Uma contribuição ao estudo de controladores robustos

Abstract

The model based predictive controller (MPC) has been successfully applied in industry, with particular emphasis on the petrochemical industry. The basic feature of MPC algorithms is based on the formulation of an optimization problem, for calculating the sequence of control moves that minimize a performance function in a time horizon with the best information available at every moment, and submitted to restrictions, operation and the plant model. The presence of uncertainties in the model can result in loss of performance of the model based control system or even instability in the closed loop. The class of robust predictive controllers (RMPC) is based on the explicit consideration of uncertainties. This research presents two RMPC algorithms. One of them adds a set of restrictions with robust stability guarantee (RMPC-MMR) to the problem of robust control, with uncertainty in time variant as min-max optimization. And the other develops RMPC using Linear Matrix Inequalities (LMI) and uncertainty as a polytope (LMI-RMPCR). The algorithms are evaluated through computer simulations, and are applied to a range of industrial process typical systems. The responses are compared with the responses of traditional controllers. The classical algorithms IHMPC (Infinite Horizon Model Predictive Control) and MPC-RS (Model Predictive Control with Reference System) are also assessed experimentally by the application in a pilot plant for controlling level and pH, built in the Experimental Unit of the Chemical Engineering College, at Federal University of Uberlândia. The results of RMPC simulation were evaluated with the simulation and experimental responses of the classical controllers implemented in the plant. The results indicate the applicability and limitations of the proposed controllers.