Download Full Issue in PDF - Academy Publisher
Download Full Issue in PDF - Academy Publisher
Download Full Issue in PDF - Academy Publisher
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
1566 JOURNAL OF COMPUTERS, VOL. 8, NO. 6, JUNE 2013<br />
• Mismatch model<br />
In order to create performance degradation condition,<br />
A disturbance, Δ A was added to the nom<strong>in</strong>al model (2),<br />
to construct a man-made mismatch model as,<br />
⎛ 0 1 0 0 ⎞<br />
⎜<br />
0 0 1 0<br />
⎟<br />
A'<br />
= A+Δ A= ⎜<br />
⎟. (36)<br />
⎜ 0 0 0 1 ⎟<br />
⎜<br />
⎟<br />
⎝−600 −300 −100 −30⎠<br />
• MPC tun<strong>in</strong>g parameters<br />
Based on the extended MVC 3 algorithm comb<strong>in</strong>ed<br />
with LQR <strong>in</strong>verse optimal, the MPC controller matrixes<br />
of above mismatch model can be gotten.<br />
R = 0.0388 , Q = diag(0.0706 0.0002 0.0357 0.000)<br />
Through MVC 3 performance evaluation, controller<br />
performance decl<strong>in</strong><strong>in</strong>g was detected. If it dropped below<br />
threshold, ψ , (here, ψ is set to 0.8), then, weight<br />
parameter of manipulated variable <strong>in</strong> MVC 3 , R ,is<br />
updated by R'<br />
= R+ ξ I ,(here,ψ is set to 0.5). Repeat the<br />
MPC weight matrix calculation process to update Q,<br />
R,<br />
the new MPC controller parameters<br />
as R = 0.0952<br />
,<br />
Q = diag(0.1703 0.0006 0.0552 0.0000) can be<br />
gotten. Apply the new controller parameters to operation<br />
to restore the desired operational performance.<br />
V. SIMULATION AND ANALYSIS OF MPC CONTROLLER<br />
A. Simulation and Comparison of MPC Controller and<br />
the LQR Controller<br />
• Simulation of LQR controller<br />
LQR controlled system <strong>in</strong> the Simul<strong>in</strong>k of Matlab was<br />
shown <strong>in</strong> Fig. 3. Parameter of LQR block was set to<br />
K = ( −31.623 − 20.151 72.718 13.155)<br />
, which was<br />
provided by Googol Technology LTD.<br />
Figure 3. Simulation block of LQR control loop<br />
• Simulation of MPC controller<br />
MPC controlled system <strong>in</strong> the Simul<strong>in</strong>k was shown <strong>in</strong><br />
Fig. 4, where, parameter of Acker block was set to<br />
stabilizer feedback ga<strong>in</strong>. Weighted matrixes of MPC<br />
block were set the parameters calculated by the nom<strong>in</strong>al<br />
model.<br />
Figure 4. Simulation block of MPC control loop<br />
• Analysis of simulations results<br />
Simulation curve charts of MPC controller and LQR<br />
controller are shown <strong>in</strong> Fig. 5, where, u denotes<br />
manipulative variable, which is cart angular velocity.<br />
angle, pos denote outputs, they are pendulum angle and<br />
cart position.<br />
The maximum deviation of MPC controller and LQR<br />
controller are shown <strong>in</strong> Table I and comparison bar chart<br />
is shown <strong>in</strong> Fig. 6.<br />
TABLE I.<br />
MAXIMUM DEVIATION COMPARISON<br />
u angle pos<br />
MPC 0.8240 0.0096 0.0093<br />
LQR 34.4589 0.3101 0.3533<br />
Obviously, due to the <strong>in</strong>troduction of steady state<br />
manipulative variable and outputs covariance constra<strong>in</strong>t,<br />
MPC controller can make the maximum deviation<br />
significantly reduced than LQR controller, which greatly<br />
improved the system dynamic performance.<br />
B. Simulation of MPC Controller Tun<strong>in</strong>g<br />
• Simulation of MPC controller tun<strong>in</strong>g<br />
MPC controller tun<strong>in</strong>g dynamic curves was obta<strong>in</strong>ed<br />
by replace orig<strong>in</strong>al weight matrix Q , R of MPC controller<br />
with mismatched and Controller-tuned parameters, shown<br />
<strong>in</strong> Fig. 7, where, the legend (good, bad and tuned) means<br />
controller runn<strong>in</strong>g under, nom<strong>in</strong>al model, mismatch<br />
model and MPC controller tuned.<br />
Through curves, if us<strong>in</strong>g orig<strong>in</strong>al controller parameters<br />
to control mismatch model, it would lead an <strong>in</strong>crease on<br />
manipulative variable and outputs deviation. After<br />
adjust<strong>in</strong>g controller parameters by controller tun<strong>in</strong>g<br />
system, the deviation reduced to some extent. This proves<br />
the feasible of MPC controller tun<strong>in</strong>g algorithm.<br />
• Extended MVC 3 performance evaluation method<br />
The extended MVC 3 performance evaluation curve is<br />
shown <strong>in</strong> Fig. 8.<br />
Set λ = 1 <strong>in</strong> the MVC 3 performance evolution object<br />
function (18) and get<br />
p m<br />
i= 1<br />
i i<br />
j=<br />
1<br />
j j<br />
−7<br />
benchmark J 3 = ∑qY + ∑ rU by LMI method,<br />
MVC<br />
here, J 3 = 1.3217 × 10 .Then actual run-time<br />
MVC<br />
variance was compared with this benchmark, get η<br />
© 2013 ACADEMY PUBLISHER