Aiming at the problem of weighted fuzzy clustering algorithm (WFCM) that the convergence speed is slow and sensitive to the initial value in transformer DGA analysis, a transformer fault diagnostic method based on weighted fuzzy clustering algorithm optimized by improved artificial fish swarm algorithm (SAAFSA-WFCM) is proposed. This method uses the artificial fish swarm algorithm (AFSA) improved by the probabilistic kick search mechanism of the simulated annealing (SA) to obtains the best initial clustering center, it takes advantage of the global optimization of AFSA and avoids getting into the local maximum through local optimization by the use of SA. By using the obtained best initial clustering center as the initiatory value, WFCM algorithm finds the final cluster center which is getting closer to the actual location through iterative computation, overcoming the shortcomings of traditional WFCM algorithm which is sensitive to the initial value, and accelerating the convergence speed. Simulation and case analysis show that this method has higher accuracy and efficiency when applied to power transformer faults diagnosis.
