Research on Fault Entity Relationship Construction and Fault Reasoning Algorithm of Substation Main Equipment with Adversarial Training

At present, the intelligent health management technology of substation main equipment is developing rapidly. In the operation and maintenance process, there are problems such as weak information relevance and low efficiency of decision generation, and the accuracy and standardization of fault handling are difficult to be guaranteed. This paper designs the construction method of fault knowledge graph of substation main equipment. The entity recognition method of BERT-FMG-BiLSTM-CRF with adversity-training is proposed, and the relationship extraction method based on sentence level and word level BiLSTM-ATT is proposed to construct the fault knowledge map of substation main equipment. The results of comparative experiments prove the superiority of this method in entity recognition and relation extraction. This model can realize the goal of intelligent fault diagnosis and decision making, so as to improve the efficiency of daily operation, maintenance and management of power grid.