Study On Nonlinear Predictive Control Based On Neural Network

Along with advancement of industrial control demand, development of control theory and computer technology, a predictive control algorithm is produced with effective control and strong robustness, It is applicable to complex industrial processes and the control system that is not easily to establish the accurate mathematics model, and is successful applied in petroleum, chemical industry, metallurgy and mechanism, and have a good prospect in application.Combined with nonlinear system in the predictive control engineering practices, the research is developed mainly in theory study and simulation, the idea that a DMC algorithm based on neural networks identification is proposed. Actually it uses the neural networks as the identified model of control plant to produce predictive signal, the control law is solved by optimized algorithm. Accordingly the predictive control of the nonlinear system is realized. BP and RBF neural networks are chosen as identified model for they can approach the function very well. Neural networks theory and DMC theory are used to solve the problem that the nonlinear system is difficult to control. for the nonlinear predictive control based on neural networks, a theory scheme is put forward which is efficient and feasible. So a new thought and method is put forward to resolve practical system control question.During the course of the research, the following achievements are obtained:Firstly, the theory of dynamic matrix predictive control (DMC) is expounded, then in this paper ,it makes a detailed analysis including its predictive model, its methods of revising feedback and receding horizon optimization, and its stability & robustness, it analyzes how the correlative parameters influence on controlling effect. Simulation shows DMC has well dynamic performance, strong tracing ability, and perfect control performance.Secondly, for the characteristicof BP and RBF network, the improved method for BP and RBF are proposed: self-adaption learning rate BP method with momentum nap and recursive k-means clustering method.Simulations show the self-adaption learning rate BP method with momentum nap has fast learning rate and good identification precision; recursive k-means clustering method improves on the modeling ability of RBF and tracing performance.Thirdly, the DMC algorithms based on the improved BP and RBF networks areproposed. MATLAB simulations show the validity and feasibility of the proposed new algorithms. The DMC algorithms based on the improved BP and RBF both obtain the due control effect.But DMC algorithms based on the improved RBF has better control performance than the algorithms based on the improved BP, and it benefits its application to real-time system.