A simplified predictive control approach for handling constraints through linear programming

Abstract

There has been a considerable interest in predictive control techniques because of their suitability for process control. Generally a constrained optimization problem needs to be solved at every sampling instant to accommodate the constraints on process variables. The on-line computational requirements of the Linear Programming (LP) formulation, resulting from the application of the general predictive control approach, have been of concern since these restrict the use of the predictive techniques for many applications. A simplified predictive control approach is presented that allows systematic handling of constraints through LP without requiring large computational effort. The performance and robustness of the proposed approach are compared with the general approach on process models exhibiting interactions, time-delays, and inverse response, through computer simulations. The results indicate that significant savings in the on-line computational effort can be realized for many applications without compromising the control performance or the useful features of the predictive control approach.