A Comparison of Multivariable Long Range Predictive Control with GMV Control in a Highly Nonlinear Enviroment

In the past few years many contributions to multivariable parameter adaptive control have been proposed, particularly in the areas of generalized minimum variance, pole placement, state feedback and LQG methodologies. In an attempt to overcome some of the problems found in their practical application, renewed attention has been focused on long-range predictive control laws. This paper examines and compares the performance of generalized minimum variance control with a new generalized predictive control algorithm. The performance of the algorithms has been studied using a comprehensive simulation of a highly nonlinear binary distillation column, where the objective was simultaneous terminal composition control. Both multivariable single rate and multivariable multirate adaptive controller configurations have been investigated.