Identification of tree-related high-impedance faults using BiLSTM model

In recent years, tree-related high-impedance faults (THIFs) on 10 kV overhead lines in forested areas have become common, posing a wildfire risk. This study analyzes the physical processes of THIFs and their effects on electrical signals. A 10 kV distribution network experimental platform was built to study discharge signal characteristics. The results show differences in zero-sequence voltage between THIFs and other faults. The BWO-VMD algorithm extracts 8 entropy features for a fault feature library. Using the BiLSTM model, fault identification accuracy reaches 99.6%, outperforming GRU and LSTM. The proposed method is effective for THIF identification and offers insights for wildfire prevention.