Condition Monitoring and Fault Prediction Optimization of Transmission Network based on Deep Learning and Edge Computing

With the continuous growth of electricity demand, the stability and reliability of the transmission network have become particularly important. This paper optimizes the state monitoring and fault prediction of the transmission network based on deep learning convolutional neural network (CNN) and edge computing methods. First, we collected the operation data of the transmission network, preprocessed it and extracted key features. Considering the computing power and resource limitations of edge computing nodes, a lightweight CNN model was designed, which adopted deep separation convolution and Transformer attention mechanism and was trained with labeled fault datasets. Finally, the trained CNN model was deployed to the edge computing node for real-time data inference and real-time fault detection. Combining the CNN model output and fault diagnosis algorithm, potential faults in the transmission network are predicted and located. The study found that the CNN model performed well in predicting electrical faults and pollution flashover faults under different weather conditions, with the highest prediction accuracy of electrical faults reaching 98.8%. In addition, the model performed well in state monitoring speed, with the minimum monitoring time being 115.1 milliseconds. The CNN model can effectively improve the accuracy and real-time performance of transmission network fault prediction.