Dual T-S fuzzy model identification with improved cooperative PSO

In this paper, an approach of identification with dual T-S fuzzy models is presented. The model proposed is based on dual T-S different in structure. The main contribution is that dual T-S fuzzy models can be constructed automatically with linear and nonlinear parts to approximate the optimal structure, and control factors are introduce to determine which T-S fuzzy model play more important role to achieve the optimal structure and parameters. The key problem is to select the control factors reasonably and identify the parameters. Firstly, To cope with the structure problem,an approach of automatically extracting fuzzy rules is exploited to achieve the optimal structure. In the identification, the fuzzy C-mean clustering based on kernel function is utilized to partition the data space and extract a set of fuzzy rules. Then, improved cooperative particle swarm optimization algorithm (ICPSO) is put forward to apply in the optimization of the parameters. The ICPSO is proposed to enhance the search the space and it employs several sub-swarms to search the space and useful information is exchange among them during the iteration process, which make the identification of dual T-S more efficient. The simulation example shows the efficacy of the proposed method.