A NOVEL TUNING STRATEGY FOR MULTIVARIABLE MODEL-PREDICTIVE CONTROL

Model predictive control (MPC) has established itself as the most popu lar form of advanced multivariable control in the chemical process ind ustry. However, the benefits of this technology cannot be realized unl ess the controller can be operated with desirable performance for an e xtended period of time. The objective of this work is to present an ea sy-to-use and reliable tuning strategy that enables the control practi tioner to maintain MPC at peak performance with minimal effort. A nove l analytical expression that computes the move suppression coefficient s. guidelines to select the additional adjustable parameters, and thei r demonstration in an overall tuning strategy are some of the signific ant contributions of this work. The compact form for the analytical ex pression that computes the move suppression coefficients is derived as a function of a first order plus dead time (FOPDT) model approximatio n of the process dynamics. With tuning parameters computed, MPC is the n implemented in the classical fashion using an internal model formula ted from step response coefficients of the actual process. Just as a F OPDT model approximation has proved a valuable tool in tuning rules su ch as Cohen-Coon, ITAE and IAE for PID implementations, the tuning str ategy presented here is significant because it offers an analogous app roach for multivariable MPC. (C) 1998 Published by Elsevier Science Lt d.