MODEL-PREDICTIVE CONTROL-BASED ON NEURAL IDENTIFICATION METHOD

This paper Presents a new approach to model predictive control (denoted as MPC) based on a new identifier utilizing the artificial neural network system (denoted as ANNS) with tapped delay lines added to the input layer of the ANNS. The identifier uses the back-propagation method in order to minimize the errors between the output of the ANNS and the output of the system to be controlled. The numerical simulation studies show that the neural identifiers have robustness in the change of operating conditions and circumstances and, hence, that the control performances of the MPC are quite satisfactory.