A fault line selection method of small current grounding system based on wavelet de-noising and improved RBF neural network

Abstract

The paper proposed a fault line selection method of small current grounding system based on wavelet de-noising and improved RBF neural network. Fault information matrix is obtained after normalization processing for maximum of absolute value of de-noised zero-sequence current, and the matrix is used as input of RBF neural network. Active value of input layer of the RBF neural network is calculated, and when the active value is in preset range, hidden layer and output layer of the RBF neural network are disconnected automatically, and neurons of the hidden layer split. After parameters such as weight, variance and central value of the RBF neural network are adjusted, the hidden layer and the output layer are reconnected and training result is output. Test sets are input into the trained RBF neural network to get fault line selection result. The example analysis result shows that the method cannot be influenced by fault phase angle and grounding resistance with accurate and reliable fault line selection.

Keywords

• small current grounding system
• fault line selection
• wavelet de-noising
• rbf neural network
• active value