Non-Linear Model Predictive Control: A Personal Retrospective†

VOLUME 85, AUGUST 2007 INTRODUCTION Model predictive control (MPC) has been the most successful advanced control technique applied in the process industries. The formulation naturally handles time-delays, multivariable interactions and constraints. Particularly in the petrochemical industry, MPC has often been tuned for robustness rather than a high level of dynamic performance. In addition to conservative tuning, performance has been limited by the use of linear models and the standard “additive output disturbance” assumption to compensate for plant-model mismatch. There is a wealth of articles on MPC in general, with many different formulations and applications of non-linear MPC in the literature; my goal is not to review these articles. Rather, my goal is to provide a personal perspective of NMPC. My primary audience is graduate students and others just beginning their foray into model predictive control research and application. The MPC literature can be quite challenging to read and it is not always clear why certain implementation decisions are made. Rarely are the particular pitfalls fully presented. One of my goals Non-Linear Model Predictive Control: A Personal Retrospective†