Comparative Analysis of Fault Identification Methods for Distribution Networks Based on Waveforms Containing White Noise

Active distribution network mostly adopts neutral point non Earthing system. When high resistance grounding fault current occurs, it is difficult for relay to detect. In addition, the presence of noise makes fault identification more difficult. In response to this issue, this article first established an IEEE119 node simulation model, obtained fault data through simulation, and established a database. Then, based on the ABLSTM method, fault identification was carried out. Finally, considering the impact of white noise in the distribution network, different data preprocessing methods were used to analyze the classification performance of various algorithms in the presence of noise. The results indicate that the ABLSTM method has certain data preprocessing capabilities, and the classification accuracy remains above 97% when adding 15dB of Gaussian white noise.