Neural Predictive Control Application to a Highly Non-Linear System

Abstract This paper proposes a new approach for constrained multivariable predictive control based on the use of a recurrent neural network as a prediction model. The model is derived in a natural way from its linear counterpart and it is a representation of the system in the state space form. The proposed predictive control scheme is able to deal with constraints in the system variables and highly non-linear dynamics. An example belonging to the sugar industry, the control of a water sulfitation plant, is presented together with a comparison with the classical PID control.