Beruflich Dokumente
Kultur Dokumente
Junzo Yamamoto
junzo@sie.idemitsu.cojp
nakane@sie.idemitsu.co.jp
Kouichi Takahashi
Tetsuji Tani
Hokkaido Refinery
ldemitsu Kosan Co., Ltd.,
25-1 Masuna-cho Tomakomai, Hokkaido, JAPAN 059-1392
45006819@sie.idemitsu.co.jp
ttani@sie.idemitsu.co.jp
Takashi Inoue
Hokkaido Refinery
Idemitsu Kosan Co., Ltd.,
25-1 Masuna-cho Tomakomai, Hokkaido, JAPAN 059-1392
16oo9244@Sie.idemitsu.coJp
Abstract
This paper describes the implementation of coopel'ative
control system for Residue Fluid Catalytic Cracking
Plant in a refinery USing a knowledge based control
system and model predictive multivariable control system
(pMC system). PMC system is effective in the area of
time-delay and inteiference process. However, due to
their poor response to large time-delay and non-linearity.
it sometimes fails to control actual plants in a refinery.
On the other hand, expert operators control their plants
with large time-delay and nonlinearity fairly well by
manipulating the set points of regulatory controllers. The
proposed cooperative control system therefore combines
their advantages. The knowledge based control system is
used to determine set points of the reactor temperature,
feed flow rate and recycle flow rate with large time
delays and non-linearity using 'if then' type rules based
on expert operators' know-how. PMC sysm sets the
targets of PID controllers for other operation variables
and keeps the process within constraints to achieve
desired process conditions, which are difficult for
operators to control because of time-delay and
inteiference among those manipulated variables. The
implemented cooperative control system showed a
satisfactory peiformance in the actual plant.
1.
Introduction
industry.
in
the refining
as PID controller
and
are
of time
realize
and
high-level
noises
fairly
well
by
operators'
experience
and
1056
Fuel Gas
Wet Gas Compressor
Gasoline. LPG
Heavy Naphtha
Main
Distillation
Column
--=====:::-.L--+
:;
k-
___
H.Recycle
Residue
Heat Exchanger
crude,
process flow
1. Gas
within
constraints to achieve desired process conditions such as
in
drawn
catalyst,
variables.
This control system has been applied to the optimum
regenerators.
alters
requirements.
2.
2.1
Process Overview
Operation
and
Problems
in
Process Overview
others
Changing
because
Some
one
of
variable
are
the
variables
simultaneously
change
in
respond
are
energy
almost
nonlinear. Main
are
reactor
and
two
regenerators
operations.
1057
Flow Rate
- RCO Recycle Flow Rate
Charge
- Reactor Temperature
30 Constraints
++- Reactor, Regenerators, Main
+- Gas Separation Plants
CVs
........
=:
L;;;J
DVs
23 Control Variables
+-
Colmnn)
'"
PMC System
Total 7 DVs
MVs
DV:
Disturbance Variable
Reactor Press
10 Control Variables
rTemp.
HCORecycle
Re
Total.5 DVs
TPAP/ADUTY
MTC PIADUTY
RCOP/ADUTY
I----BTM PIADUTY
5)
6)
non-linearity
and
large
dead-time
including
gas
separation
wet gas
as
follows:
regenerator and
1)
main
distillation column
are
system.
3.
4)
MYs.
plants'
main
overall
Overview
of
Model
Predictive
Multivariable Control (PMC) System
more
are
1058
Operation Target .
Operator
@---------------
Operation Target
Operator
1--- -------------------- -
------
--
--
Process Data
:
nnint
Predictive
Control
(LP)
Optimum Target
: MV
Predictive
..
---
!". em
__________________ ___
DV
Reasoning
Model
L-------------s;---!
Process Data
'
:---s;tp-t--------------.J
:_pb.Comro-crn----- 1 ProreiData i
I"___
...
...
_ut_r_-T-,-:-----------:- -:--------:-_:c________J
Outp
_.1i--'-i_
tput
_
r-_Ou
Reactor I Regenerators I Main Column
Plant
4.
4.1
J='
P,
(II
1=1
+ I IJI,'(U;(I+M,}-u;w)'
1='
,If
",I
(1)
,,,,I
subject to
l''
u, (I
+ M,)
(2)
(3)
where
A"qJ"t/f,
are weight coefficients of
objective function, Yp/ is the predicted value of CVs, y'",
is the target ofCVs, t.ui(t) is the move ofMVs at time
t, u;(t) is the value ofMVs at time t, u';" is the target of
MVs, Pj is the prediction horizon, M, is the control
horizon, n is the number of CVs, m is the number of
MVs, ap(k) is the step response coefficient, a;lk) aj,isj')
for all /(>Sj,'
The advantages of the PMC are as follows:
1) It easily handles multivariable problems with dead
time, time-delay and interference with constraints.
2) The step response model used in the controller is
familiar to engineers.
=
4.2
1059
For example, in case there are not any CVs violating its
MY,
constraints
MY.
or
not and,
in
such
as
Figure 4.
4.3
H <PVSH,
H,<PVSH
PV>H"
PS
ZE
NS
Nl3
PV2L
L >PV2L,
L;>PV2L
PV<L,J
ZE
ZE
ZE
ZE
Hi,: IstUpperlimitofCVi,
tlUii
output
of
MVj
PS is positive small
From
6u!;
of CVi,
final
was
1) The system is
step
3)
interference.
The knowledge based system covers the area which
is difficult for PMC system to handle satisfactorily
4)
S.
Result
The control system has been capable of
maximizing
Lngic
and the
cyclone
air
as
their
low
as
possible.
I:. uq : Output Vallie of Hh manipul1k vonbabk's xt pl.)Jnl L"3\culatcd by .i-th constraint rult'
Fig.4
1060
Ailer
Implem'. on
+2.5
30
60
90
I:)
ii"'
120
!%J
150
180
[days)
After Implementation
30
60
90
[days]
120
150
180
After Implementation
-''i;-\._-,
- (1.03 [kgicm2]
60
Table 2
90
[days]
120
ISO
180
Improvement Summery
Operation Variables
Coke Burn Ratio
Cyclone Inlet Velocity
Press. of Reactor
Diff. Press. of Slide Valve
Oxygen of Flue Gas
-:rn}piovement
+ 2.5 [%]
+ 0.3 [mls]
- 0.03 fkelcm2]
- 0.04 fkelcm2J
-0.2
['Vo]
Conclusion
References
1061