Multivariable neural network predictive control based on variable structure objective optimization

A special single neural network predictive control algorithm aimed at practical application of specific large time-delay nonlinear system is presented. This algorithm is based on a variable structure objective optimization controller to decouple the multiple control variables of the system. Consequently, the multiple single neural network predictive controller with a multi-step-ahead differential predictive cost objective function can realize the multivariable control, and guarantee the stability of system state in switching. The simulations and applications show excellent effect and great performance in practical application of uncertain large time-delay nonlinear system.