Abstract:To overcome the problem of lower control precision caused by parameters varying of the controlled object, the paper proposed a sort of predictive control algorithm based on BP neural network model. In the paper, it applies the predictive parameter of PID controller based on BP neural network on line to control the controlled object, and the system model parameter was on line predicted by means of least recursive squares algorithm. The algorithm would be based on model prediction. It first validats its control effect in the linear system, and then the non-linear problem would be treated as the linearity. The non-linear system would be controlled by use of predictive control algorithm based on BP neural network model. The simulation curves showes that it could achieve high control precision in the linear system to PID controller of BP neural network, and own the ability of adaptation and approaching arbitrary function. The simulation researches show that it is stronger in adaptation, better in stability, and higher in control precision compared with the traditional BP neural network PID controller.