Chance Constrained Model Predictive Control Based on Set-Valued Optimal Algorithm

Abstract

This paper adopts one kind of classical predictive control, namely model algorithmic control, to control the distillation process. Due to finite cognition degree for the distillation system and existence noise in the chemical industry, we present a chance constrained model predictive control algorithm to eliminate the effect of parameter pertubation and noise. In view of the linear impulse response model, we introduce set-valued optimal algorithm and orthogonal standardization method to transform the chance constrained model predictive control to a general model predictive control problem. For the determinate MPC, efficient quadratic programming exsits to solve such problem. Applying such controller to a distillation system, output constraint condition will be satisfied with a predefined probability.