Research On Intelligent Identification And Control Of Chaotic System

Chaos is a complex dynamic motion behavior, which exists in the nature widely. In recent years, chaos and other sciences permeate each other. It has been widely used in physics, biology, mathematics, electronics, psychology, information science, meteorology, economics, astronomy, even in the field of music and art. In order to utilize chaos, or eliminate its bad effects, modeling the chaos using system identification technique and perform appropriate control effort is an important problem. Recent years, intelligent control theory, including neural network and fuzzy systems, develop deeply. In this thesis, the applications of intelligent algorithm in the identification and control of chaotic systems are studied, the specific research work as follows:Firstly, a chaotic system identification method is proposed based on the interval type-II fuzzy system. The fuzzy space is divided by grid diagonal method and the type-II symmetric triangular membership function is used as the primary membership function. Under the assumption that the former parameters remain unchanged, a forgetting factor recursive least square(FFRLS) method is applied to update the consequent parameters. In order to solve the problem of noise pollution in the sampled data, the sigmoid data transform of the sampled data is carried out, and he parameters in sigmoid function and the width of type-II symmetric triangular membership function are optimized by PSO, for the purpose of avoiding to adjust the membership function and improving the accurate of identification. Simulation on Mackey-Glass chaotic system verifies the effectiveness of the proposed approach.Secondly, a Wiener-Least Square Support Vector Machine(Wiener-LSSVM) model is proposed for identifying a chaotic system,which utilizes partial structure information of chaotic system. Wiener-LSSVM model consists of linear dynamic part followed by a LSSVM. The identification of Wiener-LSSVM model is solved in a LS framework, in which the parameters of linear dynamic part and those of LSSVM are estimated simultaneous.Then, a new method is proposed for identifying a chaotic system based on a Hammerstein-Extreme Learning Machine(Hammerstin-ELM) model. Hammerstein-ELM model consists of an ELM neural network coupled with a linear dynamic part. The parameters of linear dynamic part and those of ELM neural network can be simultaneously estimated by the presented generalized ELM algorithm. The generalized ELM algorithm finds the solutions through pseudo-inverse of matrix instead of gradient descent to improve the accuracy of identification.Finally, two kinds of control and synchronization algorithms for Hénon chaotic system are proposed, based on fuzzy theory. In the first approach, the T-S model is used to identify the Hénon chaotic system, and the local dynamic linear model is obtained by the Hénon chaotic system. Based on this model, a generalized predictive control algorithm is designed to realize the tracking and synchronization control of Hénon chaotic system. In the second method, the fuzzy inverse model of Hénon chaotic system is established by using the fuzzy inverse method, and the adaptive inverse control and synchronization algorithm for Hénon chaotic system is designed based on the fuzzy inverse model. Simulation examples validate the effectiveness of the proposed method.