Synthesis of model predictive controller for an identified model of MIMO process

Abstract

Model Predictive Controller (MPC) technology has been researched and developed to meet varied demands of need to control industrial power plants and petroleum refineries. This development has paved the way for the MPC technology too many other fields like automotive, aerospace, food processing industries in this paper, primary importance has been paid to the development of a MPC for an identified model of Multiple Input and Multiple Output process. In this paper, a Four Tank System has been considered for generation of input-output data. This data i.e. generated input output data is used for the estimation of two polynomial model, name ARX model (Autoregressive exogenous) model and OE (Output Error) model. With each of model output generated, the Fit-Rates of models are compared to find out most efficient model. The model equations are now considered as plant for developing a Model Predictive Controller (MPC). Two sets of results are obtained after the development of MPC and tested. One is without noise and one is with noise. Both sets of results were a success as the output signals traces step input signals after some steady oscillations in real time with in a very short period of time which indicated a good response time. The MPC developed can be applied to any polynomial model with a good Fit-Rate, it predicts and control the process variables automatically.