Research On A PI-PD Controller Optimized By Model Predictive Control

Proportional-integral-derivative (PID) control is widely applied to various kindsof industrial processes with its simple structure and convenient implementation.However, for processes with complex behavior, uncertainties and large time delay,PID control may not always satisfy the higher requirements. Compared with PIDcontrol, proportional-integral-proportional-derivative (PI-PD) control can provideimproved control performance but is now used in a limited scope because itscontroller parameter tuning is a little bit more inconvenient than PID.Model predictive control (MPC) which is a kind of the advanced control hasbeen intensively studied in the industrial process control. On one hand, the processmodel does not need to be accurately known; on the other hand, it has a good trackingperformance and strong robustness. However, the cost of MPC is much higher thanPID control, which results in the fact that MPC strategy is less widely used than PID.If we design a new controller combining the advantage of MPC strategy and thesample structure as PI-PD controller, this can benefit the improvement of theproduction efficiency.The thesis is divided into two aspects. The first aspect is using typical modelpredictive control to optimize the PI-PD controller. Through combining typical MPCalgorithm with PI-PD control, the new PI-PD controller inherits the excellentperformance of typical MPC and the sample of traditional PI-PD control. Thesimulation results show that improved closed-loop performance is obtained comparedwith typical PID control and PI-PD control. The other aspect is that the optimalmethod is the extended non-minimal state space model predictive control. The controlperformance is better than PI-PD method which is optimized by typical MPC.