Vous êtes sur la page 1sur 5

21st European Symposium on Computer Aided Process Engineering – ESCAPE 21

E.N. Pistikopoulos, M.C. Georgiadis and A.C. Kokossis (Editors)


© 2011 Elsevier B.V. All rights reserved.

Economic Plantwide Control of C4 Isomerization Process


Rahul Jagtapa, Sonam Goenkaa, Nitin Kaisthaa
a
Chemical Engineering, Indian Institute of Technology Kanpur, Kanpur 208016, India

Abstract
Plantwide control system design for economically optimum operation of a C4
isomerization process is studied. The steady state degrees of freedom of a base case
design are optimized for a given C4 fresh feed processing rate (Mode I) and maximum
production (Mode II). At maximum production, the number of active constraints equal
the steady state degrees of freedom (dof) exhausting all the available dof. From the set
of active constraints, regulatory plantwide control structures, CS1 and CS2, that
minimize the back-off from the economically dominant active constraints are
synthesized along with a simple supervisory optimizing scheme to drive the process
operation as close as possible to the active constraints. Quantitative results for the back-
off necessary to avoid constraint limit violation during transients due to a ±10% feed
composition change are reported. Comparison with a conventional plantwide control
structure, CS3, where the fresh feed is flow controlled, shows that the maximum
achievable throughput (profit) for CS2 is higher by ~2% (> $1x106 per yr).

Keywords: Plantwide control, optimal process operation, control structure design

1. Introduction
In refinery operations, iso-butane (i-C4) is a more valuable feedstock than n-butane (n-
C4) as it is used in the production of high octane gasoline blending components,
propylene oxide and tertiary butyl alcohol. The isomerization process is commonly used
to convert the n-C4 to the more valuable i-C4. As depicted in Figure 11, it consists of a
de-isobutanizer (DIB) column that takes in the fresh C4 stream with small amounts of C3
and C5 impurities to recover i-C4 as the distillate (along with C3 impurity). The n-C4
leaves from the bottoms with some i-C4 (light key) impurity and all the C5 in the fresh
feed. This bottoms stream is further fractionated in the purge column that recovers the
heavy C5 as the bottoms with a n-C4 rich distillate. This distillate is preheated using the
hot reactor effluent in a feed effluent heat exchanger (FEHE), vaporized and further
heated to the reaction temperature in a furnace. The hot C4 stream enters an adiabatic
packed bed reactor where n-C4 isomerizes irreversibly to i-C4. The hot reactor effluent,
after losing heat in the FEHE is cooled and condensed in a flooded condenser. The i-C4
rich condensed stream is fed to the DIB above the fresh feed for recovering the i-C4.
For smooth operation of this industrially important process, Luyben et al.1 have
designed a regulatory plantwide control structure using their heuristic bottom-up design
procedure2. Of the several reasonable control structure possibilities, this procedure gives
a structure for smooth transients in the overall plantwide response to principal
disturbances such as a throughput change. Economic considerations are however
ignored in the design of the plantwide control structure.
In today’s fiercely competitive market environment, processes must be operated for
optimal economic profitability (eg to maximize throughput / operating profit or to
minimize energy consumption). The optimum steady state usually is at the intersection
of multiple process constraints. Economic operation then requires driving process
2 R. Jagtap et al.

operation as close as possible to these active constraints. The implemented regulatory


control system determines the severity of the transients in the active constraint variables
and consequently the degree of closeness of operation to the constraint limits and
economic profitability 3,4. To the best of our knowledge, there are no literature reports
that consider plantwide control system design for the industrially relevant C4
isomerization process from the perspective of economically optimal operation. This
work presents the systematic design of such a plantwide control system for the process.
nC4, iC4 Recycle

198.9°C nC4 → iC4


FEHE 45 bar
Qfur: 863 kW P2: 4.35 atm
Qcnd1: 10.1 MW Qcnd2: 3.7 MW
Qcool: 1140 kW P1: 6.4 atm
1
L1 D1: 263.1 kmol/h
0.022 C3 1
20 0.958 iC4 L2 D2: 190.1 kmol/h
Frcy: 190.1 kmol/h 0.020 nC4
0.000 iC5
30
FC4: 263.1 kmol/h 10
0.02 C3
0.24 iC4 50
Qreb1: 10.4 MW Qreb2: 3.53 MW
20
0.69 nC4
0.05 iC5
B1 B2: 13.29 kmol/h

Figure 1. C4 isomerization process schematic with base case conditions


2. Optimal Process Operation
A base case design of the C4 isomerization process for processing 263.1 kmol/h of fresh
C4 feed (2%C3, 24% i-C4, 69% n-C4 and 5% i-C5) to produce an i-C4 product stream
with 2% n-C4 impurity has been reported by Luyben et al. 1. We take this existing
design (see Figure 1 for salient design / operating parameters) and optimize the steady
state operating degrees of freedom for (a) A given fresh feed processing rate of 263.1
kmol/h (Mode I) and (b) Maximum fresh feed processing rate (Mode II).
There are a total of seven steady state degrees of freedom (dof), one for the fresh feed,
four for the two columns (two per column), one for the furnace (duty or reactor inlet
temperature) and one for the flooded condenser (duty or outlet temperature). It is
assumed that the reactor is operated at the highest possible pressure (45 bar) for
maximum reaction conversion and the reactor pressure is not counted as a degree of
freedom. The seven independent variables chosen to fully specify the process flowsheet
are the fresh feed rate (FC4), heavy key and light key impurity mol fractions in
respectively the distillate and bottoms stream of the DIB column ([xDnC4]DIB) and
[xBiC4]DIB) and the purge column ([xDiC5]Purge) and [xBnC4]Purge), the reactor inlet
temperature (Trxr) and cooler outlet temperature (Tcool). All material and energy stream
flows (except furnace duty) are constrained to between 0 and twice the base-case steady
state values. The maximum furnace duty is constrained at 1.5 the base case value to
reflect the limited overdesign of an expensive equipment. Similarly, the maximum DIB
column boilup denoting onset of flooding is taken as 1.3 times its base case value. The
corresponding factor for the purge column is 1.5. The maximum reactor temperature
Economic Plantwide Control of C4 Isomerization Process 3

and pressure limits are 200 °C and 45 atm respectively. Finally, the n-C4 impurity in the
product stream should be below 2%.
To minimize the quality give away, the impurity in the product stream must be at its
constraint value (ie 2%). Also the reactor inlet temperature should be maximum to
maximize reaction conversion for minimum recycle cost. There exists an energy
consumption versus production rate trade-off with respect to the loss of n-C4 in the C5
purge stream ([xBnC4]Purge). However since the flow rate of the purge stream is small, we
simply set [xBnC4]Purge to a small value (1%) so that the n-C4 loss is small. Lastly Tcool is
fixed at a reasonable value of 53 °C. This leaves three steady state degrees of freedom
to be optimized. The constrained
minimization of the total energy cost is Table 1. Process optimization results’ summary
performed using fmincon in Matlab
Objective Mode I: Minimum energy cost*
with Hysys as the background solver function (J) Mode II: Maximum throughput (FC4)
for the two modes of operation.
The optimization problem and its Case Mode I Mode II
results for Mode I and Mode II are FC4 263.1 kmol/hr &
334.5 kmol/h#
briefly summarized in Table 1. The Trxr 200 °C Max 200 °C Max
Mode I energy cost is $1.716x106 yr-1 Tcool 53 °C Fixed 53 °C Fixed
while the Mode II maximum [xDnC4] DIB 0.02 Max 0.02 Max
throughput for the given fresh feed [xBiC4] DIB 0.0517 0.0125
composition is 334.5 kmol/h. In Mode [xDiC5] Purge 0.0202 0.00011
[xBnC4] Purge 0.01 Fixed 0.01 Fixed
I, there are four active constraints
leaving three unconstrained dof. In Optimum J $1.716x10 6
yr -1
$334.4 kmol/h
Mode II, three additional constraints MAX
Qfur ,
Additional
namely, the maximum furnace duty - Vreb1MAX,
Constraints
(QfurMAX), the maximum DIB boilup Vreb2MAX
MAX
(Vreb1 ) and the maximum purge *: Furnace duty $9.83 GJ-1; Steam $4.83 GJ-1;
column boilup (Vreb2MAX) are active so Cooling water $0.16 GJ-1
that all dof are exhausted with seven &: FC4 is specified
active constraints. The result is typical #: FC4 is optimized for maximum throughput
of chemical processes with the process being driven to its maximum throughput limit by
exhausting all the dof to drive as many constraints to their respective limits.

3. Plantwide Control System Design and Economic Performance


3.1. Plantwide Control System Design
To design a regulatory control structure that minimizes the economic loss due to the
need for a back-off from the active constraint limit due to transients, consider the active
constraints in Mode I and Mode II. Table 2 reports the percentage loss in objective
function per unit back-off in an active
hard constraint. In Mode I, Trxr is the Table 2. Percent change in objective function J
economically dominant active constraint per percent back off in active constraint*
variable. In Mode II, the throughput is Mode I Mode II
affected most by Trxr and Qfur. To #
eliminate a back-off in Qfur, we may Trxr 0.658 Qfur 0.360
Trxr# 0.926
Vreb1 0.086 Vreb2 0.002
flow control the furnace fuel valve and
not use it as a manipulated variable (eg *: Only hard constraints considered. #: 50 °C span
to maintain Trxr). Alternatively, since both Qfur and Trxr constraint variables are located
in the reaction section, flow controlling the feed to the reactor and not using it as a
manipulated variable would eliminate the flow variability in the reactor feed and hence
4 R. Jagtap et al.

mitigate the transients (and consequently, back-off) in both Qfur and Trxr. These two
options result in plantwide regulatory control structures, CS1 and CS2.
The inventory control system for CS1 is built around the flow controlled (fixed) furnace
duty with loop pairings as in Table 3. The unavailability of Qfur for manipulation forces
Trxr control using the recycle flow rate. The purge column reflux drum level is then
controlled using the column feed.
Table 3. Plantwide control structures
The sump level is controlled using
the reboiler duty as the bottoms Regulatory control loops (Mode I)
stream is very small making it CV
MV
inappropriate for level control. The CS1 CS2 CS2
impurities [xBnC4]Purge and TPM QfurSP D2SP FC4SP
[xDiC5]Purge are maintained using Trxr D2 Qfur Qfur
Lvltop1 L1 L1 L1
the bottoms and reflux rate Lvlbot1 FC4 FC4 B1
respectively. In the DIB column, Lvl top2
B1 B1 D2
the sump level is controlled using Lvlbot2 Vreb2 Vreb2 Vreb2
D DIB
the fresh feed. The reflux drum [x nC4] [L/D] 2 [L/D]2 [L/D] 2
[xBiC4]DIB Vreb1 Vreb1 Vreb1
level is controlled using the reflux [xDiC5] Purge L2 L2 L2
rate as the reflux ratio is large (>5) [xBnC4]Purge B2 B2 B2
with a relatively small distillate. Mode II supervisory control loops
The key component impurities
D DIB B DIB Trxr Maximum Maximum Maximum
[x nC4] and [x iC4] are Qfur Maximum D2SP FC4SP
controlled using respectively the Vreb1 [xBiC4] DIB SP [xBiC4] DIB SP [xBiC4] DIB SP
distillate and the reboiler duty. Vreb2 [xDiC5] Purge SP [xDiC5] Purge SP [xDiC5]Purge SP
With the basic regulatory control
system in place, supervisory loops for Mode II operation are implemented where the
setpoints [xBiC4]DIB and [xDiC5]Purge are adjusted to maintain the boilups Vreb1 and Vreb2
near maximum. In Mode I (given throughput), the QfurSP is slowly adjusted for the
desired fresh feed processing rate. QfurSP thus is the throughput manipulator (TPM).
In CS2, the regulatory loops are built around the flow controlled recycle stream (Mode I
TPM). The purge column reflux drum level is controlled using the column feed and the
DIB sump level is controlled using the fresh feed. Trxr is controlled using Qfur. The
remainder of the regulatory control structure and the Mode II supervisory loops are
similar to CS1. For comparison purposes, Table 3 also reports a conventional control
structure where the fresh feed is flow controlled and acts as the Mode I TPM. Here, in
addition to the CS1 Mode II supervisory loops, FC4 is adjusted to maintain Qfur.
3.2. Quantitative Back-off Results
The process cannot be operated at the limit of the active constraints as ever present
disturbances would cause transient hard constraint violation which is unacceptable
(TrxrMAX QfurMAX, Vreb1MAX and Vreb2MAX constraints are considered hard). The constraint
variable control loop setpoints must be appropriately backed off from their limits for the
worst case disturbance. A ±10% step change in the fresh feed n-C4 composition with a
complementary change in the i-C4 mol fraction is considered the worst case disturbance.
Table 4 reports the back-off from the active constraints along with the economic
objective function using the three control structures for Mode I and Mode II. In both
modes, there is no back-off in Trxr for CS1 and CS2 as the flow variability in the reactor
feed is negligible while a small back-off of 0.1 °C occurs in CS3 for Mode II. In Mode
I, CS3 fails for a +10% n-C4 feed mol fraction change with the purge column reflux
drum filling up in about 10 hours. This is due to accumulation of the unreacted n-C4 in
the recycle loop (snowball effect). In Mode II, the Qfur shows a significant (7%) backoff
Economic Plantwide Control of C4 Isomerization Process 5

for CS3 while no back-off is required in CS1 and CS2. Some back-off is also necessary
in the column boilups in all the structures. This back-off causes an almost negligible
throughput loss of 0.1% in CS1 and CS2. The throughput loss for CS3 is however
much larger at 1.9% due to the back-off in Qfur. Assuming a $20 per kmol product-raw
material (including energy expense) price differential, this corresponds to a yearly
revenue loss of about $1.068x106 in CS3 compared to CS1 and CS2 which is
significant. The result shows that the implemented plantwide control structure
significantly affects the profitability of the process.
Table 4. Back-off in active constraints and economic loss for CS1, CS2 and CS3*
Mode I Mode II
Trxra FC4b Jc Trxra Qfurd Vreb1e Vreb2e Jf
Optimum 200 263.1 1.726 200.0 1294max 2522max 851.8max 334.5
CS1 200 263.1 1.726 200.0 12940% 24861.4% 845.50.7% 334.10.1%
CS2 200 263.1 1.726 200.0 12940% 24861.4% 845.50.7% 334.10.1%
CS3 Fails 199.9 12037% 24961% 820.03.7% 328.01.4%
*: Subscripts denote % backoff. a: °C; b: kmol/h; c: x106 $/yr; d: MW; e: kmol/h; f: FC4 kmol/h
The TPMs for CS1, CS2 and CS3 are respectively, QfurSP, D2SP and FC4SP. Since TrxrMAX
and QfurMAX are economically dominant Mode II active constraints and as these are
located in the reaction section, minimizing the severity of transients (back-off) in both
constraint variables requires minimizing the transient variability into the reaction
section. This is accomplished is CS1 and CS2 with the TPM located in the reaction
section eliminating reactor feed flow variability. The TPM location thus plays a crucial
role in the economic operation of a process. It should be located close to and where
possible at the economically dominant active constraint(s).

4. Conclusion
This plantwide control study of a C4 isomerization process shows that the regulatory
control structure can significantly affect process economic performance by determining
the severity of the transients in the economically dominant active constraints. To
minimize the back-off from the constraint limit and hence the economic loss, the
regulatory layer TPM should be located as close as possible to the dominant
constraint(s). A top-down bottom-up approach, where the TPM is first chosen based on
the economically dominant active constraint(s) (top-down part) followed by the
synthesis of regulatory control loops (bottom-up part) appears the most appropriate
systematic methodology for plantwide control system design.

References
1. W.L. Luyben, B.D. Tyreus, M.L. Luyben, 1999, Isomerization process, Plantwide Process
Control, McGraw Hill: New York, 273-293.
2. M.L. Luyben, B.D. Tyreus, W.L. Luyben, 1997, Plantwide Control Design Procedure,
AIChE J., 43, 12, 3161-3174.
3. R. Kanodia, N. Kaistha, 2010, Plantwide control for throughput maximization: A case study,
Ind. Eng. Chem. Res., 49, 1, 210-221.
4. R. Jagtap, N. Kaistha, S. Skogestad, 2010, Plantwide control for economic operation of a
recycle process, Compter Aided Chemical Engineering, 28, C, 499-504.

Vous aimerez peut-être aussi