Convenient T–S fuzzy model with enhanced performance using a novel swarm intelligent fuzzy clustering technique

Abstract

To automatically extract T–S fuzzy models with enhanced performance from data is an interesting and important issue for fuzzy system modeling. In this paper, a novel methodology is proposed for this issue based on a three-step procedure. Firstly, the idea of variable length genotypes is introduced to the artificial bee colony (ABC) algorithm to derive a so-called Variable string length Artificial Bee Colony (VABC) algorithm. The VABC algorithm can be used to solve a kind of optimization problems where the length of the optimal solutions is not known as a priori. Secondly, fuzzy clustering without knowing cluster number as a priori is viewed as such kind of optimization problem. Thus, a novel version of Fuzzy C-Means clustering technique (VABC-FCM), holding powerful global search ability, is proposed based on the VABC algorithm. Use of VABC allows the encoding of variable cluster number. This makes VABC-FCM not require a priori specification of the cluster number. Finally, the proposed VABC-FCM algorithm is used to extract T–S fuzzy model from data. Such VABC-FCM based convenient T–S fuzzy model extraction methodology does not require a specification of rule number as a priori. Some artificial data sets are applied to validate the performance of the convenient T–S fuzzy model. The experimental results show that the proposed convenient T–S fuzzy model has low approximation error and high prediction accuracy with appreciate rule number. Moreover, the convenient T–S fuzzy model is used to model the characteristics of superheated steam temperature in power plant, and the results suggest the powerful performance of the proposed method.