Desalination 553 (2023) 116453

Contents lists available at ScienceDirect

Desalination
journal homepage: www.elsevier.com/locate/desal

Dynamic modelling and simulation of industrial scale multistage flash

desalination process
Hasan Hasan a, Salih Alsadaie b, Mudhar A. Al-Obaidi c, Iqbal M. Mujtaba a, *
a
Department of Chemical Engineering, Faculty of Engineering and Informatics, University of Bradford, Bradford, West Yorkshire BD7 1DP, UK
b
Department of Chemical Engineering, Faculty of Engineering, University of Sirte, Sirte, Libya
c
Middle Technical University, Technical Institute of Baquba, Dayala, Iraq

H I G H L I G H T S

• A comprehensive dynamic simulation-based model of BR-MSF was developed.

• The flowrate of the distillate for each stage was modelled as the flow was withdrawn from each stage and flow though channel and not entering the next stage.
• The model was validated against data of Shuaiba South MSF plant with minor errors.
• Dynamic responses of MSF were explored under changes in some operating parameters.
• The last stage of MSF requires a longer time to settle for the tested parameters.

A R T I C L E I N F O A B S T R A C T

Keywords: Multistage Flash (MSF) desalination process is still a dominant process, especially in the Gulf region, to produce
Seawater desalination high quality freshwater. Although there has been energy price surge in recent years, MSF process will continue to
Multistage flash operate in that region for some foreseeable future. The key challenge is how to make such processes still prof­
Shuaiba-South MSF seawater desalination plant
itable. Understanding the dynamics of any processes under uncertainty and disturbances is very important to
Dynamic modelling
Dynamic simulation
make a process operationally feasible and profitable. The main aim of this work is to understand the dynamics of
industrial scale MSF process using high fidelity and reliable process model. For this purpose, a detailed dynamic
model for the MSF process incorporating key and new features is developed and validated against the actual data
of a large-scale seawater desalination plant. The model is then used to study the behaviour of large scale MSF
processes for disturbances in steam temperature, feed temperature and the recycle brine flow rate. The simu­
lation results show that the last stage requires a longer time to settle compared to the preceding stages. In
addition, steam temperature shows insignificant influence on the performance ratio compared to the inlet
seawater temperature and recycle brine flow rate. Furthermore, it is found that the productivity of plant can
increase in the winter compared to that in the summer. However, this benefit comes at the expense of increased
steam consumption in the winter, resulting in a low performance ratio.

membrane. The durability of evaporative processes makes multi-stage

flash (MSF) and multi-effect desalination (MED) ideal for the Gulf re­
1. Introduction
gion. The MED method has been extensively established, and it is
thermodynamically adapted for taking use of temperature drop per stage
Water shortage has emerged in densely populated cities and other
to less than 3 ◦ C to 4 ◦ C [2]. However, MSF and Reverses Osmosis (RO)
industrial locations as a result of large increase in water demand driven
have been the most used technologies used in desalination processes in
by fast growth of population and industrial development. This in turn
the last decades [3,4]. In this regard, the MSF is the oldest method of
has motivated many researchers to improve the water desalination
seawater desalination that is being used to produce high quality fresh­
systems in order to meet the needs of the society and fulfil the water
water in large scale compared to the membrane technology.
demand of industrial applications [1].
Due to the lowest cost fossil fuels in the past, Bennett [5] pointed that
There are several desalination technologies such as thermal and

* Corresponding author.
E-mail address: I.M.Mujtaba@bradford.ac.uk (I.M. Mujtaba).

https://doi.org/10.1016/j.desal.2023.116453
Received 18 July 2022; Received in revised form 17 January 2023; Accepted 5 February 2023
Available online 21 February 2023
0011-9164/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
H. Hasan et al. Desalination 553 (2023) 116453

Nomenclature Q The cooling brine stage heat transferred, kW

Rec The recycle brine flow rate, kg/s
AD The area of the Distillate tray, m2 TBT Top brine temperature, ◦ C
Ah The surface area of the heat transfer surface area of the Tb_in The brine heater inlet temperature, ◦ C
brine heater, m2 Tb_out The brine heater outlet temperature, ◦ C
As The surface area of the heat transfer for the stages, m2 TBin The flashing brine Temperature inlet for the stage, ◦ C
Apipe The distillate cross-section area, m2 TBout The flashing brine Temperature outlet for the stage, ◦ C
Ap The area for the stage, m2 TD The distillate tray Temperature, ◦ C
Bin The inlet brine flow rate, kg/s TDout The distillate tray Temperature outlet for the stage, ◦ C
Bout The outlet brine flow rate, kg/s TFin The cooling brine Temperature inlet for the stage, ◦ C
BPE Boiling point Elevation, ◦ C TFout The cooling brine Temperature outlet for the stage, ◦ C
CBHout The outlet NCGs Mass fraction from the brine heater, ppm TS The tube surface Temperature, ◦ C
CBe The brine NCGs mass fraction (Equilibrium), ppm TVB The flashed vapour temperature below demister, ◦ C
CBin The inlet NCGs Mass fraction to the stage, ppm TV The flashed vapour temperature in the vapour space, ◦ C
CBout The outlet NCGs Mass fraction from the stage, ppm TSteam The temperature of the steam, ◦ C
CFin The inlet NCGs Mass fraction to the HRJ, ppm U Overall heat transfer coefficient, kW/m2 ◦ C
CFout The NCGs Mass fraction for the make-up stream, ppm UBH Overall heat transfer coefficient of the brine heater, kW/m2
Cd The brine discharge coefficient ◦
C
Cc The distillate discharge coefficient Vb The volume of the brine, m3
CNCGS The NCGs Mass fraction for the seawater VB The Vapour release flow rate outlet the brine, kg/s
Cp The constant pressure specific heat, (kJ/kg ◦ C) VD The Vapour release flow rate outlet the distillate, kg/s
Crec The NCGs Mass fraction in the recycle brine, ppm Vin The inlet vapour flow rate, kg/s
CR The NCGs Mass fraction in the recycle brine, ppm Vout The outlet vapour flow rate, kg/s
CVD The overall condensate flow, kg/s Vtube The tube bundle volume of the cooling water (inside), m3
di The inside diameter of the heat exchanger tubes, m Vv The volume of vapour, m3
do The outside diameter of the heat exchanger tubes, m Wcw The brine rejected cooling, kg/s
Dout The distillate flow rate entering the stage, kg/s WHBout The brine heater brine mass flow rate outlet, kg/s
Dtotal The distillate flow rate final product, kg/s WHBin The brine heater brine mass flow rate inlet, kg/s
Flast The last stage make-up (seawater), kg/s WRin The HRS cooling brine flow inlet, kg/s
g The gravity acceleration, m/s2 WRout The HRS cooling brine flow outlet, kg/s
HBHin The inlet brine enthalpy (brine heater), kJ/kg WFin The HRS seawater flow inlet, kg/s
HBHout The outlet brine enthalpy (brine heater), kJ/kg WFout The HRS seawater flow outlet, kg/s
HBin The inlet flashing brine enthalpy (brine heater), kJ/kg Wst The stage width, m
HBout The outlet flashing brine enthalpy, kJ/kg Wsteam The steam flow rate, kg/s
HDout The outlet distillate enthalpy, kJ/kg XBin The brine salt concentration inlet, ppm
Hg The stage gate height, m XHBout The brine salt concentration outlet from the brine heater,
HNCGs The NCGs enthalpy outlet form the flashing brine, kJ/kg ppm
HNCGsin The NCGs enthalpy inlet, kJ/kg XBout The brine salt concentration outlet, ppm
HNCGsoutThe NCGs enthalpy outlet form the stage, kJ/kg Xrec The recycle brine salt concentration, ppm
HVB The vapour enthalpy for the demister, kJ/kg XRout The cooling brine in the HRS salt concentration outlet, ppm
HVD The vapour enthalpy for the distillate tray, kJ/kg XFout The cooling brine in the HRJ salt concentration outlet, ppm
HVin The vapour enthalpy inlet the stage, kJ/kg Yin The NCGs Mass fraction in the vapour (inlet), wt%
HVout The vapour enthalpy outlet the stage, kJ/kg Yout The NCGs mass fraction in the vapour (outlet), wt%
HWout The water enthalpy outlet the stage, kJ/kg
LB The brine height, m Greek letters
LD The distillate height, m ΔP Pressure difference, (Pa)
LMTD The tube condenser Logarithmic mean different ρB Brine density, (kg/m3)
temperature, ◦ C ρD Distillate density, (kg/m3)
LS The stage height, m ρv Vapour density, (kg/m3)
Lt The stage tube length, m λsteam Latent heat of steam, (kJ/kg)
dLB dLD dXBout dCBout dHBout dHWout dHV OUT dMV dMBH dY OUT dMW
MBH The brine holdup in the brine heater, kg dt , dt , dt , dt , dt , dt , dt , dt , dt , dt , dt variation of
MV The vapour holdup in the brine stage, kg (brine height, distillate height, brine salt concentration
MW The flashing holdup in the brine stage, kg outlet from the stage, outlet NCGs Mass fraction from the
NEA The allowance for the non-equilibrium, ◦ C stage, outlet flashing brine enthalpy, outlet water
kv The orifice discharge coefficient venting line enthalpy, outlet vapour enthalpy, vapour holdup in the
Nt The tube bundle number of tubes brine stage, brine holdup in the brine heater, outlet NCGs
NCGs The Non-condensable gases, kg/s mass fraction in the vapour, and water holdup in the tubes
P The stage pressure, Pa bundle) against operational time
PR The performance ratio, (kg distillate/kg steam)

2
H. Hasan et al. Desalination 553 (2023) 116453

around half of the desalination plant are situated in the gulf countries condensable gases. The model was validated against actual plant data
with large marketing share. Even with high-energy price surge in recent of Azzour desalination plant at steady state conditions with a good
years, MSF process will continue to operate in that region for some agreement. Then, the model was used to optimise the location of the
foreseeable future. The MSF desalination process is the most reliable and venting system to increase the performance ratio of the plant.
robust compered to RO and multi effect distillation (MED) due to its high A dynamic model was developed by Sowgath and Mujtaba [23] (but
productivity of high-quality water. In terms of plant issues, the MSF not validated) to analyse the influence of dynamic seawater temperature
plant achieved the minimum failure since the process was established change on the performance of a brine recirculation MSF process. The
[6,7]. In terms of specific energy consumption, the thermal desalination model comprised non-equilibrium effects, demister pressure drop and
technologies require between 2.3 and 3 kWhElec/m3 of electrical energy the brine and distillate hold up correlations. Furthermore, the model
and between 71.7 and 80.6 kWhTherm/m3 of thermal energy compared to specified the relationship between the inlet seawater temperature and
3.5 kWhElec/m3 of electrical energy consumed by seawater RO desali­ the steam temperature. The steam temperature profile subjected to a
nation system [2]. However, the desalination processes are roughly number of step changes in seawater temperature with guaranteeing a
contributing 76 million tons/year of CO2 emissions with an expectation fixed demand of freshwater at the highest allowed performance ratio.
to grow to 218 million tons/year of CO2 by 2040 [8]. Lappalainen et al. [24] developed a dynamic model for a brine
The MSF process is available in two designs: once-through MSF-OT recycling MSF desalination process using local phase equilibrium
and recirculating brine MSF unit (MSF-BR) [9]. A brine heater section calculation by Rachford-Rice equation. The model used to study the
and heat recovery section are required for the MSF-OT, while an addi­ steady state and transient behaviours besides investigating the effect of
tional heat rejection section is required for the brine recycle configu­ venting system failure on the pressure and temperature of an industrial
ration. Furthermore, the MSF desalination process can be designed with large size brine recirculation MSF plant with 21 stages in the heat re­
a minimum of 4 stages to maximum of 40 stages. The top brine tem­ covery section, and 3 stages in the heat rejection section.
perature operates with an approximate of 90 ◦ C to 110 ◦ C with a product To achieve a transient performance change quickly, a complete
around of 11 kg of distillate per kg of steam (GOR ~ 11) [10–12]. mathematical model of the brine recirculation MSF system was con­
The process modelling plays a vital role in designing the process structed by Huang et al. [25]. To adapt to dynamic changes in feed
layout, investigating the optimal inlet conditions, and controlling the seawater temperature, feed seawater concentration, recycling stream
process responses as the modelling is much cheaper than running an mass flow rate, and steam temperature, dynamic simulation and analysis
experiment [13–15]. were performed. The dynamic model was validated against a medium
Unless inherently dynamic, it is desirable that all processes run under scale actual plant data of 13 stages in the heat recovery section and 3
steady state mode (time independent). However, this is not always the stages in the heat rejection section [26] but at steady state conditions
case due to the disturbances and uncertainty associated with the process with a good agreement. The dynamic response curves for top brine
parameters and corrective measures need to be taken when the process temperature, bottom brine temperature, flashing brine temperature at
starts deviating from the steady state operating mode, which will require prior stages, and distillate mass flow rate at previous stages were
in depth understanding of process dynamics. Maniar and Deshpande developed, which accurately describe the system’s dynamic features.
[16], Gambler and Badreddin [17], Sowgath [18], Alsadaie [19] sum­ All of the previously published mathematical models have the same
marised the importance of model-based understanding of the process material and energy equations, and the only difference between them is
dynamics in relation to start-up, shutdown, fault detection and trou­ the level of sophistication. Some were simple models created to achieve
bleshooting of MSF process, in addition to designing the best control various aims, while others were detailed models created for the purpose
system to get the process back to steady state in shortest possible time. of capturing deeper insight of the process. To be more exact, all previous
A dynamic simulation model of MSF desalination process was models assumed that the product distillate from one stage is passed over
developed by Mazzotti et al. [15] that accounted the stage geometry, to the next. This assumption would lead to increased vaporization of the
variation of physical-chemical properties of seawater as a function of distillate in the next stage, resulting in an overestimation of the total
temperature and seawater salinity. The proposed MSF desalination plant product and performance ratio.
consists of 13 stages in the heat recovery section and 3 stages in the heat Based on the technical operation of MSF, the withdrawal of distillate
rejection section. The dynamic behaviour of MSF process was observed from each tray occurs through a channel that is located between the
under disturbance, which helped developing the appropriate start-up trays and the demister in parallel with the direction of the brine flow.
and shutdown procedures. This research attempts to develop a dynamic model for MSF system to
For MSF desalination with brine recirculation, Al-Fulaij [20] devel­ reflect this fact, which has not been mathematically presented in the
oped a dynamic model by including distillate flashing, non-condensable previous dynamic models available in the public domain. The inclusion
gases, and demister losses. The dynamic model developed was validated of this feature in the dynamic model of the MSF system is expected to
against actual industrial data of MSF desalination plant (21 stages in improve the accuracy of prediction of various parameters such as the
heat recovery section and 3 stages in heat rejection section) and showed product flowrate, performance ratio, etc. Furthermore, the model
a very good agreement. The dynamic simulation showed no change in development will implicitly consider the heat transfer equations, heat
the overall performance due to a variation of ±15 % of the cooling water balance and correlation between heat rejection and heat recovery. In
flow rate. However, the recycle flow rate has a considerable impact on addition, the relationship between thermodynamics and the energy ef­
the overall performance. ficiency in the MSF will be designed by considering the effect of some
Said et al. [21] optimised the design and operation parameters of influential parameters, such as the non-condensable gases, on the pro­
MSF via developing a simple dynamic model that considered distur­ cess performance. The efficiency of the model will be tested against real
bances in seawater temperature, and dynamic freshwater demand operational data collected from Shuaiba-South MSF seawater desalina­
throughout the day. The model considered the effect of non-condensable tion plant in Kuwait. Then, the model will be utilised to carry out a
gases towards the estimation of overall heat transfer coefficient for dynamic simulation for the selected desalination plant to analyse the
condensers. The model was then used to find the optimal number of performance indicators within a selected set of disturbances in the steam
stages giving the lowest freshwater production cost. temperature, seawater temperature, and the recycle brine. The model
Alsadaie and Mujtaba [22] presented an MSF dynamic model to code is written and solved using gPROMS software.
study the effect of venting system design on the performance of MSF
plant. The venting system was required to improve heat transfer rates by 2. Dynamic modelling of MSF process
eliminating non-condensable gases to the atmosphere or the evacuation
system. The model included Henry’s law to model the release of non- This section focuses on developing a comprehensive dynamic

3
H. Hasan et al. Desalination 553 (2023) 116453

modelling of the MSF. It should be noted that most of the model equa­
tions are gathered together from [14,20,25,27]. A set of assumptions are 2.1.1.2. Vapour space.
listed below:
• Mass balance
• Salt free in all stages for the distillate product [9,20], dM V
• Negligible heat of mixing, = VB + Vin + VD + NCGs − CVD + VOUT (10)
dt
• Accumulation salt in the condenser tubes is ignored,
• A lumped parameter process is used to upgrade the model,
• Water temperature and salt content are the main factors affecting the
non-equilibrium loss, • Non-condensable gases balance in vapour phase
• Constant mass cooling brine in condenser tubes, dY OUT
• Perfect mixing for the mass in the evaporators, MV = Vin Yin + NCGs(1 − YOut ) − YOut (VB + Vin + VD − CVD ) (11)
dt
• Neglecting blow through phenomenon,
• Equilibrium vapour and liquid in all stages,
• Spatial variation not measured.
• Enthalpy balance
dHV OUT MV
= (VB HVB ) + (NCGsHNCGs ) + (Vin + (1 − Yin )HVin )
2.1. Model equations dt
+ (Vin Yin HNCGsin ) + (VD HVout ) − (CVD HVD ) (12)
2.1.1. Single stage model − (VOUT YOut HNCGsout ) − (Vout + (1 − Yin )HVout ) − Q

2.1.1.1. Flash chamber.

• Vapour hold up
• Mass balance in the flashing chamber:
MV = ρV × VV (13)
dL
ρB Ap B = Bin − Bout − VB − NCGs (1)
dt

• Vapour volume space

• Salt balance in the flashing chamber: VV = Ap (LS − LB ) − AD LD − Vtubes (14)

dX Bout
ρB Ap LB = Bin XBin − XBout (Bin − VB − NCGs) (2)
dt
• Temperature of the Vapour space:
TV = TVB − ΔTDEM (15)
• Non-condensable gases balance in the brine:
dCBout
ρB Ap LB = Bin CBin − CBout ( Bin − VB ) − NCGs(1 − CBout ) (3)
dt • Vented vapour and NCGs flow rate
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )̅
VOut = KV ρV Pj − Pj+1 (16)

• Non-condensable gases stripping rate:

2.1.1.3. Product tray. Here, as mentioned in the introduction, the
NCGs = Bin ( CBout − CBe )γ (4) model of the distillate tray is slightly different from other published
works. Instead of entering the next stage, the distillate product of each
stage is withdrawn outside the stage and flows through a channel
• Enthalpy balance: located between the demister and the condensing tubes. This will be
mathematically considered in the following:
dH Bout
ρB Ap LB = Bin (HBin − HBout ) − VB (HVB − HBout ) − NCGs(HNCGs − HBout )
dt • Mass balance
(5)
dLD
The enthalpy for each stream is calculated based on ρ D AD = CVD − VD − Dout (17)
dt
H = CpΔT where Cp = f ( XB , T) (6)
• Enthalpy balance in distillate tray
• Temperature drop correlations: dH Bout
ρD AD L D = CVD (HVD − HDout ) − VD (HVout − HDout ) (18)
dt
TB = (TVB + BPE + NEA) (7)

• Temperature of the distillate

• Brine outlet flow rate
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅ VD HVout = CVD (HVD − HDout ) (19)
BOut = Cd Wst Hg 2ρb ΔP (8)
If the temperature of the vapour TV is equal to the temperature of
( ) ( )
ΔP = Pj − Pj+1 + gρB LB.j − LB.j+1 (9) distillate, TD, then the enthalpy of the released vapour HVD is equal to the

4
H. Hasan et al. Desalination 553 (2023) 116453

distillate enthalpy HDout in Eq. (19). Then there is no vapour flash from
dCBout
the distillate tray. ρB Ap LB = Bin CBin − FLast (CFout − CBout ) − Rec(CRec − CBout )
dt (35)
− NCGs(1 − CBout ) − CBout ( Bin − VB )
• Distillate outlet flow rate
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅
DOut = CC Apipe 2ρD ΔP (20)
( ) ( ) • Enthalpy balance
ΔP = Pj − Pj+1 + gρD LD.j –LD.j+1 (21)
dH Bout
ρB Ap LB = Bin (HBin − HBout ) − FLast (HBout − HFLast ) − VB (HVB − HBout )
2.1.1.4. Tubes bundle. dt
− NCGs(HNCGs − HBout )
• Mass balance: (36)
dM W
= WRin − WRout (22)
dt
Hint: In terms of heat rejection, WRin is replaced by WFin • Recycle brine concentrations of salt and non-condensable gases
RecXRec + XBOut (Flast − DTotal ) = Flast XFout + XBout (Rec − DTotal ) (37)
• Salt balance
XRin = XRout (23) RecCRec + CBOut (Flast − DTotal ) = Flast CFout + CBout (Rec − DTotal ) (38)

Hint: In terms of heat rejection, XRin is replaced by XFin 2.1.3. The brine heater model

• Enthalpy balance • Mass balance

dH Wout
MW = WRin (HWin − HWout ) + Q (24) dM BH
= WBHin − WBHout (39)
dt dt
Q = U AS LMTD (25)
• Salt balance
The log means temperature difference:
XBHin = XBHout (40)
TFout − TFin
LMTD = ( ) (26)
It should be noted that both Eqs. (23) and (40) simulate the con­
ln TTDD−− TTFout
Fin

centration of salt inside the tubes where there is no evaporation process

and thus the concentration of salt is constant.
AS = Nt × Lt × π × do (27)
• Enthalpy balance
• Mass hold up
dH BHout
MBH = WBHin (HBHin − HBHout ) + (UBH × ABH × LMTDBH ) (41)
MW = ρB Vtubes (28) dt

πdi2 UBH × ABH × LMTDBH = WSteam × λSteam (42)

Vtubes = Nt × Lt × (29)
4
Tb out − Tb in
LMTD = ( ) (43)
2.1.2. The last stage model ln TTSteam
Steam − Tb in
− Tb out

2.1.2.1. Flash chamber. 2.1.4. The multi-stage model

The equations mentioned above are applied for one single stage of
• Mass balance MSF process. To apply them in a series of stages (N), connection equa­
dLB tions should be used (as listed below) to connect the outlet of one stage,
ρ B Ap = Bin + FLast − Bout − Rec − VB − NCGs (30) j, to the next stage (j + 1). Thus, the outlet flowrate and properties of
dt
stage j are same as the inlet flowrate and properties of stage j + 1.
FLast = WFin − WCw (31)
• Brine stream
Rec = WRin (32)
Bin j = Bout j− 1, Bout j = Bin j+1, TBin j = TBout j− 1, TBout j = TBin j+1
XBin j = XBout j− 1, XBout j = XBin j+1, CBin j = CBout j− 1, CBout j = CBin j+1

• Salt balance
• Vapour and distillate streams
dX
ρB Ap LB Bout = Bin XBin − FLast (XBout − XFout ) + Rec(XBout − XRec )
dt (33) Vin j = Vout j− 1, Vout j = Vin j+1, Yin j = Yout j− 1, Yout j = Yin j+1
− XBout (Bin − VB − NCGs)
• Tube bundle streams
XFout = XFin (34)
WRin j = WRout j+1, WRout j = WRin j− 1, TFin j = TFout j+1, TFout j = TFin j− 1
XRin j = XRout j+1, XRout j = XRin j− 1, CRin j = CRout j+1, CRout j = CRin j− 1
The performance ratio (PR) of the process can be expressed as:
• Non-condensable gases balance in the brine

5
H. Hasan et al. Desalination 553 (2023) 116453

DTotal Table 1
PR = (44)
WSteam Actual plant data for Shuaiba South plant- Kuwait (Adapted from [30]).

The total distillate is the summation of the total distillate from the Variable Units Value

stages Number of stages – 25

∑N Distillate flowrate kg/s 247.13
DTotal = Dj (45) Flowrate of intake seawater kg/s 2425.64
j
Recycle brine flowrate (design parameter) kg/s 3162.93
Brine blowdown flow rate kg/s 533.8
o
3. Model validation Seawater Temperature
o
C 23.57
TBT C 91.62
Steam flowrate kg/s 30
All process models must be validated in order to assess their Steam Temperature o
C 105.2
robustness in predicting actual plant data. In this work, we considered a Make up flow rate kg/s 739
real large scale industrial MSF plant (described below) to validate our
proposed model.
the MSF plants are designed as a brine recirculation scheme due to its
less sensitivity to the variation of the seawater temperature and low
3.1. Brief background of Shuaiba-South MSF plant consumption of anti-fouling chemical additives. The actual data of the
Shuaiba South seawater desalination plant is shown in Table 1 [30].
Although the first desalination plant was installed in Kuwait in 1953,
the first MSF plant with two units installed in Shuaiba was 1965 with
unit capacity of 4546 m3/day. Other three units each with capacity of 3.2. Model validation
9092 m3/day were installed in 1968. Between 1971 and1975, Kuwait
established other MSF desalination plants in Shuaiba. The first plant The dynamic model presented in this work is coded and solved using
with one unit was installed in 1971 with a capacity of 22,730 m3/day. In gPROMS software. For validation purpose, the model is run until it
1972, another plant with four units was installed to provide a total ca­ reaches its steady state. Table 2 shows the model prediction results
pacity of 90,920 m3/day and later 1975, another plant was installed compared to the actual data from Shuaiba South plant. Table 2 also
with two units each with capacity of 22,730 m3/day [28]. shows the error between the model prediction and the actual data.
The process design considers both condensing and evaporating steps. Clearly the errors are marginal and within an acceptable accuracy.
The process is designed in three sections: brine heater, recovery section, Fig. 2 shows the model predictions of the brine outlet temperature
and heat rejection section. The boiler plays an essential role in the for 25 stages against the actual plant data. As it can be noted, there is a
process as it supplies the heat to the evaporators. In addition, the brine is great match between the actual and simulated values. The small dif­
recycled from the rejected section to the recovery section. Fig. 1 shows a ference can be attributed to the variations in the external conditions and
schematic diagram of the MSF plant with a recycled brine. measurement errors that might lead to such differences. The same re­
The rejected and recovery sections are designed as a series of stages; sults can be obtained for the distillate temperature. Fig. 3 shows the
each stage contains condenser and flash chamber. The seawater enters
the last stage through the condenser tube in the heat rejection section,
Table 2
which permits the vapour to heat up the flow water. In the next step, the
Comparison between the model predictions against actual plant data of Shuiaba
stream leaves from the heat rejection section and divided in two streams South plant- Kuwait.
where the first stream feeds into the deaerator unit and then to the
Variable Units Actual data Predicted data Error %
chamber pool of the last stage as make-up. The second stream is rejected
back to the sea. The recirculation brine stream is designed to flow out Distillate flowrate kg/s 247.13 247.34 0.085
from the last chamber of the heat rejection section and is fed to the last Brine blowdown flow rate kg/s 533.8 532.9 − 0.17
o
TBT C 91.62 91.61 − 0.01
stage in the heat recovery section. In the recovery section, the seawater Steam flowrate kg/s 30 30.46 1.53
temperature will increase gradually as it flows through the tubes inside Brine blowdown temp. o
C 39.8 42.37 6.46
the stages [9]. Furthermore, the brine is separated in the brine pool to Make up temp. o
C 39.5 42.68 8.05
flow out as blow down to be rejected back to the sea [29]. In fact, most of

Heating Section Heat Recovery Section Heating Rejection Section

Cooling seawater discharge
Condensate Tubes
Distillate
Trays Intake
seawater
Demister
Distillate
Heating Product
Steam

Brine Heater Brine Blow-

Condensate Seawater down
Recycle Brine make-up

Fig. 1. A schematic diagram of the MSF-Recycle brine.

6
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 2. Model and plant results for the brine outlet temperature for each stage.

Fig. 3. Model and plant results for the distillate outlet temperature for each stage.

model predictions of the distillate temperature in comparison to the indicates that the obtained results of the new model has close match to
actual data. Indeed, a good agreement is obtained between them. the actual data comparing to the results obtained from other models
To test the accuracy of the developed model in this study that con­ developed. Al-Shayji [31] and Alasfour and Abdulrahim [32] and many
siders the withdrawal of distillate from each tray occurs through a others (are not presented here) assumed that the distillate flow from one
channel that is located between the trays and the demister in parallel stage enters directly to the next stage. This assumption, as elucidated
with the direction of the brine flow, the model predictions have been earlier, is less accurate and would lead to overestimation of the total
tested against actual data of MSF plant and predictions of other models distillate flowrate. The distillate flows in the next stage is in higher
collected from the open literature. Table 3 shows a comparison between temperature enters to less pressure stage and therefore, a small amount
the prediction of the new model, actual plant data and simulation data of vapour evaporates releasing more heat to the recycled seawater.
from Al-Shayji [31] and Alasfour and Abdulrahim [32]. Both Al-Shayji Assuming this additional heat is accessible will cause the production rate
[31] and Alasfour and Abdulrahim [32] used the same plant data of to be overestimated because it is not there in the actual plant.
Azzour desalination plant located in Kuwait and designed with a brine To present this argument in a clearer way, the distillate equation of
recirculation mode of 24 stages. Interestingly, the comparison of Table 3 the present model is rewritten to be identical to the distillation equation

Table 3
Comparison between the model results and other simulated results of different plant.
Variable Units Actual data Al-Shayji [31] Alasfour and Abdulrahim [32] This model

Distillate flowrate kg/s 313.33 322.17 376.8 317.6

Performance ratio – 8.0 7.76 8.62 8.013

7
H. Hasan et al. Desalination 553 (2023) 116453

Table 4 These variables are the brine levels in some stages, brine heater inlet and
Comparing new feature in the distillation equation to the old equation. outlet temperatures and steam flowrate.
Variable Units This model using This model using old Fig. 4 shows the start-up behaviour of the brine level in odd number
new feature distillation equation of stages. In the absence of control system (MSF plant without control),
Distillate the brine levels in most stages reach steady state in less than 400 s.
kg/s 247.34 254.27
flowrate However, in the last five stages, the brine levels take longer time to settle
Performance
– 8.11 8.44
down (the level of the brine becomes stable). Particularly, the brine level
ratio takes longer time (around 1400 s) to reach its steady state in the last
stage. Note, the last stage has two streams in (brine from the previous
stage and makeup feed) and two streams out (recycle brine and blow
of the previous models. Then, the model was run to predict the distillate down streams). This is compared to other stages of one inlet stream and
flow rate and the performance ratio. Accordingly, this enables to make a one outlet stream. The brine level in the last stage is the key parameter
comparison between the model results of distillate flowrate and per­ that is responsible to control the brine levels in all stages. Thus, the
formance ratio for the two cases as represented in Table 4. Clearly, the controller system is installed in the last stage considering the brine level
results of Table 4 shows that using the same distillation equation of other as the controlled variable and the brine outlet flowrate as the manipu­
colleagues in our model results in higher distillate flowrate and perfor­ lated variable.
mance ratio. According to the results of Tables 3 and 4, it can be said that Fig. 5 shows the inlet and outlet temperature of the brine heater as
the MSF model with the new equation of the distillate flow rate provides well as the required steam flowrate to maintain the outlet temperature at
accurate and close results to the actual data. a desirable value. The start-up behaviour of both the inlet and outlet
temperature of the brine heater is monitored to observe their variation
4. Dynamic simulation of the Shuaiba MSF process with the steam flowrate. There is a correlation between the outlet
temperature and the required heating steam. As it can be seen, both the
The fundamental goal of any dynamic simulation is to understand brine heater temperature and the steam flowrate require the same
the process dynamics so that appropriate control systems can be amount of time to reach its steady state. It is important to mention that
designed for dealing with start-up and shut down in addition to trou­ the brine heater parameters take around 10,000 s to be stable as the
bleshooting [17]. Dynamic simulation can also be used to study the ef­ expected time in the absence of a control system.
fect of any faults that may occur in MSF process due to the failure of the For the convenience of the reader, we have provided the dynamic
process equipment such as pumps, heaters and controllers. More pre­ behaviours of distillate flowrate, salt concentrations, temperature pro­
cisely, the most frequent issues that may arise in desalination facilities file and pressure profile for the whole stages along the operational time
are failures of the distillate or brine pumps. The performance of the plant in the Appendix A.
declines as a result of blow-through events driven on by a decrease in the
brine flow rate (pressure increase and no vapour release). Another 5. Controlling MSF Process
reason for plant failure is a decrease in the amount of externally
generated steam since it reduces the amount of heat supply. Addition­ In this section, the validated model is used to highlight the process
ally, because the winter seawater temperature is regulated to be warmer, dynamics for a large scale MSF plant such as Shuaiba Plant which is
a breakdown in the control system could cause the intake seawater essential to design an appropriate control system.
temperature to decline. Due to the abundance of spare utilities and
equipment, these kinds of breakdowns are not anticipated to endure for
an extended period of time. Thus, to understand the dynamic behaviour 5.1. Control loops in MSF process
of MSF process, the dynamic simulation of the Shuaiba MSF plant is
carried out and the start-up of some important variables are observed. MSF desalination is a very complex process [33] which requires
several numbers of controlling loops to maintain the operation stable at

0.8

0.7

0.6

0.5
Brine Level (m)

0.4

0.3

0.2

0.1

0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (s)
1 3 5 7 9 11 13 15 17 19 21 23 25

Fig. 4. Start-up behaviour of the brine levels in odd number of the flash stages.

8
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 5. Start-up behaviour of the steam consumption and inlet and outlet temperatures of the brine heater.

Table 5
Most known MSF controller loops [16,33].
No. Type of Controlled variable Manipulated variable Objective
controller

Circulation flowrate of the warm water It has a great effect on the heat transferred to the rejection section and
1 PID Temperature of the intake seawater
rejected to the sea. in the performance ratio.
Temperature of the steam entering Flowrate of the steam condensate to the It dictates the latent heat of the steam and changes the steam from
2 PID
the brine heater spray system superheated to saturated steam.
Control valve position of the recycle
3 PI Flowrate of the brine recycle It is useful to keep stages brine levels constant.
stream
Control valve position of the rejected
4 PI Flowrate of the intake seawater Control the velocity of the seawater in rejection section tubes.
seawater
Control valve position of the distillate This loop ensures the level of the distillate does not overflow and avoid
5 PI Distillate level in the last stage
product flowrate pump problems.
Top Brine Temperature of the brine Flowrate of the heating steam to the
6 PID It has a direct impact on the production rate and performance ratio.
heater brine heater
Control valve position of the brine Control the level of the last stage adjusts the level in all proceeding
7 PI Brine level in the last stage
blowdown flowrate stages and avoid blow-through phenomenon.
Control valve position of the This loop ensures that the condensate tubes are not submerged in
8 PI Condensate level in the brine heater
condensate steam liquid.
Flowrate of the make-up stream Control valve position of the make-up
9 PI It is useful to keep the concentration of the brine constant.
entering the last stage stream

a minimum cost of energy. The conventional Proportional Integral De­ dynamic model. However, it is important to mention that all the control
rivative Controller (PID) is a well-known type of control that has been loops are turned off and the process is left uncontrolled. Accordingly, it
widely used in MSF industry due to its simplicity and ease of use [34]. becomes possible to observe the effect of the disturbance.
According to Alatiqi et al. [33], a well-stabilized and regulated MSF
process can be achieved by employing 13 controlling loops. Maniar and 5.2.1. Effect of inlet seawater temperature
Deshpande [16] and Ismail [35] reported that a typical MSF process MSF desalination plants face a big challenge in the fluctuation of the
could be efficiently controlled using nine controlling loops. However, weather, which significantly influences the seawater temperature.
due to the reliability and robust of MSF technology, Al-Gobaisi et al. Therefore, this section will evaluate the dynamic behaviours of the brine
[34] reported that most of the existed MSF plants can be controlled flow rate, distillate flow rate, brine level and performance ratio in a
without any problems by 4 to 6 primary loops. Table 5 presents the most dynamic simulation based on two disturbances of different tempera­
known controlling loops in MSF process with their controlled and tures, i.e., lower and upper the base case temperature as shown in
manipulated variables. It is fair to admit that all controller loops in MSF Table 6. The chosen temperatures are based on winter and summer
process are designed as a single input/single output (SISO). seasons.

5.2. Faults examination in MSF process

This section investigates the consequences of a few settings under the Table 6
assumption that some equipment will fail. To implement the failure, a The inlet seawater temperatures for the dynamic model with two disturbances
disturbance is introduced in three important variables that have the against the base case.
potential to influence the plant performance indicators. The inlet Base case Disturbance 1 Disturbance 2
seawater temperature, the steam temperature, and the recycle brine
23.6 C
◦
18 C
◦
28 ◦ C
flow rate are chosen to characterise the dynamic behaviour using the

9
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 6. Dynamic response of brine flow rate of 10th stage for the effect of inlet seawater temperature.

5.2.1.1. Brine flow rate. Fig. 6 shows the time-dependent behaviour of temperature of the water inside the tubes will result in an increase in the
the brine flow rate in the 10th stage as a response to varying the inlet vapour condensate rate. Since the tubes water in the last stage is colder
seawater temperature (Table 6). Fig. 6 shows that the model run until than stage 10th and even stage 20th, higher condensate rate is expected
having a steady state conditions for each inlet seawater temperature. Up in the last stage. The condensation process of the vapour leads to low
to t = 40,000 s, the inlet seawater temperature was 23.6 ◦ C and at t = pressure inside the flash chamber and consequently, more vapour is
40,000 s, the inlet seawater temperature is changed to 18 ◦ C (the first released from the brine resulting in sudden drop in the brine flowrate.
disturbance) and at t = 60,000 s, the inlet seawater temperature is Vice versa occurs in the second disturbance where the temperature of
changed to 28 ◦ C (the second disturbance). Fig. 6 shows that the brine the water inside the tubes increases to 28 ◦ C.
flow rate of stage 10th drops to 2943 kg/s after introducing the first Figs. 6, 7, and 8 indicate that the last stage’s brine flow rate takes
disturbance at 18 ◦ C. However, Figs. 7 and 8 specify a higher sensitivity longer time to settle than the 10th and 20th stages. This might be
to the change in the inlet temperature for the 20th and 25th stages ascribed to the higher brine flow rate in the preceding stages compared
compared to the 10th stage. to the last stage of the lowest brine flow rate.
In the occasion of the second disturbance with a temperature of
28 ◦ C, the brine flow rate drops and then immediately increased 5.2.1.2. Distillate flow rate. Fig. 9 shows the start-up period (t = 0 to
dramatically as shown in Fig. 6. Fig. 7 shows a similar response but with 4000 s) of the condensate flow rate in three specific stages (1st, 12th,
a bigger jump in the brine flow rate in the 20th stage. More importantly, and 25th stages). This period shows the transient state before attaining
Figs. 7 and 8 show a similar behaviour after introducing the first and the steady state conditions. Fig. 9 shows the variation of the condensate
second disturbances. However, Fig. 8 shows a great change in both first rate in the 1st, 12th and the last stage after utilising two different dis­
and second disturbances compared to the 10th and 20th stages. The turbances in the inlet seawater temperature (Table 6).
brine flow rate drops sharply in the last stage in the winter, whereas a As it can be seen in Fig. 9, an increase in the product flow rate in
noticeable growth in the brine flow rate is noticed in the summer. Thus, winter compared to a noticeable reduction in summer. The variation of
it can be stated that the brine flow rate drops in the winter, whereas the seawater temperature has direct effect on the condensate rate in thermal
brine flow rate increases in the summer. Basically, lowering the

Fig. 7. Dynamic response of brine flow rate of 20th stage for the effect of inlet seawater temperature.

10
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 8. Dynamic response of brine flow rate of 25th stage for the effect of inlet seawater temperature.

Fig. 9. Dynamic responses of the distillate flow rate of 1st, 12th, and 25th stages for the effect of inlet seawater temperature.

desalination plants and hence the size and capacity of the plant are causes a gradual increase of condensate rate. Also, it should be noted
practically determined based on the summer conditions. More specif­ that the condensate rate requires different periods of time to be settled at
ically, the rise of the tube condenser Logarithmic Mean Different Tem­ fixed values for the different studied stages. For instance, the 1st stage
perature (LMTD) throughout the stages can interpret the increase of requires 4300 s to have a fixed condensate rate compared to 2600 s for
productivity in winter as already confirmed by Tanvir and Mujtaba [36]. the 12th stage. In other words, the settling time of condensate rate for
The decrease of heat removed in summer can interpret the reduction of the 1st stage is longer than the settling time of the 12th stage.
condensate rate at higher seawater temperature. Another reason of the
reduced condensate rate at a high seawater temperature is that 5.2.1.3. Top brine temperature (TBT) and brine level. Top brine tem­
increasing seawater temeprature would cause an increase in the make- perature (TBT) can be defined as the temperature of the recirculation
up flow and brine level in each stage which laterlly reduces the brine after it is heated by the low-pressure steam in the brine heater. TBT
flashing efficiency. ranges between 90 ◦ C to 120 ◦ C for MSF desalination plants [37]. A high
Fig. 9 introduces two distinct condensate rate in behaviours of the TBT is mostly a pointer of high steam consumption and the reverse is
studied stages in relation to operational time for two seawater temper­ true. A high propensity of scaling and accelerates corrosion of metal
ature variations. Clearly, these behaviours were illustrated as a gradual surfaces can also accompany the high TBT [38].
increase or decrease and a spontenous increase or decrease of conden­ Fig. 10 shows that the system can be settled in less than 10000s.
sate rate to reach a stable value after introducing the disturbance. Fig. 10 illustrates that the top brine temperature decreases in the first
Specifcally, the 1st stage has characterised by the gradual increase disturbance of decreasing the seawater temperature from 23.6 ◦ C to
compared to the spontenous behaviour of the 25th stage (last stage). The 18 ◦ C, which will also decrease the brine level in the first and last stages
spontenous response of the last stage is attributed to the considerable as demonstrated in Figs. 11 and 12, respectively. Although, the
difference of inlet seawater and brine temperature which resulted in a condensate rate increases as noticed in Fig. 9, the decrease in the TBT
sharp increase in condensate rate. This is in comparison to the 1st stage below 90 ◦ C is not preferred, as it would cause an incomplete extraction
that encountered with a lower temperature difference of the recycled of non-condensable gases [39]. On the contrast, the TBT increases in the
brine from the heating rejection system and elevated vapour which summer season due to an increase in seawater temperature from 18 ◦ C to

11
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 10. Dynamic response of the top brine temperature for the effect of inlet seawater temperature.

Fig. 11. Dynamic response of the brine level of 1st stage for the effect of inlet seawater temperature.

28 ◦ C, which elevates the brine levels in the first and last stages. It can considerable energy consumption at a high cost. Therefore, studying the
also be noted that increasing the TBT would improve the performance dynamic behaviour of the brine flow rate, distillate flow rate, brine level
ratio (Fig. 13) as will be described in the next section. and performance ratio is important against the variation of steam tem­
perature. The examination of two disturbances in steam temperature
5.2.1.4. Performance ratio (PR). The performance ratio (PR) is one of will be carried out based on the base case of 105.2 ◦ C (Table 7). Table 7
the most highlighted metrics of MSF. PR can be defined as the obtainable presents the base case of steam temperature which runs for 40,000 s
distillate due to condensing one kilogram of the heated steam in the before introducing two disturbances of steam temperature, i.e., lower
brine heater. The performance ratio decreases with low temperature in and upper the base case steam temperature.
winter (Fig. 13) compared to a significant jump in the summer season.
Reducing the TBT due to reducing the seawater temperature causes a 5.2.2.1. Brine flow rate. Figs. 14, 15 and 16 depict the dynamic be­
reduction in PR due to an increase in the latent heat of vaporization of haviours of brine flow rate in the 10th, 18th, and 25th stages as a
water and an increase in the LMTD [40]. response of two disturbances in steam temperature (Table 7). As the first
Fig. 13 shows a clear variation of PR due to varying the inlet seawater disturbance (low steam temperature) is introduced, the brine flow rate
temperature. Furthermore, it can be noted that reducing the product immediately drops with a significant drop in the last stage followed by
distillate has accompanied an increase of PR. This is due to the necessity an increase of brine flow rate after a transient period. However, the
of consuming a higher amount of steam flow rate. brine flow rate decreases after implementing the highest steam tem­
perature with a significant decrease in the last stage. Thus, it can be
5.2.2. Effect of steam temperature stated that the last stage is the most effected than the first and middle
Steam temperature is one of the main variables that consume a stages due to varying the steam temperature.

12
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 12. Dynamic response of the brine level of 25th stage for the effect of inlet seawater temperature.

Fig. 13. Dynamic response of the performance ratio for the effect of inlet seawater temperature.

5.2.2.2. Distillate flow rate. Figs. 17, 18 and 19 illustrate the responses
Table 7
of distillate flow rate of the 1st, 12th, and 25th stages for the effect of
The chosen steam temperature for the dynamic study with two disturbances.
two disturbances of steam temperature provided in Table 7. The first
Base case Disturbance 1 Disturbance 2 disturbance of lowering the steam temperature causes a reduction of
105.2 100.2 120 product flow rate. However, increasing the steam temperature in the
second disturbance causes an elevated product flow rate for the whole
tested stages.
A precise estimation of Fig. 14 shows that the transient periods of the
first and second disturbances of brine flow rate in the 10th stage have 5.2.2.3. Top brine temperature (TBT) and brine level. Fig. 20 shows the
lasted around 400 s and 800 s, respectively. However, the first and dynamic response of TBT for the effect of the steam temperature of two
second disturbances of brine flow rate in the 25th stage (Fig. 16) have disturbances against the base case of 105.2 ◦ C. In low steam temperature
lasted around 1500 s and 3000 s, respectively. scenario, the TBT decreases with a low change, whereas in high steam
temperature, the TBT sharply increases. Increasing TBT causes an

13
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 14. Dynamic response of brine flow rate of 10th stage for the effect of steam temperature.

Fig. 15. Dynamic response of brine flow rate of 18th stage for the effect of steam temperature.

increase in the product flow rate (Figs. 17, 18, and 19) and lowering the temperature has a limited influence on the PR. This is considerably
brine level at a high steam temperature (Figs. 21 and 22). Choi [29] different from the influence of altering the inlet seawater temperature
confirmed an increase in the product flow rate as a response to on PR, which resulted in a considerable shift in PR (Fig. 13).
increasing the TBT that introduced after increasing the steam temper­
ature against the base case. 5.2.3. Effect of recycle brine flow rate
Figs. 21 and 22 present the variation of brine level in the 1st and 25th This section describes the effect of recycle brine flow rate on the
stages as a response to achieving two disturbances of the steam tem­ overall performance indicators. Table 8 shows the transient values for
perature. The first disturbance of lowering the steam temperature below the recycle brine flow rate for three disturbances. The first scenario is the
the base case introduces a sharp increase in the brine level of the 1st and disturbance of an increase of 10 % in recycle brine feed flow rate against
25th stages. However, a sudden decrease of brine level in the selected the base case. The second scenario is a disturbance with decreasing the
stages is noticed after increasing the steam temperature. recycle brine flow rate by 5 % against the base case. The third scenario is
a disturbance with increasing the recycle brine flow rate by 4 % against
5.2.2.4. Performance ratio (PR). Fig. 23 presents the dynamic response the base case.
of the plant performance ratio for the effect of the steam temperature of
two disturbances of steam temperature (Table 7) against the base case. 5.2.3.1. Brine flow rate. Figs. 24, 25, and 26 illustrate the dynamic re­
Fig. 23 illustrates insignificant change of the performance ratio in the sponses of the recycle brine flow rate on the brine flow rate of the 10th,
response of first and second disturbances. In other words, the steam 18th, and last stage (25th). The brine flow rate is significantly increased

14
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 16. Dynamic response of brine flow rate of 25th stage for the effect of steam temperature.

Fig. 17. Dynamic response of the distillate flow rate of 1st stage for the effect of steam temperature.

after employing the 10 % increase in the recycle brine flow rate and vice the base case. This is followed by an improvement of product flow rate
versa in the second disturbance (5 % decrease). However, the last stage due to increasing the recycle brine flow rate by 4 % of the base case
has introduced an opposite path of dynamic response compared to the (Fig. 27). The same dynamic responses can be noticed in the 12th stage
10th, and 18th stages as shown in Fig. 24. This is due to realising that the (Fig. 28). Fig. 29 illustrates the dynamic responses for the last stage,
brine is already recycled from the last stage that would introduce which is significantly showed that the product increases in a higher
different behaviour than the proceeding stages. intensity than the 1st and 12th stages.
Figs. 27, 28 and 29 also ascertain the fact that the last stage requires a
5.2.3.2. Distillate flow rate and brine level. The effect of the recycle brine longer time to be settled and attain the steady state conditions compared
flow rate on the distillate flow rate has the same dynamic behaviour of to the first and middle stages.
the brine flow rate. Fig. 27 presents the dynamic distillate flow rate for Increasing the distillate flow rate due to increasing the recycle brine
the first stage as a response to the three disturbances of recycle brine flow rate can be attributed to reducing the brine level in the flash
flow rate (Table 8). Firstly, the product is increased due to increasing the chamber of the middle and the last stages (Fig. 31) that improves the
recycle brine flow rate by 10 %. In the meanwhile, increasing the recycle condensation rate and distillate flow rate. However, this is not the case
brine flow rate can reduce the TBT and the brine level, which in turn for the first stage (Fig. 30). Figs. 30 and 31 illustrate dynamic responses
enhances the distillate flow rate. Whereas a decrease of product flow for the brine level in the first and the last stages. It clearly shows the
rate is a consequence of decreasing the recycle brine flow rate by 5 % of difference between the two graphs as the first disturbance introduced an

15
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 18. Dynamic response of the distillate flow rate of 12th stage for the effect of steam temperature.

Fig. 19. Dynamic response of the distillate flow rate of 25th stage for the effect of steam temperature.

Fig. 20. Dynamic response of the top brine temperature for the effect of steam temperature.

16
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 21. Dynamic response of the brine level of 1st stage for the effect of steam temperature.

Fig. 22. Dynamic response of the brine level of 25th stage for the effect of steam temperature.

increase of distillate in the first stage compared to the last stage of recycle brine flow rate can affect the PR compared to insignificant
sharply decrease. In the second disturbance (decrease the recycle brine change of PR due to varying the steam temperature.
flow rate by 5 % of the base case), the brine level in the first stage has
significantly decreased. However, the third disturbance of decreasing 6. Conclusions
the recycle brine flow rate causes a slight increase of brine level
(Fig. 30). The rise in population has prompted the development of viable
Fig. 31 shows a quite different dynamic behaviour of brine level in seawater desalination systems in order to avoid any water shortages. In
the last stage compared to the 1st stage of Fig. 30. This can be attributed comparison to other technologies, MSF desalination is the most cost-
to the fact that the last stage has the lowest hold-up compared to the effective and well-known production technology. Nonetheless, there is
preceding stages. still a lack of understanding of the process dynamics in terms of expected
The dynamic response of performance ratio towards the two dis­ inlet parameter disturbances. Specifically, the accurate modelling of
turbances of recycle brine flow rate are illustrated in Fig. 32. This shows product mass flowrate of each stage to enter the next stage should be
decrease/increase/decrease in the performance ratio due to increase/ considered throughout a concise dynamic model. This study aimed to
decrease/increase of recycle brine flow rate based on the base case. address this issue by developing a comprehensive dynamic simulation-
Specifically, increasing the recycle brine flow rate consumes a high based model.
quantity of steam flow rate that retards the performance ratio of the The developed model for the MSF with brine recirculation desali­
plant despite the increase of distillate flow rate (Figs. 27, 28, and 29). nation process was successfully validated against actual data of Shuaiba
Furthermore, the increase of recycle brine flow rate causes a reduction of South desalination plant. The inclusion on the new feature in the dy­
TBT that causes a reduction of PR. It can also be noted that varying the namic model of this study has clearly presented marginal errors between

17
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 23. Dynamic response of the performance ratio for the effect of steam temperature.

Table 8
The disturbance cases for the recycle brine flow rate.
Base case Disturbance 10 % Disturbance 5 % Disturbance 4 %

3089 3397.9 2934.55 3212.56

Fig. 24. Dynamic response of brine flow rate in the 10th stage for the effect of recycle brine flow rate.

the model predictions of performance indicators of MSF if compared to • The last stage requires a longer time to settle for the tested param­
the plant data and the predictions of other previous models. eters such as brine flow rate, distillate flow rate, and brine level,
The investigation of the MSF desalination dynamic responses under compared to the preceding stages.
disturbances in several operating parameters was conducted in this • Increasing the recycle brine flow rate causes a reduction in the brine
study. level and increase of product flow rate in the middle and last stages
The main findings of this study are as follows. compared to the first stage.
• Varying the inlet seawater temperature and recycle brine flow rate
• The brine flow rate drops in the winter, whereas the brine flow rate has a significant influence on performance ratio of the plant
increases in the summer. This results in more freshwater production compared to varying the steam temperature.
in winter than in summer while water demand can be quite opposite.
Thus, this study will facilitate design of appropriate control actions The dynamic model of this study can be helpful in terms of trou­
so that the seasonal freshwater demand can be met. bleshooting, operation, control and designing MSF seawater desalina­
tion plants.

18
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 25. Dynamic response of brine flow rate in the 18th stage for the effect of recycle brine flow rate.

Fig. 26. Dynamic response of brine flow rate in the 25th stage for the effect of recycle brine flow rate.

Fig. 27. Dynamic response of the distillate flow rate for 1st stage for the effect of recycle brine flow rate.

19
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 28. Dynamic response of the distillate flow rate for 12st stage for the effect of recycle brine flow rate.

Fig. 29. Dynamic response of the distillate flow rate for 25st stage for the effect of recycle brine flow rate.

Fig. 30. Dynamic response of the brine level of 1st stage for the effect of the recycle brine flow rate.

20
H. Hasan et al. Desalination 553 (2023) 116453

Fig. 31. Dynamic response of the brine level of 25st stage for the effect of the recycle brine flow rate.

Fig. 32. Dynamic response of the performance ratio for the effect of recycle brine flow rate.

Author statement Data availability

All authors contributed equally in all aspects of this paper. No data was used for the research described in the article.

Declaration of competing interest

There is no conflict of interests.

21
H. Hasan et al. Desalination 553 (2023) 116453

Appendix A

Fig. A1. Distillate flowrate of the whole stages along the operational time.

Fig. A2. Salt concentration of the whole stages along the operational time.

22
H. Hasan et al. Desalination 553 (2023) 116453

Fig. A3. Brine temperature of the whole stages along the operational time.

Fig. A4. Pressure of the whole stages along the operational time.

23
H. Hasan et al. Desalination 553 (2023) 116453

References [21] S.A. Said, M. Emtir, I.M. Mujtaba, Flexible design and operation of multi-stage flash
(MSF) desalination process subject to variable fouling and variable freshwater
demand, Processes 1 (3) (2013) 279–295.
[1] C. Ding, M. Yi, B. Liu, C. Han, X. Yu, Y. Wang, Forward osmosis-extraction hybrid
[22] S.M. Alsadaie, I.M. Mujtaba, Modelling and simulations of MSF desalination plant:
process for resource recovery from dye wastewater, J. Membr. Sci. 612 (2020),
the effect of venting system design for non-condensable gases, Chem. Eng. Trans.
118376.
39 (2014) 1615–1620.
[2] M.W. Shahzad, M. Burhan, N. Ghaffour, K.C. Ng, A multi evaporator desalination
[23] M.T. Sowgath, I.M. Mujtaba, Meeting the fixed water demand of MSF desalination
system operated with thermocline energy for future sustainability, Desalination
using scheduling in gPROMS, Chem. Eng. Trans. 45 (2015) 6.
435 (2018) 268–277.
[24] J. Lappalainen, T. Korvola, V. Alopaeus, Modelling and dynamic simulation of a
[3] M.A. Al-Obaidi, K.H. Rasn, S.H. Aladhwani, M. Kadhom, I.M. Mujtaba, Flexible
large MSF plant using local phase equilibrium and simultaneous mass, momentum,
design and operation of multi-stage reverse osmosis desalination process for
and energy solver, Comput. Chem. Eng. 97 (2017) 242–258.
producing different grades of water with maintenance and cleaning opportunity,
[25] Q.Y. Huang, A.P. Jiang, H.Y. Zhang, J. Wang, Y.D. Xia, L. He, Dynamic modelling
Chem. Eng. Res. Des. 182 (2022) 525–543.
and simulation of a multistage flash desalination system, Processes 9 (3) (2021)
[4] R. Lokk, S.M. Alsadaie, I.M. Mujtaba, Dynamic simulation of once-through
522.
multistage flash (MSF-OT) desalination process: effect of seawater temperature on
[26] M. Rosso, A. Beltramini, M. Mazzotti, M. Morbidelli, Modeling multistage flash
the fouling mechanism in the heat exchangers, Comput. Chem. Eng. 155 (2021),
desalination plants, Desalination 108 (1997) 365–374.
107515.
[27] P.J. Thomas, S. Bhattacharyya, A. Patra, G.P. Rao, Steady state and dynamic
[5] A. Bennett, 50th anniversary: desalination: 50 years of progress, Filtration+
simulation of multi-stage flash desalination plants: a case study, Comput. Chem.
Separation 50 (3) (2013) 32–39.
Eng. 22 (10) (1998) 1515–1529.
[6] I.S. Al-Mutaz, A comparative study of RO and MSF desalination plants,
[28] M. Al-bahou, Z. Al-Rakaf, H. Zaki, H. Ettouney, Desalination experience in Kuwait,
Desalination 106 (1–3) (1996) 99–106.
Desalination 204 (1–3) (2007) 403–415.
[7] M.W. Shahzad, M. Burhan, K.C. Ng, A standard primary energy approach for
[29] S.H. Choi, On the brine re-utilization of a multi-stage flashing (MSF) desalination
comparing desalination processes, npj Clean Water 2 (1) (2019) 1–7.
plant, Desalination 398 (2016) 64–76.
[8] M.W. Shahzad, M. Burhan, L. Ang, K.C. Ng, Energy-water-environment nexus
[30] H. Hasan, Dynamic Modelling And Simulation of Multi Stage Flash Desalination
underpinning future desalination sustainability, Desalination 413 (2017) 52–64.
Process, MSc Thesis, University of Bradford, UK, 2019.
[9] H.T. El-Dessouky, H.M. Ettouney, Fundamentals of Salt Water Desalination,
[31] K.A. Al-Shayji, Modeling, Simulation, And Optimization of Large-scale Commercial
Elsevier, 2002.
Desalination Plants, Virginia Polytechnic Institute and State University, 1998.
[10] I.S. Al-Mutaz, Msf challenges and survivals, Desalin. Water Treat. 177 (2020)
Doctoral dissertation.
14–22.
[32] F.N. Alasfour, H.K. Abdulrahim, Rigorous steady state modeling of MSF-BR
[11] A. Gambier, M. Fertig, E. Badreddin, Hybrid modelling for supervisory control
desalination system, Desalin. Water Treat. 1 (1–3) (2009) 259–276.
purposes for the brine heater of a multi stage flash desalination plant, in:
[33] I. Alatiqi, H. Ettouney, H. El-Dessouky, Process control in water desalination
Proceedings of the 2002 American Control Conference (IEEE Cat. No. CH37301)
industry: an overview, Desalination 126 (1–3) (1999) 15–32.
Vol. 6, IEEE, 2002, May, pp. 5060–5065.
[34] D.M.K. Al-Gobaisi, A. Hassan, G.P. Rao, A. Sattar, A. Woldai, R. Borsani, Towards
[12] M.S. Tanvir, I.M. Mujtaba, Optimisation of design and operation of MSF
improved automation for desalination processes,part I: advanced control,
desalination process using MINLP technique in gPROMS, Desalination 222 (1–3)
Desalination 97 (1–3) (1994) 469–506.
(2008) 419–430.
[35] A. Ismail, Fuzzy model reference learning control of multi-stage flash desalination
[13] S.M. Alsadaie, I.M. Mujtaba, Dynamic modelling of heat exchanger fouling in
plants, Desalination 116 (2–3) (1998) 157–164.
multistage flash (MSF) desalination, Desalination 409 (2017) 47–65.
[36] M.S. Tanvir, I.M. Mujtaba, Modelling and simulation of MSF desalination process
[14] A.M. Helal, M.S. Medani, M.A. Soliman, J.R. Flower, A tridiagonal matrix model
using gPROMS and neural network based physical property correlation, in:
for multistage flash desalination plants, Comput. Chem. Eng. 10 (1986) 327–342.
Computer Aided Chemical Engineering Vol. 21, Elsevier, 2006, pp. 315–320.
[15] M. Mazzotti, M. Rosso, A. Beltramini, M. Morbidelli, Dynamic modelling of
[37] O.A. Hamed, M.A.K. Al-Sofi, M. Imam, G.M. Mustafa, K. Bamardouf, H. Al-Washmi,
multistage flash desalination plants, Desalination 127 (3) (2000) 207–218.
Simulation of multistage flash desalination process, Desalination 134 (2001)
[16] V.M. Maniar, P.B. Deshpande, Advanced controls for multi-stage flash (MSF)
195–203.
desalination plant optimization, J. Process Control 6 (1) (1996) 49–66.
[38] S.F. Mussati, P.A. Aguirre, N.J. Scenna, Improving the efficiency of the MSF once
[17] A. Gambler, E. Badreddin, Dynamic modelling of MSF plants for automatic control
through (MSF-OT) and MSF-mixer (MSF-M) evaporators, Desalination 166 (2004)
and simulation purposes: a survey, Desalination 166 (2004) 191–204.
141–151.
[18] Md Tanvir Sowgath, Neural Network Based Hybrid Modelling And MINLP Based
[39] K.A. Al-shayji, S. Al-wadyei, A. Elkamel, Modelling and optimization of a
Optimisation of MSF Desalination Process Within gPROMS, University of Bradford,
multistage flash desalination process, Eng. Optim. 37 (2005) 591–607.
2007. PhD dissertation.
[40] S.A. Abdul-Wahab, K.V. Reddy, M.A. Al-Weshahi, S. Al-Hatmi, Y.M. Tajeldin,
[19] Salih M.M. Alsadaie, Design And Operation of Multistage Flash (MSF) Desalination:
Development of a steady-state mathematical model for multistage flash (MSF)
Advanced Control Strategies And Impact of Fouling, University of Bradford, 2017.
desalination plant, Int. J. Energy Res. 36 (2012) 710–723.
PhD dissertation.
[20] H. Al-Fulaij, A. Cipollina, H. Ettouney, D. Bogle, Simulation of stability and
dynamics of multistage flash desalination, Desalination 281 (2011) 404–412.

24