Nonlinear Systems Identification Based On Clustering Analysis And Least Squares Support Vector Machine

Most of the controlled objects have the characteristics of obvious nonlinear and uncertainty inindustrial production process of the current, and modeling is the base of the control. The key of modeling issystem identification, so it is critical to study the nonlinear system identification. The traditional systemidentification methods have good effects on the identification for the linear system, but the effects of theseidentification methods applied for the nonlinear system identification are not obvious. The identificationmethods based on neural network and support vector machine have been got sum application, but there aredisadvantages in the neural network, such as overfit-ting and local extremum etc. Although theidentification method based on the support vector machine (SVM) can overcome such disadvantages, it isnot significant to improve the system identification. On the basis of studying SVM, this paper proposed thenonlinear system identification method based on the cluster analysis and least squares support vectormachine. The main work is as follows:Firstly, several representative of clustering analysis algorithm is researched, then the basic idea and theconcrete steps of the algorithm of each cluster analysis algorithm are given. The most important algorithmis the fuzzy C-means algorithm. The advantages and disadvantages of various methods are comparedthrough the simulation experiment, and then to chose the best clustering method establishes the theoreticalbasis for the following chapters.Secondly, the inverse model system is identified by using the squares vector machine algorithm andthe least squares support vector machine, and simulates the experiment on the MATLAB application, thenanalysis of the results. The simulation results show that the identification method based on the LSSVM hasfaster calculation speed, better generalization ability and high precision of the identification than SVMmethod.Finally, the T-S fuzzy model identification is intensive studied.The data objects are clustered by usingthe clustering analysis algorithm, and the input space is divided, and the T-S fuzzy model front structure isdetermined. Finally, the consequent parameters of the T-S fuzzy model are estimated by using the riskminimization rule of LSSVM, and then the simulation results show that the algorithm can effectively solve the generalization ability,computation and robust.