Power Grid Fault Diagnosis Method Based on CNN

The existing power grid fault diagnosis methods rely on experience to design the diagnosis model, lacking the ability to extract fault knowledge from the engineering field alarm information, and cannot achieve intelligent fault diagnosis quickly and accurately. To solve this problem, this paper proposes a fault diagnosis method for power grid based on alarm information text. Firstly, two models for fault classification and key information extraction are developed based on a convolutional neural network. The fault classification model and the key information extraction model can extract the overall features and local features of the text, respectively, identify the fault types, obtain the key information corresponding to the fault in the text, and screen the suspicious fault components. Then, based on the fault type discrimination results and key information output by two classification models, a component location strategy based on fault component characteristics is proposed. Finally, we use both highly descriptive and low-noise alarm information text generated by simulation and text with high noise and simplified description collected from engineering practice in fault diagnosis. The results show that the proposed method can realize end-to-end fault identification and fault components positioning.