Análise de desempenho de controladores preditivos com estratégia Hot Vapor Bypass em um processo de Destilação com Oscilação de Pressão (PSD)

Abstract

Pressure-swing distillation (PSD) is a promising alternative for the separation of azeotropic mixtures, such as acetonitrile–water, but its operation involves significant challenges due to strong variable coupling, nonlinearities, and sensitivity to disturbances. In this context, this work evaluated the performance of different predictive control strategies in a PSD plant with full heat integration and the use of the hot vapor bypass (HVP) strategy, applied to assist pressure control and improve stability and energy efficiency. Concepts related to azeotrope formation, separation strategies, pressure control alternatives, and the characteristics of conventional and predictive controllers, particularly the Extended Infinite Horizon Model Predictive Control (EIHMPC) and its variation with implicit feedforward action (EIHMPCFF), were explored in the theoretical framework. The process was simulated using Aspen Plus™ and Aspen Dynamics™, followed by dynamic model identification through open-loop disturbances and the development of multivariable models implemented in MATLAB®, where PID, LMPC, EIHMPC, and EIHMPCFF controllers were tested. Comparative analyses considered scenarios of disturbances in the feed to the low-pressure column and setpoint changes, evaluating response time, and disturbance rejection. The results showed that LMPC achieved better performance than PID, although it still presented steady-state deviations; EIHMPC corrected these limitations, ensuring greater stability and accuracy; and EIHMPCFF stood out by anticipating disturbance effects, providing faster responses with lower accumulated error. Therefore, it is concluded that EIHMPCFF was the best-performing controller, confirming the relevance of advanced predictive strategies for complex industrial processes.