Professional Documents
Culture Documents
pubs.acs.org/IECR
ABSTRACT: This paper presents a mathematical model for the synthesis of water networks for systems consisting of process
units of different operation modes, namely, truly batch, semicontinuous, and continuous units. By treating a continuous process
as a special case of a semicontinuous process, the original problem becomes to synthesize a batch water network comprised of
truly batch and semicontinuous units operated cyclically with a fixed schedule. The model is formulated as a mixed-integer
nonlinear program based on a unit-tank superstructure including all possible network interconnections. Three modified literature
examples are used to illustrate the proposed approach, with both in-plant and interplant water integration analyzed.
© 2013 American Chemical Society 7047 dx.doi.org/10.1021/ie302521v | Ind. Eng. Chem. Res. 2013, 52, 7047−7055
Industrial & Engineering Chemistry Research Article
3. FUNDAMENTAL CONCEPTS
In this section, the fundamental concepts on which the
mathematical model is based are presented, namely, those
relating to water-using unit types, the integration of units
operating in different modes, and the representation of time.
3.1. Operation Modes of Water-Using Units. Figure 1
illustrates three common operation modes for water-using Figure 2. Continuous-time representation over a cycle.
4. MODEL FORMULATION
With all continuous units being treated as semicontinuous ones,
the original problem becomes the synthesis of a WN for truly
batch and semicontinuous units. A mathematical model is then
developed to address the remaining problem. As an extension
of the formulation of Chen et al.,21 this model consists mainly
Figure 1. Types of water-using operations. of mass-balance equations and is based on a superstructure
including all feasible network connections between water-using
units. While a continuous unit operates uninterruptedly for a units (Figure 3a,b) and storage tanks (Figure 4). Notation used
long duration (e.g., 8000 h/year) spanning many operation in the formulation is given in the Nomenclature section.
cycles, truly batch and semicontinuous units are scheduled to
operate within certain periods of time (as short as a few hours
or days) and often in a cyclic manner. In addition, the operating
period of either a truly batch or semicontinuous unit is always
shorter than the batch cycle time. Regarding water usage, a truly
batch unit takes in water at the start and discharges wastewater
at the end of its operation. A typical example for this kind of
operation is the batch reaction for agrochemical production in
which water is used as the reaction solvent and for product
washing. By contrast, water intake and discharge for continuous
and semicontinuous units take place steadily during the course
of the operation. Typical examples for such operations include
the various extraction and washing processes in the chemical
industry.
3.2. Integration of Batch and Continuous Units. For
WN synthesis involving units of different operation modes, the
main challenge would be to integrate these units. In this work,
it is proposed to treat a continuous operation as a special case
of a semicontinuous operation existing over the whole batch
cycle time, as shown in Figure 1. This approach is based on the
assumption that changes in the water supply are acceptable Figure 3. Schematics of water-using units: (a) truly batch and (b)
even for continuous water-using operations. Because water semicontinuous units.
intake and discharge for truly batch units take place at time
points and for semicontinuous units in time intervals, direct
water transfers between truly batch and semicontinuous units 4.1. Mass Balance for Truly Batch Units. Figure 3a
are not happening. Therefore, water integration between these shows a schematic diagram of a truly batch unit i ∈ 0 b. Its inlet
two types of units can only be carried out indirectly via water- water may come from other truly batch units i′ (∈0 b), storage
storage tanks. tanks s, or fresh-water sources w, and the outlet water may be
3.3. Representation of Time for Cyclic Operation. The sent to other truly batch units i′, storage tanks s, or wastewater
structure of water minimization formulations for batch disposal systems d. Equations 1 and 2 describe the inlet and
processes is largely dictated by the treatment of time. In
most formulations, the time horizon of interest, or the cycle
time (H), is divided into several time intervals, with events
taking place only at the interval boundaries.
In this work, a continuous representation of time is
employed in which a cycle is divided into T time intervals
that are not necessarily of equal duration. As shown in Figure 2,
the interval boundaries (time points) in each cycle are
numbered from t = 1 to T + 1, with the latter coinciding
with the start of the next cycle, i.e., t = 1. Using this time
representation, both time intervals and points can be Figure 4. Schematic of a storage tank.
outlet water flow balances for unit i ∈ 0 b at time point t, f iin Y itop = ∑ fi ′ it + ∑ fsit + ∑ fwit
respectively. By assuming truly batch units to operate without i ′∈ 0 sc s∈: w∈>
water losses or gains, the overall water flow balance for unit i ∈
0 b is quite simple, as given in eq 3. Note that ; SE ∀ i ∈ 0 sc, t ∈ ; (10)
i is the set of
time point pair(s) corresponding to the start and end times of
the operation of unit i ∈ 0 b. f iout Y itop = ∑ fii ′ t + ∑ fist + ∑ fidt
i ′∈ 0 sc s∈: d∈+
qitin = ∑ qi ′ it + ∑ qsit + ∑ qwit ∀ i ∈ 0 sc, t ∈ ; (11)
i ′∈ 0b s∈: w∈>
GAMS environment22 on a Core 2, 2.00 GHz processor with Table 2. Operation Parameters for Example 1
BARON as the MINLP solver.
time point/interval
5.1. Example 1: Integration of Truly Batch and
Continuous Units. The first example is adapted from parameter unit t=1 t=2 t=3 t=4 t=5 t=6
Majozi17 by adding a continuous unit (F) to the five truly YSit A 1 0 0 0 0 0
batch units (A−E). Table 1 shows the operating data for these B 1 0 0 0 0 0
C 0 0 0 1 0 0
Table 1. Operating Data for the Water-Using Operations for D 0 1 0 0 0 0
Example 1 E 0 0 0 0 0 1
YEit A 0 0 1 0 0 0
limiting B 0 0 0 1 0 0
concentration
(ppm) time (h) C 0 0 0 0 1 0
D 0 0 0 0 0 1
unit limiting flow (t) Cin,max
ic Cout,max
ic mass load start end
E 1 0 0 0 0 0
A 0 100 100 kg 0 3
Yop
it F 1 1 1 1 1 1
B 250 510 72.8 kg 0 4
C [300, 400] 100 100 0 kg 4 5.5
D 250 510 72.8 kg 2 6
E [300, 400] 100 100 0 kg 6 7.5
F 100 250 25 kg/h 0 7.5
Table 3. Limiting Water Data for Example 2 Figure 9. WN configuration for example 2.
limiting concentration
(ppm) time (h)
unit Cin,max
ic Cout,max
ic mass load (kg/h) start end
U1 0 100 2 0 1
U2 50 100 5 1 3.5
U3 50 800 30 3 5
U4 400 800 4 1 3
U5 200 400 20 0 5
use of one storage tank, the model involves 160 constraints, 255
continuous variables, and 4 binary variables. It is solved in 1
CPU s with the minimum fresh-water consumption determined
to be 356.25 t per cycle. This corresponds to a 25.78%
reduction in the fresh-water use compared to the case without
water recovery. Figure 9 shows the optimal WN configuration.
Note that most water reuse is carried out directly between
Figure 11. Gantt charts for example 3.
units; only part of the effluent from U2 is stored for reuse in
U5. Figure 10 shows the water-storage profile of the tank. The
capacity required is 47.5 t.
5.3. Example 3: Interplant Integration of All Three cycle time of 5 h is divided into eight time intervals with eight
Types of Units. To demonstrate the applicability of the time points defined. Both in-plant and interplant integration
proposed model for WN synthesis involving truly batch, scenarios are analyzed in this example.
semicontinuous, and continuous units, a combined two-plant Prior to exploration of the water reuse/recycle opportunities,
example is considered. This example is a combination of the the minimum fresh-water requirements in plants 1 and 2 are
7052 dx.doi.org/10.1021/ie302521v | Ind. Eng. Chem. Res. 2013, 52, 7047−7055
Industrial & Engineering Chemistry Research Article
Figure 13. Water flow rate profiles between continuous units and tanks.
calculated to be 2385.49 and 480 t per cycle, respectively, with a consumption and network complexity. There are also cases
total of 2865.49. This is taken as the base case. with the trade-off between the water recovery potential and the
In-plant water integration is first considered for both plants. requirement for storage or regeneration systems to be
In the presence of two storage tanks, the minimum fresh-water analyzed.24 Hence, a more comprehensive formulation would
consumption of plant 1 is found to be 1000 t per cycle. Note be required to address the conflicting objectives in WN
that optimizing water utilization for plant 1 alone needs only synthesis.
seven time intervals and seven time points. The corresponding
model has 565 constraints, 717 continuous variables, and 14 6. CONCLUSIONS
binary variables and is solved in 3 CPU s. For plant 2, the In this paper, with a continuous process treated as a special case
minimum fresh-water consumption is the same as that in of a semicontinuous process, a mathematical model has been
example 2, i.e. 356.25 t per cycle. The total of 1356.25 t per developed for the synthesis of WNs for truly batch and
cycle corresponds to a 52.67% reduction in the fresh-water use semicontinuous units in cyclic operation with a fixed schedule.
compared to the base case. Detailed results for water Three examples adapted from literature were solved to illustrate
integration in plant 1 are not shown for brevity. the proposed approach. The results show that significant
Interplant water integration is then considered. Table 5 reductions in fresh-water consumption as well as wastewater
shows the values of binary parameters (YSit, YEit , and Yop
it ) for this generation can be achieved through in-plant and interplant
case. With the use of two centralized storage tanks, the overall water integration. Future work will focus on other cases of WN
model involves 869 constraints, 1429 continuous variables, and synthesis with more different types of units, e.g., truly batch/
16 binary variables. The minimum fresh-water consumption of semicontinuous water sources and sinks and their various
1185 t per cycle is obtained in 18 CPU s. This result combinations, along with the development of suitable case
corresponds to a 58.65% reduction in the fresh-water use studies. The incorporation of regeneration processes and a
compared to the base case and shows an almost further 6% scheduling framework into the model also remains as the
reduction compared to the in-plant integration case. Figure 12 subject of future work. It should be noted that having a flexible
shows the optimal interplant WN configuration, with detailed schedule can further reduce the need for storage.
■
flow rate profiles between continuous units and tanks shown in
Figure 13. Note that in plant 1 the fresh-water intake is more AUTHOR INFORMATION
than the wastewater discharge, while a contrary situation is Corresponding Author
found in plant 2. This indicates a net cross-plant water flow of *Tel.: +886-2-33663039. Fax: +886-2-23623040. E-mail:
372.4 t per cycle from plant 1 to 2. Figure 14 shows the water- CCL@ntu.edu.tw.
storage profiles of the tanks. The required capacities of ST1 and
ST2 are 1148 and 245.19 t, respectively. Notes
The authors declare no competing financial interest.
■ ACKNOWLEDGMENTS
The authors thank the National Science Council for supporting
this research under Grants NSC101-3113-E-002-004 and
NSC99-2221-E-002-187-MY3.
■ NOMENCLATURE
Indices and Sets
c ∈ * = contaminants
d ∈ + = wastewater disposal systems
i ∈ 0 = water-using units
i ∈ 0 b ⊂ 0 = truly batch units
i ∈ 0 sc ⊂ 0 = semicontinuous units
i ∈ 0 c ⊂ 0 sc = continuous units
s ∈ : = storage tanks
t ∈ ; = time points/intervals
(t, t′) ∈ ; SE
i = start and end time points for the operation of
truly batch unit i
Figure 14. Water-storage profiles for example 3. t ∈ ; opi = operating periods of semicontinuous unit i
w ∈ > = fresh-water sources
5.4. Discussion. All of the minimum fresh-water con- Parameters
sumptions reported in examples 1−3 agree with the water Cin,max
ic = maximum inlet concentration of contaminant c for
targets obtained using the insight-based targeting methods unit i
(such as the time-dependent water cascade analysis technique23) Cout,max
ic = maximum outlet concentration of contaminant c
and have proven to be globally optimal. However, because of for unit i
the MINLP formulation, global optimality for the minimization Cwc = concentration of contaminant c in fresh-water source w
of storage capacity cannot be guaranteed. FLs = lower bound for the inlet water flow rate to tank s
Instead of taking water reduction as the overriding concern, FUs = upper bound for the inlet/outlet water flow rate of tank
the minimum fresh-water constraint may sometimes be relaxed s
to further simplify the WN for the trade-off between water Mic = mass load of contaminant c in unit i ∈ 0 b
7054 dx.doi.org/10.1021/ie302521v | Ind. Eng. Chem. Res. 2013, 52, 7047−7055
Industrial & Engineering Chemistry Research
■
Article