A PRACTICAL APPROACH TO PREDICTIVE CONTROL FOR NONLINEAR PROCESSES

This paper presents a Model Predictive Control (MPC) Technique for Nonlinear Systems. The algorithm is based on a new interpretation of the procedure used by traditional linear predictive controllers, like Dynamic Matrix Control (DMC) and Generalized Predictive Controller (GPC), to compute the predictions. The proposed technique requires no specific model structure, treating equally models based on neural networks, difference equations or ordinary difference equation systems. Comparisons to other techniques are presented in the paper applying the algorithm to nonlinear systems available in literature. The main advantages of the proposed approach are: (i) it makes no use of iterative algorithms, (ii) the control action can be obtained by the same methods used in linear MPCs and (iii) any type of nonlinear model can be used.