You are on page 1of 15

Energy 57 (2013) 236e250

Contents lists available at SciVerse ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Simultaneous synthesis of process water and heat exchanger


networks
Elvis Ahmetovi
c a, b, Zdravko Kravanja b, *
a
Faculty of Technology, University of Tuzla, Univerzitetska 8, 75000 Tuzla, Bosnia and Herzegovina
b
Faculty of Chemistry and Chemical Engineering, University of Maribor, Smetanova ul. 17, 2000 Maribor, Slovenia

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

Article history: This paper presents a novel superstructure and optimization model for the simultaneous synthesis of
Received 15 September 2012 process water and heat exchanger networks. This superstructure combines the water network and heat
Received in revised form exchanger network using interconnecting hot and cold streams. The water network has been extended
3 February 2013
for both direct and indirect heat exchanges. In addition, opportunities for heat integration between hot
Accepted 17 February 2013
Available online 6 April 2013
and cold streams, splitting and mixing of the freshwater and wastewater streams are incorporated within
the superstructure. The proposed model is formulated as a non-convex MINLP (mixed-integer non-linear
program), where the objective is to minimize the total annual costs of the network. A new convex hull
Keywords:
Simultaneous synthesis
formulation is presented for identifying the streams’ roles within the network. Three examples involving
Heat-integrated water networks single and multiple contaminant problems are presented in order to illustrate the applicability and ca-
Superstructure pabilities of the proposed superstructure and model. In all cases the resultant networks exhibit lower
MINLP (mixed-integer non-linear program) total annual costs, whilst the freshwater and utilities consumption are the same as reported in the
model literature. In addition, novel designs for heat-integrated process water networks with smaller or same
number of heat exchangers are presented.
 2013 Elsevier Ltd. All rights reserved.

1. Introduction chemical process design [6], sustainability within the process in-
dustry [7], and sustainable design through process integration [8].
Process industries consume large amounts of water and energy Over the last decade and a half, the synthesis and optimization of
for different purposes (washing operations, extraction, absorption, heat-integrated water networks have received considerable atten-
reaction, cooling, heating, etc.). Therefore, one of the main tasks of tion. The main objective was to simultaneously consider water re-use
industry has been devoted to discovering the best ways to reduce and heat exchange, and design a combined network (WN and HEN)
water and energy consumptions, whilst satisfying strict environ- with minimum freshwater, hot and cold utilities consumption, and
mental regulations. To achieve this important industrial goal, a minimum total annual cost. It should be mentioned that graphical
stronger interaction between water and energy systems needs to be methods and heuristic rules can provide good understanding of the
considered and explored. In order to systematically address this interconnections within WN and HEN. However, they are of a
problem, a conceptual method can be used based on process in- sequential nature and thus cannot simultaneously consider strong
sights and heuristic rules, pinch analysis and/or a mathematical interactions between WN and HEN, especially for multiple contami-
programming approach based on superstructure optimization. For nant problems when minimizing the total annual cost. Apart from
a deeper understanding of the above-mentioned systematic sequential procedure, in most of the published papers based on pinch
methods and tools for solving WN (water network) problems the analysis and mathematical programming several opportunities for
_
reader is referred to the review papers by Bagajewicz [1], Jezowski heat integration between streams, freshwater and wastewater split-
[2,3], Foo [4], and the HEN (heat exchanger network) by Furman ting and mixing within a network have not as yet been fully consid-
and Sahinidis [5], or books relating to systematic methods of ered and explored. An ideal case would be to have a network
superstructure which captures all possible network configurations
including direct and indirect heat recovery, as well as freshwater and
* Corresponding author. Tel.: þ386 2 22 94 481; fax: þ386 2 25 27 774.
wastewater splitting and mixing allowing for an optimization model
E-mail addresses: elvis.ahmetovic@untz.ba (E. Ahmetovi
c), zdravko.kravanja@ for finding the best network solution. However, such a model would
uni-mb.si, kravanja@uni-mb.si (Z. Kravanja). be very complex and very difficult to solve. In addition, it is worth

0360-5442/$ e see front matter  2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.energy.2013.02.061
E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250 237

pointing out that within the combined network (WNeHEN) there are and Kim [17] studied the simultaneous optimization of a combined
interconnecting hot and cold streams with unknown temperatures as WN and HEN superstructure. They used modified WN [18], heat
well as flowrates, thus making the synthesis problem more difficult to transhipment [19], and HEN superstructure [20] model. The overall
solve especially when a large number of these streams exist within MINLP problem was decomposed into two sub-problems MILP
the network. Therefore, developing methodologies and approaches to (mixed-integer linear programming) and NLP which were solved
deal with these problems is still a big challenge for the process sys- using an iterative procedure. The solution procedure was sequen-
tems engineering community. tial, and their approach can be used for solving large-scale multiple
The objectives of this work are to present a brief literature re- contaminant problems. Dong et al. [21] modified the state-space
view, a novel superstructure, and mathematical model for the superstructure [22] for simultaneously synthesizing WN and
simultaneous synthesis of process water and heat exchanger net- HEN. The problem was formulated as an MINLP. The feasible solu-
works. As will be shown, the superstructure involves additional tions were produced using randomly-generated initial estimations
opportunities for heat integration and freshwater/wastewater followed by improving the candidate solutions using perturbation
splitting and mixing. Direct heat exchange by the mixing of streams techniques, and generating alternative network structures by
and indirect heat exchange in heat exchangers is considered within shifting heat loads in loops along the utility paths. This approach
the superstructure. The model is formulated as a MINLP where the can be used for single and multiple contaminant problems. Feng
objective is to minimize the total annual costs. A set of new con- et al. [23] analyzed the reason why different WN structures have
straints is formulated for identifying the interconnecting hot and different energy performances for the same freshwater consump-
cold streams between WN and HEN. The model is tested on single tion. In their proposed model they used a variable to describe the
and multiple contaminant problems. The outline of the paper is as temperature peak or vale, and introduced some new constraints.
follows. Firstly, a brief literature review is given of those contri- The problem was solved over two steps. The first one was the
butions related to the synthesis of heat-integrated water networks. solving of the standard WN problem followed by the synthesizing
Then the problem statement is defined, and a description given of of HEN. They found that reducing the number of temperature
the superstructure and optimization model. Three case-studies fluctuations along the WN subsystems was beneficial for energy
considering single and multiple contaminant problems are solved performance. Leewongtanawit and Kim [24] presented a graphical
for demonstrating the capabilities and effectiveness of the pro- approach for the design of heat-integrated water networks. This
posed model. Finally, the last section presents the general conclu- approach was based on the Water and Energy Balance Diagram
sions, and comments on future study. which is an extension of studies by Savulescu et al. [12,13]. The
design interactions between WN and HEN were explored and
2. Literature review energy-efficient and cost-effective configurations for heat recovery
were identified. Their approach was applied on a single-
This section presents a brief literature review focused on the contaminant WN problem. Xiao et al. [25] used holistic mathe-
synthesis of heat-integrated water networks. Generally speaking, matical programming to formulate a MINLP for the heat-integrated
from 1998 until late-to-mid 2009 less than 20 papers were pub- WN problem. A hybrid optimization strategy based on stochastic
lished regarding this field [3], whilst over the last couple of years and deterministic search techniques was used for solving the pro-
the synthesis of heat-integrated water networks has been an active posed model. They presented sequential and simultaneous solution
research area. Also, it will be one of the main directions for future procedures for single and multiple contaminants. Chen et al. [26]
research in order to produce water and energy efficient and sus- developed a MINLP model based on the superstructures of WN
tainable solutions. The first works concerning the topic of heat and HEN, and proposed a set of constraints to identify cold and hot
integration within WN problems were from Savulescu and Smith streams in WN. This model was solved using a two-step sequential
[9], Savulescu et al. [10], and Bagajewicz et al. [11]. In the first two strategy. Polley et al. [27] developed a simple methodology based
studies, an insight-based approach was used, whilst in the latter on a design insight for the designing of WN and HEN. They
one an optimization based approach was presented. Later, using the demonstrated that WN and HEN can be de-coupled and solved. The
above-mentioned approaches to heat integration within WN resulting network exhibited the minimum water and energy con-
problems, a more-active research area developed. Savulescu et al. sumption, and provided simple structures for single-contaminant
[12] studied simultaneous energy and water minimization with no problems. Liao et al. [28] introduced a step-wise systematic pro-
water re-use, and with maximum re-use of the water [13]. They cedure for the synthesis of heat-integrated water networks, and
explored the conceptual approach as well as direct and indirect proposed a procedure for the identification of hot and cold streams
heat-recovery opportunities, in order to identify a network design within WN, followed by targeting and design steps. During the
with minimum freshwater and energy consumption. In addition, targeting step, the identification of the promising matches between
they introduced a new grid representation, called the two- hot and cold streams was performed, whilst in the design step, a
dimensional grid diagram. Bogataj and Bagajewicz [14,15] pro- stage-wise superstructure was used to deal with the features of
posed an approach for the simultaneous synthesis of energy- mixing and splitting inside HEN. The problem was formulated as a
efficient WN using mathematical programming and superstruc- MINLP, and single-contaminant WN problems were considered.
ture optimization. They modified the HEN introduced by Yee et al. Sahu and Bandyopadhyay [29] formulated the heat-integrated
[16] for the case of mixing and splitting streams within HEN su- water network problem as linear programming in order to deter-
perstructure and combined with WN. The combined model was mine the minimum freshwater and utilities consumption. They
solved by a two-step solution strategy where in the first step a WN analyzed both the isothermal and non-isothermal mixings of
(NLP (non-linear programming)) model was solved following by a streams for single and multiple contaminant problems and pro-
solution of a combined WN þ HEN (MINLP) model during the posed some theorems. Boix et al. [30] presented an approach based
second step. The WN model was solved in order to realize good on mathematical programming to solve WN and HEN by consid-
initialization and to classify the streams into hot and cold. The ering several objectives such as freshwater consumption, energy
proposed approach was used for solving single and multiple consumption, interconnections number, and the number of heat
contaminant problems. The resulting network had a smaller exchangers. A two-step solution procedure was used for problem
number of heat exchangers and a smaller consumption of utilities solving, and single-contaminant problems were considered. Mar-
than the network reported in Savulescu et al. [13]. Leewongtanawit tínez-Patiño et al. [31,32] studied the interactions between water
238 E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250

and energy systems, and proposed a heuristic procedure for the with the temperatures being optimization variables (Fig. 1c). For
synthesis of the heat-integrated water networks. This approach is most of these streams it is impossible to determine in advance
based on the use of a temperature vs. concentration diagram. whether they should be characterized as hot or cold streams. As
Yiqing et al. [33] studied the effect of non-isothermal mixing on shown in Fig. 1c, i.e. a stream from the mixer process unit (point M)
water network’s energy performance and how the energy penalty can be sent to HEN for heat recovery as a cold or hot stream. It is
can be eliminated or avoided. They used a problem table algorithm worth pointing out that matches between these two streams are
[6] for heat integration to analyze the energy performance of sin- forbidden (see the first and the fifth matches from left to the right in
gle-contaminant WN problem. They showed that heterogenous Fig. 1c) in HEN and they can only exchange heat with other streams
mixing of water streams can decrease the utility consumption of in HEN. After heat recovery in HEN the streams are returned to WN.
WN, which is not case for the homogenous mixing. According to the In general, when WN consists of more process units, the number
published literature, a small number of contributions have of cold and hot streams increase linearly, and the number of matches
addressed the simultaneous synthesis of heat-integrated water in HEN exponentially. In addition, if multiple contaminants are
networks using mathematical programming. However, only this considered, the synthesis problem of the heat-integrated water
approach can simultaneously explore trade-offs between fresh- network can be very complex and difficult to solve. Accordingly, the
water cost, utility cost, and the investment costs for heat ex- development of methodologies and approaches for successfully
changers. This paper presents a new contribution and results in this dealing with these kinds of problems represents a big challenge for
field using simultaneous synthesis and the mathematical pro- the scientific community. In order to develop a combined super-
gramming approach based on superstructure optimization. structure model, the recently proposed WN model [34] has now
been extended for both direct and indirect heat exchanges, addi-
3. Problem statement tional freshwater and wastewater splitting and mixing, and com-
bined with the heat exchanger network model [16,35,36].
The general formulation of the heat-integrated water network
problem can be stated as follows. For a given a set of freshwater
5. Mathematical model
sources and a set of water-using units that require water of a certain
quality and temperature, it is necessary to determine the in-
The model of the superstructure for the simultaneous synthesis
terconnections, flowrates, contaminant concentrations, and the
of a HIPWNs (heat-integrated process water networks) consists of
temperatures of each stream within the network. The freshwater
mass balance equations for water and contaminants, and heat
sources can be with different qualities and temperatures. The
balance equations. In order to circumvent the problem with
maximum inlet and outlet concentrations of contaminant, the
streams’ characterizations, each stream within the network is
contaminant load to be transferred to the water stream, and the
represented by a couple of alternative hot and cold streams, and
operating temperatures for water-using units are specified. Single
described with its convex hull formulation for identifying its role in
and multiple contaminants are considered, and the temperature
the HEN. The HIPWN model can be efficiently solved based on this
given of the discharged wastewater. The standard assumptions are
formulation and connecting constraints. The proposed model is
used as given in the literature. The number of water sources and the
formulated as a non-convex MINLP (mixed-integer non-linear
number of water-using units are specified, the water streams have
programming) problem with additional 0e1 variables included
constant heat capacities and heat-transfer coefficients, the heat
for the identification of the streams’ roles in HEN. The non-linear-
exchangers are counter-current, the network operates continu-
ities within the models appear in the mass and heat balance
ously, with one hot and one cold utility being available. The main
equations in the forms of bilinear terms (flowrate multiplied by the
goal was to simultaneously synthesize the heat-integrated water
concentration and the flowrate multiplied by the temperature). The
network with the minimum total annual cost and the optimal
other non-convexities are the concave investment cost terms
consumptions of the freshwater and utilities.
within the objective function. The objective is to simultaneously
synthesize HIPWN with the minimal TAC (total annual cost).
4. Superstructure

Fig. 1 shows the development of a combined WN and HEN su- 5.1. Mathematical model of water network superstructure
perstructure from the general representation (Fig. 1a) of a heat-
integrated water network problem. The combined superstructure Fig. 2 presents a new general superstructure for WN including
is proposed for systematically addressing the simultaneous syn- water splitting and mixing, as well as direct and indirect heat transfer
thesis of the heat-integrated process water networks, and exploring between hot and cold streams. Based on this superstructure, the
strong interactions between WN and HEN. The combined super- mathematical model has been developed and combined with HEN.
structure consists of a WN superstructure, HEN superstructure, and
their complicated interconnecting streams, which can be hot, cold, 5.1.1. Initial splitters of freshwater
or bypass streams (Fig. 1b). Within this superstructure, the fresh- The freshwater of an initial splitter SIs from freshwater source
water stream can be preheated in HEN in order to satisfy the s ˛ SW can be sent to the mixer before each process unit and to the
temperature of the process units in WN, or it can be directly freshwater splitter SSp. The overall mass balance for the initial
introduced to WN as a bypass stream. The wastewater is generated splitter is given by Eq. (1), and the flowrate and temperature
after being used as freshwater in WN. The wastewater stream from equality constraints by Eqs. (2)e(4).
WN can be sent to HEN for heat recovery or as a bypass stream X
mixed with other wastewater streams from HEN, and then dis- FWs ¼ FWout
s þ FIPs;p cs˛SW (1)
charged into the environment. Therefore, the freshwater can be p˛PU

regarded as a cold or bypass stream, whilst the wastewater as a hot


FWout
s ¼ FSSin
p cs˛SW; cp˛PU; p ¼ 1 (2)
or bypass stream. Additionally, the outlet water stream of one
process unit can be re-used in another process unit. In addition to
TFWs ¼ TSSin
p cs˛SW; cp˛PU; p ¼ 1 (3)
these streams, streams from mixers to process units or streams
from process units to splitters can be hot, cold or bypass streams TFWs ¼ TIPs;p cs˛SW; cp˛PU (4)
E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250 239

Fig. 1. Development of a combined WN and HEN superstructure.


240 E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250

Fig. 2. Extended WN superstructure including direct and indirect heat transfer and additional opportunities for freshwater/wastewater splitting and mixing.

X X
5.1.2. Freshwater splitters FPUin in
p ,xPUp;c ¼ FSSMp0 ;p ,xSSout
p0 ;c þ FIP,xWin
s;c
The freshwater from the initial splitter SIs can be sent to HEN in p0 ˛PU s˛SW
order to heat and satisfy the temperature constraints of the process X
þ FPp0 ;p ,xSPUout
p0 ;c
units. In addition, the outlet stream of freshwater from HEN can be
p0 ˛PU
split and mixed with the cold stream of freshwater from the initial
psp0 ;Rp ¼ 0
splitter or re-used stream of water from the process units. The mass X
balance for the freshwater splitters is given by Eqs. (5)e(7), the þ FPp0 ;p ,xSPUout
p0 ;c cp˛PU; cc˛CC
contaminant concentration equality constraint by Eq. (8), and the 0
p ˛PU
temperature constraints by Eqs. (9) and (10). Rp ¼ 1

X (12)
FSSin out
p ¼ FSSp þ FSSMp;p0 cp; p0 ˛PU; psjPUj (5)
p0 X X
FPUin mix
p ,Tp ¼ FSSMp0 ;p ,TSSout
p0 þ FIP,TIPs;p
X p0 ˛PU s˛SW
0
FSSin
p ¼ FSSMp;p0 cp; p ˛PU; p ¼ jPUj (6) X
p0 þ FPp0 ;p ,TSPUout
p0
0
p ˛PU
psp0 ;Rp ¼ 0
FSSout
p ¼ FSSin
pþ1 cp˛PU; psjPUj (7) X
þ FPp0 ;p ,TSPUout
p0 cp˛PU (13)
p0 ˛PU
xWin out
s;c ¼ xSSp0 ;c cs˛SW; cp0 ˛PU; cc˛CC (8) Rp ¼ 1

TSSout
p ¼ TSSin
pþ1 cp˛PU; psjPUj (9) 5.1.4. Process units
The process unit PUp consists of an inlet stream FPUin
p from the
mixer process unit and an outlet stream FPUout p from the process
TSSin out
p  TSSp cp˛PU (10) unit. The water flowrate through the process units is assumed to be
a continuous variable. The outlet stream from PUp is directed to the
splitter process unit, and can be a hot, cold or bypass stream. The
5.1.3. Mixer process units mass balance is given by Eq. (14), the mass balance for each
The outlet stream of the mixer process unit MPUp consists of a contaminant c by Eq. (15), and the equality contaminant concen-
set of inlet streams from the initial splitter, the splitter process unit, tration by Eq. (16).
and the freshwater splitter. This stream is directed to the process
FPUin out
p ¼ FPUp cp˛PU (14)
unit, and can be a hot, cold or bypass stream. The mass balance for
the mixer is given by Eq. (11), the mass balance for each contami-
nant c by Eq. (12) and the heat balance by Eq. (13). FPUin in out out
p ,xPUp;c þ LPUp;c ¼ FPUp ,xPUp;c cp˛PU; cc˛CC (15)

X X X in
FPUin
p ¼ FSSMp0 ;p þ FIPs;p þ FPp0 ;p xPUout
p;c ¼ xSPUp;c cp˛PU; cc˛CC (16)
p0 ˛PU s˛SW p0 ˛PU
psp0 ;Rp ¼ 0
X 5.1.5. Splitter process units
þ FPp0 ;p cp˛PU (11) The splitter process unit SPUp consists of an inlet stream from
p0 ˛PU the process unit, and a set of outlet streams directed to the final
Rp ¼ 1 mixer MF, the mixer process unit MPUp, and the wastewater mixer
E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250 241

X
MMFp. The mass balance for the splitter process unit is given by Eq. FMMout out
FPOp0 ,xSPUout out out
p ,xMMp;c þ p0 ;c ¼ F ,xc ; p ¼ 1;cc˛CC
(17) and the equality contaminant concentration by Eq. (18). p0 ˛PU
X X (28)
FPUout
p ¼ FSMMp;p0 þ FPOp þ FPp;p0
p0 ˛PU p0 ˛PU X
psp0 ;Rp ¼ 0
FMMout out
p ,TMMp þ FPOp0 ,TSPUout
p0
X p0 ˛PU
þ FPp;p0 cp˛PU (17)
¼ F out ,T out ; cp˛PU; p ¼ 1 (29)
p0 ˛PU
Rp ¼ 1

5.1.8. Total mass balance and contaminant mass balance for the
xSPUout in
p;c ¼ xSPUp;c cp˛PU; cc˛CC (18)
network
The total mass balance for the network is given by Eq. (30) and
5.1.6. Wastewater mixers the total contaminant mass balance by Eq. (31).
The wastewater mixer MMFp consists of a set of inlet streams X
from the splitter process unit. The outlet stream of the wastewater FWs ¼ F out (30)
s˛SW
mixer is sent to HEN for heat integration. In addition, the outlet
stream of wastewater from HEN can be mixed with wastewater X X
streams from the process units. The mass balance for the waste- FWs ,xWin
s;c þ LPUp;c ¼ F out ,xout
c ; cc˛CC (31)
water mixer is given by Eqs. (19)e(21), the mass balance equation s˛SW p˛PU
for each contaminant c by Eqs. (22) and (23), the heat balance by
Eqs. (24) and (25), and the temperature constraint by Eq. (26).
5.2. Convex hull formulation for the identification of streams for HEN
X
FMMin
p þ FSMMp0 ;p ¼ FMMout
p cp˛PU; psjPUj (19)
p0 ˛PU In WN there are certain streams i.e. after the mixer process units
or after process units, the temperatures of which are optimization
X variables and thus unknown. For these streams it is impossible to
FSMMp0 ;p ¼ FMMout
p cp˛PU; p ¼ jPUj (20)
determine whether they should be characterized in HEN as hot or
p0 ˛PU
cold streams ahead of the optimization. In order to circumvent this
problem when identifying the hot and cold streams, a new convex
FMMin out
p ¼ FMMpþ1 cp˛PU; psjPUj (21) hull formulation is presented in this section for the identification of
streams’ roles in the HEN.
X
FMMout out
pþ1 ,xMMpþ1;c þ FSMMp0 ;p ,xSPUout
p0 ;c 5.2.1. Streams from mixer to process unit
p0 ˛PU
Fig. 3 shows the superstructure of water streams (hot, cold, and
¼ FMMout out
p ,xMMp;c cp˛PU; psjPUj (22) bypass) after the mixer process unit. The inlet water streams to the
mixer process unit can be different temperatures and involved in
X direct heat transfer by mixing. The temperature of the mixed
FSMMp0 ;p ,xSPUout out out
p0 ;c ¼ FMMp ,xMMp;c cp˛PU; p ¼ jPUj
p0 ˛PU
stream of water can be:

(23)
 Tpmix > TPUin
p (hot stream),
X  Tpmix < TPUin
p (cold stream) or
FMMout out
pþ1 ,TMMpþ1 þ FSMMp0 ;p ,TSPUout
p0  Tpmix ¼ TPUin
p (bypass stream).
p0 ˛PU

¼ FMMout in
p ,TMMp cp˛PU; psjPUj (24) According to this, the temperature of the mixed stream is
optimization variable and the mathematical formulation is
X required for identifying the hot and cold streams as follows.
FSMMp0 ;p ,TSPUout
p0 ¼ FMMout in
p ,TMMp cp˛PU; p ¼ jPUj
The inequality binary constraint for the selection of hot or cold
p0 ˛PU
streams is given by Eq. (32) and the temperature constraints by Eqs.
(25) (33)e(37). Note, if a stream is unselected, then both binary vari-
ables are set to zero and the stream is sent as a bypass from the
TMMout in
p  TMMp cp˛PU (26) mixer to the process unit (from MPUp to PUp in Fig. 3).

5.1.7. Final mixer


The final mixer MF consists of a set of inlet streams from the
splitter process unit, and the outlet stream from the first waste-
water mixer. The mass balance for the final mixer is given by Eq.
(27), the mass balance equation for each contaminant c by Eq. (28)
and the heat balance by Eq. (29).
X
FMMout
p þ FPOp0 ¼ F out (27)
p0 ˛PU Fig. 3. Superstructure of streams (hot, cold, and bypass) after the mixer process unit.
242 E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250

hp cp 5.3. Mathematical model of a heat exchanger network


yp þ yp  1; p˛PU (32)
superstructure
 
cp hp
Tpmix ¼ T hot;in
i
þ Tjcold;in þ TPUin
p , 1  yp  yp ; The presented WN model is combined with the HEN model
(33) introduced by Yee et al. [16]. In WN and HEN models, the flowrates
p˛PU; i˛HP; j˛CP; i ¼ p; j ¼ p
and temperatures of the streams are treated as optimization vari-
ables, so that trade-offs can be obtained between freshwater cost,
hp
Tihot;in  TPUin;max
p ,yp ; p˛PU; i˛HP; i ¼ p (34) utility cost, and the investment cost for heat exchangers. The
mathematical model of HEN is given by Eqs. (44)e(63), as follows.
hp
Tihot;in  TPUin
p ,yp ; p˛PU; i˛HP; i ¼ p (35) 5.3.1. Overall heat balance for each stream

X X
Tjcold;in  TPUin
cp ðthini  thouti Þ,fhi ¼ qi;j;k þ qcui i˛HP (44)
p ,yp ; p˛PU; j˛CP; j ¼ p (36)
k˛ST j˛CP

  X X
Tjcold;in  TFWmin
s ,ycp
p ; p˛PU; j˛CP; j ¼ p (37) tcoutj  tcinj ,fcj ¼ qi;j;k þ qhuj ; j˛CP (45)
k˛ST i˛HP

5.2.2. Streams from process unit to splitter


Fig. 4 shows the superstructure of streams (hot, cold, and 5.3.2. Heat balance at each stream
bypass) after the process unit. The temperature of the outlet stream
of the splitter process unit SPUp can be:   X
thi;k  thi;kþ1 ,fhi ¼ qi;j;k ; i˛HP; k˛ST (46)
 TSPUout in
p > TPUp (hot stream),
j˛CP

 TSPUout in
p < TPUp (cold stream) or
  X
 TSPUmix in tcj;k  tcjþ1 ,fcj ¼ qi;j;k ; j˛CP; k˛ST (47)
out ¼ TPUp (bypass stream).
i˛HP

The mathematical formulation for identifying the hot and cold


streams after the process unit is given, as follows. The inequality 5.3.3. Assignment of superstructure inlet temperatures
binary constraint for the selection of hot or cold streams is given by
Eq. (38) and the temperature constraints by Eqs. (39)e(43). Note
again, if a stream is unselected, then both binary variables are set to thini ¼ thi;1 ; i˛HP (48)
zero and the stream is sent as a bypass from the process unit to the
splitter (from PUp to SPUp in Fig. 4). tcinj ¼ tcj;NOKþ1 ; j˛CP (49)

yhps
p þ ycps
p  1; p˛PU (38)
5.3.4. Feasibilities of temperatures
 
cps hps
TSPUout
p ¼ T ihot;out þ Tjcold;out þ TPUout
p , 1  yp  yp ; thi;k  thi;kþ1 ; i˛HP; k˛ST (50)
p˛PU; i˛HP; j˛CP; i ¼ p þ jPUj; j ¼ p þ jPUj
(39) tcj;k  tcj;kþ1 ; j˛CP; k˛ST (51)

Tihot;out  TPUout hps thouti  thi;NOKþ1 ; i˛HP; k˛ST (52)


p ,yp ; p˛PU; i˛HP; i ¼ p þ jPUj (40)

tcoutj  tcj;1 ; j˛CP; k˛ST (53)


Tihot;out  TFWmin
s ,yhps
p ; p˛PU; i˛HP; i ¼ p þ jPUj (41)

cps 5.3.5. Hot and cold utility load


Tjcold;out  TPUout;max
p ,yp ; p˛PU; j˛CP; j ¼ p þ jPUj (42)
 
cps thi;NOKþ1  thouti ,fhi ¼ qcui ; i˛HP (54)
Tjcold;out  TPUout
p ,yp ; p˛PU; j˛CP; j ¼ p þ jPUj (43)
 
tcoutj  tci;1 ,fcj ¼ qhuj ; j˛CP (55)

5.3.6. Logic constraints

qi;j;k  U,zi;j;k  0; i˛HP; j˛CP; k˛ST (56)

qcui  U,zcui  0; i˛HP (57)

Fig. 4. Superstructure of streams (hot, cold, and bypass) after the process unit. qhuj  U,zhuj  0; j˛HP (58)
E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250 243

5.3.7. Logic constraints for temperature differences fcj ¼ FPUout


p ,Cp ; cp˛PU; j˛CP; j ¼ p þ jPUj (73)
 
Dti;j;k  thi;k  tcj;k þ G, 1  zi;j;k ; i˛HP; j˛CP; k˛ST (59) tcinj ¼ TPUout
p ; cp˛PU; j˛CP; j ¼ p þ jPUj (74)

   cps 
Dti;j;kþ1  thi;kþ1 tcj;kþ1 þ G, 1  zi;j;k ; i˛HP;j˛CP;k˛ST (60) tcoutj ¼ Tjcold;out þ TPUout
p , 1  yp ; cp˛PU; j˛HP;
(75)
j ¼ p þ jPUj
Dti;CU  thi;NOKþ1  TOUTCU þ G,ð1  zcui Þ; i˛HP (61)

  5.4.3. Connecting equations for HEN for freshwater streams


Dtj;HU  TOUTHU  tcj;1 þ G, 1  zhuj ; j˛CP (62) between freshwater splitters

Dti;j;k ; Dti;CU ; Dtj;HU  EMAT; i˛HP; j˛CP; k˛ST (63)


fcj ¼ FSSin
p ,Cp ; cp˛PU; j˛CP; j ¼ p þ 2,jPUj (76)
Note that all the temperature driving forces within exchangers,
coolers, and heaters are optimization variables. Logical constraint tcinj ¼ TSSin
p ; cp˛PU; j˛CP; j ¼ p þ 2,jPUj (77)
Eq. (63) puts the EMAT (exchanger minimum allowable tempera-
ture) as a lower bound on them. Note also that Eqs. (59)e(62)
define positive driving forces as differences between the temper- tcoutj ¼ TSSout
p ; cp˛PU; j˛CP; j ¼ p þ 2,jPUj (78)
atures of hot and cold streams when matches are selected (binary
variable z ¼ 1) or make this relation also redundant for rejected
matches (z ¼ 0) for those cases when actual differences between 5.4.4. Connecting equations for HEN for wastewater streams
streams’ temperatures can be negative. between wastewater mixers

5.4. Connecting equations between WN and HEN fhi ¼ FMMout


p ,Cp ; cp˛PU; i˛HP; i ¼ p þ 2,jPUj (79)

Based on flowrates, temperatures, and binary variables from


thini ¼ TMMin
p ; cp˛PU; i˛HP; i ¼ p þ 2,jPUj (80)
WN, connecting equations, Eqs. (64)e(89), relating to the heat ca-
pacity flowrates of hot and cold streams and their corresponding
inlet and outlet temperatures can be defined in order to simulta- thouti ¼ TMMout
p ; cp˛PU; i˛HP; i ¼ p þ 2,jPUj (81)
neously solve the combined model of HIPWN minimizing the TAC
of the network.
5.4.5. Logical constraints for connecting streams in WN in HEN
5.4.1. Connecting equations for HEN for streams from mixer to
process unit
zi;j;k  yhp
p ; p˛PU; i˛HP; j ¼ CP; k˛ST; i ¼ p (82)

fhi ¼ FPUin
p ,Cp ; cp˛PU; i˛HP; i ¼ p (64) cp
zi;j;k  yp ; p˛PU; i˛HP; j ¼ CP; k˛ST; j ¼ p (83)
 
hp
thini ¼ Tihot;in þ TPUin
p , 1  yp ; cp˛PU; i˛HP; i ¼ p (65) hp
zcui  yp ; p˛PU; i˛HP; i ¼ p (84)

thouti ¼ TPUin
p ; cp˛PU; i˛HP; i ¼ p (66) zhuj  ycp
p ; p˛PU; j˛CP; j ¼ p (85)

fcj ¼ FPUin
p ,Cp ; cp˛PU; j˛CP; j ¼ p (67) hps
zi;j;k  yp ; p˛PU; i˛HP; j ¼ CP; k˛ST; i ¼ p þ jPUj (86)
 cp 
tcinj ¼ Tjcold;in þ TPUin
p , 1  yp ; cp˛PU; j˛CP; j ¼ p (68) zi;j;k  ycps
p ; p˛PU; i˛HP; j ¼ CP; k˛ST; j ¼ p þ jPUj (87)

tcoutj ¼ TPUin
p ; cp˛PU; j˛CP; j ¼ p (69) zcui  yp ;
hps
p˛PU; i˛HP; i ¼ p þ jPUj (88)

cps
5.4.2. Connecting equations for HEN for streams from process unit zhuj  yp ; p˛PU; j˛CP; j ¼ p þ jPUj (89)
to splitter

5.4.6. Objective function of the proposed model


fhi ¼ FPUout The objective function of the model, given by Eq. (90) and
p ,Cp ; cp˛PU; i˛HP; i ¼ p þ jPUj (70)
formulated to minimize TAC, consists of cost terms for freshwater,
utilities, and the investment for the heat exchangers. The cost terms
thini ¼ TPUout
p ; cp˛PU; i˛HP; i ¼ p þ jPUj (71) for the investment in heat exchangers consist of a fixed charge for
exchanger units and an area cost for heat exchangers, and these
 
hps terms are calculated as given in Eq. (90). In order to make a com-
thouti ¼ Tihot;out þ TPUout
p , 1  yp ; cp˛PU; i˛HP; i ¼ p þ jPUj
parison between ours and others’ solutions, the annualized capital-
(72) cost model for a conventional shell-and-tube heat exchanger is
244 E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250

taken from Dong et al. [21], and given as 8000zi;j þ 1200A0:6i;j where HRAT (heat recovery approach temperature), as all temperature
the heat-transfer area Ai,j is in m2. The annual interest rate is set at driving forces are now considered as optimization variables.
10% and the plant is assumed to be operated continuously for
8000 h a year.
6. Solution strategy
X X X
TAC ¼ H, FWs ,CFWs þ CCU ,qcui þ CHU ,qhuj The proposed model of HIPWN is formulated as a non-convex
s˛SW i˛HP j˛CP
X X X X MINLP (mixed-integer non-linear programming) problem with
þ CFi;j ,zi;j;k þ CFi;CU ,zcui additional 0e1 variables included for the identification of cold and
i˛HP j˛CP k˛ST i˛HP hot streams. Due to its non-convex and highly non-linear nature,
X X X X B the MINLP problem is very difficult to solve especially if the desire
þ CFj;HU ,zhuj þ Ci;j ,Ai;j;k
i;j

j˛CP i˛HP j˛CP k˛ST


is to explore all interactions within the combined network. How-
X Bi;CU
X B
ever, by using the proposed model we managed to solve the com-
þ Ci;CU ,Ai;CU þ Cj;HU ,Aj;HU
j;HU
bined HIPWN whilst simultaneously minimizing the total annual
i˛HP j˛CP costs. It is worth mentioning, that the problem can be solved using
(90) two solution approaches. The first one is by directly solving the
The A (areas), LMTD (temperature differences), and U (overall combined network as one system. In the second approach, WN is
heat-transfer coefficients) of the exchangers, coolers, and heaters solved first, where the objective is to minimize freshwater con-
are given by Eqs. (91)e(99). sumption and provide a good initial point. After that the overall
qi;j;k HIPWN is solved. The problem is modeled in GAMS (General
Ai;j;k ¼ ; i˛HP; j˛CP; k˛ST (91) Algebraic Modeling System) [37] and for model solving different
Ui;j ,LMTDi;j;k
solvers from GAMS library were used, like BARON (Branch-And-
Reduce Optimization Navigator) and SBB (Simple Branch-and-
qi;CU Bound). In the second approach BARON was used for WN and SBB
Ai;CU ¼ ; i˛HP (92)
Ui;CU ,LMTDi;CU for the combined HIPWN. To illustrate the capabilities of the pro-
posed model and solution strategy, we solved three examples
qj;HU including single and multiple contaminants. In all examples, the
Aj;HU ¼ ; j˛CP (93) resulted networks exhibited lower total annual costs compared to
Uj;HU ,LMTDj;HU
the reported results in the literature, because the superstructure
2 captured additional opportunities for heat integration.
 31=3
  Dti;j;k þ Dti;j;kþ1
6 7
LMTDi;j;k ¼ 4 Dti;j;k ,Dti;j;kþ1 , 5 ; 7. Application examples
2 (94)

i˛HP; j˛CP; k˛ST This section presents three examples involving single and
multiple contaminant problems to illustrate the applicability and
" capabilities of the proposed optimization model. The examples
  were implemented in GAMS [37] and solved on a PC (personal
LMTDi;CU ¼ Dti;CU ,ðTOUTi  TINCU Þ ,
computer) machine (2.67 GHz, 8 GB RAM (random-access mem-
 #1=3 ory)), with reasonable computation time (in less than 70 CPUs). In
Dti;CU þ ðTOUTi  TINCU Þ the first and second examples, the model consisted of 1031 con-
; i˛HP (95)
2 straints, 928 continuous variables, and 176 discrete variables. The
model of the third example consisted of 1075 constraints, 970
" continuous variables, and 176 discrete variables. The cost and
   operating parameters given in Table 1 [21] were used in all exam-
LMTDj;HU ¼ Dtj;HU , TINHU  TOUTj ,
ples. Examples have been solved by several authors using different
  #1=3 models and solution strategies based on pinch analysis or mathe-
Dtj;HU þ TINHU  TOUTj matical programming. However, in all cases, networks obtained
; j˛CP (96)
2 using the approach presented in this paper had lower total annual
costs compared to results in the literature. In addition, the resulting
1 1 1
¼ þ ; i˛HP; j˛CP (97) Table 1
Ui;j hi hj Cost and operating parameters for Examples 1e3.

Parameter
1 1 1 Freshwater cost 0.375 $/t
¼ þ ; i˛HP (98) Cooling utility (cooling water) cost 189 $/(kW a)
Ui;CU hi hCU
Heating utility (low pressure steam, 120  C) cost 377 $/(kW a)
Fixed charge for heat exchangers 8000 $
1 1 1 Area cost coefficient for heat exchangers 1200 $/m2
¼ þ ; j˛CP (99) Cost exponent for exchangers 0.6
Uj;HU hj hHU
Overall heat-transfer coefficient (individual 0.5 kW/(m2  C)
heat-transfer coefficients for streams and
This model enables the obtaining of an appropriate trade-off
utilities were assumed to be 1 kW(m2  C)
between utility consumption, water usage, and investment. In Working hours of plant per year 8000 h
cases with temperature pinches it, in principle, allows for a reduc- Temperature of freshwater 20  C
tion in energy consumption when compared to existing models, Temperature of wastewater 30  C
since the heat integration applied is performed simultaneously and The inlet and outlet temperature of cooling water 10  C and 20  C
Specific heat capacity of water 4.2 kJ/(kg  C)
the proposed model does not rely on the assumption of a fixed
E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250 245

networks had simpler designs with the same or a smaller number has been studied by several authors [9,11e13,15,21,24,25,27,28,31].
of heat exchangers. In order to compare the presented model’s In all the reported networks freshwater consumption was the same
results with the results published in the literature, data from (90 kg/s), whilst in Savulescu et al. [13] both the hot utility con-
Table 1 [21] was used to recalculate costs (freshwater cost, hot sumption (4265 kW) and the cold utility consumption (485 kW)
utility cost, cold utility cost, investment cost, and finally the total were greater than in other cases (3780 kW and 0 kW). In addition,
annual cost) for all cases taken from the literature. the reported networks had different numbers of heat exchangers
(see Table 3).
7.1. Example 1 The superstructure for this example is given in Fig. 5 and con-
sists of all possible connections between the splitting points of
Example 1 is a well-known case study given by Savulescu and sequentially heating-up freshwater, process units, and the mixing
Smith [9] and Savulescu et al. [12,13], and involves WN with four points of sequentially cooling discharged wastewater. The heat
processes and a single contaminant. Data for the processes, con- transfer between the hot and cold streams in the superstructure is
taminants, and temperatures are presented in Table 2. This problem possible through direct heat transfer (mixing of water streams), and
indirect heat transfer (through the heat exchanger area). It is worth
Table 2 pointing out that allowing for additional water splitting and mixing
Water-using operation data. increases the possibilities for direct heat transfer. Consequently,
Process Contaminant Maximum inlet Maximum outlet Temperature networks can be obtained with a reduced number of heat ex-
number load (g/s) concentration concentration ( C) changers (see Tables 3 and 4). In Case c, i.e. when one heat
(ppm) (ppm) exchanger and one heater are selected within a network with the a
1 2 0 100 40 total annual cost of 2,668,035 $/a, the heat-transfer area of the
2 5 50 100 100 exchanger is very large (about 5938 m2) and impractical because
3 30 50 800 75
4 4 400 800 50
the shell-and-tube heat exchangers usually have heat-transfer
areas [38,39] of up 4000 m2 or 4647 m2. However, it may be

Table 3
The results for the freshwater and utilities consumption, the total heat flow exchanged in the heat exchangers, and the total area of the heat exchangers.

Method proposed by the authors FW (kg/s) HU (kW) CU (kW) Equipment requirements (/) Qtotal (kW) Atotal (m2)
Savulescu et al. [13] 90 4265 485 3 Heat exchangers, 1 heater, 1 cooler 23,585 4549
Xiao et al. [25] 90 3780 0 3 Heat exchangers, 2 heaters 26,040a 4690
Polley et al. [27]
Case a 90 3780 0 3 Heat exchangers, 2 heaters 26,040 4690
Case b 90 3780 0 3 Heat exchangers, 3 heaters 23,940 4258
Case c 90 3780 0 4 Heat exchangers, 4 heaters 23,100 4088
Martínez-Patiño et al. [31] 90 3780 0 3 Heat exchangers, 1 heater 26,040 4722
Dong et al. [21] 90 3780 0 4 Heat exchangers, 1 heater 22,680 4050
Bagajewicz et al. [11] 90 3780 0 3 Heat exchangers, 1 heater 22,008 3782
Bogataj and Bagajewicz [15] 90 3780 0 3 Heat exchangers, 1 heater 22,008 3782
Leewongtanawit and Kim [24] 90 3780 0 3 Heat exchangers, 1 heater 22,260 3776
Liao et al. [28]
Case a 90 3780 0 3 Heat exchangers, 1 heater 22,008 3633
Case b 90 3780 0 4 Heat exchangers, 1 heater 25,102 5531
This paper
Case a 90 3780 0 3 Heat exchangers, 1 heater 22,400 3424
Case b 90 3780 0 2 Heat exchangers, 1 heater 22,344 3961
Case c 90 3780 0 1 Heat exchanger, 1 heater 23,688 5938
a
Note that in Xiao et al. [25] the total amount of heat flow should be 26,040 kW instead of 22,480 kW (see Fig. 11 in Xiao et al. [25]).

Fig. 5. Superstructure of HIPWN for Example 1.


246 E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250

Table 4
The results of the freshwater and utilities costs, the investment cost for heat exchangers and the total annual cost.

Method proposed by the authors FWC ($/a) CUC ($/a) HUC ($/a) IC ($/a) TAC ($/a)
Savulescu et al. [13] 972,000 91,665 1,607,905 369,236 3,040,806
Xiao et al. [25] 972,000 0 1,425,060 364,587 2,761,647
Polley et al. [27]
Case a 972,000 0 1,425,060 364,587 2,761,647
Case b 972,000 0 1,425,060 344,905 2,741,965
Case c 972,000 0 1,425,060 379,349 2,776,409
Martínez-Patiño et al. [31] 972,000 0 1,425,060 341,919 2,738,979
Dong et al. [21] 972,000 0 1,425,060 341,047 2,738,107
Bagajewicz et al. [11] 972,000 0 1,425,060 314,495 2,711,555
Bogataj and Bagajewicz [15] 972,000 0 1,425,060 314,495 2,711,555
Leewongtanawit and Kim [24] 972,000 0 1,425,060 310,293 2,707,353
Liao et al. [28]
Case a 972,000 0 1,425,060 304,644 2,701,704
Case b 972,000 0 1,425,060 375,039 2,772,099
This paper
Case a 972,000 0 1,425,060 293,165 2,690,225
Case b 972,000 0 1,425,060 255,899 2,652,959
Case c 972,000 0 1,425,060 270,975 2,668,035

possible to install multiple units, or use plate and frame exchangers and the corresponding annual costs for freshwater, utilities and
which are more compact, have lower cost and more flexibility investment in Table 4. It is interesting to note that, in all cases,
(extra plates can be added) compared to shell-and-tube heat ex- better network designs (with lower TAC and simpler design) were
changers. Compact heat exchangers have up to five times higher obtained with the proposed method in this paper compared to the
heat-transfer efficiency than shell-and-tube heat exchangers [40]. published results in the literature. All the resulting networks
However, it should be mentioned that the largest plate-shell-type exhibited the same freshwater consumption (90 kg/s) and hot
heat exchangers can have heat exchange surfaces of up to (3780 kW) and cold utility (0 kW) consumption as reported in the
10,500 m2 per single unit and have already been successfully used literature but TAC, (i.e. for Case a: 2,690,225 $/a, or Case b:
for energy saving within industries. 2,652,959 $/a) were lower due to smaller investment costs for heat
Tables 3 and 4 show the results from this case study as obtained exchangers. Accordingly, Fig. 6 presents a novel network design for
by using the approaches in the literature and the proposed this case study with the minimal TAC 2,652,959 $/a, involving three
approach given in this paper. The results for FW (freshwater) and exchangers (two heat exchangers and one heater). The selected
utilities (HU (hot utility) and CU (cold utility)) consumptions, the matches for indirect heat integration are between the inlet cold and
Qtotal (total heat flow) exchanged within the heat exchangers, and outlet hot streams of the process unit with the highest temperature
the Atotal (total area) for all exchangers, are summarized in Table 3, (100  C), and the cold freshwater and the hot wastewater streams.

Fig. 6. A novel network design with a smaller number of heat exchangers for Example 1.
E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250 247

In addition, direct heat recovery between water streams of different of scale. In order to compare their results with the results in this
temperatures is selected within the network (i.e. see mixing points paper, using the cost and operating parameters from Table 1, the
of PU1, PU2, PU3 as well as the mixing points of the wastewater freshwater cost was recalculated, together with the hot utility cost,
streams). investment cost for heat exchangers, and TAC for the network
design given in their paper (see Table 5). We solved this problem
7.2. Example 2 and obtained a simpler network with only one heat exchanger and
one heater (see Fig. 7). The selected match for heat integration
In this problem the contaminant loads are scaled-down by a within the network is between the freshwater cold stream and the
factor of 3.6, whilst all other data are the same as given in Example wastewater hot stream. The total heat duty and total area of heat
1. The network superstructure for this problem is the same as given exchangers are 6580 kW and 1725 m2. Due to economy of scale and
in Example 1. Bogataj and Bagajewicz [15] solved this problem additional opportunities for heat integration within the network, as
using the two-step solution strategy and obtained HIPWN with two well as the water splitting and mixing in the presented solution, the
heat exchangers and one heater. In their solution the in- investment costs for the heat exchanger (134,227 $/a) and TAC of
terconnections between water-using process units were the same the network (800,077 $/a) were somewhat lowered compared to
as in the network for the first example (three heat exchangers and the result in Bogataj and Bagajewicz [15] (see Table 5).
one heater) but the topology of HEN was changed due to economy
7.3. Example 3
Table 5
Results for Example 2.
This problem is an extension of the second one, but now con-
Recalculated results from This paper siders multiple contaminants (A, B, C). The data for the problem
Bogataj and Bagajewicz [15] (see Fig. 7)
(Table 6) are taken from Bogataj and Bagajewicz [15] where the
Heat exchangers cost ($/a) 146,748 134,227
HIPWN with freshwater consumption (95.527 t/h z 26.535 kg/s),
Freshwater cost ($/a) 270,000 270,000
Hot utility cost ($/a) 395,850 395,850 three heat exchangers and one heater is given. Their results were
Total annual cost ($/a) 812,598 800,077 recalculated using data from Table 1 in order to compare with the
results obtained using the proposed model (see Table 7). The

Fig. 7. Optimal design of HIPWN with a smaller number of heat exchangers for Example 2.

Table 6
Data for Example 3.

Process Contaminant load (g/s) Maximum inlet concentration Maximum outlet concentration Temperature ( C)
number (ppm) (ppm)

A B C A B C A B C
1 2 1 3 0 15 0 100 100 100 40
2 5 0 15 50 100 30 100 200 250 100
3 30 4 0 100 100 100 800 750 600 75
4 4 22 17 400 380 250 800 800 800 50
248 E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250

Table 7 costs and a reduced number of heat exchangers compared to the


Results for Example 3. results given in the literature, due to the incorporated additional
Recalculated results from This paper opportunities for heat integration within the superstructure. Future
Bogataj and Bagajewicz [15] (see Fig. 8) research will be directed towards an extension of the superstructure
Heat exchangers cost ($/a) 199,861 190,890 by allowing for both isothermal and non-isothermal heat transfer
Freshwater cost ($/a) 286,581 286,579 between the process-to-process streams in order to explore some
Hot utility cost ($/a) 420,624 420,157
additional opportunities for heat integration, e.g. multiple utility
Total annual cost ($/a) 907,066 897,626
configurations. In addition, the effect of direct and indirect heat

Fig. 8. Optimal design of HIPWN with three heat exchangers and one heater for Example 3.

obtained network design given in Fig. 8 involves three heat ex- transfer, and freshwater/wastewater splitting and mixing on heat
changers and one heater which is the same number of heat ex- recovery within the network would be interesting to explore.
changers as given in Bogataj and Bagajewicz [15]. However, the TAC
of the network in our case (897,626 $/a) was lower than the TAC Acknowledgments
(907,066 $/a) of Bogataj and Bagajewicz [15] mainly due to the
smaller investment cost for heat exchangers. It is interesting to note The authors are grateful to the Scholarship scheme for academic
that in the proposed network design the inlet and outlet streams of exchange between the EU and Western Balkan countries for a
PU4 are hot streams. Additionally, the inlet and outlet streams of JoinEU-SEE postdoctoral fellowship. Financial support is also
the process units PU1 and PU4 play an important role within the gratefully acknowledged from the Slovenian Research Agency
network’s heat integration. (Program No. P2-0032), and the Federal Ministry of Education and
Science, Bosnia and Herzegovina.
8. Conclusions
Nomenclature
This paper addressed the simultaneous synthesis of heat-
integrated process water networks. A new combined superstruc-
ture for process water and heat exchanger networks has been pre- Indices
sented involving additional opportunities for direct and indirect c contaminant
heat transfer and freshwater and wastewater splitting and mixing. i hot process stream
These opportunities have been systematically incorporated into the j cold process stream
mathematical model, which is formulated as a MINLP problem. Also, k index for stage and temperature location
a new mathematical formulation has been proposed for identifying p process unit
hot and cold streams (streams from mixers to process units, and s freshwater source
streams from process units to splitters) in HEN. Example problems
have clearly indicated that the proposed method can be successfully
used for single and multiple contaminant water network problems. Sets
In addition, the model can produce an efficient and better design of CC contaminants
heat-integrated process water networks with lower total annual CP cold process streams
E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250 249

HP hot process streams TSSin


p temperature of inlet freshwater stream to heat
PU process units exchanger, K
ST stages in the HEN superstructure TSSout
p temperature of outlet freshwater stream from heat
SW freshwater sources exchanger, K
TSPUout
p temperature of outlet water stream from splitter process
unit, K
Parameters Tpmix temperature of outlet stream from mixer process unit, K
B exponent for area cost TMMin p temperature of inlet wastewater stream to wastewater
C area cost coefficient, $/m2 mixer, K
CCU per unit cost for cold utility, $/(W a) TMMout
p temperature of outlet wastewater stream from
CF fixed charge for exchangers, $ wastewater mixer, K
CFWs cost of freshwater from source s, $/kg thi,k temperature of hot stream i at hot end of stage k, K
CHU per unit cost for hot utility, $/(W a) tcj,k temperature of cold stream j at hot end of stage k, K
Cp heat capacity of water, J/(kg K) Dti,j,k temperature approach for the match (i, j) at temperature
EMAT minimum approach temperature, K location k, K
H hours of plant operation per annum, h Dti,CU temperature approach for the match of hot stream i and
h individual heat-transfer coefficients, W/(m2 K) cold utility, K
LPUp,c load of contaminant c in process unit p, kg/s Dtj,HU temperature approach for the match of hot utility and
Rp local recycle around process unit p (Rp ¼ 0 does not exist, cold stream j, K
Rp ¼ 1 if exists) qi,j,k heat flow exchanged between hot process stream i and
TFWs temperature of freshwater source s, K cold process stream j in stage k, W
TPUinp temperature at inlet of process unit p, K qcui heat flow exchanged between hot process stream i and
TPUout
p temperature at outlet of process unit p, K cold utility, W
TIN inlet temperature of utility stream, K qhuj heat flow exchanged between hot utility and cold process
TOUT outlet temperature of utility stream, K stream j, W
TIPs,p temperature of freshwater stream from source s to mixer xout
c concentration of contaminant c in discharge stream to the
process unit, K environment, ppm
Tout temperature of outlet stream from final mixer, K xMMout
p;c concentration of contaminant c in outlet stream from
TFWmins minimum temperature of freshwater source s, K wastewater mixer, ppm
U overall heat-transfer coefficient, W/(m2 K) xPUin
p;c concentration of contaminant c in inlet stream to process
xPUin;max
p:c maximum concentration of contaminant c in inlet stream unit p, ppm
to process unit, ppm xPUout
p;c concentration of contaminant c in outlet stream from
xPUout;max
p;c maximum concentration of contaminant c in outlet process unit p, ppm
stream from process unit, ppm xSPUin
p;c concentration of contaminant c in inlet stream to splitter
xSSout
p0 ;c concentration of contaminant c in freshwater in outlet process unit p, ppm
stream from splitter, ppm xSPUout
p;c concentration of contaminant c in outlet stream from
xWins;c concentration of contaminant c in freshwater source s, ppm splitter process unit, ppm
G upper bound for temperature difference
U upper bound for heat exchange
Binary variables
hp
yp existence of hot stream from mixer to process unit
ycp
p existence of cold stream from mixer to process unit
Continuous variables
hps
fc, fh heat capacity flowrate of cold and hot stream, W/K yp existence of hot stream from process unit to splitter
FIPs,p mass flowrate of water stream from freshwater source s to ycps
p existence of cold stream from process unit to splitter
process unit p, kg/s zi,j,k existence of match (i, j) in stage k
FMMin p mass flowrate of inlet stream to wastewater mixer, kg/s zcui existence of match between cold utility and hot stream i
FMMout p mass flowrate of outlet stream from wastewater mixer, zhuj existence of match between hot utility and cold stream j
kg/s
FSSinp mass flowrate of inlet freshwater stream to heat
exchange, kg/s Subscripts, superscripts, abbreviations
FSSout
p mass flowrate of outlet freshwater stream from heat CU cold utility
exchanger, kg/s CUC cold utility cost
FSSMp;p0 mass flowrate of freshwater stream from splitter to mixer FWC freshwater cost
process unit, kg/s HEN heat exchanger network
FPp0 ;p mass flowrate of water stream from process unit p0 to HIPWN heat-integrated process water network
process unit p, kg/s HU hot utility
FPOp0 mass flowrate of water stream from process unit p’ to final HUC hot utility cost
mixer, kg/s IC investment cost
FPUin p mass flowrate of inlet water stream to process unit p, kg/s in inlet stream
FPUoutp mass flowrate of outlet water stream from process unit p, max maximum
kg/s min minimum
Fout mass flowrate of outlet wastewater stream from final MINLP mixed-integer non-linear programming
mixer, kg/s out outlet stream
FWs mass flowrate of water for freshwater source s, kg/s TAC total annual cost
tc, th temperature of cold and hot stream, K WN water network
250 E. Ahmetovic, Z. Kravanja / Energy 57 (2013) 236e250

References [21] Dong H-G, Lin C-Y, Chang C-T. Simultaneous optimization approach for in-
tegrated water-allocation and heat-exchange networks. Chemical Engineering
Science 2008;63(14):3664e78.
[1] Bagajewicz M. A review of recent design procedures for water networks in
[22] Bagajewicz MJ, Pham R, Manousiouthakis V. On the state space approach to
refineries and process plants. Computers and Chemical Engineering
mass/heat exchanger network design. Chemical Engineering Science
2000;24(9e10):2093e113.
_ 1998;53(14):2595e621.
[2] Jezowski J. Review and analysis of approaches for designing optimum industrial
[23] Feng X, Li Y, Shen R. A new approach to design energy efficient water allo-
water networks. Chemical and Process Engineering 2008;29(3):663e81.
_ cation networks. Applied Thermal Engineering 2009;29(11e12):2302e7.
[3] Jezowski J. Review of water network design methods with literature annota-
[24] Leewongtanawit B, Kim J-K. Improving energy recovery for water mini-
tions. Industrial and Engineering Chemistry Research 2010;49(10):4475e516.
misation. Energy 2009;34(7):880e93.
[4] Foo DCY. State-of-the-art review of pinch analysis techniques for water
[25] Xiao W, Zhou R-j, Dong H-G, Meng N, Lin C-Y, Adi VSK. Simultaneous optimal
network synthesis. Industrial and Engineering Chemistry Research
integration of water utilization and heat exchange networks using holistic
2009;48(11):5125e59.
mathematical programming. Korean Journal of Chemical Engineering
[5] Furman KC, Sahinidis NV. A critical review and annotated bibliography for
2009;26(5):1161e74.
heat exchanger network synthesis in the 20th century. Industrial and Engi-
[26] Chen C-L, Liao H-L, Jia X-P, Ciou Y-J, Lee J-Y. Synthesis of heat-integrated
neering Chemistry Research 2002;41(10):2335e70.
water-using networks in process plants. Journal of the Taiwan Institute of
[6] Biegler LT, Grossmann IE, Westerberg AW. Systematic methods of chemical
Chemical Engineers 2010;41(4):512e21.
process design. New Jersey: Prentice-Hall; 1997.
[27] Polley GT, Picón-Núñez M, López-Maciel JdJ. Design of water and
[7] Klemes J, Friedler F, Bulatov I, Varbanov P. Sustainability in the process in-
heat recovery networks for the simultaneous minimisation of water
dustry: integration and optimization. New York: McGraw-Hill; 2011.
and energy consumption. Applied Thermal Engineering 2010;30(16):
[8] El-Halwagi MM. Sustainable design through process integration, fundamen-
2290e9.
tals and applications to industrial pollution prevention, resource conservation,
[28] Liao Z, Rong G, Wang J, Yang Y. Systematic optimization of heat-integrated
and profitability enhancement. Oxford: Butterworth-Heinemann; 2012.
water allocation networks. Industrial and Engineering Chemistry Research
[9] Savulescu LE, Smith R. Simultaneous energy and water minimisation. In:
2011;50(11):6713e27.
Presented at the 1998 AIChE annual meeting, Miami Beach, FL; 1998.
[29] Sahu GC, Bandyopadhyay S. Energy optimization in heat integrated water
[10] Savulescu LE, Sorin M, Smith R. Direct and indirect heat transfer in water
allocation networks. Chemical Engineering Science 2012;69(1):352e64.
network systems. Applied Thermal Engineering 2002;22(8):981e8.
[30] Boix M, Pibouleau L, Montastruc L, Azzaro-Pantel C, Domenech S. Minimizing
[11] Bagajewicz M, Rodera H, Savelski M. Energy efficient water utilization systems
water and energy consumptions in water and heat exchange networks.
in process plants. Computers and Chemical Engineering 2002;26(1):59e79.
Applied Thermal Engineering 2012;36:442e55.
[12] Savulescu L, Kim J-K, Smith R. Studies on simultaneous energy and water
[31] Martínez-Patiño J, Picón-Núñez M, Serra LM, Verda V. Systematic approach for
minimisationdpart I: systems with no water re-use. Chemical Engineering
the synthesis of water and energy networks. Applied Thermal Engineering
Science 2005;60(12):3279e90.
2012;48(15):458e64.
[13] Savulescu L, Kim J-K, Smith R. Studies on simultaneous energy and water
[32] Martínez-Patiño J, Picón-Núñez M, Serra LM, Verda V. Design of water and
minimisationdpart II: systems with maximum re-use of water. Chemical
energy networks using temperatureeconcentration diagrams. Energy
Engineering Science 2005;60(12):3291e308.
2011;36(6):3888e96.
[14] Bogataj M, Bagajewicz MJ. Design of non-isothermal process water networks.
[33] Yiqing L, Tingbi M, Sucai L, Xigang Y. Studies on the effect of non-isothermal
In: Valentin P, Paul Şerban A, editors. Computer aided chemical engineering.
mixing on water-using network’s energy performance. Computers and
Elsevier; 2007. p. 377e82.
Chemical Engineering 2012;36(0):140e8.
[15] Bogataj M, Bagajewicz MJ. Synthesis of non-isothermal heat integrated water
[34] Ahmetovi c E, Grossmann IE. Global superstructure optimization for the
networks in chemical processes. Computers and Chemical Engineering
design of integrated process water networks. AIChE Journal 2011;57(2):
2008;32(12):3130e42.
434e57.
[16] Yee TF, Grossmann IE, Kravanja Z. Simultaneous optimization models for heat
[35] Yee TF, Grossmann IE, Kravanja Z. Simultaneous optimization models for
integrationdIII. Process and heat exchanger network optimization. Com-
heat integrationdI. Area and energy targeting and modeling of multi-
puters and Chemical Engineering 1990;14(11):1185e200.
stream exchangers. Computers and Chemical Engineering 1990;14(10):
[17] Leewongtanawit B, Kim J-K. Synthesis and optimisation of heat-integrated
1151e64.
multiple-contaminant water systems. Chemical Engineering and Processing:
[36] Yee TF, Grossmann IE. Simultaneous optimization models for heat integra-
Process Intensification 2008;47(4):670e94.
tiondII. Heat exchanger network synthesis. Computers and Chemical Engi-
[18] Alva-Argáez A, Kokossis AC, Smith R. Wastewater minimisation of industrial
neering 1990;14(10):1165e84.
systems using an integrated approach. Computers and Chemical Engineering
[37] Rosenthal RE. GAMS e a user’s guide GAMS development corporation.
1998;22(Suppl. 1):S741e4.
Washington, DC, USA 2012.
[19] Papoulias SA, Grossmann IE. A structural optimization approach in process
[38] Smith R. Chemical process design and integration. West Sussex, England: John
synthesisdI: utility systems. Computers and Chemical Engineering 1983;7(6):
Wiley & Sons Ltd.; 2005.
695e706.
[39] Covey R. Waste heat recovery. www.alfalaval.com; 2010 [accessed 10.04.2012].
[20] Floudas CA, Ciric AR. Strategies for overcoming uncertainties in heat
[40] Alfalaval. Heat exchanger designs. http://www.alfalaval.com; [accessed
exchanger network synthesis. Computers and Chemical Engineering
10.04.2012].
1989;13(10):1133e52.

You might also like