Predictive Control Of Nonlinear Systems Based On Intelligent Error Compensation

Dynamic Matrix Predictive Control was widely used because of its modeled convenience and robustness, while neural network as the excellent approximation ability also get the favor of many scholars. Based on the understanding of the present situation of Dynamic Matrix Control and neural network, We focusing on the analysis of the dynamic matrix control principle, RBF neural network model function approximation, and combine it with the DMC. The concrete content includes:First, the article introduce the principle, characteristics, usages and current situation of Dynamic matrix predictive control algorithm, Made clear the advantages of areas in need of improvement in predictive control algorithm. Then, through analysis principle of neural networks, the algorithm for RBF neural network is introduced in this paper. Also the application of neural networks in control systems is analyzed. To lay a solid foundation for a combination of neural networks with dynamic matrix predictive control algorithm.Second, we propose a solution to the model predictive control for the case when the practical industrial objects have model mismatch which is caused by unknown disturbance, parameter variation or nonlinearity. A new Dynamic matrix control based on RBF neural network is presented. RBF use a class of past inputs and outputs as the training data to forecast current errors. Then combine the RBF error model with general DMC algorithm, the combination turned out to be very robust.Third, we introduce the principle and form of the backlash nonlinear model, and analyzed its effect on the control system. In view of this, this paper present the solution of using RBF neural network algorithm to compensate the backlash nonlinearity, through specific industry objects by matlab simulation experiments, the simulation results prove the validity of this method, make for the possibility in engineering application.