Adaptive Critic Designs for Optimal Event-Driven Control of a CSTR System

Abstract

This article presents an optimal event-driven control scheme for a continuous stirred tank reactor (CSTR) system. The CSTR system differs from most of studied plants in that its equilibrium point is nonzero. In order to obtain the optimal event-driven control without coordinate transformations, we first introduce a discounted cost for such a system. Then, under the framework of adaptive critic designs, we use a single critic network to solve the Hamilton–Jacobi–Bellman equation related to the discounted-cost optimal event-driven control problem. To update the critic network weights, we employ the gradient descent method together with the experience replay technique. An advantage of the experience replay technique is that it brings about an easy-checked persistency of excitation-like condition. The stability analysis of all the signals in the closed-loop system is conducted via the Lyapunov approach. Finally, we validate the present optimal event-driven control strategy through simulations of the CSTR system.