater temperature () gradually increases, and the demand for distillate () also increases. Therefore, higher flash brine volume and steam consumption () are required, resulting in higher operating costs (TOC). During the period of 11:00–12:00, the demand for distillate () reaches the peak in a day, and the temperature of seawater () is also at a high level. At the same time, the operating cost (TOC) of the device also reaches the maximum in this period. After 12:00, the temperature of seawater () gradually decreases, and the distillate demand () maintain a downward trend from 12:00 to 17:00, while it increased from 17:00 to 20:00, and began to decline after 20:00. The operating cost of the device is the same as the trend of distillate demand ().
Figure 17 shows the relationship between TOC and steam temperature (
) and steam flow (
) in each time period.
Figure 18 shows the profile of optimal the recycle stream mass flowrate (
) and the reject seawater splitter (
) under the given ambient parameter change to minimize the TOC.
6. Conclusions
The multi-stage flash desalination process is one of the most important technologies to obtain fresh water on a large scale. However, its operational cost is relatively high and its performance is significantly affected by seawater temperature, salt concentration, steam quality and other operational factors. In this paper, the sensitivity analyses of these parameters on the performance of MSF system were carried out based on the elaborated established process model, and then two kinds of operational optimization problems were studied.
To obtain a more comprehensive and more elaborate model, this paper considers the effects of factors such as boiling point elevation, influence of heat loss of condenser tubes, the unequal temperature drops in each flash stage, and effect of the reject seawater recycle mass flowrate. Simulation results of an MSF system with a 16-stage flash chamber demonstrate that the model has similar accuracy with those from Rosso and Mujtaba. The sensitivity analysis of seawater temperature shows that higher seawater temperature can cause higher BBT and lower distillate flowrate, and there is a ‘best temperature’ to maximize GOR (gained output ratio). If the feed seawater is quite low, the increase of reject seawater recycle mass flowrate(S) will decrease the distillate flowrate, but will increase the GOR to a maximum until the mixed seawater reach the ‘best temperature’. Steam temperature is the only factor which has a significant and positive effect on distillate flowrate and GOR.
The optimal operation of MSF was studied with two different objective functions. The optimization problem to maximize the GOR under given daily seawater temperature and fresh water demand shows that with limited upper steam temperature of 97 °C, the maximum GOR can be achieved with given demand and feed seawater temperature. However, at a high freshwater demand time interval, the rejected seawater recycle mass flowrate(S) is quite large, so the rejected seawater recycle mass flowrate should not be ignored and set to zero. Also, the recycle stream mass flowrate and the rejected seawater mass flowrate should be adjusted as ambient conditions change. To minimize the daily operational cost, the objective function was reformulated considering the pump power cost of other mass flowrate and the price of steam in China. Computing results of the optimization shows that, with the same seawater and freshwater demand, the optimal operation is quite different from that from OPT1. Optimal value of the rejected seawater recycle mass flowrate(S) is always zero, the steam flowrate and temperature increase as the demand for freshwater increase, but the rejected seawater mass flow decreases at the same time. With the established process model and objective function, the optimal values for those manipulated variables can be obtained, which is helpful to guide the economical operation of the MSF system.