Study on Identification of Off-line Arcing Based on Wavelet Packet Decomposition and Neural Network

Abstract

Electromagnetic radiation field generated by off-line arcing will disturb sensitive equipment during the operation of a high-speed train. To quickly identify the pantograph and catenary off-line arcing signal, this paper proposes a method based on wavelet packet decomposition and Back Propagation (BP) Neural Network. Firstly, wavelet packet decomposition is carried out to analyze the characteristics of electromagnetic radiation field of offline arcing; secondly, eigenvectors are extracted based on signal analysis; finally, a signal classifier is set up by BP neural network. The eigenvectors are used as the input of the signal classifier to identify the signal. Simulation results show that the proposed model can quickly and effectively identify the signals to achieve an early warning effect