Mode Control Based On Support Vector Machines Non-linear System

Statistical Learning Theory (SLT) is a novel approach and research filed in small scale data machine learning, which is based on Structure Risk Minimization and VC dimension. Based on the SLT, Support Vector Machine (SVM) has been introduced as an effective currency learning machine method. By solving convex optimization problems to do pattern recognition, classification and regression problems, SVM avoids the problems, such as dimension trouble, over learning and has stronger generalization characteristic than neural networks. Now, SLT and SVM have become new hotpots in the area of machine learning.The theory of SVM and a new method of nonlinear system identification by using support vector machine regression algorithm have been studied in this paper. Three identification methods are given in this part. Compared to the identification results of LS-SVM, RBF andε-SVM,ε-SVM can be used in function approximation effectively just by using the input and output data of nonlinear system and its generalization is more valuable, which proves that the method of SVM used in internal model control of nonlinear system is powerful.The principle and characteristic of the tradition IMC have been researched. A new method of one kind of nonlinear system IMC has been presented because of the excellent regressive ability of SVM. The algorithm of SVM is used to build the model and inverse model. Then, IMC is designed in the end. Simulation results demonstrate that the system designed has good set-point tracking characteristic, disturbance rejection characteristic and robustness. A new mind of nonlinear system control and research is supplied.