Model Predictive Control Design of Pressure Swing Distillation for Methanol-C5 Separation

Abstract

The azeotropic mixture that is greatly influenced by pressure is usually separated by
pressure-swing distillation (PSD). This paper investigates the control selection of the
PSD system. The PSD scheme was developed by Luyben. Model predictive control (MPC)
is used as an alternative control method because it had a better response compared to the
proportional-integral (PI) controller. MPC utilizes a dynamic model to make a control
action toward the process. Generating the model is a crucial step for the MPC to work
accurately. In this work, the model representing the process was approached by the first
order plus dead time (FOPDT) model. The steady-state and dynamic simulations were
carried out with HYSYS V11. The fine-tuning method was used to find the optimum tuning
parameter of MPC meanwhile, the PI controller was tuned by auto-tuner. The results
show that MPC was successfully implemented to the process and the control performance
has increased by 21-52% in the set-point tracking test and 5-42% in disturbance rejection
test compared to the PI controller.