AN ADAPTIVE TECHNIQUE FOR ON-LINE STEADY-STATE OPTIMIZING CONTROL

摘要

This study is concerned with the development of an adaptive technique for on-line steady-state optimizing control. The technique uses an approximate on-line control algorithm with adjustable parameters which are adapted to maximize expected system performance. The technique is applicable to the general multivariable nonlinear system.nConvergence conditions for the adaptation technique are derived. Statistical techniques are applied to the problem of predicting the statistics of the system disturbance inputs which are generally nonstationary. Two approximate methods of implementing this technique are developed to ease the computational burden. The factors involved in designing a control system using this technique are discussed from the point of view of trying to maximize an overall system profit function; the function includes the cost of control. A testing procedure is developed for updating on demand; this procedure calls for adaptation only when it is econom-ically desirable. Some of the advantages and limita¬tions of the adaptation technique are demonstrated by an application of the technique to the control of a simulated stirred-tank reactor.