Predictive Control: Some Propositions for the Design Methodology

Abstract

This paper deals with predictive controller tuning. The study is made through some simulations of second order linear or non linear processes, corresponding to industrial furnaces models. The concept and properties of predictive controller are quite attractive. This results in its widespread use in the industry. However, many parameters have to be tuned for achieving efficient control laws. Then, the goal of the present work is to derive nearly automatic design strategies for Model Based Predictive Controllers (MPC). In a first time, relations between the main design parameters and some closed-loop performances have been underlined. In a second time, new design parameters have been introduced : they improve the results when the reference signal is known in advance. In a third time, some propositions have been made for a semi-automatic tuning methodology and for a tuning method based on expert rules.