Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Pergamon

PII: SOOOS-2509(97)00186-3

Effluent treatment system design


Wen-Chu Janice Kuo and Robin Smith
Centre for Process Integration, UMIST, PO Box 88, Manchester M60 1QD. U.K

(Received 3 June 1996; in revised form 4 February 1997; accepted 29 May 1997)

Abstract-This paper addresses the design of distributed effluent treatment systems. In the case
of single contaminants, targets are first set for the minimum flowrates in a distributed effluent
treatment system. Design methods then allow the targets to be achieved in practice. Previously
published methods failed to address important features of the design for multiple treatment
processes. In the case of multiple contaminants the treatment network is developed in a staged
approach by repeated use of targets and design. Minimum flowrate is not guaranteed for
multiple contaminants, but the designer is guided towards the best solutions. Overall, the paper
presents improved methods for the design of distributed effluent treatment systems and extends
the concepts to retrofit cases. \c 1997 Elsevier Science Ltd

INTRODUCTION representation which allowed targets to be set for


the minimum flowrate in a distributed effluent treat-
Effluent treatment is most often carried out by collect-
ment system. These targets were set before the design.
ing aqueous process effluents into a common sewer
Design rules were also developed to allow the targets
along with utility effluents such as cooling tower
to be achieved in practice. The design rules were
blowdown and then treated in a central facility before
based on the location of the pinch for the system.
being discharged to the receiving water. This central-
Streams starting above the pinch for the treatment
ised treatment might include primary, secondary and
process are fully treated, those starting at the pinch
tertiary treatment processes. The stages needed de-
are partially treated and partially bypass treatment
pends on the contaminants, their concentrations and
and those starting below the pinch completely bypass
the discharge regulations of the effluent. However, the
treatment.
characteristic of centralised treatment systems is that
Although the general methodology presented by
they treat large volumes of wastewater with low con-
Wang and Smith (1994) provides valuable insights
centrations of contaminants resulting from the mixing
into the design of distributed effluent treatment sys-
of wastewater streams before treatment.
tems, there are some drawbacks associated with the
Eckenfelder et al. (1985), Lankford et al. (1988) and
method. The method fails to predict the lowest pos-
Higgins (1989) highlighted that segregated wastewater
sible target for the treatment flowrate in some cases.
treatment could have significant advantages over cen-
The reason for this will be discussed in detail later. In
tralised wastewater treatment. McLaughlin et al.
addition, important features of the design for multiple
(1992) pointed out that capital and operating costs of
treatment processes in both single and multiple con-
most wastewater treatment operations are propor-
taminant cases were not addressed.
tional to the total flowrate of wastewater which flows
The objectives of the methodology for the design of
through the treatment. Also, costs increase with the
distributed effluent treatment systems are to:
decreasing concentration for a given mass load of
contaminant.
Study of the optimal design of wastewater treat- (i) Choose the most appropriate type and number
ment systems has focused on the optimisation of spe- of treatment operations.
cific units or small groups of units. Few studies have (ii) Design a network for the treatment operations
addressed the optimal structure of wastewater treat- which segregates streams for treatment where appro-
ment systems. What is needed is an approach to the priate and mixes them where appropriate.
design of effluent treatment systems which segregates (iii) The resulting distributed effluent treatment
effluent streams for treatment in the first instance and network should bring the effluent streams to their
only combines them if it is appropriate. Takama et al. required consent limits for discharge at minimum
(I 980) presented a non-linear programming approach cost. Because the capital and operating costs of efflu-
to optimise distributed wastewater treatment sys- ent treatment operations are most often dominated
tems. Wang and Smith (1994) presented a graphical by flowrate, the methodology in the first instance
4273
4274 W.-C. J. Kuo and R. Smith

will seek to minimise the flowrate of effluent to be For example 1, the performance of the treatment
treated. processes are specified by removal ratios, Table l(b).
Applying the treatment processes in Table l(b) will
The methodology presented in this paper thus pro- not allow the treatment task to be solved with a single
vides an initial network design which must then be treatment process because of concentration and per-
subjected to detailed simulation and costing before formance constraints. Figure l(c) shows one possible
the final design is accepted. Such detailed evaluation solution with effluent treatment lines for Treatment
might lead to features of the initial design being unac- Process I (TPI) and Treatment Process II (TPII)
ceptable, requiring iteration back to the targeting and matched against the composite effluent curve (Wang
network design. Previous work also did not address and Smith, 1994). Treatment lines have been specified
the case of retrofit of the existing system. Upgrade of with the steepest slope in order to minimise the flow-
existing treatment systems to comply with new regu- rate of effluent to be treated. The slope is limited both
lations is an important consideration for which no by the performance specifications for the unit and the
systematic methods are currently available. This pa- pinch for the system (Wang and Smith, 1994). Alterna-
per presents improved methods for the design of dis- tive solutions are possible by changing the distribu-
tributed effluent treatment systems and extends the tion of mass loads removed between the two
concepts to retrofit cases. treatment processes. Moreover, the treatment pro-
cesses might have been put the other way around in
the alternative sequence. These two issues of different
NEW TARGETING PROCEDURE FOR SINGLE
mass load distributions and different possible se-
CONTAMINANTS
quences will be addressed later. First, let us accept the
Background
The method introduced by Wang and Smith (1994)
for targeting single contaminants starts by represent-
ing the data on a plot of concentration vs con- Table 1. Wastewater stream and treatment process data fort
taminant mass removed. Table l(a) presents data for Example 1
(a) Wastewater stream data
a set of effluent streams.
This is a single contaminant problem for which the
Stream Flowrate Concentration
environmental discharge limit is 30ppm. The data
Number (t/h) @pm)
from Table l(a) are presented graphically in Fig. l(a).
These individual effluent streams are then combined 1 20 800
to produce a composite of the effluent streams as 2 30 400
shown in Fig. l(b) (Wang and Smith, 1994). An efflu- 3 50 200
ent treatment line then needs to be matched against
this curve. The performance of the treatment process (b) Treatment process data
can be defined either in terms of a specified outlet
Process Number Removal ratio (%) Gin,,,_ (ppm)
concentration or a removal ratio defined by

I 90 600
RR =.LCin -.LwtCout
(1) II 99 200
.hnCin ’

800 . . . . . . ~ . . ..__. _

m (kg/h)

(a) Individual effluent streams (b) Composite effluent curve (c) Possible effluent treatment
targets

Fig. 1. The individual effluent streams can be represented by a composite effluent curve.
Effluent treatment system design 4272

distribution of mass loads and the sequence and ex- the problem will not behave like the line FG shown in
plore whether the target predicted by the method of Fig. 2(a). To obtain the true picture after placing TPI
Wang and Smith (1994) as used in Fig. l(c) actually we need to replot the new effluent treatment curve
gives the lowest possible target. based on new concentration order, C3 > C’, Figure
3(a) starts by placing TPI and then constructing the
Turyeting multiple treatment processes composite effluent curve after TPI has been placed,
If we now tear the diagram in Fig. l(c) into two which is line HJ in Fig. 3(b). Adopting this approach
parts, the effluent treatment designs for TPI and II allows the minimum treatment tlowrate for TPII’ to
can be designed from the procedure presented by be reduced when compared with Fig. l(c). The orig-
Wang and Smith (1994). Streams starting above the inal composite effluent curve is also shown in Fig. 3(b)
pinch for the treatment process are fully treated, those as the dashed curve. The new targeting line for TPII’
starting at the pinch are partially treated and partially actually crosses the original composite effluent curve
bypass treatment and those starting below the pinch which would not be allowed by the previous method.
completely bypass treatment. The resulting design for If there is only one treatment process, the treatment
TPI and TPII are shown in Figs 2(b) and (c). Combin- targeting line should always operate beneath the efflu-
ing the networks for TPI and TPII, the resulting ent treatment curve. If the treatment line crosses the
network is shown in Fig. 2(d). In Fig. 2(c) the outlet of effluent treatment curve it is infeasible to achieve the
TPI with lower concentration C’, is mixed with waste- mass transfer task, for reasons presented by Wang
water stream 3 and fed directly to TPII. We shall and Smith (1994). However, with multiple treatment
designate this kind of arrangement to be treatment processes. there are multiple treatment targeting lines
processes operating in series. It is a characteristic of and we need to look at the overall treatment system.
the previous targeting method for multiple treatment This means that we should construct a composite of
processes that the design structures which emerge are al1 treatment lines to judge the overall feasibility.
always series in nature. Yet, it is conceivable that There might be some treatment targeting lines which
parallel structures might sometimes be attractive, in cross the composite effluent curve on an individual
which the outlet of one treatment does not necessarily basis without the composite treatment curve crossing.
feed the next treatment process, and so on. The composite treatment curve must not cross the
In Fig, 2(b), after TPI the concentration and flow- composite effluent curve.
rate of wastewater streams have been changed. Figure 4 compares the individual treatment lmes
Instead of the original concentration order of with the composite treatment line based on the new
C, > Cz > C3, the new concentration order after the targeting line for TPII’. The composite treatment
TPI placement switched to C3 > C’,, where C’, is the curve shown in Fig. 3(b) lies below the effluent treat-
outlet concentration of TPI. Thus, the remainder of ment curve and is therefore feasible. In addition. it

(b)

WI c,

Fig. 2. The design implications of placing multiple treatment processes


4216 W.-C. J. Kuo and R. Smith

Fig. 3. Basing targets on unmixed streams leads to a result which looks infeasible by the original composite
effluent curve.

Effluent Treatment Curve

Curve)

0 m
Fig. 4. The composite treatment line does not cross the effluent composite curve and is therefore feasible.

confirms that TPII only needs the 49.5t/h capacity operate in parallel since the outlet of TPI does not
shown in Fig. 4(a) instead of 77.4 t/h flowrate require- feed to TPII directly. Such structures could not be
ment shown in Fig. l(c). obtained from the previous targeting method.
Based on the target in Fig. 4(a), the effluent treat-
ment network for TPII can be obtained from the Distribution of mass load for multiple treatment
previous design rules and is shown in Fig. 5(a). The processes
network for the complete system can be obtained by Consider again the data in Table 1. Figure l(c)
combining Fig. 2(b) and 5(a) as shown in Fig. 5(b). In shows possible effluent treatment targets to be 50 t/h
Fig. 5(b), it can be seen that the treatment processes for TPI and 77.4 t/h for TPII. This result has now
Effluent treatment system design 3277

been revisited and the true targets shown to be 50 t/h One way to optimise the mass load for a system
for TPI and 49.5 t/h for TPII. As mentioned pre- which requires multiple treatment processes is to first
viously, alternative solutions are possible by changing fix the relative loads arbitrarily. Having set the tar-
the distribution of mass loads removed between the gets, this would lead to a structure for the treatment
two processes. Wang and Smith (1994) suggested that system (Wang and Smith, 1994). Once the structure of
the maximum mass load should be removed by the the treatment system has been specified we can vary
cheaper treatment process. This result is straightfor- the load on each treatment process to minimise the
ward if the ratio of costs for TPII and TPI is extreme- cost. If we optimise a fixed structure the flow ratio for
ly high. However, should the mass load allocated to each bypass stream might change, or the whole by-
the cheaper treatment process (TPI) always be maxi- pass stream might be removed, but the basic structure
mised? Figure 6 shows the option which allocates the will remain unchanged. Consider the initial structure
maximum mass load to TPI. Comparing the two in Fig. 7. The optimisation might lead to the whole of
options in Figs 4(a) and (6), the second shown in Fig. 6 Stream 1 being fed to Treatment Process 1 as shown in
involves a flowrate increase of 50 t/h in TPI while the Fig. 7(a) or Stream 3 being fed to Treatment Process
flowrate required for TPII decreases by 28.2t/h. If 11 completely shown in Fig. 7(b). However, we would
there is not a large cost differential between TPI and not be able to obtain the structure shown in Fig. 7(c)
TPII then the first solution shown in Fig. 4(a) might since this involves a fundamental change to the basic
be economically more attractive. structure. Optimisation of the initial structure can, in
principle, remove features such as bypasses but struc-
tural features cannot be added. Therefore, optimisa-
tion of the initial structure will be able to access some
structures but others will not be able to be accessed.
Thus, fixing the loads arbitrarily early in the proced-
(a) -1 Dfscharge ure means that attractive designs might be missed
even after extensive optimisation.
I I An alternative approach shown in Fig. 8 would be
to optimise targets instead of optimising a single
structure. In optimising targets, the approach would
be to vary m, and thus the mass load removed by each
treatment process, Fig. 8. At each setting, the target
(b) flowrate for each treatment process allows the total
cost to be estimated through cost functions which
express total costs as a function of treatment flowrates
(and mass load if the cost model requires). Thus, in
Fig. 8 the cost is optimised based on targets. This
shows that total costs vary when the mass load re-
moved from each treatment process is redistributed
Fig. 5. Completing the design leads to a network in which depending on the cost ratio of TPl to TPII. Four
treatment processes operate in parallel. different cost curves are shown in Fig. X. The cost

21.3 t/h

Fig. 6. Maximizing mass load on the cheaper treatment process for Example 1
4278 W.-C. J. Kuo and R. Smith

Optimize the structure

Fig. 7. Continuous optimisation can change the bypass flowrates but not the structure.

/1 TPI/TPII=l
/ i

TP I /TP II = 0.8

TP I I TP II = 0.3
c2 :
‘-,_ - -’ i TPI/TPII=0.5
:Nc_#M I

1’2 14.4 25.2 34.2 m,(kgjh)


Fig. 8. The targets can be optimised instead of optimising a design.

curves in Fig. 8 do not represent the absolute cost mise the mass load removal in the cheaper treatment
relating to the different cost ratios, but show the process. Most importantly, whatever relative costs we
trends. The curves in Fig 8 indicate that different cost are going to use, different effluent treatment structures
ratios for TPI to TPII will produce different locations can be explored in the optimisation by optimising
for the optimum. Thus, we should not always maxi- targets.
Effluent treatment system design 4279
Each point on the curves shown in Fig. 8, in prin- handle the higher oil concentrations. Some treatment
ciple, represents one treatment structure. Alterna- process sequences are fixed by these kinds of treat-
tively, there might be a range of conditions leading to ment process constraints. Later we shall consider a re-
the same structure. As long as the setting of the mass finery case study and revisit this problem.
load which divides the two treatment processes is In most cases the treatment process constraints will
located within the same region of the composite efflu- have a major influence on the sequence. However, if
ent curve, i.e. between kinks on the composite effluent process constraints do not dictate the sequence of
curve. the resulting structure will turn out to be the treatment processes then one way to solve the prob-
same. Thus, in Fig. 8. as long as the setting which lem would be to carry out an exhaustive search across
divides the two processes is between points E and the feasible options. Later we shall introduce
F the same structures will emerge even though bypass a method based on thermodynamics which can help
flowrates will change. Once a kink in the composite to choose the best sequence.
effluent curve is crossed then different structures will
emerge as a result of the change in the effluent stream TARGETING AND DESIGN FOR MULTIPLE
population comprising that part of the composite CONTAMINANTS
effluent curve. Figure 8 shows that some cost ratios
are likely to lead to a local optimum if an inappropri- Background
ate initial setting is chosen. Optimising targets avoids So far we have restricted consideration to single
this problem allowing the global optimum to be contaminants such as chemical oxygen demand or
obtained. suspended solids. However, most wastewater prob-
lems involve multiple contaminants. Wang and Smith
Treatment process sequence (1994) suggested an approach which initially targeted
Figure 9 shows that there could be two alternatives for each contaminant in isolation. The highest flow-
for the same mass load distribution. TPI can be rate obtained across all contaminants for a treatment
located either prior to or after TPII. Both options process was taken to be the target for the treatment
should in principle be examined at the targeting stage. process for the multiple contaminant case. Design for
However, the performance of some treatment pro- multiple contaminants was carried out by first design-
cesses will decline with the decreasing concentration ing a network for each contaminant in isolation and
for a given mass load of contaminants. For example, if then merging the sub-networks. By including all fea-
TPI shown in Fig. 9(a) has a maximum inlet concen- tures from all sub-networks in the final design a mul-
tration constraint, Cin.max, Fig. 9(b) would not be tiple contaminant design was obtained.
acceptable since the sequence would need TPI to Table 2 presents data for Example 2 which consists
operate at high concentration and it conflicts with the of three wastewater streams containing three con-
constraint. A practical example of this could be an taminants A, B, C at different concentrations.
API separator which achieves a removal ratio of typi- The wastewater concentrations for each con-
cally 70% for oil removal but it will only be effective taminant must each be brought down to IOOppm
at higher oil inlet concentrations. Therefore API sep- before discharge to the receiving water. Three effluent
arators are used upstream of wastewater treatment to treatment processes are available in Table 7. Each

Fig. 9. Concentration constraints might dictate the treatment process sequence


4280 W.-C. J. Kuo and R. Smith
treatment process can only remove one contaminant. TPII for removal of contaminant B. On the other
The removal ratios of the treatment processes are hand, Fig. 10(c) suggests that we should take part of
specified in Table 2. Following the methodology of stream 3 to mix it with stream 1 and the combined
Wang and Smith (1994) we target and design each wastewater stream sent to TPII for removal of con-
individual contaminant. The result is shown in taminant C. Those two proposals have inherently
Fig. 10. The next step is to merge the three sub-net- incompatible and is a problem fundamental to the
works to obtain the final treatment network for all previous procedure.
contaminants together. But this is not straightfor-
ward. The sub-network shown in Fig. 10(b) suggests
that stream 3 should combine with part of Stream Superstructure approach
1 and the combined wastewater stream passed to One approach to the solution of this problem is to
create a superstructure for the final network contain-
ing all structural features from the individual sub-
networks. The superstructure should contain all
Table 2. Stream and treatment process data for Example 2 structural features which are candidates for the global
(a) Stream data
optimum network. Optimisation of this superstruc-
ture in a combined structural and parameter opti-
Contaminant
misation will then remove unnecessary features from
Stream Flowrate A B C the design.
number (t/h) (ppm) (ppm) (ppm) Figure 11 shows the superstructure of Example
2 based on the three sub-networks from Fig. 10. Six
1 20 600 500 500 pathways (paths P, Q, R, S, T, U) are included in the
2 15 400 200 100 structure and they provide all possible treatment pro-
3 5 200 1000 200 cess sequences. This enables us to embed all possibili-
ties in the superstructure. In principle, there should
(b) Treatment process data not be more than two pathways left after optimisation
in the final network for a three treatment process
Removal ratio (%) problem. For example, only paths U and Q will be
needed in the final design if the treatment process
Contaminant sequence turns out to be TPII followed by TI followed
Process by TPIII. Optimisation of the superstructure would
number A B C remove unnecessary features. However, it is clear that
even for this very simple problem we are left with
I 90 0 0
a complex superstructure to be optimised. This is
II 0 99 0
III 0 0 80
a difficult mixed integer non-linear optimisation
problem. The difficulties in carrying out such

Contaminant A Contaminant B Contaminant C

0 mA 0 m6 0 mc

(a) (W
Fig. 10. Multiple contaminant problems start by targeting and design of each contaminant in isolation.
Effluent treatment system design 428 I

t Discharge

Fig. Il. A superstructure approach to multiple contaminant problems

optimisations are well known, the greatest problem treatment processes should be included but we can
being that of the optimisation finding local optima only accept one proposal a time. Once the wastewater
rather than the global optimum. has been sent to the treatment process selected first,
Figure 11 shows the superstructure which, although the wastewater flowrates and concentrations of differ-
complex, is simpler than a superstructure for the orig- ent contaminants will be changed, since some of
inal problem which would contain all possible struc- wastewater streams will have been combined. In total
tural features. The dashed lines in Fig. 11 are the extra there are six different sequences for the three treat-
connections required to include all possible structural ment processes. One way to deal with the problem
features. The superstructure in Fig. 11 is based on the would be to explore all six different sequences and
sub-networks for the individual contaminants and then choose the best. Let us instead develop a better
thus allows some structural features to be discarded way based on thermodynamics.
through the physical insight from the targeting pro-
cedure for single contaminants. Whilst the superstruc- Single contaminant targeting for multiple
ture in Fig. 11, based on single contaminant targeting, contaminunt problems
brings some simplification it is not too significant. It is For each contaminant, targeting tells us what hap-
still a complex network for optimisation. If the treat- pens to that specific contaminant and how wastewater
ment process sequence can be determined, the pos- streams should be segregated for the removal of that
sible combinations can be reduced further. contaminant. Unfortunately, we do not know what is
happening to other contaminants while we are target-
Staged approach to building (I network ing for the contaminant under consideration. For
In principle, the streams detailed in Table 2 can be example, Fig. 12 shows the target and sub-network
fed to TPI, TPII or TPIII in any sequence. Single design. The sub-network design in Fig. 12 shows
contaminant targeting tells us the combinations of stream 1 mixing with part of stream 2 to optimise
wastewater streams to be treated for different con- contaminant A removal. But does it create problems
taminants, but the sequence of treatment processes for the treatment for contaminant C? The concentra-
must be decided through the merging of sub-net- tion of contaminant C in stream 2 is 100 ppm which is
works, which often encounters conflicts. The problem the same as the environmental limit. We should there-
seems to be one of a lack of guidelines for merging. On fore be able to discharge stream 2 directly as far as
the other hand, the superstructure approach does, in contaminant C is concerned. If we accept the sub-
principle, provide a universal solution but the opti- network shown in Fig. 12, we have carried out degra-
misation is difficult and does not allow the designer to dation as far as contaminant C is concerned, caused
interact with the solution. by inappropriate stream mixing. In other words.
The sub-networks shown in Fig. 10 are based on each sub-network shown in Fig. 10 is proposed by
minimum treatment flowrate required for removal of considering contaminant only and minimising
a certain contaminant. It is important that the three treatment flowrate for each treatment process for that
4282 W.-C. J. Kuo and R. Smith

where AH is heat of mixing; ni is molar flowrate


Contaminant A
(kmol/h) and Xi is mole fraction for each component.
Since wastewater streams are usually dilute solutions
and if assumed to be an ideal solutions, AH can be
omitted. Therefore, this equation can be rewritten as

AEx = - RTl,iniln$
I I
OI

Mixing

(4)

where ni is measured in kilomoles per hour. If the


contaminants are known, then the mole fractions and
mole flowrates can easily be calculated. In this case we
Fig. 12. Targeting for contaminant A does not account for
the mixing of contaminants B and C. might not know the exact molecular weight for each
contaminant. Here we shall assume that all con-
taminants have the same molecular weight as water
(Mu20 = 18). Clearly, if information is available on the
contaminant. However, each sub-network incorpor-
individual contaminants then it is not necessary to
ates some degradation in the concentrations of the
introduce this approximation. The main concern is
other contaminants which can result in the treatment
the value of the exergy loss as a measure of extent of
flowrate in the following treatment processes to be
wastewater degradation simply to choose between
increased.
competing options. To choose the sequence of treat-
If the total cost for the effluent treatment system is
ment processes we do not need the exact value of
proportional to the total treatment flowrate for the
exergy loss but simply the correct relative magnitudes
system, then the lower the total treatment flowrate,
between options. Taking an average molecular weight
the lower the total cost of effluent treatment. The
the above equation can be rewritten as
lowest total treatment flowrate will be the sum of all
targeting flowrates in the sub-networks. However, we
have already concluded that it is not always possible
AEx = -%imilnd 1.m. (5)
I mi
to maintain the features of each sub-network due to
the wastewater degradation occurring in each sub- where Mi = Mj = M.
network. No matter which sub-network we accept, As an example, the targeting of contaminant A leads
wastewater degradation will happen simultaneously. to a design involving stream 1 mixing with part of
If wastewater degradation is unavoidable for each stream 2 before treatment in TPI as shown in
treatment process placement, it makes sense to mini- Fig. 13(a). Contaminants B and C will be degraded
mise the wastewater degradation as much as possible through the mixing and the mixing exergy loss,
whilst selecting the treatment process sequence. AExuiC,, should be calculated based on the degrada-
tion of these contaminants. In another situation, if
Evaluation of wastewater degradation contaminant B is being targeted, the mixing exergy
In qualitative terms it seems desirable to avoid loss should be calculated for contaminants A and C as
degradation as much as possible. But before the con- AEx(,,o> as shown in Fig. 13(b). Similarly, AEx~~,~), is
cept can be made quantitative we need a method to calculated if contaminant C is being targeted, as
allow degradation to be evaluated. Since wastewater shown in Fig. 13(c). Comparing across options after
degradation is caused by inappropriate mixing, we targeting for each contaminant, the option with the
need some characteristic which can quantity the ex- minimum exergy loss, min{AEx}, will indicate the
tent of mixing. option with minimum wastewater degradation. How-
Mixing exergy loss can be used to express how ever, the mixing exergy loss we calculate is based on
much potential the system loses through mixing and the ‘untargeted’ contaminants and the calculation
might provide a measure of the extent in the waste- basis varies between each contaminant. Therefore,
water degradation. Generally, the mixing exergy loss instead of considering the absolute value of the exergy
for ideal solutions can be expressed as loss, AEx, the mixing exergy loss should be evaluated
in terms of percentage change, %AEx, defined as

AEx=AH-RRT,iniln& (2) %AEx = g * 100%


1 I I”
Effluent treatment system design 4283

(a) ON
Contaminant A Contaminant B

1 !,, ‘:j$Y~

cl + cl .
.-
I ’ TPII
3 \ 3
_.

%AEx IW cl %AEx ,,(A, c, %AEx ,,,(A, ~1

sl u bt

Choose mjn. {%AEx,}


as the first TP placement.

Fig. 13. Calculation of the mixing exergy loss for each contaminant target dictates the first treatment
placement.

where AEx is the exergy change across a mixing junc- Contaminant A Contaminant 6
tion. defined by AEx = Ex,,[ - Exi,. The sub-net-
work with minimum %AEx should be chosen for the
first treatment process placement. In Example 2,
TPIII has the minimum mixing exergy loss and hence
we place TPIII first to remove contaminant C.

Re-targeting the remaining problem


Once a treatment process has been placed, TPIII in
our example, some mixing has also been accepted to
achieve the target for the contaminant. The remainder
of the effluent treatment problem still needs to be
solved, but because of the mixing which has been
\ Y
accepted due to the placement of the treatment pro-
Choose min. {%AExi}
cess, the remaining wastewater stream data are
I
changed. Therefore, we should re-target the remaining
problem. Based on the new stream population and Fig. 14. The procedure of targeting and calculation of the
concentrations shown in Fig. 14, the new %AEx’ can mixing exergy loss is repeated on the remaining problem for
the other contaminants.
be obtained for the remainder of the treatment prob-
lem The sub-network with minimum %AEx’ is
placed next. Figure 15 shows the network after the
placement of TPII. Having placed TPII another new
stream population and new concentrations will be practice, it is likely that treatment processes will be
created. The new set of stream data will provide us capable of removing more than one contaminant
information for the next layer sub-network design. a time. For example, Fig. 18 shows a problem involv-
Following this design procedure, the final design for ing two contaminants and three treatment processes.
Example 2 is shown in Fig. 16. But will the design in The sub-network for contaminant A suggests that
Fig. 16 have lowest total treatment flowrate? Figure TPI and TPII should operate in parallel, but that for
17 shows all possible treatment sequences for contaminant B suggests that TPI and TPIlI should
Example 2 together with their total treatment flow- operate in series. For treatment processes operating in
rates. In Fig. 17 it can be seen that the design with the series there is no point calculating the exergy loss
lowest overall treatment flowrate is Fig. 17(f), which associated with the second process. This results from
is the one obtained from the procedure. The wrong the fact that the mixing losses of the second process
sequence, Fig. 17(a), will result in a 14% increase in depend on the placement of the first processes which
total treatment flowrate. has not yet been fixed. %AExlll,, should not be evalu-
ated at this stage. Therefore, we only evaluate
Treatment processes which can remove more %AEx,,,, %AEx,,, and %AEx,,,, and choose the
than one contaminant minimum exergy loss as the basis for the first treat-
In Example 2, it was specified that each treatment ment process placement. The difference between
process was only capable of removing a single con- %AEx,,~ and %AExl,, is that TPl can accept different
taminant. This was done for clarity of explanation. In stream combinations where different contaminant
4284 W.-C. J. Kuo and R. Smith

‘_____-_____-_________--- __--

Completed previously

Fig. 15. After targeting the final contaminant the final network can be completed.

Fig. 16. The final design for Example 2.

(81.7tfh)

(4

(85.311h)
(90.4ffh)

(b)

(EIS.OUh) (80&/h)

6)
Fig. 17. Comparison of different designs for Example 2

removal is concerned. After placement of the first ment processes, in principle involving multiple
treatment process, there is a new stream population as contaminants. There are two basic elements which
seen in Example 2. Therefore, the remaining problem bind the entire superstructure. These are the stream
should be solved by following the same procedure as combinations and treatment process sequence.
before. The present paper suggests the targeting concept
to select the right stream combination for each treat-
Discussion ment process and the mixing exergy loss as a measure
Figure 19 shows a complete superstructure for of the wastewater degradation to determine the treat-
a problem involving three streams and three treat- ment process sequence. These two concepts work
Effluent treatment system design 3285

c*:w
. c*:k”L
Contaminant A Contaminant B

(Composite treatment targeting curve)


0 m 0 m-

u u
%AEx, (,,) %AEx, caj

Discharge
.

Choose Min. {%AEx, (bj; %A%, cbj ; ohAExl taj 1


as the first TP placement

u
Retarget the remaining problem as before
and find the 2nd TP placement,
.... and so on.
Fig. 18. Calculation of the mixing exergy loss for problems in which the treatment processes affect more
than one contaminant.

(Targeting)
stream
combination

u
(Mixing Exergy Loss)
TP sequence

Fig. 19. The superstructure can be decomposed into two parts, one which dictates the stream combination
and the other which dictates the treatment process sequence.

together to resolve the whole superstructure layer by guarantee optimality. Degradation of different con-
layer. taminants will have different implications for different
Here the concept of wastewater degradation has treatment processes, which is not necessarily reflected
been introduced to determine the correct sequence to in the mixing exergy loss. Similarly, the mixing
deal with each contaminant. Whilst such a measure is exergy loss cannot translate directly into cost which
useful in that it is a physical parameter, its use cannot depends on issues outside those considered. Thus, the
4286 W.-C. J. Kuo and R. Smith

approach does not guarantee optimality in a math- C (wm)


ematical sense. It does, however, always guide the
designer towards the best solutions. In fact, in the 1000 -
majority of cases it is capable of identifying the best
solution. The approach can also be used for initialisa-
tion of a superstructure if an approach based on the
optimisation of a superstructure is to be used.

APPLYING THE NEW METHOD TO RETROFIT

With environmental regulations becoming stricter,


the retrofit of existing treatment systems to meet
more stringent discharge regulations is becoming an .
0 4.0 12.0
important problem. Apart from complying with stric-
m (kg/h)
ter regulations, an increased effluent load might be
created due to changes in capacity or new plants v
coming on-stream.
Whether to accommodate changes in effluent aris-
ing or to comply with stricter regulations, the waste-
water stream data will change. The obvious solution
Fig. 20. An existing design is operating on target.
to such problems is to install a new treatment process
downstream of the installation. But can the treatment
process be used in a more effective way? Let us ex-
plore what the methodology can offer in revamping
cases. C (fwm)
t

An example
Suppose a wastewater treatment plant exists and
was designed to meet an environmental limit set some
time previously with the wastewater stream and oper-
ating data given in Table 3.
The treatment process, TPI, is running at its fully
capacity of 26.7 t/h and cannot be increased further.
The removal ratio of TPI is fixed at 90% as shown in
Fig. 20. To meet new effluent regulations, a new treat-
ment process must be installed to remove an addi- m(kg/h)
tional mass load of 2 kg/h. Where should the new r I
treatment process be installed? 40tIh
. 5Owm
Discharge
Proposal 1. To install a new treatment process down-
stream of the existing treatment process. Fig. 21. The first proposal installs a new treatment unit
Figure 21 shows the new composite effluent curve downstream of the existing unit.
based on the new environmental limit together with
the treatment line for 26.7 t/h is the existing treatment
process. The instinctive change is to install another
treatment process, TPII, downstream of TPI as flowsheet in Fig. 21 shows TPI and II operating in
shown in Fig. 21. There would be 22.2 t/h treatment series.
flowrate needed in TPII to meet the requirement. The
Proposal 2. Install a new treatment process which will
operate in parallel with the existing treatment process.
The second option is to let the targeting line of TPII
Table 3. Wastewater data for an exiting plant cross the composite effluent curve and start from the
pinch point as shown in Fig. 22. As long as the
Stream number Flowrate (t/h) Ci, (ppm) composite treatment curve does not cross the com-
posite effluent curve the solution will be feasible. This
1 10 1000
is shown as a flowsheet in Fig. 22. Stream 2 which
2 30 200
causes the pinch is split into three and then fed to TPI
Note: Existing environmental limit = 100 ppm. and TPII separately. This scheme enables TPI and
Removal ratio of TPI = 90%. TPII to operate in parallel with the result that TPII
New environmental limit = 50 ppm. only needs 10.53 t/h capacity instead of the 22.2 t/h in
Min. outlet cont. of TPII = 10ppm. Proposal 1.
Effluent treatment system design 42x7

Proposal 3. As Proposul2 hut the new treatment pro- No suggestion is made here that any one of these
cess will remoae more mass load and hence reduce the proposals is better than the others. Detailed examina-
burden on the existing treatment process. tion would be required in a retrofit situation. The
Proposal 2 suggested that treatment line starts at example demonstrates that the methodology offers
the pinch point. However, the treatment line for TPII insights into revamping cases and provides a system-
can go even further across the pinch. TPII has been atic way to identify different options. The new treat-
assumed to be always capable of bringing the outlet ment process can be generally installed wherever it is
concentration down to a fixed outlet concentration. found to be most appropriate after more detailed
say IOppm in this case. The maximum inlet concen- examination, as long as the resulting composite treat-
tration is assumed to be far higher than the treated ment curve lies below the composite effluent curve.
wastewater streams. If so, we can install TPII which
operates across the pinch and reduces the mass Multiple contaminant problems
load on TPI. This would be good for flexibility of In both gross-root design and revamping, the prob-
operation. lems are multiple contaminant in nature. For most
Figure 23 shows TPII placed across the pinch and multiple contaminant problems, the bottleneck will be
it will only need 7.39 t/h to be treated instead of the associated with a certain contaminant. Therefore, it is
10.53 and 22.2 t/h in the previous cases. The flowsheet possible to revamp a multiple contaminant process by
shown in Fig. 23 shows TPI and TPII operating in applying the single contaminant concept.
parallel, as Proposal 2, but with the difference that
stream 1 with its high concentration has been split CASE STUDY
and not stream 2. It should be noted that this propo- Three wastewater streams are produced by a re-
sal will only be feasible if TPI it subjected to the finery site and must be treated before discharge. The
removal ratio but not the minimum outlet concentra- flowrates of the streams and the concentrations of the
tion constraint. three contaminants involved (HIS, oil. suspended
solids) are given in Table 5.
Results of the example The environmental limits of the concentrations of
A summary for the three proposals is given in three contaminants are 5, 20, lOOppm, respectively.
Table 4. No treatment facilities exist and the problem can be

C (pm) C (pw)
t t
lOOO} 1000

;;EF//i?
, 14.0 0 2.0 6.0 14.0
m(kg/h) m WW

Fig. 77. The second proposal places a new unit in parallel Fig. 23. The third proposal places a new unit in parallel with
with the existing unit. the existing unit and takes load from the existing unit.

Table 4. Summary for the retrofit example

TPI ‘I-PI1

Flowrate Mass load Flowrate Mass load


(t/h) (kg/h) (t/h) (kg/h)

Proposal 1 26.7 12.0 22.2 2.0


Proposal 2 26.7 12.0 IO.53 2.0
Proposal 3 26.1 6.684 7.39 7.316
4288 W.-C. J. Kuo and R. Smith

treated as a new design. Three new treatment pro- The cost functions for treatment processes are ex-
cesses can be used. Treatment Process I (TPI) is a foul pressed as functions of treatment flowrates and are
water stripper which only has an effect on the H2S given in Table 1.
removal, Treatment Process II (TPII) is a combined These data are taken from previously published
process of coagulation, sedimentation and filtration information (Takama et al., 1980) but some changes
which can treat all three contaminants with different have been made in order to make them more realistic.
removal ratios and Treatment Process III (TPIII) is
an API Separator which mainly treats oil and sus-
pended solids. The removal ratios for the three con-
taminants are given in Table 6. Even though it is a multiple contaminant problem
with multiple treatment processes, we start by target-
ing each contaminant in isolation to set up an in-
itialisation. For single contaminant removal, it is
Table 5. Wastewater stream data for the refinery case study obvious that H2S removal cannot be achieved by
a single treatment process, either TPI or TPII. As
Contaminant concentrations discussed previously, for multiple treatment processes
@em) in single contaminant targeting, the mass load allo-
cated to each treatment process must be optimised at
Stream Flowrate Suspended the targeting stage. Based on the mass load optimista-
number (t/h) HIS Oil solids
tion, Fig. 24(a) shows the treatment network to re-
move the H$. Figure 24(b) shows the corresponding
1 13.1 390 10 250
2 32.1 16 780 110 400
network for oil treatment. Since there is only 70%
3 56.5 25 100 350

Table 7. Cost functions for treatment processes


Table 6. Removal ratios for treatment processes
TPI Capital (S) 16,800* j”,’
Removal ratios (X) Operating (S/h) 1.0* j
TPII Capital ($) 12,600* ,j”.’
Treatment Suspended Operating (S/h) 0.0067* ,j
processes Oil solids TPIII Capital (S) 4800* ,j”-’
HzS
Operating (S/h) 0
TP I 99.9 0 0
TPII 90 70 98 Note: Annual rate of return = 10%.
TPIII 0 70 50 Operating hours = 8600 h/yr.
f= flowrate treated t/h.

(a) H,S removal

TP I = 37.8 t/h.
TP II = 102.3 t/h.

(b) Oil removal

TP III = 89.2t/h.
TP II = 38.6 t/h.

TP II = 38 t/h.
TP III = 89.2 t/h.

(c) S.S.

TP II = 70.9 t/h.

I3 ??

Fig. 24. Designs for each contaminant in isolation for the case study
Effluent treatment system design 4289

removal ratio for TPII and TPIII neither on its own is only be calculated for TPI in HIS removal, TPIII in
enough to meet the environmental limit. After opti- oil removal and TPII in suspended solids removal.
misation of the mass load for both treatment pro- Moreover, a certain treatment process sequence has
cesses, there are two alternative sub-networks for oil been embedded in the single contaminant targeting.
removal shown in Fig. 24(b) with only 1% cost differ- such that TPII should operate after TPI and TPIII as
ence between them, meaning that either could be it was suggested in H2S and oil removal. Hence,
accepted in principle. However, in practice the re- wastewater degradation should only be evaluated in
moval ratio of an API separator (TPIII) will deterior- terms of TPI and TPIII placement as shown in Figs
ate with decreasing inlet concentration. Therefore, the 24(a) and (b).
option with the sequence TPIII followed by TPII is
taken. Finally, let us consider suspended solids re-
moval. Similar to oil removal. suspended solids can be Ke-targeting
removed by TPII or TPIII. TPIII alone cannot per- Let us consider Figs 24(a) and (b) in more detail as
form the task but TPII is capable on its own as shown shown in Figs 25(a) and (b). The sub-networks suggest
in Fig. 24(c). The contaminants can be targeted in that TPII should operate either after TPI or TPIII,
a different order and the resulting sub-networks will but we should not calculate the wastewater degrada-
be unchanged since each sub-network is targeted as tion for TPI and TPIII based on the sub-networks
a single contaminant sub-problem at this stage. shown in Figs 25(a) and (b). For H2S removal, TPII
must treat all wastewater streams in order to meet the
environmental limit. However, a treatment flowrate of
only 38.6 t/h is required for TPII which will incorpor-
Before we evaluate wastewater degradation and ate TPIII to achieve the oil content removal. There
decide the treatment process sequence, practical con- are two different targeting treatment flowrates for
straints need to be considered. In principle, we must TPII, and we must choose the one which can achieve
evaluate the wastewater degradation for TPI, TPII the task for removal of both contaminants, i.e.
and TPIII. It should be emphasised that the waste- 102.3 t/h for TPII. Therefore, the oil removal targets
water degradation must be evaluated in an appropri- must be readjusted as shown in Fig. 25(b) to 33.4 t/h
ate way or it can lead to errors in the process for TPIII and 102.3 t/h for TPII. Having done so,
configuration. There might be more than one treat- all we have to do is to determine the sequence of TPI
ment process required for a single contaminant re- and TPIII. Based on the stream combinations for
moval such as H2S or oil. If this is the case, it is TPI and TPIII, the relative wastewater degradation,
misleading to calculate the wastewater degradation of S/~AEX,,~,,~,,~,~.,
and ?/oAEx, ,,,, HIs,S.S.,, can be evaluated
TPII in terms of HIS and oil removal, since the stream and hence the sequence of treatment processes can be
data for the evaluation of wastewater degradation of settled. In this case, TPI should be placed first due to
TPII cannot be obtained without TPI or TPIII place- the lower wastewater degradation. After the TPI
ment. Therefore, the wastewater degradation should placement. there will be a new stream population and

(a) H,S removal (b) Oil removal

EWE-

%AEx,, ,a,,,s.sj

u u Re-targeting

(b’) Oil removal

Choose the min { %AEx, }


as first TP placement

Fig. 25. Evaluation of the degradation for the case study.


4290 W.-C. J. Kuo and R. Smith

Fig. 26. Final design for the case study.

the remaining problem must be solved. There is a con- RR removal ratio of treatment unit
straint in that TPII should be centralised treatment TPI Treatment Process I
and it is therefore straightforward to set the target of TPII Treatment Process II
TPIII from overall mass balance. The final waste- TPIII Treatment Process III
water treatment network is shown in Fig. 26. xi mole fraction for component i
AEx exergy loss
CONCLUSIONS AEx(,, exergy loss in terms of the mixing degrada
The general methodology for the design of distributed tion of contaminant A
effluent treatment systems previously presented Wang %AEx the percentage exergy loss
and Smith (1994) provided valuable insights into the AH heat of mixing
problem. However, the method failed to address im-
portant features of the design for multiple treatment Subscripts
processes in both single and multiple contaminant H.C. hydrocarbon
cases. This paper presents an extension of the meth- HIS hydrogen sulphide
odology. An improved method has been presented for in inlet concentration to an operation
targeting treatment flowrate. The distribution of load out out concentration from an operation
between multiple treatment processes has also been S.S. suspended solids
addressed. The concept of wastewater degradation 1,2 3, . wastewater stream number
has been introduced to account for treatment process
sequence in multiple contaminant problems. REFERENCES
Previous work also did not address the case of Eckenfelder, Jr. W. W., Patoczka, J. and Watkin, A. T.
retrofit of an existing system. Upgrade of existing (1985) Wastewater treatment. Chem Engng 60-74.
treatment systems to comply with new regulations is Higgins, T. E. (1989) Hazardous Waste Minimizution
an important consideration for which no systematic Handbook. Lewis Publishers inc.
methods are currently available. The methods de- McLaughlin, L. A., McLaughlin, H. J. and Groff,
veloped have been applied to the case of retrofit. K. A. (1992) Develop an effective wastewater treat-
ment strategy. Chem. Engng Progr. 3442.
Mishra, P. N., Fan, L. T. and Erickson, L. E. (1975)
NOTATION Application of mathematical optimization tech-
C concentration of contaminant niques in computer aided design of wastewater
environmental limit treatment systems. In Water-1974, AIChE Sympo-
C,
sium Series, Vol. 71, (145) pp. 136-153.
Ex exergy
Takama, N., Kuriyama, T., Shiroko, K. and Umeda,
f targeting flowrate for treatment process
T. (1980) Optimal water allocation in a petroleum
Mi molecular weight of component i refinery. Comp. Chem. Engng 4, 251-258.
mi mass of component i Wang, Y. P. and Smith, R. (1994) Design of distrib-
ml contaminant mass load removed by TPI uted effluent treatment systems. Chem. Engng Sci.
ni molar flowrate (kmol/h) 49(18), 3127-3145.

You might also like