model-driven pid control system, its properties and ...apc2004apc2004 toshiba toshiba it & control...
TRANSCRIPT
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Model-Driven PID Control System, its properties and multivariable application
ModelModel--Driven PID Control System, its Driven PID Control System, its properties and multivariable applicationproperties and multivariable application
Takashi Takashi ShigemasaShigemasa, Masanori , Masanori YukitomoYukitomo, , Power and Industrial Systems Research and Development Center , Power and Industrial Systems Research and Development Center ,
Toshiba CorporationToshiba CorporationToshiba IT & Control Systems CorporationToshiba IT & Control Systems Corporation
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
OutlineOutline
◇ Introduction
◇ Model Driven PID Control Systemstructure and properties
◇ Multivariable Model Driven PID Control System structure and simulations
◇ Conclusions
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
IntroductionIntroduction
PID control system: widely used as a basic control technology,
however, is not always easy tuning and has control limitation for long
dead-time processes.
What is a simple, widely applicable process controller with easy to tune?
Based on the motief, we developed Model Driven PID Control System byextending the Model Driven control concept proposed by Kimura.
Firstly, Model Driven PID Control System, its structure, properties.
Secondary, the Model Driven PID Control System for multivariable processes on regulatory control level and simulation results.
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Model Driven PID Control SystemModel Driven PID Control System
Model Driven Control Concept
Model Driven PID control system
Structure and Properties
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Model-Driven Control ConceptModel-Driven Control Concept
Definition:A control system architecture which uses a model of the plant as a
principal component of Controller is called a model-driven control.
Kimura(CDC2000Sydney)
Features of MDC
● Simple Structure
● Easy Tuning
● Proven Stability
and robustness
iff P(s) is stable
-
-
Controller Plant
IMC Architecture Morari, Zafirou,1989
Q(s) P(s)
(s)P̂
IMC
r y
ŷ
e u
Model-
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Structure of Model Driven PID Control SystemStructure of Model Driven PID Control SystemStructure of Model Driven PID Control System
・PD Feedback
・Main control ;
Gain , second order Q-filter and normalized first order delay
model with dead-time for inner loop
・Set-point filter
cK
sL
c
cesT
−
+11
Model
Q Filter Gain
r v + -
+
+
Main Controller
2)1()1)(1(
sTsTsT
c
cc
λα
+++
sTsT
c
c
αλ
++
11
Set Point Filter
sT
sTK
f
ff
κ+
+
1
)1(
)(sPProcess
PD Feedback
u y - +
d
+ +
)(sG
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Step 1 PD Feedback F(s)
Compensate inner loop dynamics G(s) to a first order system with dead time by using PD feedback F(s).
Design methods:partial model matching method,frequency region method,pole placement,simulation.
Wide applicability, for not only first order delay process with dead time but also integral process, oscillatory process and unstable process
sT 1
s)T(1KF(s)
f
ff
κ+
+=
y+
−P(s)v
F(s)
Design stepsDesign stepsDesign steps
TsLsK
sPsFsPsG
+−
≅
−= −
1)exp(
)()]()(1[)( 1
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
0 20 50 60 900
1
2
4
5
6
sexp(
)s(P10
=-10s)
sexp(exp(
)s(G40+
4=
-10s)-10s)
s.exp(
)s(Ĝ428+1
4=
-11.8s)
4
sexp(
)s(P10
=-10s)
sexp(exp(
)s(G40+
4=
-10s)-10s)
s.exp(
)s(Ĝ428+1
4=
-11.8s)
250.KF(s) f ==
Example: A integral process can be converted to a first order delay system with dead time
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Step 2:Main controller and set-point filter
1) Kc=1/K
2) Tc=T
3) Lc=L
Step 3:TDOF property by adjusting λand α.
λ: response speed α:Disturbance regulation
]ds)T (1
s)Ts)(1T(1[1
s)T(1Ks)exp(-L
rsT1s)exp(-L
y 2c
cc
cc
c
c
c
λα
λ +++
−+
++
=
2c
cc
s)T(1s)T s)(1T(1
Q(s)λ
α+
++=
sT1sT 1
c
c
αλ
++
Design steps(continued)Design steps(continued)
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
0 20 40 60 80 100 120 140 160 180 2000
0.5
1
1.5
2 responces
PV SV
0 20 40 60 80 100 120 140 160 180 2000
0.5
1
1.5
2
time
MV D
V
1=?
50=? .
51=? .
1=?
51=? .
50=? .
SV
DV
ses
sG 1051
1)( −
+=
dT
esTsTK
er
sTe
yc
sLc
cc
sL
c
sL ccc
+
+−
++
+=
−−−
2)1()1(
1)1(1 λ
αλ
Adjustable response speed by λ
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
0 20 40 60 80 100 120 140 160 180 2000
0.5
1
1.5
2 responces
PV
SV
0 20 40 60 80 100 120 140 160 180 200-0.5
0
0.5
1
1.5
time
MV
DV
SV
1=a
3=a
DV
751=a .
3=a
751=a .1=a
ses
sG 1051
1)( −
+=
dT
esTsTK
ey
c
sLc
cc
sL cc
+
+−
+=
−−
2)1()1(
1)1( λ
α
Adjustable Disturbance regulation speed by α
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Nyquist curves
s7.281s)21(exp714.0
G(s)
4.0F(s)s501
s)20(expP(s)
+−
=
=+
−=
Robustness
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
+-
+
sTK
cc
11
r u y)( sP
0,1,0 ===== cff LTK αλPI Control:
+-
+
sTK
cc
11
r u y)( sP
0,1,0 ===== cff LTK αλPI Control:
+-
+
sTK
cc
11
r u y)( sP+-
+
sTK
cc
11
+
sTK
cc
11
r u y)( sP )( sP
0,1,0 ===== cff LTK αλPI Control: PI-PD Control: 0,1 === cLαλ
sT
sTK
f
ff
κ+
+
1
)1(
+-
+
sTK
cc
11
r u y)( sP+-
PI-PD Control: 0,1 === cLαλPI-PD Control: 0,1 === cLαλ
sT
sTK
f
ff
κ+
+
1
)1(
+-
+
sTK
cc
11
r u y)( sP+-
sT
sTK
f
ff
κ+
+
1
)1(
sT
sTK
f
ff
κ+
+
1
)1(
+-
+
sTK
cc
11
+
sTK
cc
11
r u y)( sP )( sP+-
sL
c
cesT
−
+11
cKr u y)( sP+
-
++
1,0 ==== αλff TKPIDτd Control:
sL
c
cesT
−
+11
cKr u y)( sP+
-
++
1,0 ==== αλff TKPIDτd Control:
sL
c
cesT
−
+11
cKr u y)( sP+
-
++
sL
c
cesT
−
+11 sL
c
cesT
−
+11
cK cKr u y)( sP )( sP+
-
++
1,0 ==== αλff TKPIDτd Control:
+- sT
sT
c
c
λ++
11
sL
c
cesT
−
+11
cKr u y)( sP
++
1,0 === αff TKIMC:
+- sT
sT
c
c
λ++
11
sL
c
cesT
−
+11
cKr u y)( sP
++
1,0 === αff TKIMC:
+- sT
sT
c
c
λ++
11
sL
c
cesT
−
+11
cKr u y)( sP
+++
- sTsT
c
c
λ++
11
sTsT
c
c
λ++
11
sL
c
cesT
−
+11 sL
c
cesT
−
+11
cK cKr u y)( sP )( sP
++
1,0 === αff TKIMC: 1,0 === αff TKIMC:
cK
sL
c
cesT
−
+11
Model
Q Filter Gain
r v + -
+
+
Main Controller
2)1()1)(1(
sTsTsT
c
cc
λα
+++
sTsT
c
c
αλ
++
11
Set Point Filter
sT
sTK
f
ff
κ+
+
1
)1(
)(sPProcess
PD Feedback
u y - +
d
+ +
)(sG
Upper compatibility
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Summery of Properties
1. Wide Applicability PD feedback
2. Two degree of freedom characteristicsKc=1/K Tc=T Lc=L
λ tuning;Response speed, α tuning:Disturbance regulation speed
3. Robustness can be examined by well-known nyquistchart
4. Upper compatibility from conventional PID control systems
s)T? (1s)T(1K
F(s) s T 1
) s L Kexp() s G(
f
ff
++
=+
−≅
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
Control performance - PV distribution -
-5 -4 -3 -2 -1 0 1 2 3 4 50
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Before
AfterOld SV
New SV
Reported reduce operator’s over-riding actions and fuel cost, obtain high quality of product though some field applications.
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
1r
2r
+
-
-
+
sTsT
cc
cc
11
11
11
αλ
++
sTsT
cc
cc
22
22
11
αλ
++ +
=
2221
1211
dd
dd
KKKK
K
)1( 11 sTK ff +
Decoupler
PD Feedback2
)1( 22 sTK ff +
-
-+
+
sTe
c
sLc
11
1
+
−
211
111
)1()1)(1(
sTsTsT
c
cc
λα
+++
222
222
)1()1)(1(
sTsTsT
c
cc
λα
+++
sTe
c
sLc
21
2
+
−
+
Model1
Model2
Q1
Q2
PD Feedback1
2y
1y
MIMO Process
)(sP
+
+1cK
2cK
Multivariable Model Driven PID Control SystemMultivariable Model Driven PID Control System
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
1.Decoupler(Kdec)
the control process
1
2221
12111
22
22
2221
21
21
12
12
1211
11
11
)0(
11
11)(−
−
−−
−−
==⇒
++
++=KKKK
PKe
sTK
esT
K
esT
Ke
sTK
sP decsLsL
sLsL
2.Other MD control systems
=
2221
1211
ˆˆˆˆ
)(PPPP
sPKdec
Design stepsDesign stepsDesign steps
A decoupler
Designed for the diagonal element of KdecP(s)
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
0 200 400 600 800 10000
1
2
3
4
5
6
G11
0 200 400 600 800 10000
1
2
3
4
5
6G
12
0 200 400 600 800 10000
1
2
3
4
5
6
G21
0 200 400 600 800 10000
1
2
3
4
5
6
G22
SimulationSimulationSimulationControlled object : Shell heavy oil fractionator
++
++=−−
−−
ss
ss
es
es
es
essP
1418
2827
60127.5
50139.5
60177.1
50105.4
)(
IEEE Control System MagazineShell heavy oil fractionator
By Daniel E. Rivera
Step responses
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
1.Distributed MD PID Control
Design resultsDesign resultsDesign results
−
−=
3431.04567.01500.04465.0
Decoupler
ss es
es
1427
6011
model2,5011
model1 −−+
=+
=
2.Decoupling PID Control by Model-matching method by Kitamori
=
−
−=
6909.581015.516909.471015.45
7372.03455.02618.02982.0
I
p
T
K
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
0 200
400
600
800
1000
1200
1400
1600
1800
2000
-0.50
0.51
1.52
0 200
400
600
800
1000
1200
1400
1600
1800
2000
-0.20
0.20.4
0.60 200
400
600 800 1000 1200 1400 1600 1800 2000-0.5
0
0.5
1
1.5
2
0 200 400 600 800 1000 1200 1400 1600 1800 2000-1
-0.5
0
0.5
time
SV1 Step
DV1 Step
SV2 Step
DV2 Step
1y
2y
1u
2u
Green:Decoupling PIDBlue:MD-PID
Modelling error 0%
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
0 200 400 600 800 1000 1200 1400 1600 1800 2000-0.5
0
0.5
1
1.5
2
出力
1
0 200 400 600 800 1000 1200 1400 1600 1800 2000-0.4
-0.2
0
0.2
0.4
0.6
入力
1
0 200 400 600 800 1000 1200 1400 1600 1800 2000-0.5
0
0.5
1
1.5
2
出力
2
0 200 400 600 800 1000 1200 1400 1600 1800 2000-1
-0.5
0
0.5
1
入力
2
time
Modelling error 25%
SV1 Step
DV1 Step
SV2 Step
DV2 Step
Green:Decoupling PIDBlue:MD-PID
-
APC2004APC2004
TOSHIBA IT & CONTROL SYSTEMS CORPORATIONTOSHIBA
ConclusionsConclusions
PID
MD-PID
Large L/TSmall L/TUnstableIntegralOscillationZero
Delay with dead time
★We discussed a Model Driven PID control system,its properties and a multivariable application.
★PD feedback give us wide applicability of controlled processes:
★MD-PID control system shows Good control performances like a TDOF control system with easy tuning.
★ MD- PID control system has upper compatibility from conventional PI, IMC, PIDτdand PI-PD
★ Multivariable MD PID Control system shows useful results in spite of simple control structure.