With the accelerated construction of new power systems, the scale and complexity of transmission systems are increasing, while transmission line fault diagnosis algorithms for complex power timing data urgently need to be studied and tested.In this paper, a transmission line fault diagnosis model fusing convolutional neural network (CNN) and long short-term memory network (LSTM) is proposed. The model realizes automatic extraction of local spatio-temporal features of fault signals through convolutional neural network, captures the long-term dependence of power system operation state by using long and short-term memory network, and introduces an attention mechanism to enhance the characterization of key fault features, which effectively improves the identification accuracy of the fault types of the transmission line under the complex working conditions. It adapts to the requirements of smart grid real-time monitoring scenarios on algorithm robustness. Also, it considers the use of the synthetic few over-sampling technique (SMOTE) to realize the data enhancement to improve the imbalance of data distribution. The results show that compared with the traditional connection neural network, back propagation neural network, and LSTM model, the CNN-LSTM model is only 0.0497 in the loss value, which is reduced by about 70%, and the fault identification accuracy and precision rate are more than 97%, which is improved by about 2%. It provides theoretical support for solving the problem of real-time fault diagnosis of large-scale transmission systems, and has significant engineering practice value for reducing the cost of grid operation and maintenance and improving the reliability of power supply.
Hao et al. (Mon,) studied this question.