A power line fault cause identification method based on CNN-BiLSTM and attention mechanism

To accurately identify various power line fault causes such as lightning strike, bird pest, and crane touch line, A power line fault cause identification method based on convolutional neural network-bidirectional long short-term memory network and attention mechanism (CNN-BiLSTM-SE) is proposed. Firstly, the measurement data after a fault occurs in the power line, i.e. the three-phase voltage, three-phase current, zero-sequence voltage, zero-sequence current and the calculated harmonic data, are used as fault characteristic data samples. Secondly, considering that the combination of convolutional neural network, bidirectional long short-term memory network and attention mechanism has advantages of processing complex non-stationary time series data and enhancing data correlation, a power line fault cause identification model based on CNN-BiLSTM-SE is constructed. Finally, the fault cause identification model based on CNN-BiLSTM-SE is trained and tested by using the power line fault measurement data samples. Results show that the average accuracy of the power line fault cause identification model proposed in this paper exceeds 99%, which verifies the effectiveness and accuracy of the proposed method.