MODEL PREDICTIVE CONTROL USING NEURAL NETWORKS

Abstract

In the present fast moving industry, an accurate and fast control mechanism has become a critical factor for successful production. To cater these needs, Model based Predictive Conti l is becoming increasingly popular everyday which in turn provides huge advantages over conventional control mechanisms. Almost every industry is based on nonlinear plant which is rather complicated and difficult to model & control. In such scenarios neural networks seem to provide an unmatched solution to such complicated problems. This project report focuses to describe the advantages of using neural network to model nonlinear plants. The efficacy of the neural predictive control with the ability to perform comparably to the nonlinear neural network strategy in both set point tracking and disturbance rejection proves to have less computation expense for the neural predictive control. Neural based MPC has advantages like multivariate control, control over safety constraints and physical constraints without much calculation, optimization of control variable at each sampling instant etc. MPC is being used in refining, petrochemical, pulp& paper, power, and food industries.