Abstract:

RBF neural network is effective for nonlinear fault diagnosis,but such method has the problem of local minimum and slow convergence.Evolutionary algorithm has strong global search ability,and it can realize parameter optimization for the RBF neural network through intersection,selection and aberrance;but during parameter optimization,covergence is slow.This paper proposes a distribution network fault diagnosis model based on improved RBF nueral network through combination of gradient descent algorithm and evolutionary numerical hybrid method.Application expriments show that this model has a high diagnostic accuracy.

Key words: RBF neural network, improvement, distribution network, fault diagnosis