Automatic Extraction Method of Power Grid Fault Disposal Strategy Based on Deep Learning

In order to enhance the precision of the analysis and fault disposal ability of power grid fault disposal plan, an automatic extraction method of power grid fault disposal strategy based on deep learning was proposed. Firstly, the treatment strategy in the power grid fault disposal plan is labeled based on syntactic analysis and event elements, and the fault disposal plan is encoded using a pretrained dynamic word vector model based on the BERT protocol to generate word vectors of power grid fault disposal strategies in different professional contexts. In order to extract deep features of fault disposal strategy entity information, attention mechanism and bidirectional long shortterm memory (BiLSTM) were combined and conditional random field (CRF) was used to calculate the optimal annotation information of fault disposal strategy. Therefore, an automatic fault disposal strategy extraction model based on BERT+BiLSTM-CRF+Attention hybrid neural network is constructed. Ultimately, the effectiveness of the proposed method is demonstrated through its application in the fault disposal plan of a control center. This method successfully extracts relevant treatment strategy information from the fault disposal plan with high precision, exhibiting superior recognition accuracy and greater generalization capabilities when compared to alternative approaches.