Multivariable feedforward control design using disturbance cost maps and a genetic algorithm

Abstract

The addition of feedforward action to a regulatory process control system is meant to reduce the effect of a measurable disturbance on the process outputs, thus reducing the amount of feedback action required. Disturbance cost maps can be used to quantify how well a given feedforward control design achieved this. In this paper, a genetic algorithm is used to optimize objective functions derived in terms of the feedforward-modified disturbance cost. Since the genetic algorithm operates on a population of designs, statistical hypothesis testing permits the critical feedforward control parameters to be identified and their optimal values computed. The use of the proposed method is illustrated on a typical multivariable control design problem.