An improved decoupling structure based state space MPC design with improved performance

Abstract

In this paper, an improved decoupling structure is first proposed for multivariable processes, where the resulted decoupler can avoid zero-pole cancelation and simultaneously guarantee feasibility by tuning of additional parameters. A subsequent single-input–single-output design of model predictive control (MPC) is then proposed using a new state space realization of the independent single loops, where we adopt a new extended non-minimal state space structure to enhance control performance. Through this new design model, dynamics of processes’ state variables can be directly taken into account. The decoupler property, closed-loop control performance, the relationship with traditional state space MPC and robust stability issues are fully analyzed. The efficiency of the proposed is evaluated through a typical process in the recent literature, in comparison with a typical non-minimal state space MPC recently developed.