| As we know, the T-S fuzzy models have drawn particular attention in the area of nonlinear modeling, due to their simple structures as well as powerful approximation capabilities of nonlinear functions. They have been studied and widely applied in system identification. A lot of research work on the identification of the structures and parameters of the T-S fuzzy models has been done. In this thesis, we explore the identification of the high-order T-S fuzzy models using the particle swarm optimization algorithm.Firstly, a novel approach for the T-S fuzzy model identification is proposed in this thesis. In more details, the consequent of the fuzzy rules of our T-S fuzzy systems is set to be a polynomial model instead of linear ones. The particle swarm optimization algorithms are employed to estimate the parameters of this model. A few time series benchmark datasets are used to evaluate the proposed method. Numerical simulations show that the number of fuzzy rules in the T-S fuzzy models can be significantly reduced, and the identification procedure is also accelerated.It is rather difficult to model those nonlinear systems, which are corrupted with noise. Secondly, the polynomial T-S fuzzy models are applied to attack the nonlinear noise cancellation problem. They can cancel the disturbing noise by approximating the unknown noise transfer functions without any prior knowledge of the noise characteristics. Simulation results demonstrate the effectiveness of the proposed algorithm.Thirdly, the parameters of the premise fuzzy membership functions in the T-S fuzzy models are optimized together with their consequent parameters based on the particle swarm optimization method in this thesis. Computer simulations illustrate that the performance of these T-S fuzzy models is drastically improved. Moreover, the simplification of the fuzzy rule base is considered here. The optimal number of fuzzy rules can be obtained according to the firing strength of each rule. |
PDF Full Text Request