e Pergamon

Computers and Chemical Engineering Supplement (1999) S157-S160

«:> 1999 Elsevier Science Ltd. All rights reserved
PI!: S0098·1354/991OO125-8

Optimal Water Use and Treatment

Allocation
N. Benk6*, E. Rev, Z. Szitkai and Z. Fony6
Department of Chemical Engineering, Technical University of Budapest; 1521 Budapest, Hungary
*OLAITERV Contracting and Engineering Co. Ltd., 1519 Budapest, P.O. Box 270, Hungary

Abstract
Water consumption and emission can be decreased by optimal allocation of water re-use, regeneration and re-use.
regeneration and recycle in water consuming mass exchange networks and by applying optimally distributed
effluent treatment processes in case of several emission sources of different contamination levels. Several example
problems taken from the literature have been reproduced and solved by GAMS/MINOS/CONOPT package. In most
cases our optimal solution is identical to those of others; in some other cases, however, we found better solutions.
We also found our method more flexible than those based on targeting and conceptual design. Some example
problems involve constraints and/or specifications of those kinds the referred methods cannot deal with. Namely,
the constant mass load specification is substituted by either mass load proportional to the water flow rate (involving
constant concentration shift) or some other, even more realistic, non-linear functions. The optimal systems have
different structures according to the applied mass load relations.

Keywords: Optimal water allocation, mathematical programming, wastewater minimisation

Introduction superstructure and removing some of its elements by

optimisation (either NLP or LP) is applied.
Use of fresh industrial water accompanied by
The terminology 'Mass-Exchange Networks' (MEN-
emission of used, and therefore contaminated, water
s) and methodology was then introduced (El-Halwagi
is a steady attribute of many chemical processes all
and Manousiouthakis; 1989, 1990a, 1990b; EI-
over the world. The contaminated water should be
Halwagi, 1997) patterned after tile pinch theory of
treated and then either be emitted or used again in
HEN-s (Heat Exchange Networks), developed earlier
some chemical process.
(LinnhojJ, B. et al, 1982). HEN-analogous terms, like
Water using chemical operations can be characterised
Problem Table, Composite Curve,' Pinch are
by several quantitative properties that specify their
accordingly elaborated.
target and provide constraints to the water they can
A similar, but different, approach has also been
accept and the water they may let out. These
developed with the terms of 'limiting water profiles'
specifications and constraints are inherently
for water consuming processes and 'treatment lines'
determined by the type and tile capacity of the
for treatment processes. (Wang and Smith, 199-1a,
process, though in some cases they may be variables
1994b; Olesen and Polley, 1997).
involving costs. For the sake of unique handling of
the different chemical operations in tile mass
Application of Mathematical Programming
exchange network synthesis (MENS) problem, they
Approach
are characterised by the properties of tile water
streams they use or emit. The pinch-oriented approaches, preferably that of
The structural tools for nunmusmg water Wang and Smith, provides us witil clear
consumption and emission are re-use, regeneration understanding of the problem and its draft overall
re-use, regeneration recycling, and distributed solution, as well as targeting. However, tile complex
effluent treatment. TIle above picture is further nature of multiple contaminant, multiple regeneration
complicated by tile possibility of applying several re-use and recycling possibility, and multiple
regenerating and/or treatment processes with treatment processes is not really caught by these
different capacities and performance. approaches but calls for an adequate tool that can
Optimal assignment and design of water consuming, deal with it. This is tile reason we apply the
regenerating, and treatment system is therefore a GAMSIMINOS/CONOPT package (Brook, A.' et ai,
complicated task, and can be mathematically 1988) for tile problem sketched above. Although
formulated as mixed integer non-linear programming MINLP includes integer variables, and in fact there
(MINLP) problem. are tools in GAMS to involve them, here we restrict
ourselves to apply the synthesis strategy based on
Solution Methodologies superstructures and NLP.
Early studies on the solution of the water network
Application of superstructure and NLP consists of
problem applied the method of structure variables
tile following steps:
(Takama et al, 1980a) and successive linear
I, Construction of a superstructure; 2, Assigrunent of
progranuning (Takama et ai, 1980b, 1981). In both
variables in the structure; 3, Compilation of equations
methodologies the technique of fixing a
SI58 Computers and Chemical Engineering Supplement (1999) SJ57-SJ60

describing the balances and restrictions; 4, For all the units, the inlet (and outlet) concentrations
Construction of a set of initial values; 5, Compilation are usually limited from above.
of additional equations describing the set of initial For regenerating and treatment processes, the inlet
values: 6, Compilation of additional equations concentrationsare also usually limited from below.
describing the objective function; 7, Optimisation TIle final emission should satisfy environmental
over the variables for minimal or maximal value of limitations. These are usually given as limit
the objective function: 8, Interpretation of the results concentrations, but may also be given as limit flow
and decision on eliminating some units. rates, as well.
In case of elimination,jump back to step 3.; 9, Finish. Either one or both of these limitations on the
emission can be active.
The mathematical model consists of the following Arbitrary limitations may become necessary to apply
relations: for computationstability.
TIle water consuming, regenerating, and treating TIle objective function is usually an economic
processes are all considered as units. There are two pointer, for example sum of all the computable costs
additional units: one for the fresh water source and projected to one year, like:
Z =I.fw (IJ(fut)+ ~ [loAJV;in) + PJinv(H;"')]
one for the used water sink. TIIC output water stream
(3)
from any real unit can, theoretically, branch out to I
any other unit except water source. Freshwater cost and operating cost are generally
Equality constraints arc constituted by component proportional to the load, e.g.:
transfer balances, mass conservation at branching and
unification points, as well as component conservation fjw(x)=dx ; f.,p(x)=cx (4)
at mixing (calculation of new concentrations). and the investment cost can be estimated by a power
Loss is usually not accounted for; but its possibility is function, e.g.:
not excluded. limo (x)-a/ (5)
Some specific reformulation is also applied for
regenerators and treatment processes. Namely, the
Resolve the "Illustrative example" of Takarna et
specifications of these processes can be given in
aI,1980a.
several ways. One is to specify the removal rate, r~k;
an other one is specifying the output concentrations, This is a small network in a refinery with 3
cI.kout • contaminants (H2S, Oil, and Suspended Solid), 3
Component transfer takes place at any process in the water-using processes (Steam strippers block, HDS
networkaccording to Figure 1 high pressure block. Desalters block), and 3
process stream regenerating/treating processes (Foul water stripper.
~- ------- -------- --- Oil separator, a block of Coagulating, Sedimentation.
, m and Filtrating units). The original data are presented
in Table 1. Here the inlet flow rates of the water
wash ing water L-- --....J
using processes and the contamination rates or the
removal ratios. respectively. are fixed. TIle flowrates
Figure 1 should be fixed, because of the technology. E.g. the
rate of the stripping steam must be in a strong
Only the water stream side is considered in our connection with the feed rate of the distillation tower.
model; and the component transfer rate is counted for TIle method of Wang & Smith (wastewater
by a variable m (for contaminant mass rate), if loss is minimisation with varying flowrate in each item) is
not considered: rather specific. TIle inlet concentrations of the water
in in out out using processes are limited; concentrations about the
(1)
1It i •k =H; c i •k -J~ ci,k
treatment processesare not.
In case of multiple contaminants the target may be TIle resolve of the case study (Resolve.) give a result
related to one or only a partial group of pollutants with lower fresh-water consumption due to the
while some non-targeted (or side effect) utilisation of regeneration re-cycle. This opportunity.
contamination rates may be complicated functions of and even regeneration re-use, is missing from the
the circumstances. For example a given rate of earlier solution although it was not intentionally
contaminant A is to be washed out, by water, from excluded. The original solution is just a re-use and
the process streams while washing out of end-of-pipe treatment structure. TIle reason why they
contaminant B depends on the inlet flowrate and inlet do not use the regeneration recycle must be the
concentrations in a complex way. In this case the rate problem of salt contamination increase due to the
of washing out contaminant B can be expressed as a intensive recirculation. TIle problem should be
complicated function: extended with the salt as a contaminant (151
m I. , B =f(W"i n ,CiinA ,Ciin)
B
••
(2) modification). The values can be found in the Wang
& Smith article, where the suspended solid as a
Inequality constraints are constituted by limitations contaminant is missing. It is also an irrealistic
according to the unit operations' and processes' treatment of the problem because the desalter is very
physical chemistry, operability, safety, as well as by sensitive to the suspended solids.
environmental regulations.
 Computers and Chemical Engineering Supplement (1999) S157-S160 5159
We also update the concentrntion limits of the stream. We solved the problem with a number of
nd
desalter with the specification of Nalco-Exxon (2 values of «; but practising industrial experts
modification). The oil and salt limit is so tight, but recommend the hypothesis of considering the outlet
the limit for SS is harder than we found in the oil concentrntion of the desalter as a fix value that
literature. can be 100ppm for a very well operated unit; and
We also build functions for the oil concentration of 200ppm can be good for calculations. (3rd
outlet water from the desalter. The hybrid function is: modification)
w We resolved the problem. with these modifications.
lit = a- lit <. + (1- a )n l j i . " , -
J- Wfix and got more realistic results (resolve-). All the
where m is fixed as it is found in the literature if 0.=1 optima are summarised in the Table 2.; the optimal
network after resolve, can be seen in Figure 2.
and 0.=0 means: the higher water flowrate in desalter,
the higher transferred oil to the outlet washing

T able 1 Sspec ifircatiIOn 0 fth ecasest udy

Water-using W M ern• x
In

processes H2S Oil SS Salt* H2S Oil SS Salt*

(tIh) (kg/h) (kg/h) (kg/h) Kg/h (ppm) (ppm) (ppm) (ppm)
Strippers 45.8 17.9 0.5 1.2 1.58* 0 0 0 0*
HDS highP 32.7 536 3.3 0.5 6.12* 500 20 50 -l5*
Desalters 56.5 1.3 5.7 2 532* 20 120 50 200*
cout=200'" 1000** 10** 2000**
Treatment R A b C
processes H2S Oil SS Salt* $/(t/h)b - $/h
FWS 99.9 0 0 0* 16800 0.7 1.0
OS 0 95 20 0* 4800 0.7 0
CSF 90 90 97 0* 12600 0.7 0.0067
Fresh water 0.3 $/t em..x
env
specific cost
Annual 8000 h/a H 2S Oil SS Salt
operation time
Annualisation 13.5 % (ppm) (ppm) (ppm) (ppm)
factor (B)
Environmental 2 2 5 -
limits
*.. 1st modification, **.. 2,na modification, ***.. 3,111 modification
Table 2: Resu ts 0 fth e case stu diy
Original state Takama etal Resolve, Resolve-
Freshwater flow rate (tIh) 135.0 102.3 45.8 60.2
Freshwater cost (M$/a) 323 246 110 144
Operating cost (M$/a) 1087 371 372 339
Capital cost (M$) 1060 676 658 907
Total annual cost (M$/a) 1550 707 568 603

3.6

0.2

1.2. 25.9
14.4 )
L . - - -. . .f-----.+...J·28.61----,..---~_l___l,

6.5
Figure 2.
Conclusions
No generally accepted approach for specifications found different specification approaches in the
has yet been developed in tile literature. We have literature studied. These differences may origin from
5160 Computers and Chemical Engineering Supplement (1999)SI57-S160
altering approaches to the problem and/or from the Considering the variety of the specifications that may
applied methodologies. occur simultaneously in a single problem it seems
One of the water using processes, in our case study, that no conceptual (targeting based or graphical)
is a block of steam strippers, At steam stripping the method can cope with the general case. We discussed
steam flow rate is specified by the targeted water some particular details in the application of the
using process (e.g. specified necessary boil-up), GAMS/CONOPTIMINOS package. ' In spite of thc
therefore it is well fixed. On the other hand, there are MINLP formulation of the problem itself: the
water-using processes whose main target is to wash approach of Cover & Eliminate (cover by
out some contaminants from the process stream; and superstructure and eliminate by NLP) can be
their operation can use more or less water in some successfully used . However, we think, the conceptual
range. In this later case both upper and lower bounds methodologies provide the user with indispensable
can be specified. A prop er methodology should tools for analysing, understanding, and interpreting
simultaneously deal with both specification types. the problems and their solutions. Use of integer
In our review of a case study we used this kind of decision variables (MINLP methodology) may be
methodology. which is related on up-to-date preferable in case of large problems with many units
industrial practice. and connections.
An even more interesting problem is the We have to say thank to the Nalco-Exxon for the help
interdependency of contamination rates. It is an they gave us. We acknowledge the financial support
actual research task to develop appropriately tailored provided by Hungarian Scientific Research Fund
approximation formulas for specific water using unit (OTKA) grant TOI6851.
operations.

Notation
Subscripts
a investment cost factor
fw fresh-water cost
b investment cost exponent
i running index for units
c operating cost factor (in eq. x)
inv .investment cost
c concentration
op operating cost
m mass load
w component flow rate
Superscripts
IV water flow rate
in inlet
Z objective function
out outlet
P factor for projecting investment
cost for a unit time period
References
Takama, N., Kuriyama, T ., Shiroko, K., Umeda, T. : "Optimal water allocation in a petroleum refinery", Camp.
Chem. Eng ., 4(4) 251-258 (1980)
Takarna, N., Kuriyama, T ., Shiroko, K., Umeda, T .: "Opt imal planning of water allocation in industry", J. Chem.
Eng. Japan, 13(6) 478-483 (1980)
Takama, N., Kuriyarna, T ., Shiroko, K., Umeda, T.: "On the formulation of optimal water allocation problem by
linear programming", Compo Chem. Eng., 5(..) 119-121 (1981)
EI-Halwagi, M.M., Manousiouthakis, V.: "Synthesis of Mass Exchange Networks", AIChE Journal 35(8) 1223-
1244 (1989)
El-Halwagi, M.M., Manousiouthakis, V.: "Automatic Synthesis of Mass-Exchange Networks with Single-
Component Targets", Chem. Eng. Sci., 45(9) 2813-2831 (1990)
El-Halwagi, M.M., Manousiouthakis, V.: "Simultaneous Synthesis of Mass-Exchange and Regeneration Networks",
AIChEJourna/36(8) 1209-1219 (1990)
El-Halwagi, M.M.: Pollution Prevention through Process Integration. Systematic Design Tools., Academic Press,
San Diego, 1997.
Linnhoff, B.,Townsend, D.W., Boland, D., Hewitt, G.F., Thomas, REA, Guy, A.R., Marsland, R.H.: "A User
Guide on Process Integration for the Efficient Use of Energy", IChemE, Rugby, 1982.
Wang, Y-P., Smith, R.: "Wastewater minimisation", Chem. Eng. Sci.; 49(7) 981-1006 (1994)
Wang, Y-P ., Smith, R.: "Design of distributed effluent treatment systems", Chem. Eng. Sci., 49(/8) 3127-3Iot5
(1994)
Wang, Y. P., Smith, R.: "Wastewater minimisation with flowrate constraints ", Trans. IChemE, 73A(/ I) 889-904
(1995)
Olesen, S.G., Polley, G.T. : "A simple methodology for the design of water networks handling single contaminants",
Trans. IChemE, 75A(May) 420-426 (1997)
Brooke, A., Kendrick, D., Meeraus, A., Rosenthal, RE.: "GAMS. A User's Guide (reI. 2.25)" , boyd & fraser publ.
co., Danvers, MA, after Scientific Press, San Francisco, 1992 .