Neural network based nonlinear model predictive control for an intensified continuous reactor

Abstract

Intensified continuous reactor is designed to replace conventional batch reactor for three-phase catalytic slurry hydrogenation. The control of intensified continuous reactor imposes difficulties due to complex three-phase reaction kinetics, inherent process nonlinearities and stringent temperature limitation. In this work, two nonlinear model predictive control (NMPC) algorithms based on neural networks are proposed and implemented for the intensified continuous reactor. One is the NMPC with nonlinear optimization that is solved by sequential quadratic programming (SQP). The other is the NMPC with local linearization that is applied at each sample instant, a linear model with parameter varying is obtained and the optimization in NMPC can be written as a quadratic programming (QP) problem like linear MPC. Simulations results show that the NMPC with local linearization presents satisfactory performance.