Fault diagnosis for generator unit based on RBF neural network optimized by GA-PSO

Abstract

PSO (Particle Swarm Optimization)-RBF is widely used in intelligent fault diagnosis for generator unit. Since PSO has slow convergence rate, low accuracy, and early-maturing problem which effect training speed and diagnosis accuracy of PSO-RBF, the operations of crossover and variation of genetic algorithm (GA) are introduced into PSO such that the performance of PSO can be improved. GA-PSO is employed to optimize the RBF neural network with concrete steps, then GA-PSO-RBF is applied in fault diagnosis for generator unit. Simulation results show that GA-PSO-RBF is superior to PSO-RBF in training speed, convergence accuracy, and diagnosis accuracy, thus, it is a new efficient diagnosis approach.