Adaptive Model Predictive Control with Successive Linearization for Distillate Composition Control in Batch Distillation

This paper investigates the application of adaptive model predictive control (MPC) with successive linearization for the control of top product purity of a batch distillation column. Adaptive MPC with successive linearization can overcome the prediction inaccuracies associated with linearization of highly non-linear and dynamic mathematical model of a batch distillation column, with a lower computational load than nonlinear MPC. A binary mixture of methanol and water was selected to demonstrate the controller development, and its performance was investigated by varying MPC tuning parameters in the MATLAB/Simulink simulation environment. Results indicated that the choice of tuning parameters had a considerable influence on the MPC's ability to track a constant set-point for the output. With the correct choice of tuning parameters, however, it is possible for the controller to track a constant set-point. The present approach is compared with nonlinear MPC in order to gain a quantitative understanding on accuracy and computational effort.