Design of an Auto-Tuning PID Controller by a Generalized Predictive Control Method

In this work, an auto-tuning PID controller based on the generalized predictive control (GPC) method is proposed for on-line tuning of the controller parameters due to the changes in operating conditions. The proposed scheme is applied to the control of a distillation column, with identification of the parameters of an ARMAX model determined from the operating data of an industrial column. An evaluation of the maximum likelihood (ML) and Kalman filter (KF) methods of identification showed that at a 10-second sample interval, reliable model parameters could be predicted using either the ML or KF method. However, at a 60-second sample interval simulation results showed that only the KF method generated reliable parameters. Column control was studied for an objective of maintaining constant liquid temperature on an intermediate tray, in the presence of a disturbance in the feed flow rate, with the top product flow rate being a manipulated variable. The simulation results for the auto-tuning PID control scheme using the GPC formulation are compared to the control performance which was obtained using a tuned conventional PID controller and also GPC, at a 60-second sample interval.