17.

Water optimisation in process industries

Elvis Ahmetovića,b,c,1, Ignacio E. Grossmannb , Zdravko Kravanjac, Nidret Ibrića

a
University of Tuzla, Faculty of Technology, Univerzitetska 8, 75000 Tuzla, Bosnia and
Herzegovina; [email protected], [email protected]
b
Carnegie Mellon University, Department of Chemical Engineering, 5000 Forbes Avenue,
Pittsburgh 15213, Pennsylvania, [email protected]
c
University of Maribor, Faculty of Chemistry and Chemical Engineering, Smetanova ulica 17,
2000 Maribor, Slovenia, [email protected]

Contents
17. Water optimisation in process industries ................................................................................. 1

17.1. Abstract ............................................................................................................................... 2

17.2. Introduction ......................................................................................................................... 2

17.3. Concepts of water use and water networks within an industrial process ............................ 5

17.4. Systematic methods for water network design ................................................................... 8

17.5. Steps of mathematical programming approach for water network design.......................... 9

17.6. Case study ......................................................................................................................... 19

17.7. Conclusions ....................................................................................................................... 33

17.8. Acknowledgments ............................................................................................................. 33

17.9. Nomenclature .................................................................................................................... 34

17.10. References ......................................................................................................................... 38

1
Corresponding author. Present address: Faculty of Technology, University of Tuzla, Univerzitetska 8, 75000 Tuzla, Bosnia and
Herzegovina. Phone: +387 35 320756; Fax: +387 35 320741. E-mail address: [email protected] (E. Ahmetović).
 2

17.1. Abstract

Global consumption of natural resources has been significantly increased over recent decades.

Consequently, the research regarding sustainable utilisation of natural resources including water

and energy has received considerable attention throughout academia and industry. The main

goals have been to find promising solutions with reduced water and energy consumption within

different sectors (i.e., domestic, agricultural, and industrial). Those solutions are also beneficial

from the aspects of wastewater and emission minimisation and protection of the environment.

The focus of this chapter is on optimisation of water consumption within the industrial sector

including process industries (i.e., chemical, food, petrochemical, pulp and paper). This chapter

first briefly presents the global water consumption, and water usages within the process

industries. Then, a concept of process water networks involving wastewater reuse, wastewater

regeneration and reuse/recycling is explained, followed by a brief description of systematic

methods based on water pinch analysis and mathematical programming. An illustrative large-

scale case study of the total water network including multiple contaminants is used to

demonstrate a superstructure-based optimisation approach. The results of the optimal water

network show that significant savings of freshwater consumption and wastewater generation can

be obtained when compared to a conventional water network design.

17.2. Introduction

Global water consumption. Water is a valuable natural resource, which is used for different

purposes in daily life and various sectors, namely, domestic, agricultural, and industrial.

Average global water consumption within those sectors varies, and it depends on the

development level of countries. In developed and industrialised countries (i.e., United States of

America, Germany, France, and Canada) the average industrial water usage varies between
 3

4680%, while in developing countries (i.e., India and Brazil) it is between 417% (Davé 2004).

The predictions show that the populations’ needs for freshwater as well as different products

obtained within industrial and agricultural sectors will be increased. According to the predictions

of global water consumption, the important issues that can be addressed in the future are rational

utilisations of natural resources and sustainable water management within the above-mentioned

sectors.

Water use within the process industries. In a typical industrial process, water is used for

different purposes (i.e., washing, extraction, absorption, cooling, and steam production). After

using water within a process, wastewater is generated and discharged into the environment.

Generally speaking, large amounts of freshwater are consumed and, consequently, large amounts

of wastewater are generated within industrial processes. Accordingly, freshwater usage and

wastewater generation should be minimised for achieving sustainable and more efficient

processes. Some of the works focused on freshwater and wastewater minimisation considering

separate networks (Wang and Smith 1994a, b, Wang and Smith 1995, Galan and Grossmann

1998, Kuo and Smith 1997, Castro, Teles, and Novais 2009), namely, process water-using

network and/or wastewater treatment network. Also, an overall network consisting of process

water-using and wastewater treatment networks (Takama et al. 1980, Huang et al. 1999,

Gunaratnam et al. 2005, Karuppiah and Grossmann 2006, Ahmetović and Grossmann 2011,

Faria and Bagajewicz 2011) was the focus of the research in order to explore additional

freshwater and wastewater minimisation opportunities. In other words, in the process its overall

water system was studied for identifying process units, which consume water and/or generate

wastewater. The main research challenges were devoted to systematically exploring all water

integration opportunities within the process in order to achieve solutions with a reduced water
 4

usage and wastewater generation. It is important to mention that the synthesis problem of an

overall or total water network is more complex, and this problem has been addressed in fewer

papers compared to the synthesis problem of separate networks (water-using network or

wastewater treatment network).

In order to address the synthesis problems of separate and total water networks, different tools

and targeting techniques based on water pinch analysis, namely, the limiting composite curve

(Wang and Smith 1994a, b), the water surplus diagram (Hallale 2002), the water cascade

analysis (Manan, Tan, and Foo 2004) as well as methods based on mathematical programming

(Takama et al. 1980, Doyle and Smith 1997, Alva-Argáez, Kokossis, and Smith 1998, Yang and

Grossmann 2013) have been proposed. Over recent decades, it has been shown that water-pinch

technology (Foo 2009) and mathematical programming (Bagajewicz 2000, Jeżowski 2010) are

very useful approaches for solving industrial water reuse and wastewater minimisation problems.

Those approaches are used for determining the minimum freshwater usage and wastewater

generation before designing the network (targeting phase) as well as for synthesizing the water

network (design phase). It has been demonstrated that by applying systematic approaches very

promising solutions can be obtained, with the increased water reuse (18.6-37%) within different

processes and very short payback times (0-10 months) (Mann and Liu 1999). Accordingly, it is

worth pointing out that a significant progress has been made within this field. The reader is

referred to several review papers (Bagajewicz 2000, Yoo et al. 2007, Foo 2009, Jeżowski 2010,

Klemeš 2012, Grossmann, Martín, and Yang 2014, Khor, Chachuat, and Shah 2014, Ahmetović

et al. 2015), and books (Mann and Liu 1999, Smith 2005, Klemeš et al. 2010, El-Halwagi 2012,

Klemeš 2013, Foo 2013) for more details about water network synthesis and recent progress

within this field.

 5

17.3. Concepts of water use and water networks within an industrial process

Figure 17.1 shows typical water users and water treatment within an industrial process (Mann

and Liu 1999). Raw water usually taken from lakes, rivers or wells is firstly treated in raw water

treatment units in order to be purified and used within the process in water-using units, steam

boiler and cooling tower (deaeration, demineralisation, deionisation, dealkalisation, and pH

control, etc.). After using freshwater within process water-using units, cooling tower, and steam

boiler, wastewater is generated and directed to a final wastewater treatment. The main role of the

wastewater treatment is to remove contaminants from wastewater and to satisfy environmental

constraints on the effluent discharged into the environment. Note that also water from

wastewater treatment can be reused or recycled within the process (not shown in Figure 17.1)

and that in this way freshwater consumption and wastewater generation can be significantly

reduced.

Freshwater Steam Wastewater

boiler

Raw water Raw water Freshwater Process Wastewater

water-using Wastewater Wastewater
treatment treatment discharge
units

Water loss by
evaporation

Freshwater Cooling Wastewater

tower

Figure 17.1. Water use and wastewater treatment within an industrial process.

Figure 17.2 shows those concepts of water networks. A water network problem is a special case

of a mass exchanger network (MEN) problem (El-Halwagi and Manousiouthakis 1989). Figure

17.2a presents a process water network consisting of water-using units. Within this network mass
 6

loads of contaminants are transferred from process streams to water streams, and wastewater is

generated within a process water network. Mass loads of contaminants transferred to the water

streams are usually too small when compared to the water flow-rate. In such cases it can be

assumed that the inlet and the outlet water flows of water-using units are the same. Also it should

be mentioned that wastewater can be generated by utility systems, a boiler and a cooling tower

(Figure 17.1). In those systems, a part of wastewater, namely, boiler blow-down and cooling

tower blow-down, must be periodically discharged (Allegra, da Silva, and Al Goodman 2014)

and make-up water introduced, in order to keep the water quality at certain acceptable levels.

There is loss of water within the cooling tower due to water evaporation (Figure 17.1), which

cannot be reused again within a process. In the cooling tower, there is no direct contact between

process and water streams. The contaminants’ concentrations within this unit are increased due

to water evaporation. Within the process water network, freshwater consumption and wastewater

generation can be minimised by water reuse and the local recycling of water. Figure 17.2b shows

a wastewater treatment network. It consists of wastewater treatment units that remove

contaminants from wastewater. The removed mass loads of contaminants from wastewater are

usually too small when compared to the wastewater flow-rates, and in such cases it can be

assumed that the inlet and the outlet wastewater flows of wastewater treatment units are the

same. A wastewater treatment network can be a centralised or a distributed system (Galan and

Grossmann 1998, Zamora and Grossmann 1998). In the centralised system, consisting of more

wastewater treatment units, all wastewater streams from different processes are mixed and

directed to wastewater treatment. In this case, the total flow of wastewater streams goes through

each wastewater treatment unit. However, in the distributed system, consisting of more

wastewater treatment units, all flows of wastewater streams are not treated within each
 7

wastewater treatment unit. Consequently, the total cost of the distributed wastewater treatment

network can be decreased compared to the centralised wastewater treatment network.

Mass load of contaminants Mass load of contaminants

Process Wastewater Wastewater

Freshwater Wastewater Wastewater
water network treatment network discharge

a) b)

Mass load of contaminants Mass load of contaminants

Process Wastewater Wastewater

Freshwater Wastewater
water network treatment network discharge

Regenerated wastewater reuse/recycling

c)

Mass load of contaminants Mass load of contaminants

Regenerated
wastewater
recycling
PU1 TU1
Wastewater Wastewater
Freshwater reuse discharge
PU2 TU2

Wastewater
recycling Regenerated
wastewater
Process reuse Wastewater
water network treatment network
PU-Process unit d) TU-Treatment unit

Figure 17.2. Concepts of water networks: a) process water network, b) wastewater treatment
network, c) combined process water network and wastewater treatment network, d) extended
case c) to show different water integration options within the combined network.

It should be mentioned that different technologies can be used for removing contaminants, and in

most cases wastewater treatment models are based on the fixed removal ratios of contaminants.

However, note also that more realistic models (Yang, Salcedo-Diaz, and Grossmann 2014) have
 8

been recently proposed for wastewater treatment units.

A process water network (Figure 17.2a) and a wastewater treatment network (Figure 17.2b) can

be integrated thus producing an overall or total network (Figure 17.2c). The following water

integration opportunities are enabled within this network (see Figure 17.2d):

 wastewater reuse (direct wastewater reuse from process unit 1 (PU1) to process unit 2

(PU2)),

 wastewater recycling (wastewater recycling within PU2),

 wastewater regeneration and recycling (wastewater from PU1 is regenerated within

wastewater treatment unit 1 (TU1) and recycled to the same unit (PU1), where it has

previously been used) and

 wastewater regeneration and reuse (wastewater from PU1 is regenerated within TU1 and

reused in PU2).

Note that in the case of wastewater regeneration and reuse, wastewater from TU1 does not enter

the same unit (PU1), where it has previously been used. By solving the overall network (Figure

17.2c, d) an improved solution can be obtained simultaneously when compared to the overall

network obtained by sequential solutions of networks shown in Figure 17.2a, and Figure 17.2b.

This will be shown later in this chapter, as demonstrated on an illustrative case study.

17.4. Systematic methods for water network design

This section presents a brief description of systematic methods based on water pinch analysis and

mathematical programming, which are used water network design. Water pinch

technology/analysis (Wang and Smith 1994a, b, Wang and Smith 1995) is a graphical method,

which represents an extension of pinch analysis for heat integration (Linnhoff and Hindmarsh

1983). It consists of two phases, namely, targeting and design. Assuming a single contaminant,
 9

the main goal of the targeting phase is to determine the minimum freshwater consumption

(maximum water reuse) (Doyle and Smith 1997) before a water network design, while within the

design phase a water network is constructed satisfying the minimum freshwater consumption.

Also, water targeting models for multiple contaminants have been proposed based on

mathematical programming in order to perform simultaneous flow-sheet optimisation (Yang and

Grossmann 2013).

The mathematical programming approach is based on a water network superstructure

optimisation (Takama et al. 1980). A superstructure includes all feasible alternatives from which

the best is selected. The main steps within the mathematical programming approach (Biegler,

Grossmann, and Westerberg 1997) are to develop a superstructure of alternative designs, develop

a superstructure optimisation model, solve the model in order to extract the optimum design from

the superstructure, and analysis of the obtained results. This approach can easily deal with

multiple contaminant problems, different constraints (i.e., forbidden connections), and trade-offs

between investment and operational costs when compared to water pinch analysis, which can

have difficulties when addressing those issues, especially for large-scale problems. In some

cases, it can be very useful to combine both approaches in order to solve the overall water

network synthesis problem.

17.5. Steps of mathematical programming approach for water network design

As will be shown later, the mathematical programming approach was applied for the synthesis of

water networks consisting of process water-using and wastewater treatment units. Accordingly,

this section describes the main steps of the mathematical programming approach for the

synthesis of water networks, namely, problem formulation, superstructure development,

optimisation model formulation, and solution strategies’ development. The reader is referred to
 10

the paper (Ahmetović and Grossmann 2011) for more details regarding the superstructure,

model, and solution strategy described within this chapter. Here, only a brief description is given.

Problem formulation. The first step in the application of the mathematical programming

approach for the synthesis of water networks is a problem formulation. The problem formulation

of the water network synthesis problem can be stated as follows. Given is a set of water sources,

a set of water-using units, and a set of wastewater treatment operations. For the set of water

sources, concentrations of contaminants within water sources, and the cost of water are specified.

Water-using units can operate with fixed or variable flow-rates. In the case of the fixed flow-

rates through water-using units, for the set of process water-using units, given are the fixed flow-

rates, maximum concentrations of contaminants within inlet streams at process units, and mass

loads of contaminants within the process units. However, in the case that the flow-rate of water is

not fixed, given are the maximum concentrations of contaminants within inlet and outlet streams

of process units, and mass loads of contaminants within process units. For a set of wastewater

treatment units, percentage removals (removal ratios) of each contaminant within wastewater

treatment units are specified and corresponding cost relationships between investment and

operation cost, which depend on the flow-rates of the wastewater treated within treatment units.

Wastewater discharged from the network has to satisfy environmental constraints specified by

regulations. Accordingly, the maximum acceptable level of concentrations of contaminants

within effluent is given. The problem formulation can be extended to include a set of water

demand units and a set of water source units (non-mass transfer operations) as presented in

Ahmetović and Grossmann (2011). In that case, flow-rates and contaminant concentrations of

those units are usually specified.

The main goal of the water network synthesis problem is to determine the minimum freshwater
 11

usage and wastewater generation, the interconnections, flow-rates and contaminants’

concentrations within each stream of the water network. The objective can be formulated as the

minimisation of the total annualised cost of the water network consisting of the cost of

freshwater usage, the cost of wastewater treatment, the cost of piping, and the cost of pumping

water through pipes. It is assumed that the water network operates continuously under isothermal

and isobaric conditions. However, those assumptions can be easily relaxed and the model

extended (Ahmetović, Ibrić, and Kravanja 2014).

Superstructure development. On the basis of problem formulation, a superstructure should be

synthesised including all feasible network alternatives. Figure 17.3 shows a superstructure

consisting of freshwater sources, process water using units, wastewater treatment units, internal

water source units and water demand units (Ahmetović and Grossmann 2011). This

superstructure incorporates both the mass transfer and non-mass transfer operations. In the mass-

transfer process operations (process units PUp) there is a direct contact between a contaminant-

rich process stream and a contaminant-lean water stream. In that case, during the mass-transfer

processes, the contaminants mass load LPUp, j (pollutant) is transferred from the process streams

to the water. The contaminant concentration within the process stream is reduced, while the

contaminant concentration increases within the water stream. The main purpose of water use in

mass-transfer operations (i.e. absorption, washing) is to remove contaminants. However, in non-

mass-transfer operations (i.e. cooling tower, boiler, reactor), the main purpose of water use is not

to remove contaminants. In those cases, water can be required, for example, in a cooling tower

due to a water loss by evaporation or in a reactor due to a water demand by chemical reaction. In

the case that there is a loss of water in a unit, water cannot be re-used from that unit in other

water-using operations. That unit is represented by a water demand unit DUd in Figure 17.3.
 12

Water can be produced in a unit as by-product (i.e. in reactor) and be available for re-using

within other operations. That unit is represented in Figure 17.3 by an internal water source unit

SUr, which is a typical non-mass-transfer operation. There is also wastewater treatment unit TUt

within the superstructure in order to remove contaminants (percentage removal RRt , j is specified)

from wastewater stream FTU tin .

The superstructure considers all feasible connections between units including options for water

integration (water reuse, regeneration reuse and regeneration recycling). The freshwater from the

freshwater splitter SIs can be directed towards process water-using units (stream FIPs , p ),

wastewater treatment units ( FITs ,t ) in order to enable freshwater pre-treatment if required, final

wastewater mixer (stream FIFs ) and demand units ( FIDs ,d ). Streams from other process units

(stream FPp ', p ) and source units ( FSPs , p ) are directed to process unit mixer MPUp, enabling water

reuse options. Also water regeneration reuse and recycling are enabled by the existence of stream

FTPt , p connecting treatment unit t with process unit p. The wastewater stream FPp , p ' leaving

process unit p (from splitter SPUp) is directed to another process unit p’, demand unit d (stream

FPD p ,d ), wastewater treatment unit t ( FPTp ,t ) for wastewater regeneration or directly discharged

into the environment ( FPO p ). The inlet streams to demanding unit mixer (MDUd) are those

directed from the external freshwater source and internal source ( FIDs ,d and FSDr , d )

respectively, as well as streams leaving process and treatment units ( FPD p ,d and FTDt , d ). From

the internal water source r water can be sent towards process unit p ( FSPr , p ), demand unit d

( FSDr , d ), treatment unit t ( FSTr ,t ) or discharged directly into the environment ( FSOr ). Streams

from all the splitters within the network are directed to the treatment unit mixer MTUt enabling
 13

freshwater pre-treatment if required ( FITs ,t ) and wastewater regeneration ( FPTp ,t , FSTr ,t ). Also,

streams from other treatment units FTt ',t are directed to mixer MTUt.

Optimisation model formulation. For a given superstructure, an optimisation model is formulated

in order to perform optimisation and extract the optimal solution embedded within the

superstructure. The optimisation model of water network synthesis problem consists of an

objective function and constraints. The objective function can be formulated using various

economic criteria (Pintarič et al. 2014). The objective functions used in this chapter represent the

minimisation of total annualised cost of the networks (Ahmetović and Grossmann 2011)

consisting of:

a) The freshwater cost, the investment and operating costs of wastewater treatment units, the

cost of piping and the operational cost of water pumping through pipes (integrated water

network involving process water-using and wastewater treatment units)

b) The freshwater cost, the cost of piping and the operational cost of water pumping through

pipes (water network involving only process water-using units)

c) The investment and operating costs of wastewater treatment units, the cost of piping and

the operational cost of water pumping through pipes (water network involving only

wastewater treatment units)

Note that in the first case (a), the objective function is used for the integrated water network,

while in other cases for its separate networks, namely, water-using network (b) and wastewater

treatment network (c).

In addition to the economic nature of the objective function of the water network problem, it can

also be formulated as the minimisation of the total freshwater consumption, or the minimisation

of the total flow-rate of the freshwater consumption and wastewater treated within treatment
 14

units because the cost of the wastewater treatment units depends on wastewater flow-rates

treated within wastewater treatment units.

LPUp, j
FTPt,p FSPr,p
FIFs FPOp
in
FWs FIPs,p FPUp FPUpout FPTp,t
SIs MPUp in PUp SPUp
xPUp, j xPUp, jout FPDp,d
FIDs,d
FPp’,p FPp,p’
FITs,t

in
FTDt,d FDUd
MDUd in DUd
xDUd, j
FPDp,d

FSDr,d
FSPr,p
out
FSUr FSOr
SUr out
SSUr
xSUr, j
FSTr,t

FIFs
FSTr,t FTDt,d
bTU t , j
FITs,t FTPt,p
in out FPOp
FPTp,t FTUt FTUt FTOt F out
MTUt in TUt out STUt MF
xTUt, j xTUt, j xj
out
FTt’,t
FTt,t’

Figure 17.3. The superstructure of an integrated water network (Ahmetović and Grossmann
2011).

The model constraints can be formulated as equalities and inequalities. The equalities are, for

example, the overall mass and contaminant mass balance equations, while inequalities can

represent constraints on variables, for example, flow-rates, contaminant concentrations, or design

constraints corresponding to the existence of pipes or wastewater treatment technologies, etc.

The variables within an optimisation model of a water network synthesis problem can be

continuous and binary. Continuous variables are, for example, flow-rates, contaminant

concentrations, while binary variables can only have values 0 or 1 and they are used for
 15

selecting, for example, a wastewater treatment technology or a piping connection. If a binary

variable has the value 1 in that case the wastewater treatment technology or the piping

connection is selected and vice versa. The optimisation problem can be formulated as linear

programming (LP), nonlinear programming (NLP), mixed-integer linear programming (MILP)

or mixed-integer nonlinear programming problem (MINLP). The latter one is the most difficult

to solve due to the nonlinear nature of the problem and many design alternatives within the

network. For the purposes of solving the case study presented in this chapter, an MINLP

optimisation model of water network proposed by Ahmetović and Grossmann (2011) was used.

This model was implemented within the General Algebraic Modeling System (GAMS)

(Rosenthal 2014). It is worth mentioning that this model is generic and independent of initial

data, and it can be used for solving different types of water network synthesis problems. For

example, process water networks or wastewater treatment networks can be optimised separately

(sequentially) or as an integrated network (simultaneously). Also, the proposed model can be

used for the synthesis problems with fixed or variable flow-rates through process water-using

units, and to establish trade-offs between the network cost and network complexity. The generic

nature of the model enables easy manipulation within the proposed model, for example,

excluding some units from the superstructure can be done only by specifying an empty set for the

given units. Figure 17.4 shows different options of the water network model starting from a data

input to the optimal network design solution. The following data input is required, for example,

in the case that an overall network consisting of a process water network (fixed flow-rates

through process water-using units) and a wastewater treatment network should be synthesised:

 a number of freshwater sources, a number of contaminants and corresponding

concentrations within the freshwater source stream, and the cost of freshwater;
 16

 a number of process water-using units, mass flow-rates of process units, mass loads of

contaminants transferred to water streams within the process units, maximum inlet

concentrations of contaminants at the inlet streams to the process units;

 a number of wastewater treatment units, removal percentages of contaminants within the

wastewater treatment units, operation and investment cost coefficients, cost function

exponents, and an annualised factor for investment on treatment units;

 cost coefficients corresponding to existence of pipes, investment cost coefficients, cost

function exponent for the investment cost of pipes, and an annualised factor for investment

on pipes;

 operating cost coefficient for pumping water through pipes;

 hours of network plant operation per annum; and

 limiting the concentrations of contaminants within the wastewater stream discharged into

the environment.

Accordingly, only the number of units has to be specified next to data input in order to solve the

model using MINLP optimisation solvers, for example, BARON (Tawarmalani and Sahinidis

2005) in order to obtain the global optimum solution of the network design. The presented model

can be also applied for multiple pinch water network problems (Hallale 2002, Manan, Tan, and

Foo 2004, Parand 2014).

 17

Data input

Yes Multiple water No

sources? Single
source

Yes Contaminants No
in water source(s)?
Clean water

No Superstructure No Superstructure No Superstructure

consists only of process consists of process and consists only of treatment
units? treatment units? units?

Yes Yes Yes

Number of Number of process Number of
process units and treatment units treatment units

Superstructure Yes Superstructure No

consists of pre-treatment/ Local recycles
consists of demand/
process/treatment around units?
source units?
units?
No Yes
Yes
Number of pre-treatment/ Number of source
process and/or treatment units and/or demand units

MINLP Global or local

GAMS optimisation solvers
optimisation model

Optimal design
of network

Figure 17.4. Different options of the water network model proposed by Ahmetović and
Grossmann (2011).
 18

Solution strategies development. The overall synthesis problem consisting of process water-using

and wastewater treatment networks can be solved sequentially and simultaneously. In a

sequential approach, a process water-using network is firstly solved in order to determine the

minimum freshwater usage and the water network design, followed by solving a wastewater

treatment network in order to determine a wastewater network design (see Figure 17.2a and

Figure 17.2b as well as Figure 17.4). On the basis of the obtained network design solutions, an

overall network consisting of a water-using network and a wastewater treatment network can be

constructed. In a simultaneous approach, process water-using and wastewater treatment networks

are solved together as an overall synthesis problem (see Figure 17.2c as well as Figure 17.4) in

order to extract the optimal design of integrated network from the superstructure. Water pinch

technology and mathematical programming can be used for sequentially solving the overall

water network synthesis problem. However, only the mathematical programming approach can

address the overall water network synthesis problems simultaneously. The obtained solution by

the sequential strategy is sub-optimal, while the simultaneous strategy enables obtaining locally

optimal as well as globally optimal solutions depending on the type of solver that is used. In the

cases of using local optimisation solvers a good initialisation point should be provided as well as

tights bounds for optimisation variables, while an initialisation point does not need to be

provided for the global optimisation solver, for example BARON. The proposed MINLP

optimisation model can be solved by global optimisation solver BARON directly or by using a

two-stage solution strategy. The reader is referred to works (Ahmetović and Grossmann 2010,

2011) for more details about the MINLP model and those strategies. For solving of case studies

in this chapter, MINLPs were solved directly with BARON, and global optimal solutions of all

case studies were found within the specified optimality tolerance (1%) and reasonable
 19

computational times. It is worth pointing out that the special redundant constraint for the overall

contaminant mass balance (Karuppiah and Grossmann 2006) and good variable bounds were

incorporated within the model, and this had a very big impact in improving the strength of the

lower bound for the global optimum, as well as reducing the CPU time for BARON.

17.6. Case study

The case study considered in this section represented a large-scale example involving five

process water‒using units, three treatment units and three contaminants (A, B and C). The

contaminants concentrations within the wastewater stream discharged into the environment were

limited to 10 ppm. Data for the process units (flow-rates, discharge load/contaminant mass load

and maximum inlet concentrations of contaminants) were taken from the literature (Karuppiah

and Grossmann 2006) and are given in Table 17.1. Table 17.2 shows data for the process units

modified to address a case with variable flow-rate through water-using units. Table 17.3 shows

data for wastewater treatment units (percentage removal of contaminants, operating and

investment cost coefficients, cost function exponent).

The freshwater cost is assumed to be $1/t, the annualized factor for investment 0.1 (1/y), and the

total time for the network plant operation in a year 8000 hours. The fixed cost coefficient

pertaining to the pipe is assumed to be $6, while the investment cost coefficient for each

individual pipe is taken to be $100/(th-1), and operating cost coefficient for pumping water

through pipe $0.006/t. Data for piping and water pumping costs were taken from the literature

(Karuppiah and Grossmann 2008).

Several cases are presented in this section for the same case study addressing the issues of a

conventional water network, centralised and distributed wastewater treatment systems, water

reuse and recycling, sequential and simultaneous synthesis of process water and wastewater
 20

treatment networks, fixed and variable flow-rates through process water-using units. For all cases

optimality tolerances were set at 0.01, and MINLPs of all cases were directly solved by BARON.

Table 17.1. Data for process units (fixed flow-rate problem) (Karuppiah and Grossmann 2006).

Process Flow-rate Discharge load Maximum inlet

unit (t/h) (kg/h) concentration (ppm)
A B C A B C
PU1 40 1 1.5 1 0 0 0
PU2 50 1 1 1 50 50 50
PU3 60 1 1 1 50 50 50
PU4 70 2 2 2 50 50 50
PU5 80 1 1 0 25 25 25

Table 17.2. Data for process units (variable flow-rates through the process units) (Ahmetović and
Grossmann 2011).

Process Discharge load Maximum inlet Maximum outlet Limiting water

unit (kg/h) concentration (ppm) concentration (ppm) flow-rate (t/h)
A B C A B C A B C
PU1 1 1.5 1 0 0 0 25 37.5 25 40
PU2 1 1 1 50 50 50 70 70 70 50
PU3 1 1 1 50 50 50 66.67 66.67 66.67 60
PU4 2 2 2 50 50 50 78.57 78.57 78.57 70
PU5 1 1 0 25 25 25 37.5 37.5 25 80

Table 17.3. Data for treatment units (Karuppiah and Grossmann 2006).

Treatment % removal of Investment cost Operating cost Cost function

units contaminant coefficient coefficient exponent
A B C
TU1 95 0 0 16,800 1 0.7
TU2 0 0 95 9,500 0.04 0.7
TU3 0 95 0 12,600 0.0067 0.7

Firstly, a conventional water network without water reuse and with centralised wastewater

treatment was considered as a base-case 1 (BC1). In this case, the freshwater was used in all
 21

process units (see Figure 17.5). The freshwater consumption for this case was 300 t/h and the

concentrations of contaminants (A, B and C) within the mixed wastewater stream were 20, 21.67

and 16.67 ppm. The total annual cost (TAC) of the network including freshwater, water pumping

and piping cost is 2,429,964.5$/y (see BC1 in Table 17.4). The concentrations of the

contaminants in the wastewater stream should not exceed the limiting concentrations of 10 ppm.

In order to satisfy this constraint, in the next base case 2 (BC2), the wastewater stream was firstly

treated within the centralised wastewater treatment system with the same wastewater flow-rate

going through all treatment units subsequently (see Figure 17.5).

A B C (kg/h)
1 1.5 1

40 40
PU1
25.0 A
37.5 B
25.0 C A B C (%)
1 1 1 95 0 0

50 50
PU2 TU1
20.0 A 1.00 A
20.0 B 21.67 B
20.0 C 16.67 C
1 1 1 0 0 95

300 60 60 300
PU3 TU2
0A 16.67 A 20.00 A 1.00 A
0B 16.67 B 21.67 B 21.67 B
0C 16.67 C 16.67 C 0.83 C
(ppm) 2 2 2 0 95 0

70 70 300
PU4 TU3
28.57 A 1.000 A
28.57 B 1.083 B
28.57 C 0.833 C
1 1 0

80 80
PU5
12.5 A
12.5 B
0.0 C

Figure 17.5. Base case of process water-using and wastewater treatment network (cases
BC1+BC2)

The centralised wastewater treatment system is characterised by very high operating and
 22

investment cost of treatment units (see BC2 in Table 17.4) due to maximum wastewater flow-rate

through all three treatment units. The TAC of the centralised wastewater treatment for the BC2 is

2,781,739.0 $/y. As can be seen from Figure 17.5, the network solutions of the base cases BC1,

and BC2 were merged representing an overall network (BC1+BC2) (Figure 17.5). The TAC of the

merged base case networks shown in Figure 17.5 was 5,211,703.5 $/y.

Table 17.4. Base case results of conventional networks.

Base case networks

BC1 BC2 BC1+BC2
FW (t/h) 300 0.0 300
CostFW ($/y) 2,400,000.0 0.0 2,400,000.0
ICTU ($/y) 0.0 210,831.0 210,831.0
OCTU ($/y) 0.0 2,512,080.0 2,512,080.0
ICPipes ($/y) 1,164.5 1,228.0 2,392.5
OCPumping ($/y) 28,800.0 57,600.0 86,400.0
TAC ($/y) 2,429,964.5 2,781,739.0 5,211,703.5
CPU (s) 0.1 0.1 0.2
BC1Base case process water network, which uses freshwater in all process units; BC2Base case
centralized wastewater treatment network.

In the previous base case (BC1), water reuse options were unconsidered within the network and

consequently the network required the maximum amount of freshwater (300 t/h). In order to

consider water reuse options, on the basis of the general superstructure (Figure 17.3), the

superstructure was constructed for 5 process units (Figure 17.6). This superstructure included all

possible connections between a freshwater source splitter, process units, and a final wastewater

mixer. The objective in this case (C1) was to find the optimal network design by minimising

TAC of the network. From the initial data (Table 17.1) regarding the maximum inlet

concentrations of contaminants (0 ppm for A, B, and C) within the water stream directed to the

process unit 1 (PU1), it was noted that water reuse options from all process units to PU1 were

infeasible because the PU1 required only clean freshwater without contaminants (0 ppm).
 23

Accordingly, these connections could be removed from the superstructure and a simplified

network could be obtained (5 connections could be removed). The optimal network design

obtained by using the MINLP model (Ahmetović and Grossmann 2011) is given in Figure 17.7.

The network design (case C1) exhibited the minimum freshwater consumption of 84.286 t/h. The

freshwater consumption was reduced by approximately 71.9 % (84.286 vs. 300 t/h) compared to

the base case BC1. Note that the investment cost for piping (see solution C1 in Table 17.5) was

also significantly reduced compared to the base case design even though the base case design

exhibited a lower number of connections. The reason for this was that piping costs were directly

proportional to the water flow-rate which was at the maximum through all connections in the

base case. The TAC of the network, when comparing BC1 and C1 designs, was reduced from

2,429,964.5 $/y to 693,655.5 $/y.

PU1

PU2

out
FW1 F
PU3

PU4

PU5

Figure 17.6. Superstructure of process water network (case C1).

 24

40 40 40 25.641
PU1

14.359

50 50 14.286
PU2
10.073
25.641

50

84.286 60 60 84.286
PU3
0A 71.186 A
0B 77.119 B
0C 10 59.322 C

70 70 70
PU4

20

44.286 80 80 60
PU5

Figure 17.7. Optimal design of process water network (case C1).

Two wastewater streams leaving the process units PU2 and PU4 were collected and discharged as

a single wastewater stream (84.286 t/h) and given contaminants concentrations as shown in

Figure 17.7. As can be seen contaminants’ concentrations within the wastewater stream were

above their maximum allowable concentrations of 10 ppm. Accordingly, the wastewater needed

to be treated before it was discharged into the environment. Two cases were studied, the first one

(case C2) when the wastewater stream was directed as a single stream to a wastewater treatment

network and the second one (case C2a) when two separate wastewater streams leaving the

process units PU2 and PU4 were directed to a wastewater treatment network. The superstructure

representation of the wastewater treatment network for these two cases is given in Figure 17.8.

The superstructures presented includes options of distributed wastewater treatment, where

wastewater streams can be distributed amongst different treatment units in order to reduce
 25

wastewater treatment cost that depends on wastewater flow-rate through the treatment units. The

objective was to find the optimal wastewater treatment design by minimising the total annual

cost. Figure 17.9 shows the optimal network designs obtained for two cases of single (Figure

17.9a) and two separate (Figure 17.9b) wastewater streams. As can be seen, wastewater streams

were distributed within three different treatment units compared to base-case design (BC2),

where a centralised wastewater treatment was used. The TACs of the wastewater treatment

networks for cases C2 and C2a were 733,987.5 and 727,958.1 $/y (see Table 17.5). An additional

reduction in wastewater treatment cost for the case C2a was possible due to the existence of more

options for distributive wastewater treatment caused by an increased number of wastewater

connections. Note that by not merging the wastewater streams enabled more than a half of the

first wastewater stream to be directed to the effluent, thereby reducing the loads of the treatment

units. The wastewater treatment operating costs, for cases C2 and C2a, were 637,813.5 and

632,454.8 $/y, respectively. The centralised wastewater system (BC2), which treated wastewater

stream generated within the process units had a substantial increase in operating treatment cost

(2,512,080.0 $/y) compared to both cases of distributive wastewater treatments (C2 and C2a).

The solutions presented for cases C1 and C2/C2a were sequential solutions, where a process water

network and a wastewater treatment network were synthesised separately and sequentially. This

approach could not produce the best solutions because all interactions were not considered

between the process water network and the wastewater treatment network. Table 17.5 shows the

operating, investment and total annual costs of the individual networks cases (cases C1, C2, C2a)

as well as merged individual networks (C1+C2 and C1+C2a) in order to produce overall network

designs. Note that this approach could not produce best solutions because only one-way

interactions were considered – those from the process water network to the wastewater treatment
 26

network. Nevertheless, the TACs of both total network designs (C1+C2 and C1+C2a) were

decreased almost to one quarter when compared to the one of the conventional network

(BC1+BC2).

TU1

FW1 Fout
TU2

a)
TU3

TU1

FW1

Fout
TU2

FW2

b)
TU3

Figure 17.8. Superstructures of wastewater treatment networks a) C2 and b) C2a.

 27

7.068 76.259 76.259

TU1

84.286 73.766 73.766 73.766 84.286

TU2
71.186 A 10 A
77.119 B 10 B
59.322 C 10 C

69.190
3.451 77.218 77.218 8.027 a)
TU3

7.787
75.635 75.635
TU1

14.286 6.499
65.426 A
70 B
51.116 C 72.742 72.742 84.286
TU2
10 A
10 B
70 10 C
72.742
72.364 A
78.571 B 2.893
60.997 C 76.499 76.499 0.864 b)
TU3

Figure 17.9. Optimal designs of wastewater treatment networks a) C2 and b) C2a.

Table 17.5. Sequential solution results of integrated water network.

Sequential solution 1 Sequential solution 2

C1 C2 C1+C2 C2a C1+C2a
FW (t/h) 84.286 0 84.286 0 84.286
CostFW ($/y) 674,285.7 0.0 674,285.7 0.0 674,285.7
ICTU ($/y) 0.0 80,602.3 80,602.3 80,041.9 80,041.9
OCTU ($/y) 0.0 637,813.5 637,813.5 632,454.8 632,454.8
ICPipes ($/y) 924.0 618.3 1,542.4 621.8 1,545.8
OCPumping ($/y) 18,445.7 14,953.4 33,399.1 14,839.7 33,285.4
TAC ($/y) 693,655.5 733,987.5 1,427,642.9 727,958.1 1,421,613.6
CPU (s) 0.28 0.12 0.40 0.16 0.44
C1optimal solution of waterreuse network; C2optimal solution of wastewater treatment network
(wastewater streams from process water units are mixed and directed to wastewater treatment network.
C2aoptimal solution of wastewater treatment network (wastewater streams from process water units are
distributed to wastewater treatment network.
 28

In the next case (C3), a superstructure of integrated water and wastewater treatment networks

(Figure 17.10) was constructed in order to explore all interactions between the two different

networks. Additional interactions, those directed from the wastewater treatment network to the

process water network, enabled additional options of wastewater reuse and recycling within the

process and, hence, possible reductions in freshwater consumption and the total annualised cost

of the network. Similarly, as explained earlier in this chapter, due to the specified contaminants’

concentrations (0 ppm A, B, and C) within the inlet water stream to PU1, all connections from all

process units (PU1-PU5) and all treatment units (TU1-TU3) to PU1 could be excluded from the

superstructure because they were infeasible in this case study. Accordingly, the superstructure

given in Figure 17.10 can be simplified by 8 connections.

PU1

PU2 TU1

FW1 Fout
PU3 TU2

PU4 TU3

PU5

Figure 17.10. Superstructure of integrated process water and wastewater treatment networks
(case C3).
 29

The optimal network design (Figure 17.11) exhibited a minimum freshwater consumption of 40
t/h, which was reduced by 52.5 % when compared to the solution obtained with the sequential
approach. An additional reduction of freshwater consumption was enabled by the existence of
treatment units in which the contaminants were removed from the wastewater streams, thus
increasing the potential for water reuse.

40 40 11.024
PU1
40

28.976

5.639
50 50 75.639 75.639 3.434
PU2 TU1
5.436
4.343

72.205
50

40 60 60 10 68.329 68.329 36.524 40

PU3 TU2
0A 31.805 10 A
0B 4.307 10 B
0C 34.582 10 C
68.329

6.768
70 70 88.665 88.665 0.042
PU4 TU3
9.219
31.111

80 80 48.889
PU5

Figure 17.11. Optimal solution of integrated process water and wastewater treatment networks
for fixed flow-rates through water-using units (case C3).

The dotted lines in Figure 17.11 represent the regenerated wastewater streams reused in the

process units. The freshwater consumption of 40 t/h was the theoretically minimum consumption

determined by the existence of process unit PU1 requiring only freshwater without contaminants

(see Table 17.1). The optimal network design obtained by the simultaneous optimisation

approach using the integrated superstructure exhibited lower operating costs as well as

investment cost in treatment units and piping installations (see solution C3 in Figure 17.11 and

Table 17.6). The TAC was reduced by approximately by 25 % when compared to the sequential
 30

solution (1,062,675.6 vs. 1,427,642.9 $/y). This clearly showed that the simultaneous approach

integrating process water network with wastewater treatment networks was the better approach

to synthesising water networks. Note that the wastewater stream (0.042 t/h), leaving the

wastewater treatment unit TU3 and directed to the final wastewater mixer, was too small and

could be impractical. Accordingly, the same model (C3) was solved again when a wastewater

stream of 0.042 t/h was fixed at zero. As can be seen from Table 17.6 in this case (C4), an

insignificantly increase in TAC (1,062,709.2 vs. 1,062,675.6 $/y) was obtained by excluding the

impractical flow-rate. In addition, the integrated model of a process water network and a

wastewater treatment network was solved for the case of variable flow-rates throughout the

process units (fixed mass load problem). Figure 17.12 shows the optimal network design.
7.3
40 40 40
PU1 27.291
0.922
5.285
5.409

23.732 23.732 13.158 76.342 76.342

PU2 TU1
8.347

62.709

40 40.288 40.288 62.709 62.709 31.456 40

PU3 TU2
0A 10 A
0B 10 B
0C 76.342 10 C

10.575
70 70 83.158 83.158
PU4 TU3
1.244
4.65
19.137

31.253 31.253 12.116

PU5
31.253

Figure 17.12. Optimal solution of integrated process water and wastewater treatment networks
for variable flow-rates through water-using units (case C5).

As can be seen, water flow-rates through the process units PU2, PU3 and PU5 were significantly
 31

reduced when compared to the case of the fixed flow-rates (see Figure 17.11 and Figure 17.12).

However, this did not affect the minimum freshwater consumption of 40 t/h. On the other hand,

treatment units’ operating and investment costs as well as pumping and piping costs were

reduced due to decreased water flow-rates within the network (see comparison between cases C3

and C5 in Table 17.6). The TAC was reduced only by about 0.4 % in this case. In the previous

case studies considered, local recycles around the process and treatment units were disallowed.

Figure 17.13 shows the optimal network design for the integrated water network (variable flow-

rates through process units) when local recycles were allowed (case C6) within the

superstructure. However, local recycle options were not selected by optimisation and most of the

network connections were the same as in the previous case (C5).

4.896
40 40 40
PU1 25.082
1.708
4.097
10.022

24.761 24.761 13.158 76.342 76.342

PU2 TU1
9.535

62.709
1.288

40 29.356 29.356 62.709 62.709 32.982 40

PU3 TU2
0A 10 A
0B 10 B
0C 28.069 10 C
76.342
11.603
70 70 70 83.158 83.158
PU4 TU3
0.834
4.274
20.793

29.727 29.727 8.934

PU5
29.727

Figure 17.13. Optimal solution of integrated process water and wastewater treatment networks
for variable flow-rates through water-using units (local recycling allowed but not selected) (case
C6).
 32

The dotted lines in Figure 17.13 represent connections not existing within the network design

given in Figure 17.12. The TAC of the network for the case C6 was only very slightly reduced

compared to network design C5 (1,058,273.8 vs. 1,058,828.8 $/y). Note that both solutions were

within the selected optimality tolerance of 1% when solving the model by using global

optimisation solver BARON. As can be seen from Table 17.6, all simultaneous solutions were

obtained within reasonable computational times.

Table 17.6. Simultaneous solutions of integrated process water and wastewater treatment network.

Simultaneous solutions
C3 C4 C5 C6
FW (t/h) 40 40 40 40
CostFW ($/y) 320,000.0 320,000.0 320,000.0 320,000.0
ICTU ($/y) 82,080.5 82,088.4 79,963.0 79,963.0
OCTU ($/y) 631,730.4 631,743.9 635,257.6 635,257.6
ICPipes ($/y) 1,378.3 1,388.4 1,169.0 1,162.6
OCPumping ($/y) 27,486.4 27,488.4 22,439.2 21,890.5
TAC ($/y) 1,062,675.6 1,062,709.2 1,058,828.8 1,058,273.8
CPU (s) 38.5 41.1 75.5 39.0
C3, C4optimal network solutions with fixed flow-rates through process units; C5, C6optimal
network solutions with variable flow-rates through the process units;

Table 17.7 shows the summarised results of the various presented cases for the fixed flow-rate

problem. Those results clearly show significant savings in freshwater consumption (FW) and

total annual cost (TAC) obtained by the simultaneous optimisation of the integrated network.

Table 17.7. The summarised results of freshwater consumption and total annual cost.

No optimisation Sequential optimisation Simultaneous optimisation

(BC1+BC2) (C1+C2a) (C3)
FW (t/h) 300 84.286 40
FW reduction (%) 0 71.9 86.7
TAC ($/y) 5,211,703.5 1,421,613.6 1,062,675.6
TAC reduction (%) 0 72.7 79.6
Reductions of FW and TAC are given when compared to the base case (no optimisation).
 33

17.7. Conclusions

This chapter has presented water optimisation in process industries as a challenging problem that

should be addressed to achieve sustainable solutions with the minimum freshwater consumption

and wastewater generation. It has been shown that using systematic methods and considering

water integration opportunities (wastewater reuse, wastewater regeneration reuse/recycling) a

significant reduction in freshwater consumption and wastewater generation can be achieved.

Different cases of a large-scale example involving total water network and multiple contaminants

have been solved in order to present the development of increasingly better network solutions

obtained by applying sequential and simultaneous strategies, as well as solutions of a fixed flow-

rate and a fixed mass load problem. The results showed that by using the simultaneous approach

more than 50 % of savings in water usage and wastewater generation can be obtained when

compared to the design obtained by the sequential approach, and more than 86 % when

compared to the conventional network design. The obtained results in all cases correspond to

global optimal solutions. The applied model is general, data independent and could be used for

solving different water network problems to global optimality.

17.8. Acknowledgments

The authors are grateful to the Swiss National Science Foundation (SNSF) and the Swiss Agency for

Development and Cooperation (SDC) for providing financial support within the SCOPES 2013‒2016

(Scientific Co-operation between Eastern Europe and Switzerland) joint research project (CAPE‒EWWR:

IZ73Z0_152622/1) as well as support from the Slovenian Research Agency (Program No. P2‒0032), and

Bilateral Project (No. 05‒39‒116‒14/14) between Bosnia and Herzegovina and Slovenia. Also, the

supports from Center for Advanced Process Decision-making (CAPD) at Carnegie Mellon University,

and different mobility programmes (Fulbright, EM JoinEU SEE Penta, EM STEM, DAAD, and

CEEPUS) are appreciated.

 34

17.9. Nomenclature

Abbreviations

BARON Branch and Reduce Optimisation Navigator

CPU central processing unit time

FW freshwater

GAMS General Algebraic Modeling System

ICPipes investment cost for piping

ICTU investment cost for treatment units

LP linear programming

MEN mass exchange network

MILP mixed-integer linear programming

MINLP mixed integer nonlinear programming

NLP nonlinear programming

OCTU operating cost for treatment units

OCPumping pumping operating cost

PU process unit

TU treatment unit

TAC total annual cost

Sets and Indices

j contaminant

DU set of demand units

d demand unit

PU set of process units

 35

p process unit

SU set of source units

r source unit

SW set of freshwater sources

s freshwater source

TU set of treatment units

t treatment unit

Parameters

FDU din mass flow-rate of inlet water stream in demand unit d

FPU inp mass flow-rate of inlet water stream in process unit p

FSU rout mass flow-rate of outlet water stream from source unit r

LPUp,j mass load of contaminant j in process unit p

RRt , j percentage removal of contaminant j in treatment unit t

x out
j
, max
maximum concentration of contaminant j in discharge stream to the environment

xDUdin. ,jmax maximum concentration of contaminant j in inlet stream into demand unit d

xPU inp.,jmax maximum concentration of contaminant j in inlet stream into process unit p

xSU rout
.j concentration of contaminant j in outlet stream from source unit r

xWsin. j concentration of contaminant j in freshwater source s

Continuous variables

F out mass flow-rate of outlet wastewater stream from final mixer

FIDs ,d mass flow-rate of water stream from freshwater source s to demand unit d
 36

FIFs mass flow-rate of water stream from freshwater source s to final mixer

FIPs, p mass flow-rate of water stream from freshwater source s to process unit p

FITs ,t mass flow-rate of water stream from freshwater source s to treatment unit t

FPp ', p mass flow-rate of water stream from other process unit p’ to process unit p

FPDp ,d mass flow-rate of water stream from process unit p to demand unit d

FPOp mass flow-rate of water stream from process unit p to final mixer

FPTp ,t mass flow-rate of water stream from process unit p to treatment unit t

FPU pout mass flow-rate of outlet water stream from process unit p

FSDr ,d mass flow-rate of water stream from source unit r to demand unit d

FSOr mass flow-rate of water stream from source unit r to final mixer

FSPr , p mass flow-rate of water stream from source unit r to process unit p

FSTr ,t mass flow-rate of water stream from source unit r to treatment unit t

FTt ',t mass flow-rate of water stream from other treatment unit t’ to treatment unit t

FTDt ,d mass flow-rate of water stream from treatment unit t to demand unit d

FTOt mass flow-rate of water stream from treatment unit t to final mixer

FTPt , p mass flow-rate of water stream from treatment unit t to process unit p

FTUtin mass flow-rate of inlet water stream in treatment unit t

FTUtout mass flow-rate of outlet water stream from treatment unit t

FW s mass flow-rate of water for freshwater source s

x out
j concentration of contaminant j in discharge stream to the environment
 37

xDUdin, j concentration of contaminant j in inlet stream into demand unit d

xPU inp, j concentration of contaminant j in inlet stream into process unit p

xPU pout, j concentration of contaminant j in outlet stream from process unit p

xSPUtout
,j concentration of contaminant j in outlet stream from splitter process unit p

xSTUtout
,j concentration of contaminant j in outlet stream from splitter treatment unit t

xTU tin, j concentration of contaminant j in inlet stream into treatment unit t

xTUtout
,j concentration of contaminant j in outlet stream from treatment unit t

Subscripts/superscripts

in inlet stream

out outlet stream

 38

17.10. References

Ahmetović, E., and I.E. Grossmann. 2010. "Strategies for the Global Optimization of Integrated Process

Water Networks." In Computer Aided Chemical Engineering, edited by S. Pierucci and G. Buzzi

Ferraris, 901-906. Elsevier.

Ahmetović, E., and I.E. Grossmann. 2011. "Global superstructure optimization for the design of

integrated process water networks." AIChE Journal no. 57 (2):434-457. doi:

http://dx.doi.org/10.1002/aic.12276.

Ahmetović, E., N. Ibrić, and Z. Kravanja. 2014. "Optimal design for heat-integrated water-using and

wastewater treatment networks." Applied Energy no. 135:791-808. doi:

http://dx.doi.org/10.1016/j.apenergy.2014.04.063.

Ahmetović, E., N. Ibrić, Z. Kravanja, and I.E. Grossmann. 2015. "Water and energy integration: A

comprehensive literature review of non-isothermal water network synthesis." Computers &

Chemical Engineering no. 82:144-171. doi: 10.1016/j.compchemeng.2015.06.011.

Allegra, K., P.E. da Silva, and P.E. Al Goodman. 2014. "Reduce water consumption through recycling."

Chemical Engineering Progress no. 110 (4):29-35.

Alva-Argáez, A., A.C. Kokossis, and R. Smith. 1998. "Wastewater minimisation of industrial systems

using an integrated approach." Computers & Chemical Engineering no. 22 (0):S741-S744. doi:

10.1016/s0098-1354(98)00138-0.

Bagajewicz, M. 2000. "A review of recent design procedures for water networks in refineries and process

plants." Computers & Chemical Engineering no. 24 (9-10):2093-2113. doi: 10.1016/s0098-

1354(00)00579-2.

Biegler, L.T., I.E. Grossmann, and A.W. Westerberg. 1997. Systematic methods of chemical process

design. New Jersey: Prentice-Hall.

Castro, P., J. Teles, and A. Novais. 2009. "Linear program-based algorithm for the optimal design of

wastewater treatment systems." Clean Technologies and Environmental Policy no. 11 (1):83-93.
 39

doi: 10.1007/s10098-008-0172-5.

Davé, B. 2004. "Water and Sustainable Development: Opportunities for the Chemical Sciences: A

Workshop Report to the Chemical Sciences Roundtable. Washington (DC): National Academies

Press (US). 2004. www.ncbi.nlm.nih.gov/books/NBK83724/. Accessed 29.08.2013."

Doyle, S.J., and R. Smith. 1997. "Targeting Water Reuse with Multiple Contaminants." Process Safety

and Environmental Protection no. 75 (3):181-189. doi: 10.1205/095758297529020.

El-Halwagi, M.M. 2012. Sustainable Design Through Process Integration, Fundamentals and

Applications to Industrial Pollution Prevention, Resource Conservation, and Profitability

Enhancement. Amsterdam, The Netherlands: Butterworth-Heinemann/Elsevier.

El-Halwagi, M.M., and V. Manousiouthakis. 1989. "Synthesis of mass exchange networks." AIChE

Journal no. 35 (8):1233-1244. doi: 10.1002/aic.690350802.

Faria, D.C., and M.J. Bagajewicz. 2011. "Global Optimization of Water Management Problems Using

Linear Relaxation and Bound Contraction Methods." Industrial & Engineering Chemistry

Research no. 50 (7):3738-3753. doi: 10.1021/ie101206c.

Foo, D.C.Y. 2009. "State-of-the-Art Review of Pinch Analysis Techniques for Water Network

Synthesis." Industrial & Engineering Chemistry Research no. 48 (11):5125-5159. doi:

10.1021/ie801264c.

Foo, D.C.Y. 2013. Process Integration for Resource Conservation. Boca Raton, Florida, USA: CRC

Press.

Galan, B., and I.E. Grossmann. 1998. "Optimal Design of Distributed Wastewater Treatment Networks."

Industrial & Engineering Chemistry Research no. 37 (10):4036-4048. doi: 10.1021/ie980133h.

Grossmann, I.E., M. Martín, and L. Yang. 2014. "Review of optimization models for integrated process

water networks and their application to biofuel processes." Current Opinion in Chemical

Engineering no. 5 (0):101-109. doi: http://dx.doi.org/10.1016/j.coche.2014.07.003.

Gunaratnam, M., A. Alva-Argáez, A. Kokossis, J.K. Kim, and R. Smith. 2005. "Automated Design of

Total Water Systems." Industrial & Engineering Chemistry Research no. 44 (3):588-599. doi:
 40

10.1021/ie040092r.

Hallale, N. 2002. "A new graphical targeting method for water minimisation." Advances in

Environmental Research no. 6 (3):377-390. doi: http://dx.doi.org/10.1016/S1093-0191(01)00116-

2.

Huang, C.H., C.T. Chang, H.C. Ling, and C.C. Chang. 1999. "A mathematical programming model for

water usage and treatment network design." Industrial and Engineering Chemistry Research no.

38 (7):2666-2679.

Jeżowski, J. 2010. "Review of Water Network Design Methods with Literature Annotations." Industrial

& Engineering Chemistry Research no. 49 (10):4475-4516. doi: 10.1021/ie901632w.

Karuppiah, R., and I.E. Grossmann. 2006. "Global optimization for the synthesis of integrated water

systems in chemical processes." Computers & Chemical Engineering no. 30 (4):650-673. doi:

10.1016/j.compchemeng.2005.11.005.

Karuppiah, R., and I.E. Grossmann. 2008. "Global optimization of multiscenario mixed integer nonlinear

programming models arising in the synthesis of integrated water networks under uncertainty."

Computers & Chemical Engineering no. 32 (1–2):145-160. doi:

http://dx.doi.org/10.1016/j.compchemeng.2007.03.007.

Khor, C.S., B. Chachuat, and N. Shah. 2014. "Optimization of Water Network Synthesis for Single-Site

and Continuous Processes: Milestones, Challenges, and Future Directions." Industrial &

Engineering Chemistry Research no. 53 (25):10257-10275. doi: 10.1021/ie4039482.

Klemeš, J., F. Friedler, I. Bulatov, and P. Varbanov. 2010. Sustainability in the Process Industry:

Integration and Optimization. New York, USA: McGraw-Hill.

Klemeš, J.J. 2012. "Industrial water recycle/reuse." Current Opinion in Chemical Engineering no. 1

(3):238-245. doi: 10.1016/j.coche.2012.03.010.

Klemeš, J.J., ed. 2013. Handbook of Process Integration (PI): Minimisation of energy and water use,

waste and emissions. Cambridge: Woodhead Publishing Limited.

Kuo, W.-C.J., and R. Smith. 1997. "Effluent treatment system design." Chemical Engineering Science no.
 41

52 (23):4273-4290. doi: http://dx.doi.org/10.1016/S0009-2509(97)00186-3.

Linnhoff, B., and E. Hindmarsh. 1983. "The pinch design method for heat exchanger networks."

Chemical Engineering Science no. 38 (5):745-763. doi: http://dx.doi.org/10.1016/0009-

2509(83)80185-7.

Manan, Z.A., Y.L. Tan, and D.C.Y. Foo. 2004. "Targeting the minimum water flow rate using water

cascade analysis technique." AIChE Journal no. 50 (12):3169-3183. doi: 10.1002/aic.10235.

Mann, J.G., and Y.A. Liu. 1999. Industrial water reuse and wastewater minimization. New York:

McGraw Hill.

Parand, R. 2014. "Water and wastewater optimization through process integration for industrial processes.

http://espace.library.curtin.edu.au/R/?func=dbin-jump-

full&object_id=203501&local_base=GEN01-ERA02. Accessed 01.05. 2016."

Pintarič, Z.N., N. Ibrić, E. Ahmetović, I.E. Grossmann, and Z. Kravanja. 2014. "Designing optimal water

networks for the appropriate economic criteria." Chemical Engineering Transactions no.

39:1021-1026. doi: 10.3303/CET1439171.

Rosenthal, R.E. 2014. GAMS - A User's Guide. Washington, DC, USA: GAMS Development

Corporation.

Smith, R. 2005. Chemical process design and integration. Chichester, England: John Wiley & Sons Ltd.

Takama, N., T. Kuriyama, K. Shiroko, and T. Umeda. 1980. "Optimal water allocation in a petroleum

refinery." Computers & Chemical Engineering no. 4 (4):251-258. doi:

http://dx.doi.org/10.1016/0098-1354(80)85005-8.

Tawarmalani, M., and N.V. Sahinidis. 2005. "A polyhedral branch-and-cut approach to global

optimization." Mathematical Programming no. 103 (2):225-249. doi: 10.1007/s10107-005-0581-

8.

Wang, Y.P., and R. Smith. 1994a. "Design of distributed effluent treatment systems." Chemical

Engineering Science no. 49 (18):3127-3145. doi: http://dx.doi.org/10.1016/0009-2509(94)E0126-

B.
 42

Wang, Y.P., and R. Smith. 1994b. "Wastewater minimisation." Chemical Engineering Science no. 49

(7):981-1006. doi: http://dx.doi.org/10.1016/0009-2509(94)80006-5.

Wang, Y.P., and R. Smith. 1995. "Wastewater minimization with flowrate constraints." Trans. IChemE

no. 73 (Part A):889–904.

Yang, L., and I.E. Grossmann. 2013. "Water Targeting Models for Simultaneous Flowsheet

Optimization." Industrial & Engineering Chemistry Research no. 52 (9):3209-3224. doi:

10.1021/ie301112r.

Yang, L., R. Salcedo-Diaz, and I.E. Grossmann. 2014. "Water Network Optimization with Wastewater

Regeneration Models." Industrial & Engineering Chemistry Research no. 53 (45):17680-17695.

doi: 10.1021/ie500978h.

Yoo, C., T. Lee, J. Kim, I. Moon, J. Jung, C. Han, J.-M. Oh, and I.-B. Lee. 2007. "Integrated water

resource management through water reuse network design for clean production technology: State

of the art." Korean Journal of Chemical Engineering no. 24 (4):567-576. doi: 10.1007/s11814-

007-0004-z.

Zamora, J.M., and I.E. Grossmann. 1998. "Continuous global optimization of structured process systems

models." Computers & Chemical Engineering no. 22 (12):1749-1770. doi:

http://dx.doi.org/10.1016/S0098-1354(98)00244-0.
 43

List of Figures

Figure 17.1. Water use and wastewater treatment within an industrial process. .......................................... 5

Figure 17.2. Concepts of water networks: a) process water network, b) wastewater treatment network, c)

combined process water network and wastewater treatment network, d) extended case c) to show different

water integration options within the combined network. .............................................................................. 7

Figure 17.3. The superstructure of an integrated water network (Ahmetović and Grossmann 2011). ....... 14

Figure 17.4. Different options of the water network model proposed by Ahmetović and Grossmann

(2011). ......................................................................................................................................................... 17

Figure 17.5. Base case of process water-using and wastewater treatment network (cases BC1+BC2) ....... 21

Figure 17.6. Superstructure of process water network (case C1). ............................................................... 23

Figure 17.7. Optimal design of process water network (case C1). .............................................................. 24

Figure 17.8. Superstructures of wastewater treatment networks a) C2 and b) C2a. ..................................... 26

Figure 17.9. Optimal designs of wastewater treatment networks a) C2 and b) C2a. .................................... 27

Figure 17.10. Superstructure of integrated process water and wastewater treatment networks (case C3). . 28

Figure 17.11. Optimal solution of integrated process water and wastewater treatment networks for fixed

flow-rates through water-using units (case C3). .......................................................................................... 29

Figure 17.12. Optimal solution of integrated process water and wastewater treatment networks for

variable flow-rates through water-using units (case C5). ............................................................................ 30

Figure 17.13. Optimal solution of integrated process water and wastewater treatment networks for

variable flow-rates through water-using units (local recycling allowed but not selected) (case C6). ......... 31
 44

List of Tables

Table 17.1. Data for process units (fixed flow-rate problem) (Karuppiah and Grossmann 2006). ............ 20

Table 17.2. Data for process units (variable flow-rates through the process units) (Ahmetović and

Grossmann 2011). ....................................................................................................................................... 20

Table 17.3. Data for treatment units (Karuppiah and Grossmann 2006). ................................................... 20

Table 17.4. Base case results of conventional networks. ............................................................................ 22

Table 17.5. Sequential solution results of integrated water network. ......................................................... 27

Table 17.6. Simultaneous solutions of integrated process water and wastewater treatment network. ....... 32

Table 17.7. The summarised results of freshwater consumption and total annual cost. ............................. 32