Research on Transmission Line Fault Prediction and Identification Method Based on TCN-BiLSTM-Multihead-Attention

The reliable and effective method of line fault prediction is an important guarantee to further improve the reliability of transmission system. The existing prediction methods cannot make full use of the waveforms of fault forward. This paper proposes a TCN-BiLSTM-Multihead-Attention mechanism fault prediction method for transmission lines based on sample expansion and feature selection. Through the fault forward wave current data collected by transmission line distribution device, an effective waveform database containing pre-fault information of transmission lines is sorted out. Borderline-SMOTE algorithm was used to expand the sample, and the pre-fault waveform features including time domain, frequency domain and time-frequency domain were extracted in various ways. Lasso regression method was used to optimize the pre-fault features. Finally, TCN-BiLSTM-Multihead-Attention mechanism was used to predict the fault of transmission lines. The practical application results show that the proposed method has higher accuracy.