| Abstract: | The particle swarm optimization (PSO) algorithm is widely used in identifying Takagi-Sugeno (T-S) fuzzy system models. However, PSO suffers from premature convergence and is easily trapped into local optima, which affects the accuracy of T-S model identification. An immune coevolution particle swarm optimization with multi-strategy (ICPSO-MS) is proposed for modeling T-S fuzzy systems. The proposed ICPSO-MS consists of one elite subswarm and several normal subswarms. Each normal subswarm adopts a different strategy for adjusting the acceleration coefficients. A Cauchy learning operator is used to accelerate the convergence of the normal subswarm. During the iteration step, the best individual in each normal subswarm is added to the elite subswarm. Using adaptive hyper-mutation, the immune clonal selection operator is used to optimize the elite subswarm while the individuals in the elite subswarm migrate to the normal subswarms. This shared migration mechanism allows full exchange of information and coevolution. The performance of the proposed algorithm is evaluated on a suite of numerical optimization functions. The results show good performance of ICPSO-MS in solving numerical problems when compared with other recent variants of PSO. The performance of ICPSO-MS is further evaluated when identifying the T-S model, with simulation results on several typical nonlinear systems showing that the proposed method generates a good T-S fuzzy model with high accuracy and strong generalizability. |
