On-line Optimal Regulator for Multi-Variable Non-Linear Systems by Dynamic Programming

Abstract

Although the optimal regulator is widely known as a typical optimal-feedback controller, the optimal problem for multi-variable non-linearsystems is hard to solve in real-time. In this paper, a new method for solving the optimal regulator is proposed. Optimal values of manipulation variable are calculated by evaluating an objective function on state-variable space which is iteratively solved by a dynamic programming. As a result of simulation for level-control of a tank and temperature-control of a CSTR, the proposed method provides real-time calculation of the optimal problem, and a good controller can be realized. Consecutive calculation of optimal regulator realizes tracking controller of a batch process with quasi-steady state condition. Simulation of reflux-ratio control of a batch distillation shows good control performance.