Basics of Model Predictive Control

P.E. Orukpe Supervisor: Dr Imad Jaimuokha

Imperial College,London

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 1/27

Outline

s s s s s s s

Introduction Problem Formulation Solution to problem Examples Direction for future work Conclusion References

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 2/27

Introduction

What is Model Predictive Control?

MPC is a form of control in which the current control action is obtained by solving on-line, at each sampling instant, a nite horizon open-loop optimal control problem, using the current state of the plant as the initial state; the optimization yields an optimal control sequence and the rst control in this sequence is applied to the plant.

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 3/27

Introduction contd

Early Industrial MPC Application

1 Model Predictive Heuristic Control by Richalet et al. 1976 (Adersa)... 2 Dynamic Matrix Control (DMC) by Cutler and Ramaker 1979 (Shell Oil)... 3 Quadratic-Program Dynamic Matrix Control (QDMC) by Cutler et al. 1983 (Shell Oil)...

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 4/27

Introduction contd

Academic Research s Few early theoretical investigations: Klieinmann 1970, Thomas 1975, Chen and Shaw 1982 etc.
s

Predictive control theory: Keerthi and Gilbert 1988, Mayne and Michalska 1990 etc.

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 5/27

Introduction contd

Industrial Technology
Company Aspen Tech Adersa Product name DMC IDCOM HIECON PFC Honeywell Promatics Setpoint Inc. Treiber Controls Shell Global ABB Pavillion Technologies Inc. Simulation Sciences RMPCT PCT SMCA IDCOM-M OPC SMOC-II 3dMPC PP Connoisseur Process Perfecter Control and Identication Package Description Dynamic Matrix Control Identication and Command Hierarchical Constraint Control Predictive Functional Control Robust Model Predictive Control Technology Predictive Control Technology Setpoint Multivariable Control Architecture Multivariable Optimum Predictive Control Shell Multivariable Optimizing Control

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 6/27

Introduction contd

Variations of MPC
s s s

Robust MPC - guaranteed feasibility and stability Feedback MPC - mitigate shrinkage of feasible region Pre-computed MPC - Piecewise-linear solution stored in database or Solve off-line using parametric (linear or quadratic) programming Decentralised MPC as used in autonomous air vehicle Speed up computation.

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 7/27

Introduction contd

Basic structure of MPC

Figure 1: Basic Structure of MPC

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 8/27

Introduction contd

Components of MPC
s s s

Prediction model Objective function Obtaining the control law

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 9/27

Introduction contd

What makes MPC successful in industry

1. It handles multivariable control problems naturally 2. It can take account of actuator limitations 3. It allows operation closer to constraints, hence increased prot 4. It has plenty of time for on-line computations 5. It can handle non-minimal phase and unstable processes 6. It is an easy to tune method and 7. It handles structural changes.

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 10/27

Introduction contd

Characteristics of MPC
s s s s

Moving horizon implementation Performance oriented time domain formulation Incorporation of constraints and Explicit system model used to predict future plant dynamics.

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 11/27

Introduction contd

Types of MPC
Linear MPC 1. Uses linear model:
s

Nonlinear MPC 1. Uses nonlinear model:

s

x = Ax + Bu

x = f ( x, u)

2. Quadratic cost function: 2. Cost function can be nonquadratic:F (x,u) T T F = x Qx + u Ru 3. Nonlinear constraints:h(x,u) < 0 3. Linear constraints: 4. Nonlinear program Hx + Gu < 0 4. Quadratic program

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 12/27

Introduction contd

Applications of MPC s Distillation column

s s s s

Hydrocracker Pulp and paper plant Servo mechanism Robot arm ...

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 13/27

Problem Formulation

Model of Plant in State Space x (k + 1) = Ax (k) + Bu(k) y(k) = Cy x (k) z(k) = Cz x (k) (1)

where x R n , u R l , y R my , z R mz
s

A Basic Formulation of Cost function

Hp

V (k) =

i = Hw

z(k + i \k) r (k + i \k)

2 Q (i ) +

Hu 1

u(k + i \k)

i =0

2 R (i )

(2)

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 14/27

Problem Formulation contd

x (k + Hu \k ) x (k + Hu + 1\k ) . . . x (k + H p \k) B

x ( k + 1\ k ) . . .

A Hu = A Hu +1 . . . AHp ... ... .. . 0 0 ... B

A . . .

Hu 1 Ai B i =0 x (k) + Hu Ai B i =0 . . . H p 1 i =0 A i B
past

B . . .

u ( k 1)

AB + B . . . + Hu 1 i i =0 A B iHu0 Ai B = ... H p 1 i i =0 A B
ICM,EEE-CAP,April 14, 2005

... ... ... ...

AB + B ...
H p Hu i =0

Ai B

u(k\k) . . . u(k + Hu 1\k )

(3)

future
Basics of MPC - p. 15/27

Problem Formulation contd

The predictions of z is z ( k + 1\ k ) . = . . z(k + H p \k)

Cz 0 . . . 0

0 Cz . . . 0

... ... .. . ...

0 0 . . . Cz

x ( k + 1\ k ) . . . x (k + H p \k)

(4)

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 16/27

Solution to problem

Unconstrained problem with measured state :The cost

function to be minimized is
Hp

V (k) =

i = Hw

z(k + i \k) r (k + i \k)

2 Q

2 Q (i ) +

Hu 1

u(k + i \k)

i =0 2 R
u(k + i \k) . . . u(k + Hu 1\k )

2 R (i )

(5)

Z (k) T (k)
Z (k) = z(k + Hw \k ) . . . z(k + H p \k) T (k) =

+ U (k)

r (k + Hw \k ) . . . r (k + H p \k)

U (k) =

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 17/27

Solution to problem contd

Z (k) = x (k) + u(k 1) + U (k)

(k) = T (k) x (k) u(k 1)

V (k) = U (k) (k)

= [U (k)T T (k) ]Q[U (k) (k)] + U (k) T RU (k)

2 Q+ T

U (k)

2 R

= (k)T Q (k) 2U (k)T T Q (k) + U (k)T [ T Q + R]U (k)

U (k)opt = 0.5H1 G

(6)

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 18/27

Solution to problem contd

Constrained problem with QP formulation : The

system is subject to constraints of the form:
E U (k) 1 F

(7)

U (k)
1

(8)

Z (k)
1

(9)

The constraints are converted to a single linear inequality of the form:

F F1 u(k 1) f U (k ) 0 [x (k ) + u(k 1)] g W w

(10)

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 19/27

Solution to problem contd

The constrained optimization problem to be solved is

U (k)

min U (k) T H U (k) G T U (k)

(11)

subject to the inequality

U (k) 0

F1 u(k 1) f [x (k) + u(k 1)] g w

(12)

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 20/27

Example

A randomly generated system with state space model parameters:

0.1267

0.3357
0.3487 0.3482

0.0958 0.0511

0.1723
0.6393

0.4877 A= 0.0367 0.0842

0.0547
0.2125

0.0110

, B = 0.2311 0.0399 0.6068 0.0334 0.4860

0.9501

0 1 0 0

0 0 1 0

0 Cy = 0 0

0 , Cz = 0 1

H p = 5; Hu = 3; Hw = 1; Q > 0; R > 0 and using the following constraints

10 u(k) 10 2 u(k) 2 3 z ( k ) 5
with initial state x0 = [0.5 0.5 0.2 0.2] T
ICM,EEE-CAP,April 14, 2005 Basics of MPC - p. 21/27

Example contd

9 8 7 6 5 4 3 2 1 0 x1,x2,x3,x4

1.4 1.2 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0 20 40 time 60 80 100 0.6 0 20 40 time 60 80 100 state 1 state 2 state 3 state 4

Figure 2: Cost function for unconstrained case

cost

Figure 3: States for unconstrained case

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 22/27

Example contd

25

1.4 1.2 state 1 state 2 state 3 state 4

20

1 0.8 x1,x2,x3,x4 0 20 40 time 60 80 100

15 cost

0.6 0.4 0.2 0

10

0.2 0.4

0.6

20

40 time

60

80

100

Figure 4: Cost function for constrained case

Figure 5: States for constrained case

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 23/27

Direction for future work

Consider normbounded uncertain system of the form: x (k + 1) = Ax (k) + Bu(k) y(k) = Cx (k), A = A0 + and B = B0 + (13) (14)

It is not possible to minimize the cost and as such we will consider an upper bound on the cost Use LMIs which are of the form F ( x ) = F0 + xi Fi > 0,
i =1 l

(15)

to minimize this upper bound

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 24/27

Conclusion

s s s s

The importance of MPC has been shown We have mentioned some practical application of MPC It can handle constraints We have shown the components of MPC Model Objective function Obtaining the control law We have outlined a future direction for research concerning Robust Model Predictive Control

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 25/27

References

[1] Garcia C.E, Prett D.M. and Morari M., Model Predictive Control: Theory and Practice a Survey. Automatica, Vol.25, No.3, pp.335-348, 1989. [2] Mayne D.Q.,Rawlings J.B., Rao C.V., Scokaert P.O.M., Constrained model predictive control: Stability and optimality. Automatica, Vol.36,pp. 789-814,2000. [3] Lee J.H, Morari M. and Garcia C.E., State-space Interpretation Of Model Predictive Control. Automatica, Vol.30, No.4, pp. 707-717, 1994. [4] Rawlings J.B., Tutorial Overview of Model Predictive Control. IEEE Control Systems Magazine, pp.38-52, June 2000. [5] Qin S.J., Badgwell T.A., A survey of industrial model predictive control technology. Control Engineering Practice, Vol.11, pp.733-746, 2003. [6] Morari M., Lee J.H., Model predictive control: past, present and future. Computers and Chemical Engineering, Vol.23, pp.667-682, 1999. [7] Maciejowski J.M., Predictive control with constraints. Prentice Hall, 2002. [8] Camacho E.F. andd Bordons C., Model Predictive Control. Second Edition, Springer, 2004. [9] Rawlings J.B. and Muske K.R., The Stability Of Constrained Receding Horizon Control. IEEE Transactios on Automatic Control, Vol.38, No. 10, October 1993.

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 26/27

Thank You

ICM,EEE-CAP,April 14, 2005

Basics of MPC - p. 27/27