Nonlinear Model Predictive Control Based on Multi-linear Models of an Unstable Chemical Reactor

Abstract

It is well known that operating a process under unstable conditions is a challenging control problem. In this article, by use of basic concepts of the weighted space distance, a set of locally linearized models is simply and effectively combined into a global description of a nonlinear plant. To reduce the computational load, these multiple linear models are then used as prediction equations in an MPC framework. Simultaneously, some parameter tuning strategies are presented to guarantee low overshoot and good robustness for the predictive control system. The effectiveness of the proposed multiple linear model predictive control with state estimation is demonstrated through its application to the exothermic chemical reactor, which is a typical nonlinear unstable process.