Research on Neural Network Predictive Control of Induction Motor Servo System for Robot

Abstract

Neural network control has some applications in many areas; the neural network has strong capability of self-learning, adaptive and fault tolerance, and predictive control for complex systems that has strong adaptability. Through combining the approximation ability of neural network for nonlinear objects and optimization strategy of predictive control, predictive control scheme based on BP neural network has been proposed. In this paper, predictive control algorithm design idea based on BP neural network is proposed: Firstly, by the use of BP neural network model predictive control, the controlled object prediction model can be established, then by taking advantage of the prediction model, based on the input and output information of the current system and the future output values of predict objects, by the use of feedback correction, so as to overcome the model prediction error due to other uncertain disturbance in the system, more accurate predictive value of the object can be obtained. On this basis, based on the future corrected predicted value of the object, with given system output values, the control variable can be scrolling optimized to obtain future system control sequence according to the defined quadratic performance standard. The predictive control has achieved good control effect based on BP neural network; it has proved the feasibility and superiority of this control scheme.