MODEL-PREDICTIVE CONTROL OF A PVC BATCH REACTOR

This paper discusses a method to improve the quality of temperature co ntrol of the polyvinyl chloride (PVC) suspension batch reactor by impl ementing two Nonlinear Model Predictive Controllers (NMPCs) for this e xponentially unstable process with very nonlinear behavior. The first method is based on recomputing the step response matrix at each sampli ng time with a double model based prediction and solving twice the opt imization problem using the linear model of the process described by t he step response matrix. The second method uses the rigorous model for both prediction and optimization. These methods were tested for diffe rent disturbances, and their performances were compared with those obt ained with proportional-integral-derivative (PID) control. Significant improvement of temperature control was achieved using NMPC. The proce ss model, used in both NMPC methods, is presented. (C) 1997 Elsevier S cience Ltd.