Chapter 6 Model Predictive Control Prof. Shi-Shang Jang National Tsing-Hua University Chemical Engineering Department Historical Development Questions: – Given a system what are the absolute limitations to output control? – How can one make use of a process model and account for model error? • Smith Predictor(1955) • Feedforward Control • Inferential Control (1975, 1979) • Dynamic Matrix Control (Shell, 1979) • Model Algorithmic Control (France, 1979) • Internal Model Control (Garcia and Morari, 1982) 2 Criteria for Controller Quality Regulatory Behavior – Compensation for (unmeasured disturbances) Servo Behavior- Follow set point changes (fast, smooth, no offset) Robustness-Controller should be effective when there are modeling errors (both structure and parameters) 3 Criteria for Controller QualityContinued Constraints- Ability to deal with constraints on inputs and states (no windup) Remark: 90% of loops can be handled by PID type controllers. 4 Internal Model Control (IMC) Structure d ys + - e u GI GP Gm + + ym - y + dˆ e=ys-y+ym 5 Analysis of Internal Model Structure From the block diagram: yd u G p GI y s d 1 GI (G p Gm ) ys d GI 1 GI (G p Gm ) 6 Properties of IMC (Principles of Internal Model Control) 1. (Dual Stability) If the model is perfect, stability of controller and plant is sufficient for overall system stability Proof: If Gp=Gm, then y=GpGI(ys-d)+d u=(ys-d) GI Use IMC only on stable systems, unstable systems can be stabilized by feedback control Constraints on inputs has no effect on stability 7 Properties of IMC (Principles of Internal Model Control)- continued 2. (Prefect Control) If model is perfect and invertible, and GI=Gp-1, then y=ys for any d Notes: (1) This is “optimal control”. (2) Suppose Gp-1 is not realizable, then it is recommended to factor this transfer function into two terms: Gp(s)=G+(s)G-(s) where G+(s) is not realizable contains all time delays and RHP zeros. In this case the “best” controller possible is: GI(s)=G--1(s), this controller minimizes sum of the square of the errors in output. (3) This suggests the design of F(s) as GI=Gp-1 F(s) such that GI(s) realizable. 8 Example G p (s) N ( s ) Ds e D( s) GI ( s) G p1 ( s) F ( s) D( s ) Ds e F ( s) N ( s) Which is realizable if we choose: e Ds F ( s) s 1n Where n=degree of D(s)-degree of N(s)>0, is chosen 9 Properties of IMC (Principles of Internal Model Control)- continued 3. Zero Offset There is no offset if we choose: GI(0)=1/Gm(0) Pf: y (0) 1 Gm (0)GI (0)d (0) G p (0)GI (0) ys (0) 1 GI (0) G p (0) Gm (0) y s ( 0) a. This means the output will attain the set point exactly in presence of persistent disturbances and set-point changes integral feedback b. Note that this is true even if the model is imperfect. 10 Properties of IMC (Principles of Internal Model Control)- continued 4. Comparison with Feedback Controller If we choose G s I Gc 1 GcGm Then we get classical output feedback control. Note that GI(s) in the closed loop transfer function of the following: e(s) + - Gc(S) m(s) Gm(S) 11 Properties of IMC (Principles of Internal Model Control)- continued Joining this with the IMC block diagram we get ys + - e(s)+ - Gc(S) m(s) Gp(S) Gc(S) d Gp(S) y(s) + - which reduced to a classical feedback control: ys(s) + - Gc Gp d 12 Properties of IMC (Principles of Internal Model Control)- continued Notes: (a) G s I Gc 1 GcGm may be looked upon as a “poor” approximation to Gm-1(s). For large Kc, this approximation gets better. (b) Note the similarity of this structure to Smith Predictor also approximates the invertible part of Gm-1(s). (c) This feedback structure implies use of a filter: F s Gc s Gm s 1 Gc s Gm s 13 On-Line Tuning of IMC 1. 2. 3. Choose a process model through plant tests Choose a filter F s e Ds to make GI(s) n s 1 realizable Decrease untill system becomes oscillatory. Brosilow recommends the following time constant: (where u=min. filter constant; Pu=period of oscillation) 2 4. 2 n 4 u Pu2 Pu2 2 1/ 2 2 If /D<1, then IMC will yield better performance than a PID controller. If 1< /D<2; then IMC and PID are competitive. 14 Computation of Approximate Inverses In practice, it is easier to find an approximate inverse of the process transfer function in the time domain (using discrete models). Time Domain View: Given the past history of inputs to the process and current estimate of the disturbance, compute the current and future inputs which will make the output follow the desired set point. Limitation in Practice: 1. The future will be limited to a finite time horizon (3-4 times time-constant of system. 2. Attention must be limited to values of output at discrete times. 15 Summary of MPC MPC consists of three blocks: • • • Process model A controller (approximate inverse) A filter Advantages: • • • • • Quality of response depends on controller design Robustness depends on filter Stability is not an issue Implementation is straight forward On-line tuning can be provided by the filter time constant 16 Computation of Approximate Inverses - Continued 3. Values of ‘all’ future inputs may be limited to a few in the immediate future. 4. Problem must be solved every so often (at discrete sampling times) when new estimates of disturbance become available. 5. We must limit the size and velocity of control input variations. 6. On-line computations should be kept to a minimum. 7. Smooth transfer between auto/manual should be possible. 8. It should be recognize constraints on inputs. 9. There should be operator adjustable constant(s) to account for plant/model mismatch. 17 Review of least-square problem Given a set of equations: Ax=b+e We seed a solution which minimizes iei2 The solution is given by: x=(ATA)-1ATb We term (ATA)-1AT to be pseudo inverse of matrix A 18 A Discrete Input Plant Model y( z) N ( z ) z m( z ) d ( z ) D( z ) Let N ( z) h1 h2 z 1 h3 z 2 hN z N 1 z 1 D( z ) Note that N/D is actually the impulse response of the system m(z)=1 without delay y(t) h3 h4 h1 h2 h5 h6 h7 t 19 Example: G(s)=1/(s+1)3 global m m=1;TSPAN=[0 1]; Y0=[0 0 0]; Y_real=[];Y_sample=[];T_real=[]; T_sample=[0];Y_model=[0]; for i=1:29 [T,Y] = ODE45('model_3',TSPAN,Y0); TSPAN=[TSPAN(2),TSPAN(2)+1]; Y0(1)=Y(end,1); Y0(2)=Y(end,2); Y0(3)=Y(end,3); TT=T(end); Y_real=[Y_real;Y(:,1)];T_real=[T_ real;T]; m=0; T_sample=[T_sample,TT]; Y_model=[Y_model,Y0(1)]; end function dy=model_3(t,y) global m dy(1)=y(2); dy(2)=y(3); dy(3)=-3*y(3)-3*y(2)-y(1)+m; dy=dy'; 20 Example: G(s)=1/(s+1)3 continued TSPAN=[0 1];mm=zeros(1,29); Y0=[0 0 0]; Y_real=[];Y_sample=[];T_real=[]; T_sample=[0];Y_pred=[0]; for i=1:50 m=randn(1,1); for i=1:28 mm(29-i+1)=mm(29-i); end mm(1)=m; [T,Y] = ODE45('model_3',TSPAN,Y0); TSPAN=[TSPAN(2),TSPAN(2)+1]; Y0(1)=Y(end,1); Y0(2)=Y(end,2); Y0(3)=Y(end,3); TT=T(end); Y_real=[Y_real;Y(:,1)];T_real=[T_real;T]; yp=Y_model*mm'; Y_pred=[Y_pred,yp]; T_sample=[T_sample,TT]; end 21 A Discrete Input Plant Model yk 1 h1mk h2 mk 1 hN mk N 1 Let 0 for example,four step ahead forecast yk 4 h1 y 0 k 3 yk 2 0 y k 1 0 h2 h3 h1 0 0 h2 h1 0 h4 mk 3 hN h3 mk 2 h2 mk 1 h3 h1 mk h2 0 mk 1 0 mk 2 hN 0 h3 hN mk N 0 hN 0 0 ˆ Λm yˆ Am 22 A Discrete Input Plant ModelContinued y( z ) z 1H ( z )m( z ) d ( z ) G( z )m( z ) d ( z ) This expresses y in terms of past inputs m; i.e.: Then y( z ) z 1 N h z i 1 l l 1 m( z ) d ( z ) yk 1 h1mk h2 mk 1 hN mk N 1 d yˆ k 1 d 23 Approximate Inversion Since we cannot make y(t)=yd(t) exactly, we pose the following least square minimization problem: P min m ( k ), m ( k 1),...,m ( k M 1) 2 2 2 2 ˆ y k l y k l l d l m k l 1 i 1 subject to the above process model: y( z) z 1 N h z i 1 l l 1 m( z ) d ( z ) m(k M 1) m(k M ) m(k P 1) No control changes beyond M 24 The Solution The previous problem can be solved based on a Quadratic Programming solver or using previous pseudo-inverse of matrix approach. 25 MPC-Servo Control (A Feed-forward Approach) Want yk+1=yk+2=…=yd yd h1 h2 y 0 h 1 d yd 0 0 yd 0 0 y d Am Λ m h3 h2 h1 0 h4 mk 3 hN h3 mk 2 h2 mk 1 h3 h1 mk h2 0 mk 1 hN 0 0 mk 2 hN 0 h3 hN mk N 0 0 ˆ A 1 y d Λ m m P=4; M=4 26 MPC-Servo Control (A Feedforward Approach) -Example Y_real Y_sample time time P=4; M=4 27 MPC-Servo Horizon Control (A Feedforward Approach) Want yk+1=yk+2=…=yd, but mk+1=mk+2=mk+3 yd h1 h2 h3 h4 hN m y h h h 1 2 3 k 1 d yd h1 h2 mk h3 y 0 h d h2 1 y d Am Λ m ˆ pinv( A)y d Λ m m 0 mk 1 hN 0 0 mk 2 hN 0 h3 hN mk N 0 0 P=4; M=2 28 MPC-Servo Horizon Control (A Feedforward Approach) Want yk+1=yk+2=…=yd, but mk+1=mk+2=mk+3 Response Time P=4; M=2 29 MPC-Regulation Control (A Feedback Approach) yd h1 h2 h3 y 0 h h 1 2 d yd 0 0 h1 yd 0 0 0 ˆ Λm d y d Am h4 mk 3 hN h3 mk 2 h2 mk 1 h3 h1 mk h2 0 mk 1 d 0 mk 2 d hN 0 h3 hN mk N d 0 hN 0 0 ˆ A 1 y d Λ m d m d yk yˆ k P=4; M=4 30 MPC-Regulation Control (A Feedback Approach) yd h1 h2 h3 h4 hN m y h h h 1 2 3 d k 1 yd h1 h2 mk h3 y 0 h d h2 1 ˆ Λm d y d Am ˆ pinv( A)y d Λ m d m 0 mk 1 hN 0 0 mk 2 hN 0 h3 hN mk N 0 0 d yk yˆ k P=4; M=2 31 MPC-Regulation Control (A Feedback Approach) Response Response Time M=2 Time M=4 32 Multi-variable Discrete Input Plant Model 1 12 2 12 2 12 2 y1k 1 h111m1k h211m1k 1 h11 N mk N 1 h1 mk h2 mk 1 hN mk N 1 Let 0 for example,four step ahead forecast y1k 4 h111 h211 h311 1 11 11 yk 3 0 h1 h2 y1k 2 0 0 h111 1 0 0 yk 1 0 2 y h 21 h 21 h 21 2 3 k2 4 1 21 21 yk 3 0 h1 h2 y2 0 0 h121 k2 2 yk 1 0 0 0 0 0 0 h512 h12 h412 m1k 3 h511 h11 N N 1 11 11 12 12 12 12 hN 0 0 h4 0 h1 h2 h3 mk 2 h4 1 11 11 12 12 hN 0 h312 0 0 h1 h2 mk 1 h3 h311 h11 h212 0 0 0 h112 m1k h211 N h122 h222 h322 h422 mk23 h521 hN21 0 0 0 h522 hN22 0 h122 h222 h322 mk2 2 h421 hN21 0 0 h422 0 0 h122 h222 mk21 h321 hN21 0 h322 0 0 0 h122 mk2 h221 h321 hN21 h222 h411 h112 h311 h211 h111 h421 h321 h221 h121 h212 h312 m1k 1 h12 0 0 m1k 2 N h12 0 N 1 h312 h12 mk N N 0 0 0 mk21 hN22 0 0 mk2 2 hN22 0 h322 hN22 mk2 N 0 0 0 ˆ Λm yˆ Am 33 Examples of Multivariable Control: Control of a Mixing Tank Hot Cold LT TT MV’s: Flow of Hot Stream Flow of Cold Stream CV’s: Level in the tank Temperature in the tank 34 Example- Mixing Tank Problem Height Time 35 Example- Mixing Tank Problem Temperature Time 36 Dynamic Matrix Control (DMC) Response Response a4,…. a1 a2 a3 h1 h2 h3 Time Step response h4,…. Time Pulse response 37 Dynamic Matrix Control (DMC)Continued y( z) z 1 N h z i 1 l 1 l m( z ) d ( z ) 1 1 z 1 1 h3 z 2 hN z N 1 1 h1 h2 z y( z) z d ( z) 1 1 z z 1 h1 h2 z 1 h3 z 2 hN z N 1 1 z 1 z 2 Let m( z ) z 1 z 1 h h h z h h h z d z a a z a z a z 1 1 1 1 2 2 2 1 1 2 3 2 3 hl al al 1 al (1 z 1 ) N 1 N 38 Dynamic Matrix Control (DMC)Continued y( z) z 1 N l 1 1 a z 1 z m( z ) d ( z ) l i 1 But 1 z m( z ) mz 1 Finally y( z) z 1 N l 1 a z l m( z ) d ( z ) i 1 y 1 a1m1 a2 m0 a3 m1 a N m N 1 d Effect of the past disturbance 39 Dynamic Matrix Control (DMC)Continued yd a1 y 0 d yd 0 yd 0 a2 a1 a3 a2 0 0 a1 0 a4 mk 3 aN a3 mk 2 a2 mk 1 a3 a1 mk a2 0 aN 0 0 aN a3 0 mk 1 d 0 mk 2 d 0 aN mk N d ˆ Λm d y d Am ˆ A 1 y d Λm d m d yk yˆ k P=4; M=4 40 Dynamic Matrix Control (DMC)Continued yd y a a d 1 3 yd 0 yd aN a2 a4 mk 1 a1 a3 mk a3 a2 0 0 0 aN 0 aN a3 0 mk 1 d 0 mk 2 d 0 aN mk N d mk 2 mk 3 ˆ Λm d y d Am ˆ pinv( A ) y d Λm d m d yk yˆ k P=4; M=2 41 Tuning Procedures Sampling time (T): stability is not affected by T. Larger T leads to less variations in m, but deteriorates system performance in presence of frequent disturbances Horizon for m (M): Choosing M=P (perfect control) leads to severe oscillation in m(t). Reducing M, leads to a more desired response 42 Tuning Procedures - Continued Input penalty parameter : Increasing makes system more sluggish and nonzero lead to offset, but can be compensated by adding integral control algorithm itself. Optimization horizon, P: increasing P gets better inverse for system of order n, P>2n is generally sufficient. 43 Summary - Continued Model Predictive Control (MPC) is the major existed advanced process control in chemical engineering industrial The modeling in the MPC is crucial The tuning of MPC using M (horizon of suppression) is the most effective for stability. All other parameters may also frequently implement to improve the control quality 44