Predictive Learning Control for a Batch Polymerization Reactor

Abstract

Absract A new control algorithm which we call predictive learning control algorithm (PLCA) is proposed for the control of repetitive unit operations. The algorithm uses information from previous operations to correct the predicted future errors in the current operation. By applying the predictive learning controller, we can improve the learning speed with less amount of information than the other learning algorithms require. From the study of convergence condition, we prove that the algorithm can asymptotically converge under certain conditions. The proposed control scheme has been applied to a Styrene-Acrylonitrile copolymer reactor control.