Abstract: Focused on the structure and parameter identificantions in TS fuzzy system modelling, a hybrid clustering method SOMK, which combines the SelfOrganizing Map (SOM) and Kmeans approaches, is used to cluster the input samples, and the DaviesBouldin (DB) validity criterion is used to determine the optimal number of clusters, which is taken as the optimal number of rules of the TS fuzzy system. Then the premise parameters are determined by the number of rules and the cluster centers. Furthermore, a least square algorithm is applied to obtain the consequent part of each rule. A case simulation is used to illustrate the proposed medolling method and the algorithm. Simulation results prove that the method has high modelling precision and high ability of power generalization, and it is valuable for practical application.

Key words: system engineering, SOM algorithm, K-means algorithm, FCM algorithm, T-S fuzzy system