Research On Direct Generalized Predictive Control Based On RBF Neural Network And Its Convergence

With the successful application in the process industry, Generalized Predictive Control(GPC) is one of the most promising control methods in academia and in engineering. It has received considerable attention in the last decade. Today, it is universal phenomenon for multivariable nonlinear system or nonlinear system with large delay time during many complex industrial processes, so it is imperative to extend the GPC algorithm to these processes. Recent years, people have assumed large amount of research in this respect, among them, because of the wide applicability, the method which leads neural network into GPC has got thorough research.Basing on algorithm and the international current status of the application of neural network in GPC, Radial Basis Function(RBF) neural network and its applications in GPC are researched in this paper. First, for the complicated online calculation and not suitable for nonlinear system of GPC, direct GPC based on RBF neural network method is presented, and its convergence is analyzed thoroughly using mathematic method, this proves that the method is stable and can made the function index converge to the best value. Second, obtaining from linear SISO system, the method is generalized to nonlinear SISO system, linear MIMO system and nonlinear MIMO system combining with median theorem and spine function. The stability and convergence of the method in the three systems are proved. Third, using matlab to simulate the four kinds of systems, and analyzes the simulation results, it demonstrates the effectiveness of this method.