Nonlinear System Identification Based On T-S Fuzzy Model

Nonlinear systems exist in the objective world widely,so it is important to study the theory of nonlinear systems. The other hand, the modeling for a system is the base and premise of a systematic analysis, design, forecasting, control and decision-making , so it has very important significance to investigate the identi- fication of nonlinear systems.T-S fuzzy model is a mathematical model which indicates features of the nonlinear system, using this method can characterize a complex nonlinear system into simple linear subsystems, Thus it is possible to use traditional linear control theory to control and analyze stability for nonlinear systems.Nonlinear system identification that based on T-S fuzzy model include struc- ture identification and parameter identification. Fuzzy clustering and its improve- ment is used to identify the premise structure and parameters of rules,and the co- nsequent parameters of rules are calculated by the least squares method and its improved methods .This paper starts to discuss and research closely around nonlinear system identification that based the T-S fuzzy model.For identification of T-S fuzzy model,To determine the number of fuzzy rul- es, namely, the number of fuzzy clustering is one of the most important issue of creating fuzzy classification subsystem. Finding the optimal number of clusters is a problem that is in the clustering validity areas, compared to traditional methods of clustering analysis, this paper summarizes a simple and convenient method of clustering analysis-biobjective clustering,which can determine the optimal num- ber of clusters conveniently,and optimize the structure of fuzzy system so that to achieve nonlinear system identification.The other hand, taking into account good accuracy and interpretability of T-S fuzzy model, multiobjective genetic algorithm NSGA2 is used to optimize T-S fuzzy model.It uses systematic mean square error and the number of fuzzy rules as the objective functions,and uses the real-coded genetic methods.Using the method can get a set of optimal solution that mean less rules and smaller error.Finally based on balanced evaluation,We choose appropriate solution to get T-S fuzzy model with good accuracy and interpretability so that to approximate nonlinear systems .