Parameter Identification of T-S Fuzzy Models Based on Particle Swarm Optimization Algorithms

Most of the T-S fuzzy models commonly used in the identification of nonlinear processes have linear or affine consequents. More specifically, the local mathematical models in the consequents of fuzzy rules are taken to be linear or affine. However, it can always be observed that the number of fuzzy rules of the resultant T-S fuzzy models is very large. In order to reduce the number of fuzzy rules and keep the model accuracy unchanged, a special class of T-S fuzzy models is taken to be the candidate models in this study. In more detail, the consequent of the fuzzy rule in this research is polynomial models instead of linear or affine ones. Based on this candidate T-S fuzzy model, the particle swarm optimization algorithms are employed to estimate the parameters in this model. Numerical simulations demonstrate that the number of fuzzy rules is significantly reduced while the model accuracy is still unchanged. This advantage comes to be more prominent with the increase of input variables.