Joseph Lu
Honeywell
Gas ($x1)
Kerosene ($x4)
Fuel ($z)
Residue ($x7)
penalty
penalty
60 60
Degrees of freedom 40 40
20 20
~
Output prediction: y Au y
The set-point y r is changed into a ―set-range‖ y LO y y HI u* arg min Au ~
y s u
2 2
u , s Q R
A slack variable s is introduced in the MPC formulation
y LO s y HI
Additional degrees of freedom are used for 120
100 Range
Local disturbance rejection/absorption 80
Control
Optimization: e.g., y LO y y HI
penalty
min y; 60
40
0
-2 -1 0 1 2
control variable
u , w
current time A free w free Q
time
U k RkVkT
1
When the solution is not unique, a
(k )
Let Q 2 Aact
secondary optimization can be performed
arg min U k RkVkT u wact
2
u*
to find the minimum effort solution. u 2
Dewaxed Oil
WFO Atmos WFO Press
Flash Tower Flash Twr
Htr
Mix
WFO
Reflux
Reflux
Double Double
Pipe Pipe
Exchanger Exchanger
Pipe Pipe
Exchanger Exchanger
Wet
Reflux
Reflux
Steam
Vacuum
WFO
Dehydrator
Double Double
Pipe Pipe
Chiller Chiller
Ketone
Stripper
Slop
Refrig
Solvent
Refrig
Wax
Steam
Refrig
Reflux
Reflux
Steam
Refrig
PW Final Double
PW Stripper Pipe
Flash Twr Double
Chiller
Pipe
Chiller
Reflux
Primary
Reflux
Repulp
Steam
Filtrate
Repulp
Filtrate
Primary
Page 8
PW Mix
Surge
Honeywell
Solvent Dewaxing Unit Illustration Honeywell
Filtering
Wax + Oil Wax
Mixture
1 2 3
Solvent
Lubricant Oil
6 5 4
10
4
5
3 5
0
6 2
-5
-10
1
-15
0 100 200 300 400 500 600 700 800 900 1000
Manipulated Variables
100%
Time[Min]
2 5
50
3
1
0
4
-50
5
-100%
0 100 200 300 400 500 600 700 800 900 1000
Min
CAUSTIC
ETHYLENE
BOX
SPLITTER
OIL FUEL GAS
WATER
► Furnace and converter decoking FURNACES QUENCH QUENCH ACETYLENE
RECOVERY
► Frequent furnace switching UNIT
DEMETHANIZER
DEETHANIZER
Site specific: ETHANE
► Feed quality variations Feed
System CGC
COCRACK
CONDENSATE
DEPROPANIZER
DISTILLATE
STRIPPER
PROPYLENE
STRIPPER
SPLITTER
► Periodic switching (e.g., dryers)
Advanced Control and Optimization Goals
► Stabilize operation MIXED
DEBUTANIZER
DEBUTANIZER
C4
► Minimize product quality give away
PROPANE
► Maximize selectivity & yield TCG COCRACK
► Minimize converter over-hydrogenation
► Minimize ethylene loss to methane/ethane recycle. Profit
Controllers Fig 2. Process overview – flow diagram
TCG
C3R
C2R
RMPCT 9
Cold
Quench
Box
C2H2
C2 Splitter
Converter
RMPCT 1 RMPCT 12
Fractionator
RMPCT 15
DeC1
DeC2
RMPCT 2
C3 Splitter
DeC3
RMPCT 8 RMPCT 10
MAPD
Converter
RMPCT 14
C2R
C3R
RMPCT 9
Cold
Quench
Box
C2H2
Converter
P
RMPCT 1 RMPCT 12
Fractionator
DeC1
DeC2
RMPCT 2
C3 Splitter
DeC3
RMPCT 10
RMPCT 8
MAPD
Converter
RMPCT 14
RMPCT 15
+500
Commissioning
Profit Controllers are commissioned first to provide stable unit operations. Profit
Bridge is then commissioned to provide gain updating for the furnace controllers.
+400
Next, Profit Optimizer is commissioned, and Bridge Model predictions are
verified against data. Finally, a plantwide objective function is populated and +300
optimization moves are validated. Continuous operation of the controllers and
optimizer starts after operator training. +200
Maintenance +100
Typically little maintenance is required to keep the ethylene APC/Optimization
running. Clients either dedicate a parttime control engineer to monitor and Base Before APC After APC
perform minor services as needed, or depend on Honeywell’s BGmax quarterly-
visit services.
Semiconductor 3.1
Asm1 28
4.1
Test1 33
5.1
Fuse1
Box1
Multi-factory 34 37
7.1
18 3.2
Supply Chain B 39 43 7.2
20 24 6.1 40
7.4 7.3 44
vend
11 vend8
3.3 pp 3.4 pp
7
Fab1 7 Sort1 41 pp pp 46
2.1 6.2
P1 P1 vendor3 vendor4 C 42 45
vendor1 21 7.5
12 25
1
1.1 si 3.5 35 38
2 A 19
3 13 7.6 Box2
Fab2 8 Sort2 Asm2 29 Test2 Fuse2
14 4.2 5.2
2.2 3.6 36
4
P1,P2 P1,P2
15
Semiconductor30 F
5
1.2 si 3.5 3.7
vendor2
6
16
Blue = Intel
Fab3 9 Sort3
2.3
P2
vendor5 22 26 vendor6
Red = Mat. Sub.
P2 3.8 pp 3.9 ram
10
17
27
Green = Cap. Sub.
23
= Mats Mfg
3.10 3.11 31
= Inv Hold D
wafer
test
sort
C1 Fab/Sort
M1 T1
I10 Die/Package Inventory
Simple Supply
C2 Assembly/Test
Chain Sequence M2 T2
Semi-Finished Goods Inv.
I20
Finish/Pack
1.2 0.25 C3
TPT Distribution
1
0.2
M3 T3 Components Warehouse
0.8 I30 Shipping
Data
0.15
Model
0.6 C4
DataPDF
0.1 Demand
0.4 ModelPDF M4 T4
0.05
(Orders)
0.2
0 0
D1 t
D3 D2
Inventory
1500
1000
500
Low Limit
0
5/21/06 5/28/06 6/4/06 6/11/06
Die xxx_27 6/18/06
Inventory 6/25/06 7/2/06 7/9/06
3000 Date
High Limit
2500
2000
K Die
Inventory
1500 APC used
1000 Low Limit
500
0
5/21/06 5/28/06 6/4/06 6/11/06
Die xxx_47 6/18/06
Inventory 6/25/06 7/2/06 7/9/06
1500 Date
1250
High Limit
1000
K Die
750
APC used
Inventory
500
250
Low Limit
0
5/21/06 5/28/06 6/4/06 6/11/06 6/18/06 6/25/06 7/2/06 7/9/06
Date
From Process Unit to Plantwide Control & Optimization Page 20
Refinery-wide (Production) Control/Optimization
Honeywell
Production Plan
Supply Side (monthly) Customer Demand
(upstream What Feed to buy? What products to sell? (downstream
supply chain) supply chain)
Production Schedule
(weekly)
What Feed to use? What products to deliver? Product to
Feed Received be Deliveried
Multivariable
APC
Control
month ahead days ahead hours ahead now hours ahead days ahead months ahead
1-to-n MPC cascade – One Master MPC and multiple Slave MPCs
The Master MPC employs a dynamic model which can be, at user’s discretion, a coarser-scale model
e.g., similar to the scale-level of a planning model
The Slave MPCs employs a dynamic model which can be a finer-scale one
If Slave MPCs are deployed at every process unit, then the Master MPC can solve the plantwide
production control problem dynamically and in closed loop
Currently, a steady-state version of this problem is controlled by planning solutions in open-loop
Crude Imports
Unit Light Moving Avg
1 Distillate daily delivery:
Crude Feed 1 Hydrotreater
Component
Tank 1
5 KBarrels
Vac
Unit
8 KBarrels
1
Crude
Component
Unit Heavy Tank 2
2 Distillate 10 KBarrels
Hydrotreater
Crude Feed 2 7 KBarrels
Component
Tank 3
10 KBarrels
Crude Vac
Unit Unit Hydrocracker Blending
3 2
Crude Feed 3
50~300
Crude Imports
Sweet ppm S Light
Unit
1 Distillate
Hydrotreater
4~20 ppm S
0.1% Sulfur
Component
Tank 1
12 ppm S
Vac
Unit
10 ppm S
1
Crude 100~400 Component
Medium Unit ppm S Heavy 4~20 ppm S Tank 2
2 Distillate 10 ppm S
0.3% Sulfur Hydrotreater
10 ppm S
Component
Tank 3
10 ppm S
Crude Vac
Sour Unit Unit Hydrocracker 1~2 ppm S Blending
3 2 100~600
1.2% Sulfur ppm S
Few if any (closed-loop) control solutions are available for this class of problems!
CompTank2.Level
Production amount
CompTank2.Sulfur
Product quality
CompTank2.Cloud
CompTank3.Level
CompTank3.Sulfur
CompTank3.Cloud
1)
2)
More CVs…
as bs 1
2
e or s (s 1)
e would suffice
Notice that there is no setpoint in this control problem – just constraint control
CVs
Conv Sulfur deliver
y
Master MPC
Tank2.Level
Economic
Tank2.Sulfur Optimization
Tank2.Cloud Multivariable
Tank3.Level MPC Control
Tank3.Sulfur
Tank3.Cloud -
More CVs…
violations). CV2
CV3
• Quality control
• Cost and profit optimization
Conjoint
MVs
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.a 2.b
Slave MV index
Slave1 CVs
• Local Optimization
Slave MPC 2
The relationship is a bit convoluted, so let’s first look at the simplest case: 1 conjoint variable per Slave MPC
MV 2 High
Master MVs
k k+1
MV 2 Low
8 inquiry calls:
min c t xconj
x
Call # ct dt
1 -1 0 0 1 bLO Ac xconj A f x free bHI
2 1 0 0 1
3 0 -1 1 0 d t xconj 0 Slave MPC (i)
4 0 1 1 0
5 -1 1 1 1
6 1 -1 1 1
7 1 1 1 -1 Where, A
G
8 -1 -1 1 -1 I
From Process Unit to Plantwide Control & Optimization Page 31
Proxy Limits – Customer Example Honeywell
To what extent is the Master MPC allowed to change the Slave unit’s operations?
95.000
90.000
85.000
80.000
75.000
70.000
57
15
22
29
36
43
50
64
71
78
85
92
99
1
8
106
260
309
113
120
127
134
141
148
155
162
169
176
183
190
197
204
211
218
225
232
239
246
253
267
274
281
288
295
302
316
323
330
337
344
351
358
365
From Process Unit to Plantwide Control & Optimization Page 32
Proxy Limits – Customer Example (2) Honeywell
134
169
204
239
106
113
120
127
141
148
155
162
176
183
190
197
211
218
225
232
246
253
260
267
274
281
288
295
302
309
316
323
330
337
344
351
358
365
From Process Unit to Plantwide Control & Optimization Page 33
Master’s Control and Optimization Results:
Component Quality Changes in Tank #3 Honeywell
Tank3.Cloud Point in Deg C Tank3.Flash Point in Deg C
-40
-48 48
50 100 150 200 250 300 350 50 100 150 200 250 300 350
Sulfur is controlled at
Tank3.Qualities (After - Master MPC)
its high limit which
Historical Operation Data (Before)
means much lower
Tank3.90pct Point in Deg C Tank3.Sulfur in PPM
11 catalyst/utility costs
10
260
9
3
50 100 150 200 250 300 350 0 100 200 300 400
Average crude
saving is ~ 0.5% for -1
meeting the same
product demand.
-2
Huge cost saving!
-3
-4
-5
0 50 100 150 200 250 300 350 400
Level-3 Master
Economic
Optimization
Multivariable
MPC Control
two-way connection
two-way connection
two-way connection two-way connection two-way connection two-way connection two-way connection two-way connection
Questions?