Deep Particle Swarm Optimization on Fault Power Flow Diagnosis

The particle swarm optimization (PSO) method is designed on the principle of bird flock foraging, which can be effective in constrained multi–objective optimization problems. In the classifier design, the applications of PSO make the results obtained from the training process with higher confidence. In this study, a deep PSO (DPSO) is proposed. The PSO is combined with the Bayesian classifier, and the feature learning structure is further optimized by using the gcForest algorithm to obtain the DPSO classifier. The DPSO improves the reference value of the training results and can overcome the overfitting phenomenon in the training process to a certain extent. This paper concludes by comparing the DPSO classifier, Bayesian classifier and gcForest algorithm for the identification of data on IEEE 30–Bus system fault power flow, and verifies the effectiveness of DPSO, whose judgment accuracy is improved by about 5% compared with the other algorithms.