Neural Wiener Based Model Predictive Control (NWMPC) for MTBE Catalytic Distillation Using Reduced Sequential Quadratic Programming (RSQP)

Abstract

Reactive distillation is a process that combines both reactor and distillation column in one unit process. The reactive distillation is normally applied in MTBE production in order to achieve high reaction conversion and purity of the MTBE. Controlling such reactive distillation is a challenging task due to its highly nonlinear behavior and the existence of strong interactions among control variables. In this work, a Neural Wiener based model predictive control (NWMPC) is designed and implemented to control the tray temperature of MTBE reactive distillation. The Reduced SQP (RSQP) has been embedded as an optimizer in the NWMPC proposed. The MTBE reactive distillation has been modeled using aspen dynamic and the control study has been simulated using Simulink (Matlab) which is integrated with Aspen dynamic model. The results achieved show that the NWMPC is able to maintain tray temperatures at desired set points, able to reject the disturbance and robust toward robustness test conducted.