Research On Intelligent Identificaton And Control Of Chaotic System

Recently, research of chaos theory and chaos control methods have become an attractive field of nonlinear system research. As everyone knows, model of chaotic system is the basis of its analysis, design and control, so model identification from observations is worth studying. This paper focuses on intelligent identification of chaotic system and optimal design of chaos control. In this paper, interpretation of model, practicability and robustness of control algorithm are the basic requirements; while intelligentizing identification and systematizing control design are the goals expected to reach. GA, GP (Genetic Programming) and fuzzy-neuron net work, etc artificial intelligence approachs, and OLS (Orthogonal Least Square), H∞synthesis, LMI (Linear Matrix Inequality), etc optimistic algorithms are synthetically used in study of model structure identification, parameter estimation, optimal design and reduction of controller. Some methods on evolution modeling, fuzzy modeling of chaotic system, H∞loop shaping control and fuzzy control are proposed:1. Given a modeling method based on muti-objective optimization GP for difference equation of SISO discrete system. Polynomial NARMAX model is represented in tree structure of GP, and the incorporation of a multi-objective approach has enabled the separate consideration of different objectives related to model complexity, model performance and chaotic dynamic. So the identified model can make good balance among precision, complexity and generalization, and have similar chaotic attractor as original system. Addressed a identification method using GA and coevolution multi-population GP for differential equations of MIMO continuous chaotic system. Benefiting from coevolution, probability of trapping in local minimum is reduced and convergence rate is improved.2. Proposed a linear T-S fuzzy modeling technique which utilizes improved Gath-Geva fuzzy clustering. After clustering, model reduction is implemented using OLS and modified Fischer's interclass separability criteria in order to obtain compact and transparent model, then optimize parameters of reduced fuzzy model by constraint Levenberg-Marquardt method to improve its precision, while preserving its interpretability. Linear T-S fuzzy model is adopted instead of affine T-S fuzzy model because it is more convenient to the former to carry on stability analysis and controller design, and simulation result has proved this too.3. A weight selection approach based on a combination of GA and pseudo...