Model Predictive Controller

1
Model Predictive Controller

• Emad Ali
• Chemical Engineering Department
• King Saud University

2
Review

• Major Control Elements
• Instrumentation
• Control algorithm
• Process model

3
Review

• Control Algorithms
• Classical
• PID, cascade, override, ratio, split range,
  inferential
• Advanced
• Adaptive control
• Fuzzy logic control
• Internal model control
• Optimal control
• Neural network control
• Globally linearizing control
• Model predictive control

4
Benefits of MPC
5
Industrial MPC Technology

• IDCOM (Identification and Command)
• Model type Impulse response
• Optimization is solved by QP approach
• DMC (Dynamic Matrix Control)
• Model type Step response
• Optimization is solved by LP approach
• OPC (Optimum Predictive Control)
• Use step response, solves LP problem
• Model building, controller design and simulation
• tasks are carried out on PCs
• PCT (Predictive Control Technology)
• Combines the aspects of IDCOM and DMC
• HMPC (Horizon Multivariable Predictive Control)
• For proprietary reasons, information is
  unavailable

6
Multivariable vs. Multi-loops
7
General Application Concept
8
Receding Horizon Concept(Prediction)
9
Process Model
Model
Y(k/k) M Y(k-1/k) S Du(k-1/k)
Y(k/k) y(k/k) y(k1/k) y(kn-1/k)
Y(k-1/k) y(k-1/k) y(k/k) y(kn-2/k)
Prediction
Y(k1/k) MP Y(k/k) SP DU(k/k)
Y(k1/k) y(k/k) y(k1/k) y(kP-1/k)
DU(k/k) Du(k/k) Du(k1/k) Du(kM-1/k)
10
Process Model Correction

• Output Feedback

Y(k1/k) Y(k1/k) N yp(k)-ym(k)
11
Methods of Solution
1. Algebraic Equation
R(k1)Y(k1) MP Y(k) SP DU(k)
2. Linear Programming (LP)
R(k1)-Y(k1) R(k1)-MP Y(k) SP DU(k)
0
12
Methods of solution
3. Quadratic Programming
min R(k1)-Y(k1)T G R(k1)-Y(k1) DUT(k) L
DU(k) DU(k)
4. Constrained QP
min R(k1)-Y(k1)T G R(k1)-Y(k1) DUT(k) L
DU(k) DU(k)
Ul U Uu DUl DU DUu
13
Methods of solution
14
Tuning Parameters
Output weights
Input weights
Prediction horizon
Control horizon
15
Tuning Guidelines
16
Generating Step Response Model
1. Step testing
17
Generating Step Response Model
2. PBRS Testing
18
Step Vs. PBRS
19
Implementation requirement

• DCS system
• Personal Computer
• Step response model
• Tuning

20
Special Features of MPC

• Feed-forward capability
• Inferential control
• Output Constraints

Y(k1/k) M Y(k/k) S DU(k/k) W d(k/k)
Y(k1/k) M Y(k/k) S DU(k/k) Z(k1/k)
CY(k1/k)
Yl (k1) Y (k1) Yu(k1)
21
Special Features of MPC

• Variable set point
• or constraints

22
Simulation Example
23
F200/C2, P100/P2
24
P100/C2, F200/P2
25
(No Transcript)
26
(No Transcript)
27
Thank You

• Questions ?