An experimental study of model predictive control based on artificial neural networks

Abstract

Practical implementations of two typical types of artificial neural networks (ANNs), feedforward networks and external recurrent networks, as the model for model predictive control (MPC) were performed on the dual temperature control problem of two distillation columns, a pilot scale i-butane and n-butane distillation column and a bench scale ethanol and water column. The superiority of MPC based on an ANN models over conventional proportional-integral controllers and over dynamic matrix control were testified through experiments.