A Nonlinear Predictive Model Based on Multilayer Perceptron Network

Predictive model is the foundation of model predictive control algorithm. Most of the model predictive control algorithms applied in practice are based on linear predictive model. But a linear predictive model is not suitable to the process which is highly nonlinear and has long time delay. This paper analyzed the advantages and defects of some nonlinear models and proposed a kind of nonlinear predictive model based on multilayer perceptron network through consulting to NARMAX model and making use of the function approximation capability of multilayer perceptron network. Simulation experiment indicated that the nonlinear predictive model proposed in this paper can excellently predict the output information of a nonlinear system.