Journal of Cleaner Production 189 (2018) 878e886

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

New superstructure-based optimization of property-based industrial

water system
Chun Deng a, *, Wei Jiang a, 1, Wenjin Zhou a, 1, Xiao Feng b
a
State Key Laboratory of Heavy Oil Processing, China University of Petroleum-Beijing, 18 Fuxue Road, Changping, Beijing, 102249, China
b
School of Chemical Engineering & Technology, Xi'an Jiaotong University, No.28, Xianning West Road, Xi'an, Shaanxi, 710049, China

a r t i c l e i n f o a b s t r a c t :

Article history: Traditional industrial water network only consists of water-using processes and water treatment pro-
Received 26 December 2017 cesses. The practical water-using processes may use freshwater, desalted water, condensation water,
Received in revised form steam, circulating cooling water etc. and those are typically water utilities. It is not considered in the up-
29 March 2018
to-date model of industrial water network. To overcome the limitation, this paper proposed a novel
Accepted 31 March 2018
Available online 5 April 2018
superstructure of property-based industrial water system and it consists of water utility, water-using and
water treatment sub-systems. The developed mathematical model includes the relevant equations
among different water utilities, flowrate and mass balance constraints and property constraints. The
Keywords:
Mathematical programming
water system of a certain coal-based chemical complex in China is optimized to illustrate the proposed
Water network model. The results show that the total annualized cost of the water system is reduced from 1.825
108
Property integration CNY/y (preliminary design) to 1.494
108 CNY/y (optimum design) and the flowrate of water resource is
Wastewater minimization decreased from 1369.855 t/h to 1030.498 t/h.
Superstructure © 2018 Elsevier Ltd. All rights reserved.

1. Introduction et al. (1999). Considerable contributions, i.e. improved problem

The population growth and rapid development of urbanization
and industrialization results in the increase of freshwater intake. It
is projected that the global water demand will increase by 55%
globally between 2000 and 2050, as reported by Organization for
Economic Co-operation and Development (OECD, 2012). The major
increment of water demand will be caused by manufacturing,
electricity and domestic. The optimum water management and
reutilization is very helpful for water minimization of manufacture
industries. Enormous pinch techniques and mathematical pro-
gramming optimization approaches are proposed for the optimum
design of water network.
On the basis of the concept of synthesis of mass exchange
network (El-Halwagi and Manousiouthakis, 1989), Wang and Smith
(1994) firstly proposed the limiting composite curve and water
supply line to locate the flowrate targets (i.e. minimum flowrate of
freshwater, regenerated water) and it is considered as pioneering
work of water pinch technique. The mass problem table as an
alternative technique for flowrate targeting is proposed by Castro
* Corresponding author.
E-mail address: chundeng@cup.edu.cn (C. Deng).
1
Wei Jiang and Wenjin Zhou contributed equally to this work.
table as well as limiting composite curve and water supply line for
regeneration recycling (Feng et al., 2007) and regeneration re-use
(Bai et al., 2007), zero liquid discharge (Deng et al., 2008), have
been addressed. Numerous graphical and tabular water pinch
techniques are proposed, i.e. Water Surplus Diagram (Hallale,
2002), Material Recovery Pinch Diagram developed separately by
El-Halwagi et al. (2003) and Prakash and Shenoy (2005), Water
Cascade Analysis (Manan et al., 2004), Source Composite Curve
(Bandyopadhyay et al., 2006), Material Surplus Composite Curve
(Saw et al., 2011), Composite Table Algorithm and improved
limiting composite curve (Agrawal and Shenoy, 2006) and its
extension improved problem table for flowrate targeting water
network with multiple water resources (Deng and Feng, 2011) and
multiple partitioning interception units (Deng et al., 2016). The
progress of the pinch technique for water system optimization is
addressed by Foo (2009). However, there are difficulties for pinch
techniques to handle the optimum design of water network with
multiple impurities or contaminants, and the annualized operating
and investment costs as the objectives.
Takama et al. (1980) initially proposed a superstructure of re-
finery water system which includes water-using units and water-
regeneration/treatment units. Huang et al. (1999) improved the
previous superstructure proposed by Takama et al. (1980) and it

https://doi.org/10.1016/j.jclepro.2018.03.314
0959-6526/© 2018 Elsevier Ltd. All rights reserved.
 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886 879

Notation fs;k flowrate from water source s to water sink k, t/h

fs;t flowrate from water source s to wastewater
treatment unit t, t/h
Sets and indices fe flowrate of evaporation, t/h
Nk set of process water sources fw flowrate of splash loss, t/h
Ns set of process water sources fr flowrate of circulating cooling water, t/h
SNs sub-set of Ns Dt temperature difference between the inlet and outlet
Nt set of water treatment system circulating cooling water,  C
Np set of water properties
ftin inlet flowrate of wastewater treatment unit t, t/h
s index for water sources
k index for water sinks ftprod regenerated water flowrate of wastewater treatment
t index for water treatment system unit t, t/h
e index for environment ftresd residual water flowrate of wastewater treatment unit
p index for water properties t, t/h
ftloss water loss flowrate of wastewater treatment unit t, t/
Parameters h
aFWS freshwater production ratio of FWS cin
t;p inlet property p of wastewater treatment unit t, t/h
aPS boiler blowdown ratio of PS prod
ft;k regenerated water flowrate allocated from
aDWS desalted water production ratio of DWS
at production ratio of regenerated water of wastewater wastewater treatment unit t to water sink k, t/h
prod
treatment t ft;t0 regenerated water flowrate allocated from
CN cycle concentration of cooling tower wastewater treatment unit t to wastewater treatment
K temperature coefficient unit t’, t/h
cUB
k;p
upper limit value of water property p of water sink k fe water flowrate discharged to MOE, t/h
Efresh unit cost of water resource prod
ft;e regenerated water flowrate from wastewater
EFWS unit treatment cost of FWS treatment unit t to MOE, t/h
EDWS unit treatment cost of DWS resd
ft;e residual water flowrate from wastewater treatment
EPS unit treatment cost of PS
unit t to MOE, t/h
ECWS unit treatment cost of CWS
ce;p property p of water stream discharged to MOE, t/h
Et unit treatment cost of WTS t
EMOE unit treatment cost of effluent sent to MOE cresd
t;p property p of the residual water for wastewater
AWH annual working hours treatment unit t, t/h
Af annual factor
Superscripts/Subscript
Variables out outlet
out
fFWS outlet flowrate of pretreated freshwater, t/h in inlet
in FW fresh water
fFWS inlet water flowrate of FWS, t/h
DW desalted water
FW
fFWS;k freshwater flowrate allocated from FWS to water sink CDW condensation water
k, t/h loss water loss
DW
fs;k desalted water flowrate allocated from DWS to water steam steam
sink k, t/h prod regenerated water
prod UB upper bound
ft;k flowrate of regenerated water from water treatment
resd residual water
unit t to water sink k, t/h
fkin inlet water flowrate of water sink k, t/h Abbreviations
cin
k;p
inlet property p of water sink k, t/h FWS fresh water station
DWS desalted water station
cFW
p property p of fresh water, t/h
CWS cooling water system
cDW
p property p of desalted water, t/h PS power and steam system
prod
ct;p property p of regenerated water, t/h WTS wastewater treatment system
cout outlet property p of water source s, t/h MOE municipal wastewater treatment system or the
s;p
environment
fsout outlet flowrate of water source s, t/h

included multiple fresh water sources and water generation and
loss. Bagajewicz and Savelski (2001) proposed the linear mathe-
matical models for the design of water system on the basis of the
proposed the necessary conditions (Savelski and Bagajewicz, 2000).
Gunaratnam et al. (2005) proposed the superstructure of total
water network and the water-using system and treatment system
could be optimized simultaneously. Karuppiah and Grossmann
(2006) proposed the superstructure of integrated water network
and the global optimization strategy was developed. Ng et al.
(2009a, b) presented the automated targeting model for the opti-
mization of single-contaminant water network with direct reuse/
recycling (Ng et al., 2009a) and single-pass or partitioning
880 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886

regeneration reuse/recycling (Ng et al., 2009b). Ahmetovi c and

Grossmann (2011) proposed the superstructure of water network
with fixed mass load and flowrate water-using processes. Zhu et al.
(2017) proposed a mathematical model for the optimization of
water network addressing the technical and economic selection of
the high-concentration wastewater treatment processes in coal-
based chemical plant.
Property integration which can fully consider water quality (i.e.
pH, conductivity, COD, hardness, toxicity and color) is firstly pro-
posed via El-Halwagi et al. (2004). Considerable work has been
conducted on the design of property-based water network, such as,
direct reuse/recycling and treatment processes (Ponce-Ortega et al.,
2009), in-plant (Napoles-Rivera et al., 2010) and interplant (Rubio-
Castro et al., 2013) interception reuse/recycling, batch processes
(Ng et al., 2008), combined batch and continuous processes (Chen
et al., 2010), economic and environmental objectives (Ponce-
Ortega et al., 2011).
Worthy to mention, the types of water only include fresh water,
regenerated water and wastewater in the previous model. Huang
et al. (1999) categorized the water sources into primary water
(i.e. freshwater) and secondary water (i.e. effluent of water-using
process). Faria and Bagajewicz (2010) firstly proposed the com-
plete water system and it includes the pretreatment system, water-
using system and wastewater treatment system. The water utilities
(i.e freshwater, desalted water, cooling water, condensation water)
and their relevant equations have never been addressed in the
previous mathematical model. Deng et al. (2018) firstly proposed
the general superstructure of refinery water network and intro-
duced the generalized water-using process with multiple types of
water. The general superstructure includes the water pretreatment
systems, such as freshwater station, desalted water station, steam
power station. The proposed model (Deng et al., 2018) is used for
the optimization of a coal-based chemical complex (Jiang et al.,
2017). However, the previous model is complex, and the contami-
nant mass balance constraints are ignored with the introduction of
water replacement ratio. Zhang et al. (2018) introduced the unit
models of indirect open circulating cooling water system and
desalination system for water system optimization in steel
industry.
To overcome the limitations of previous approaches, this paper
categorizes industrial water system into water utility, water-using
and water treatment sub-systems. It is important to distinct and
model the water utilities and address the relevance among
different utilities. A new superstructure-based optimization model
of industrial water system is proposed, and water system of coal-
based chemical system is optimized to show the applicability of
the proposed approach.
2. Methodology
2.1. Problem statement
Fig. 1 shows a new superstructure for the optimum design of
industrial water system. It consists of three parts: utility system,
water-using system and treatment system. The utility system is the
major feature of the proposed new superstructure. It includes fresh
water station (FWS), desalted water station (DWS), power and
steam system (PS), cooling water station and cooling system (CWS).
The freshwater generated from FWS is mainly allocated to DWS and
CWS as make-up water. The desalted water from DWS is mainly
converted to steam in PS. The generated condensate water from PS
is typically recovered to DWS as feed water, or used by water-using
processes, or discharged to wastewater treatment system if
contaminated. Cooling water station is typically used for the
Fig. 1. New superstructure of an industrial water system.
cooling down and treatment for the circulated cooling water. Note
that, freshwater, desalted water, steam and cooling water are
typical water utilities of industrial water system. Water-using sys-
tem includes a series of water-using processes. Their inlets can be
considered as process sinks and the outlets can be taken as water
sources. The sinks can receive streams from process water sources
and may use fresh water from FWS and desalted water from DWS.
The treatment system (WTS) consists of several distributed water
treatment units. The treated or regenerated water can be reused/
recycled to process water sinks, DWS and CWS as make-up water.
The residual water streams of WTS and part of treated or regen-
erated water would mix and discharge to municipal water treat-
ment system or environment (MOE) if the discharge limitation is
satisfied. This paper aims to develop a new superstructure and the
associated mathematical model for the optimization of industrial
water system.
2.2. Mathematical model
Based on the proposed superstructure and the problem state-
ment in the previous section, we proposed the mathematical model
for the optimum design of industrial water system. A list of indices,
sets, parameters and variables are presented in the Notation, where
all the parameters are denoted as upper-case symbols and all the
variables are denoted as lower-case symbols.
The effluents of utility system (i.e. DWS, PS, CWS) and the water-
using units are treated as a set of process (or internal) water sources
(s2Ns). Several water-using processes directly use the steam for
reaction, stripping, etc. and their effluents are not considered as
condensation water and can be specified as sub-set SNs (SNs3Ns).
Note that, freshwater, desalted water, cooling water, condensation
water and regenerated water generated from WTS are taken as
water utilities and they are not included in the set of water sources
(Ns). The inlets of water-using processes and utility system (i.e.
DWS, PS, CWS) are treated as a set of water sinks (k2Nk). Water
treatment system (WTS) includes a set of treatment units (t2Nt). A
set of properties (p2Np) (i.e. conductivity, chemical oxygen de-
mand (COD), turbidity, total suspended solid (TSS), total organic
carbon (TOC), pH, color) is used to measure the quality of water,
instead of contaminants or impurities.
 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886 881

2.3. Constraints DWS to water sink k. Worthy to mention, CWS should not use
desalted water as made-up water and the self-utilization of desal-
The fresh water station (FWS) is used as the pre-treatment unit ted water in DWS is embedded in the production of desalted water.
for the water resources outside the industrial plant (e.g., river, lake, Power Station (PS) converts desalted water into steam and the
and reservoir). The choice of the pre-treatment process technolo- steam is utilized for reaction, heating, power and stripping, etc. The
gies depends on water quality of resource characterized by different inlet water flowrate balance of PS is given by,
properties. For example, if the water comes from the municipal
water plant, we can use sand filtration, ozone exposure and acti- fkin ¼ fDWS;k
DW
ck ¼ PS (7)
vated carbon absorption as pre-treatment techniques. Water
flowrate balance of the FWS is given by,
where fkin denotes the inlet water flowrate of PS. fDWS;k
DW represents
the desalted water flowrate allocated from DWS to PS.
out
fFWS ¼ aFWS $fFWS
in
(1)
Several water-using processes (i.e. reaction, stripping) use the
out is the outlet flowrate of pretreated freshwater. a generated steam directly and part of steam will enter the material
where fFWS FWS
stream and the effluents can be considered as water sources. Major
in represents
denotes the freshwater production ratio of the FWS. fFWS part of steam acts as the heating medium for the reboilers and
the inlet water flowrate of FWS. power driving medium for the turbines and it generates conden-
The freshwater (FW) can be allocated to desalted water station, sation water. The discharge of surplus steam and steam loss
cooling water station and other water-using sinks. The outlet commonly exists in the industrial plant. The blowdown of boilers
flowrate balance of FWS is given by, can be considered as water sources.
X
out
fFWS ¼ FW
fFWS;k (2) X
fkin ¼ fssteam þ fsCDW þ fsloss þ fsout ck ¼ PS; cs ¼ PS (8)
k2Nk&ksPS
s2SNs

FW denotes the freshwater flowrate allocated from FWS to

where fFWS;k
where fkin denotes the flowrate of inlet desalted water of PS. fPS
CDW
water sink s. Note that, freshwater cannot directly feed to the denotes the flowrate of steam condensation water produced by PS.
boilers of power station and steam system (PS). out denotes the flowrate of blowdown of boiler in PS. Note that,
fPS
Desalted water station (DWS) utilizes freshwater, steam
there may be several boilers for waste heat recovery in the main
condensate and regenerated water to produce desalted water. The
production processes units (i.e. fluidized catalytic cracking unit in
typical treatment techniques include ultrafiltration, reverse
refinery, coal gasification unit in coal-based plant). They would use
osmosis and ion exchange bed. The inlet water flowrate balance of
desalted water and generate the steam and blowdown. They are
DWS is given by, P out
assumed to be in the power station and steam system. fs
X prod
s2SNs
fkin ¼ fFWS;k
FW
þ ft;k CDW
þ fPS;k ck ¼ DWS (3) represents the summation flowrate of wastewater which is con-
t2Nt verted from the steam when it can be used in the production
processes.
FW denotes the flowrate of freshwater allocated from FWS
where fFWS;k The correlation of the boiler blowdown and the inlet water
prod
to DWS. ft;k denotes the flowrate of regenerated water from water flowrate of PS is shown in equation (8),
CDW is the flowrate of steam condensate
treatment unit t to DWS. fPS;k
fsout ¼ aps $fkin cs ¼ k ¼ PS (9)
from PS to DWS.
The cleaning and regeneration of desalination process generates where aps denotes the boiler blowdown ratio of PS and it is
wastewater effluent. The water flowrate balance of DWS is given by, generally taken as 5%e10%.
Cooling water station (CWS) is part of the cooling water system
fkin ¼ fsDW þ fsout ck ¼ s ¼ DWS (4) which is an important part of industrial water system, and the
cooling water is utilized to cool down the process streams in the
where fsDW represents the flowrate of desalted water produced by production processes (e.g. top stream of distillation column, outlet
DWS and fsout is the flowrate of wastewater effluent generated by stream of compressor). The make-up water of CWS takes up a
DWS. substantial proportion of industrial water consumption. Because of
The desalted water generated in the outlet of DWS can be water evaporation, sewage, etc., certain cooling water will be lost in
related with the fresh water in the inlet of DWS, and it is given by, CWS and generally the fresh water, regenerated water generated by
WTS or boiler blowdown can be used as the make-up water for
fkin ¼ aDWS $fsDW ck ¼ s ¼ DWS (5) CWS. The inlet water flowrate balance of CWS is given by,

X
where aDWS denotes the ratio of fresh water to desalted water fkin ¼ fFWS;k
FW
þ fPS;k þ
prod
ft;k ck ¼ CWS (10)
during desalted water production. t2Nt
Major part of the desalted water generated in DWS is assigned to
the power system to generate steam for heating, power and strip- where fkin denotes make-up water flowrate of CWS. fFWS;k
FW denotes
ping etc. Several water-using processes use desalted water. The
the fresh water flowrate allocated from FWS to CWS. fPS;k denotes
outlet water flowrate balance of the DWS is given by, prod
the boiler blowdown flowrate from PS to CWS. ft;k represents the
X
fsDW ¼ DW
fs;k cs ¼ DWS (6) flowrate of regenerated water directed from water treatment unit t
k2Nk&ksDWS&ksCWS to CWS.
The make-up water flowrate equals the flowrate summation of
DW denotes the flowrate of desalted water allocated from
where fs;k evaporation, splash water loss and blowdown. It can be given by,
882 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886

p of the process water sink k, fresh water, desalted water, regen-

fkin ¼ fek þ fwk þ fsout cs ¼ k ¼ CWS (11) erated water and water source s respectively.
The inlet water properties of water sink should fulfil its upper
where fek denotes the evaporation flowrate of cooling tower. fwk limit, which is given by,
denotes the water flowrate of splash loss. fsout denotes the flowrate
of sewage in CWS, which can be regard as a process source s. cin UB
k;p  ck;p ck2Nk&ksDWS&ksCWS&ksPS ; p2Np (18)
The theoretical make-up water flowrate for CWS depends on the
evaporation of cooling tower and concentration multiple or rate. where cUB
k;p
denotes the upper limit value of water property p of the
The correlation is given by, water sink k.
In practical, there are a lot of wastewater treatment units in
fek $CNk
fkin ¼ ck ¼ CWS (12) industrial wastewater treatment system, i.e. wastewater stripper,
CNk  1 recirculating pool, grease trap, flotation tank and advanced
wastewater treatment units (i.e. ultrafiltration, reverse osmosis). To
where CNk denotes the concentration multiple or rate. simplify the model, we only consider two wastewater treatment
The evaporation loss depends on the flowrate of circulating units T1 and T2 in WTS, and they are placed in series, that is, the
cooling water, temperature difference between the inlet and outlet regenerated water produced by the treatment unit T1 can be sent to
of circulating cooling water and temperature coefficient. The rele- the treatment unit T2 for further treatment.
vant formulation is given by, Wastewater treatment unit T1 receives the wastewater from
water-using processes and utility system，the inlet water flowrate
fek ¼ Kk $Dtk $frk ck2CWS (13)
balance for T1 is given by,
where Kk denotes the temperature coefficient. Dtk is temperature X
ftin ¼ fs;t ct ¼ T1 (19)
difference between the inlet and outlet of circulating cooling water.
s2Ns
frk represents the flowrate of circulating cooling water.
The splash loss also depends on the flowrate of circulating where ftin denotes the inlet water flowrate of wastewater treatment
cooling water and drifting parameter. The correlation is given by, t.
Wastewater treatment unit T2 receives the wastewater from
fwk ¼ DF$frk ck2CWS (14)
production processes and utility system， and T2 also receives the
product water from T1. The inlet water flowrate balance for T2 is
where DF denotes the drifting parameter.
given by,
Effluents of water-using processes can be reuse/recycled to
water sinks or discharged directly to WTS for treatment. In addi- X X
tion, effluents from DWS, CWS and PS in utility system are also
ftin ¼ fs;t þ ftprod
0 ;t ct ¼ T2 (20)
s2Ns t 0 2Nt
taken as process water sources. The outlet water flowrate balance
of process sources is given by, P
where ftprod
0 ;t denotes product water from wastewater treatment
X X t0
fsout ¼ fs;k þ fs;t þ fs;e cs2Ns (15) T1 to T2.
k2Nk t2Nt After the wastewater is treated by the wastewater treatment
unit, regenerated water and residual water are generated, and part
where fs;k denotes the water flowrate allocated from process source of the water is lost in the form of mud. The outlet flowrate balance
s to process sink k. fs;t denotes the water flowrate from process at of wastewater treatment t is given by,
source s discharged to wastewater treatment unit t.
The water-using processes can use fresh water, desalted water, prod
ftout ¼ ft þ ftresd þ ftloss ct2Nt (21)
regenerated/treated water, boiler blowdown, or the effluents of
other water-using processes. Note that, the use of cooling water in where ftout represents the outlet water flowrate of wastewater
the production process belongs to the cooling utility system and the
treatment t. ftprod represents the flowrate of regenerated water. ftresd
use of steam as the heating medium and power driving medium
should be included in the steam system. Therefore, the inlet water represents the flowrate of residue water. ftloss represents the water
flowrate balance for the water-using process does not consider the loss flowrate during the treatment process.
cooling water and steam. The inlet water flowrate balance and The correlation between regenerated water and the inlet water
properties balance are given by, of wastewater treatment t is given by,

X prod
fkin ¼ fFWS;k
FW DW
þ fDWS;k þ
prod
ft;k ft ¼ at $ftin ct2Nt (22)
t2Nt
X where at denotes the production ratio of regenerated water of
þ fs;k ck2Nk&ksDWS&ksCWS&ksPS (16)
wastewater treatment t.
s2Ns
The regenerated water of WTS can be sent to water-using pro-
X cesses, further treatment or sent to municipal wastewater treat-
prod prod
cin in FW FW DW DW
k;p $fk ¼ cp $fFWS;k þ cp $fDWS;k þ ct;p $ft;k ment system or environment (MOE). The constraint is given by,
t2Nt
X X
þ cout
s;p $fs;k ck2Nk; p2Np ftprod ¼ prod
ft;k prod
þ ft;t0 prod
þ ft;e ct2Nt (23)
s2Ns k2Nk&ksPS
(17)
prod prod prod
where ft;k ; ft;t0 ; ft;e denotes regenerated product water flowrate
prod out
where cin ; cFW ; cDW
k;p p p ; ct;p ; cs;p ; denote the value of water property from wastewater treatment unit t to water-using processes, further
 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886 883

treatment or MOE.  0:547 X 0:547

The surplus residual water and regenerated water are discharge Cpipe ¼ 1289:338$ in
fFWS þ FW
fFWS;k
to MOE, which is given by, k
X 0:547 X 0:547 XX 0:547
X prod

þ DW
fDWS;k þ CDW
fPS;k þ fs;k
resd
fe ¼ ft;e þ ft;e ce ¼ MOE (24)
k k s k
t2Nt
XX 0:547 XX 0:547 X 0:547
þ fs;t þ ft;k þ fs;e
where fe denotes the water flowrate from the water system to MOE. s t t s
k
prod
ft;e denotes the regenerated water flowrate from wastewater X 
prod 0:547
X 0:547  L
resd
þ ft;e þ ft;e $
resd denotes the residue water flowrate
treatment unit t to MOE. ft;e t t
r0:389
from wastewater treatment unit t to MOE. (29)
The associated inlet water properties balance for waste water
treatment unit T1 and T2 are given by, where L represents the length of water pipe. r denotes the density
X of water in the pipes.
cin in
t;p $ft ¼ cout
s;p $fs;t ct ¼ T1; p2Np (25)
s2Ns
2.5. Model summary
X X prod prod
cin in
t;p $ft ¼ cout
s;p $fs;t þ ct 0 ;p $ft 0 ;t ct ¼ T2; p2Np (26) P e Total annual cost (TAC) as objective function.
s2Ns t 0 2Nt min OBJcost given in Equation (28).
s.t. mass balance constraints for utilities (1)e(14).
where cin
t;p denotes the value of property p at the inlet of wastewater mass balance constraints for water-using processes (15), (16).
treatment unit t. cout
s;p is the value of property p of process water property balance constraints for water-using processes (17).
bounding constraint for water sources (18).
source s. cprod
T1;p denotes the value of property p of regenerated
mass balance constraints for wastewater treatment system
product water of wastewater treatment unit T1. (19)e(24).
The residual water and surplus regenerated water are dis- property balance constraints for wastewater treatment system
charged to MOE and the water flowrate balance and properties (25)e(27).
balance are given by, cost formulation of pipelines (29).
X prod prod

ce;p $fe ¼ ct;p $ft;e þ cresd resd
t;p $ft;e ce ¼ MOE; p2Np It is worthy to mention that there are no binary variables but
t2Nt
nonlinear terms (i.e. cin ,f in , cin
k;p k
in resd resd 0:547 ) in the
t;p ,ft , ce;p ,fe , ct;p ,ft;e , f
(27)
model P and it results in a nonlinear programming (NLP) problem.
All the models are coded in GAMS 24.2.2 on a PC with Intel® Core™
where fe denotes the water flowrate to municipal wastewater
resd denotes the residual water
i5-3330 3.2 GHz and 4.00 GB RAM, running Windows 10, 64-bit
treatment system or environment. ft;e
operating system. The NLP model is solved using BARON solver
flowrate from wastewater treatment unit t to MOE. ce;p denotes the for global optimization. The absolute optimality tolerance for all
value of water property p to MOE. cresd
t;p denotes the value of residual solvers is set as 106.
water property p of wastewater treatment unit t.
3. Case study

2.4. Objective function To illustrate the application of the proposed model, we analyzed
the water system for a certain coal-based chemical complex in
The objective function with total annualized cost (TAC) is China. The water system can be categorized into water utility,
introduced for the consideration of economic analysis. TAC includes water-using and treatment sub-systems. The utility system in-
the cost of fresh water, treatment cost of utility system and cludes fresh water system (FWS), desalted water system (DWS),
wastewater treatment system, and investment cost of pipelines. power station (PS) and cooling water system (CWS). They generate
The objective function can be expressed by, and provide water utilities (freshwater, desalted water, steam and
cooling water) for production units in the chemical complex. The
  water-using system includes raw material preparation, gasification,
in in in
min OBJcost ¼ fFWS $ Efresh þ EFWS þ fDWS $EDWS þ fPS $EPS methanol, urea, acetic anhydride branch plants, etc. The treatment
! system includes one wastewater treatment station (WTS) for the
X reduction of the properties TSS and TOC. Note that, TSS and TOC are
þ frCWS $ECWS þ ftin $Et þ fe $EMOE $AWH
t
representative properties which represent the degree of turbidity
and contamination.
þ Af $Cpipe
Since the industrial water system is complicated, we simplify it
(28) to demonstrate the problem better. The flowrate and utilization of
cooling water and steam is not optimized in the model. Based on
where Efresh denotes the unit cost of water resource. EFWS , EDWS , EPS , the water-saving analysis of current water system, the flowrates
ECWS , Et , EMOE denotes the unit cost of water treatment in FWS, and properties of five process water sources (i.e. SR1, SR2, SR3, SR4,
DWS, PS, CWS, WTS and MOE. AWH denotes annual working hours. SR5) are extracted as shown in Table 1. Besides, the flowrate of
Af denotes the rate of depreciation. Cpipe denotes the investment effluents of DWS, CWS and PS are also considered as process water
cost of pipes. The total cost of pipes can be expressed based on sources and listed in Table 1. Those water sources are allocated to
(Alnouri et al., 2014), municipal wastewater treatment system and they would be re-
884 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886

Table 1
Extracted data of utilities, water sources and sinks.

Item Water source Outlet TSS Outlet TOC Item Water utility Outlet TSS (mg/L) Outlet TOC (mg/L)
flowrates (t/h) (mg/L) (mg/L) flowrates (t/h)

Effluent of DWS e 28 35 FW e 5 6
Effluent of CWS e 20 22 DW e 1 0
Effluent of PS e 10 8 CDW e 2 2
Treated water e 5 5

SR1 15.5 8 5 Item Water sinks flowrate (t/h) Upper bound of TSS (mg/L) Upper bound of TOC (mg/L)

SR2 2 10 8 SK1 35 2 2
SR3 20 40 90 SK2 20 12 30
SR4 30 4 27 SK3 5.5 45 80
SR5 20.5 1 5 SK4 32.5 22 24

used by other water sinks or treated for reuse and recycle. Four
process water sinks (i.e. SK1, SK2, SK3, SK4) use water utilities (i.e.
FW, DW) and they would reuse process water sources or treated
water to reduce the intake flowrate for water utilities. The required
flowrates and upper bounds of inlet properties (i.e. TSS and TOC)
are extracted according to the history properties data of the intake
water and engineers' experience and listed in Table 1. The proper-
ties for three water utilities, FW, DW and CDW are listed in Table 1.
The flowrate allocation for water sources and sinks in the pre-
liminary design is shown as Table 2. As shown, water sinks SK1, SK3
and SK4 are initially allocated by desalted water (DW) and water
sink SK2 uses freshwater (FW). The condensate water (CDW)
generated by power station (PS) are completely reused and recy-
cled to desalted water station (DWS). The circulating flowrate of
CWS is extracted as 42,812 t/h, and the flowrates of make-up water,
evaporation water and splash of CWS can be calculated as
1006.938 t/h, 755.204 t/h and 1.284 t/h via solving equations
(11)e(14). Note that, all of make-up water are freshwater and the
wastewater discharge from CWS can be calculated to 250.45 t/h (¼
1006.938 e 755.204 e 1.284). The power and steam system (PS) is
fed by 233.33 t/h of desalted water (DW) and generates the 210 t/h
of steam and 23.333 t/h of wastewater with the specified discharge
ratio of 0.1 for the boilers. In the power and steam system (PS),
steam can be used as heat medium for re-boilers and power driver
for turbines. It generates the condensate water (CDW) up to 40 t/h
and CDW can be recovered and recycled back to DWS. Other pro-
306.333
1.25 e 306.333). The total flowrate of FW can be deter-
mined as 1369.855 t/h via the summation of inlet FW flowrate of
DWS (342.917 t/h), CWS (1006.938 t/h) and SK2 (20 t/h). In the
preliminary design, the effluents of DWS, CWS and PS and process
sources SR1e5 are allocated to municipal wastewater treatment
system or environment (MOE) and the total flowrate is summed up
to be 438.367 t/h. The properties of discharged wastewater are
25.511 mg/L (TOC), 19.324 mg/L (TSS).
We estimate the TAC of the water system in preliminary design.
The cost parameters are obtained from the industrial plant as
shown in Table 3. The unit price of water resource (Ewater) is 1 CNY/t.
The municipal wastewater treatment cost EMOE is 15 CNY/t. The
treatment costs of utility systems are EFWS 0.2 CNY/t, EDWS 2.8 CNY/t,
EPS 50 CNY/t, ECWS 0.2 CNY/t. Assuming the length of all pipes and
water density in the pipes as the same values, the TAC of water
system in the preliminary design can be determined 1.825
108
CNY/y via solving equations (28) and (29).
While the water system is optimized with the objective function
with minimum TAC, the process water sources can be not only sent
to municipal wastewater treatment system but also can be sent to
Table 3
Parameters and values.
Parameter Value Parameter Value Unit

cesses use steam for stripping, reactant (i.e. coal gasification), etc. aFWS 1 Dt 12 
C
The steam loss is 60 t/h and other 110 t/h of steam is finally con- aDWS 1.25 AWH 8000 h
aPS 0.1 r 1000 kg/m3
verted to water effluent which is typically discharged to the
at 0.7 L 500 m
wastewater treatment system. The desalted water (DW) is fed to CNk 4 Ewater 1 CNY/t
the inlets of PS (233.33 t/h) and SK1 (35 t/h), SK3 (5.5 t/h) and SK4 DF 0.00003 EMOE 15 CNY/t
(32.5 t/h). The flowrate summation of DW is 306.333 t/h. The ratio K 0.00147 EFWS 0.2 CNY/t
of fresh water to desalted water is specified as 1.25 and the FW Af 0.231 EDWS 2.8 CNY/t
EPS 50 CNY/t
intake for DWS can be calculated as 342.917 t/h (¼ 306.333
1.25 ECWS 0.2 CNY/t
e 40 (CDW)). The effluent of DWS is determined as 76.583 t/h (¼ Et 4.68 CNY/t

Table 2
Water flowrate allocation in the preliminary design (t/h).

DWS CWS PS SK1 SK2 SK3 SK4 MOE

FW 342.917 1006.938 20
DW 233.333 35 5.5 32.5
CDW 40
Effluent of DWS 76.583
Effluent of CWS 250.45
Effluent of PS 23.333
SR1 15.5
SR2 2
SR3 20
SR4 30
SR5 20.5
 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886
WTS for treatment, even can be directly reused by water sinks. It
will decrease the flowrate allocated from process water sources to
municipal wastewater treatment system. Besides, the intake flow-
rate of water utilities for water sinks will be reduced with the reuse
of process water sources or treated water. The unit treatment cost
of WTS is (Et) 4.68 CNY/t. The water system is optimized via solving
the model P in 0.01 CPUs via GAMS 24.2.2 using BARON solver (the
PC specification: Intel® Core™ i5-3330 3.2 GHz and 4.00 GB RAM,
Windows 10, 64-bit operating system). One optimum water
network shown in Fig. 2 is plotted according to the optimization
results.
As shown in Fig. 2, water sources SR1, SR2, SR4 and SR5 with low
properties can be directly reused to process water sinks. However,
SR3 is with relatively high properties. Only partial of SR3 can be
reused and the rest of SR3 is sent to WTS for treatment. SR5 is
diluted via 21 t/h of water utility DW to fulfil the requirement of
SK1. The effluents of DWS and CWS in utility system are with high
properties and they are sent to WTS for treatment. The effluent of
PS (23.333 t/h) is with low properties thus it can be directly reused
to CWS as make-up water and the intake FW flowrate for CWS
would decrease 23.333 t/h. Note that, the intake flowrates of DW
Fig. 2. Optimum water network for the coal-based chemical complex.
1600
1400
1200
Flowrates (t/h)
1000
800
600 24.8
400
200
0 0.0
Water resource
Prelimilary design
Fig. 3. Flowrate comparison between water resource, desalted water and discharged wastewater.
885
for SK1, SK3 and SK4 are reduced from 35 t/h, 5.5 t/h and 32.5 t/h to
21 t/h, 0 t/h and 0 t/h, respectively. It leads to the flowrate reduction
for DW by 52 t/h with the reduction ratio of 17.0% and it indicates
DWS would use less FW to produce DW and the corresponding
effluent of DWS will decrease accordingly. The intake FW flowrate
for SK2 is reduced from 20 t/h to 0 t/h. In addition, 231.024 t/h of
treated water from WTS is reused to DWS and it leads to the sharp
flowrate reduction for FW usage in DWS. Therefore, the flowrate
reduction for FW is summed up to 339.357 t/h (¼ 23.333 t/h þ 52 t/
h
1.25 þ 231.024 t/h þ 20 t/h). It also indicates the flowrate of
water resource is finally decreased from 1369.855 t/h to 1030.498 t/
h, with a reduction ratio of 24.8%. Because the treatment cost of
MOE (15 CNY/y) is much higher than that (4.68 CNY/y) of in-plant
WTS, wastewater streams from process water sources are allocated
to in-plant WTS. The residual 99 t/h of wastewater streams with the
properties (i.e. TSS 63.370 mg/L, TOC 81.004 mg/L) is discharged to
MOE and the flowrate reduction ratio reaches 77.4%. The TAC of
water system is reduced from 1.825
108 CNY/y to 1.494
108
CNY/y, reaching a reduction ratio of 18.1%. The comparison for
water flowrates of FW, DW and discharged wastewater between
preliminary design and optimized design are illustrated in Fig. 3.
If the traditional mathematical model (i.e. water sources/sinks
with fixed flowrates) is utilized to optimize the practical industrial
water system, the inlet flowrates of process sinks, and outlet
flowrates of process sources are fixed values after data extraction.
However, in practical, with the flowrate reduction of DW, the
flowrate of generated effluent during the production of DW in DWS
is reduced from 76.583 t/h to 63.583 t/h. In addition, the optimi-
zation of power and steam system and cooling water system will
lead to the flowrate reduction of intake DW and FW, effluent of PS
and CWS. Therefore, the traditional mathematical model cannot be
directly applied on the optimization of practical industrial water
system.
Conclusion
This paper proposed a new superstructure of property-based
industrial water system which consists of water utility, water-
using and wastewater treatment sub-systems. We developed cor-
responding mathematical model considering flowrate and mass
balance constraints and property constraints as well as the rele-
vance among different water utilities which include FW, DW and
CDW. The water system for a certain coal-based chemical complex
in China is analyzed. The minimum TAC can be achieved and it is
decreased from 1.825
108 CNY/y to 1.494
108 CNY/y, reaching a
90.0
77.4
80.0
70.0
Reduction ratio (%)
60.0
50.0
40.0
30.0
17.0
20.0
10.0
Desalted water Discharged wastewater
Optimized design Reduction ratio(%)
886 C. Deng et al. / Journal of Cleaner Production 189 (2018) 878e886
reduction ratio of 18.1%. The flowrate of water resource is reduced
from 1369.855 t/h to 1030.498 t/h, as well as that of wastewater
discharged to MOE is reduced from 438.367 t/h to 99 t/h. The case
study shows that the proposed approach is applicable for optimi-
zation of practical industrial water system. Moreover, the reduction
of water resource and discharged wastewater will be very helpful
for the sustainable development and environment protection of
petroleum, chemical and electricity industries.
Acknowledgements
National Natural Science Foundation of China (No. 21576287
and 71373285) is gratefully acknowledged. The research is also
supported by Science Foundation of China University of Petroleum,
Beijing (No. 2462018BJC003, 2462015BJB02 and 2462015YQ0305).
References
Agrawal, V., Shenoy, U.V., 2006. Unified conceptual approach to targeting and
design of water and hydrogen networks. AIChE J. 52 (3), 1071e1082.
Ahmetovi c, E., Grossmann, I.E., 2011. Global superstructure optimization for the
design of integrated process water networks. AIChE J. 57 (2), 434e457.
Alnouri, S.Y., Linke, P., El-Halwagi, M., 2014. Optimal interplant water networks for
industrial zones: addressing interconnectivity options through pipeline merg-
ing. AIChE J. 60 (8), 2853e2874.
Bagajewicz, M., Savelski, M., 2001. On the use of linear models for the design of
water utilization systems in process plants with a single contaminant. Chem.
Eng. Res. Des 79 (A5), 600e610.
Bai, J., Feng, X., Deng, C., 2007. Graphically based optimization of single-
contaminant regeneration reuse water systems. Chem. Eng. Res. Des. 85 (A8),
1178e1187.
Bandyopadhyay, S., Ghanekar, M.D., Pillai, H.K., 2006. Process water management.
Ind. Eng. Chem. Res. 45 (15), 5287e5297.
Castro, P., Matos, H., Fernandes, M.C., Nunes, C.P., 1999. Improvements for mass-
exchange networks design. Chem. Eng. Sci. 54 (11), 1649e1665.
Chen, C.-L., Lee, J.-Y., Ng, D.K.S., Foo, D.C.Y., 2010. A unified model of property
integration for batch and continuous processes. AIChE J. 56 (7), 1845e1858.
Deng, C., Feng, X., 2011. Targeting for conventional and property-based water
network with multiple resources. Ind. Eng. Chem. Res. 50 (7), 3722e3737.
Deng, C., Feng, X., Bai, J., 2008. Graphically based analysis of water system with zero
liquid discharge. Chem. Eng. Res. Des 86 (A2), 165e171.
Deng, C., Jiang, W., Feng, X., 2018. Deciphering refinery water system design and
optimization: superstructure and generalized mathematical model. ACS Sus-
tainable Chem. Eng. 6 (2), 2302e2315.
Deng, C., Shi, C., Feng, X., Foo, D.C.Y., 2016. Flow rate targeting for concentration-
and property-based total water network with multiple partitioning intercep-
tion units. Ind. Eng. Chem. Res. 55 (7), 1965e1979.
El-Halwagi, M.M., Gabriel, F., Harell, D., 2003. Rigorous graphical targeting for
resource conservation via material recycle/reuse networks. Ind. Eng. Chem. Res.
42 (19), 4319e4328.
El-Halwagi, M.M., Glasgow, I.M., Qin, X.Y., Eden, M.R., 2004. Property integration:
componentless design techniques and visualization tools. AIChE J. 50 (8),
1854e1869.
El-Halwagi, M.M., Manousiouthakis, V., 1989. Synthesis of mass exchange networks.
AIChE J. 35 (8), 1233e1244.
Faria, D.C., Bagajewicz, M.J., 2010. On the appropriate modeling of process plant
water systems. AIChE J. 56 (3), 668e689.
Feng, X., Bai, J., Zheng, X.S., 2007. On the use of graphical method to determine the
targets of single-contaminant regeneration recycling water systems. Chem. Eng.
Sci. 62 (8), 2127e2138.
Foo, D.C.Y., 2009. State-of-the-Art review of pinch analysis techniques for water
network synthesis. Ind. Eng. Chem. Res. 48 (11), 5125e5159.
Gunaratnam, M., Alva-Arg aez, A., Kokossis, A., Kim, J.K., Smith, R., 2005. Automated
design of total water systems. Ind. Eng. Chem. Res. 44 (3), 588e599.
Hallale, N., 2002. A new graphical targeting method for water minimisation. Adv.
Environ. Res. 6 (3), 377e390.
Huang, C.H., Chang, C.T., Ling, H.C., Chang, C.C., 1999. A mathematical programming
model for water usage and treatment network design. Ind. Eng. Chem. Res. 38
(7), 2666e2679.
Jiang, W., Deng, C., Feng, X., 2017. Water system optimization in coal-based chemical
complex. Chem. Eng. Trans. 61, 355e360.
Karuppiah, R., Grossmann, I.E., 2006. Global optimization for the synthesis of in-
tegrated water systems in chemical processes. Comput. Chem. Eng. 30 (4),
650e673.
Manan, Z.A., Tan, Y.L., Foo, D.C.Y., 2004. Targeting the minimum water flow rate
using water cascade analysis technique. AIChE J. 50 (12), 3169e3183.
Napoles-Rivera, F., Ponce-Ortega, J.M., El-Halwagi, M.M., Jimenez-Gutierrez, A.,
2010. Global optimization of mass and property integration networks with in-
plant property interceptors. Chem. Eng. Sci. 65 (15), 4363e4377.
Ng, D.K.S., Foo, D.C.Y., Rabie, A., El-Halwagi, M.M., 2008. Simultaneous synthesis of
property-based water reuse/recycle and interception networks for batch pro-
cesses. AIChE J. 54 (10), 2624e2632.
Ng, D.K.S., Foo, D.C.Y., Tan, R.R., 2009a. Automated targeting technique for single-
impurity resource conservation networks. Part 1: direct reuse/recycle. Ind.
Eng. Chem. Res. 48 (16), 7637e7646.
Ng, D.K.S., Foo, D.C.Y., Tan, R.R., 2009b. Automated targeting technique for single-
impurity resource conservation networks. Part 2: single-pass and partitioning
waste-interception systems. Ind. Eng. Chem. Res. 48 (16), 7647e7661.
OECD, 2012. OECD Environmental Outlook to 2050. OECD Publishing.
Ponce-Ortega, J.M., Hortua, A.C., El-Halwagi, M., Jimenez-Gutierrez, A., 2009.
A property-based optimization of direct recycle networks and wastewater
treatment processes. AIChE J. 55 (9), 2329e2344.
Ponce-Ortega, J.M., Mosqueda-Jime nez, F.W., Serna-Gonza lez, M., Jimenez-
Gutierrez, A., El-Halwagi, M.M., 2011. A property-based approach to the syn-
thesis of material conservation networks with economic and environmental
objectives. AIChE J. 57 (9), 2369e2387.
Prakash, R., Shenoy, U.V., 2005. Targeting and design of water networks for fixed
flowrate and fixed contaminant load operations. Chem. Eng. Sci. 60 (1),
255e268.
Rubio-Castro, E., Ponce-Ortega, J.M., Serna-Gonz alez, M., El-Halwagi, M.M.,
Pham, V., 2013. Global optimization in property-based interplant water inte-
gration. AIChE J. 59 (3), 813e833.
Savelski, M.J., Bagajewicz, M.J., 2000. On the optimality conditions of water utili-
zation systems in process plants with single contaminants. Chem. Eng. Sci. 55
(21), 5035e5048.
Saw, S., Lee, L., Lim, M., Foo, D., Chew, I., Tan, R., Klemes, J., 2011. An extended
graphical targeting technique for direct reuse/recycle in concentration and
property-based resource conservation networks. Clean Technol. Environ. Pol. 13
(2), 347e357.
Takama, N., Kuriyama, T., Shiroko, K., Umeda, T., 1980. Optimal water allocation in a
petroleum refinery. Comput. Chem. Eng. 4 (4), 251e258.
Wang, Y.P., Smith, R., 1994. Wastewater minimisation. Chem. Eng. Sci. 49 (7),
981e1006.
Zhang, K., Zhao, Y., Cao, H., Wen, H., 2018. Multi-scale water network optimization
considering simultaneous intra- and inter-plant integration in steel industry.
J. Cleaner Prod. 176, 663e675.
Zhu, Q., Zhang, B., Chen, Q., Pan, M., Ren, J., He, C., 2017. Optimal synthesis of water
networks for addressing high-concentration wastewater in coal-based chemical
plants. ACS Sustainable Chem. Eng. 5 (11), 10792e10805.