A nonlinear predictive model based on BP neural network

MPCs have been widely applied in industrial process control field because of the excellent control effect. The classic MPCs, which are all based on linear predictive models, are unfit for the strong-nonlinearity control systems. In these cases, NMPCs must be constructed if a model predictive controller wants to be used. Nonlinear predictive model is the foundation of NMPC, and should be established firstly. This paper proposed a one-step nonlinear predictive model based on BP neural network by combining NARMAX model and neural network, and supplied a calculation method of the hidden-layer-neuron number of the two-layer BP neural network used in the one-step predictive model.