Abstract: An approach of Self-Gravitational Search Algorithm (SGSA) was proposed for the identification of T-S fuzzy model. In the identification of T-S model, structure parameters and consequent parameters were encoded into a particle of SGSA. Based on new strategies, all the individuals in SGSA dynamically adjusted the distance and force between particles according to the intensity of the swarm, and the gravitational constant was adaptively alerted. As a result, the global search ability of the proposed SGSA was enhanced in the earlier stage of the search process to prevent the algorithm from trapping into local regions, in the mean while; the local search capability was improved in the latter iterations of optimization to reduce the probability of the algorithm fruitlessly searching around the local optima. The simulation results and comparative analysis demonstrate the good performance of the proposed method with high accuracy and great robustness.

Key words: T-S model, system identification, self-adaption, GSA