Partial predictors for the supervision of a multimodel direct generalized predictive control of highly non stationary systems

The multimodel approach is recently developed in order to resolve the problems of the increasing complexity of many industrial processes. In this paper we propose a multi- model generalized predictive control based on a commutation algorithm. This commutation is controlled by partial predictors associated to the local controllers. These predictors constitute the supervisor. The control law is established to ensure the desired performances in closed loop in the presence of a highly non stationary system. The obtained results are very satisfactory and show a very good precision relatively to the case in which the classical direct adaptive generalized predictive control is adopted.