Improved State Estimation in an MPC Algorithm Based on Fuzzy Decision

Abstract

Due to the difficulties arising in state estimation in Model Predictive Control (MPC) algorithms, Kalman filtering and dynamic matrix control (DMC) estimation approaches were combined in the current work. Then a weighting average of both estimated states was passed to the algorithm. To determine the weighting coefficient of the mentioned average, a fuzzy supervisor was designed to control the combined estimation. An industrial process 'heavy oil fractionator' was used for simulation. The results demonstrated the improved performance of the approach particularly in better disturbance rejection capability.