INDUSTRIAL APPLICATIONS OF MODEL-BASED PREDICTIVE CONTROL

In this paper we describe two classical applications of MBPC which enhance the advantages of the method: feed-forwarding, constraints handling, no-lag error on dynamic set points, easy trade-off between robustness and dynamics specifications. We insist more on the project procedure than on the control algorithms which are presumed to be known from references. Practical difficulties appear to come more from the selection of the model and from the formulation of the specifications than from the strict controller design. To be efficient with a good economic pay-back, MBPC should be used at the optimization level, embedded in an appropriate environment. These requirements nowadays, due to efficient control packages, demand more time and effort than the control algorithm per se.