As the size of power facilities grows and expands, op-erational monitoring is an important part of system management, and a key step in power system operational monitoring is the prediction of equipment defects. Power equipment defects refer to abnormalities or hidden dangers that occur during the use of the equipment, and these hidden dangers may cause the power system to malfunction. To solve this problem, artificial intelligence technology is combined and taken to the field of knowledge graph, through knowledge fusion, using the relationship-aware graph neural network to get the representation of entities and relationships. A series of data collected on power equipment defects are input into the jieba participle tool for processing, and then the pre-processed data are subjected to entity recognition and relationship extraction, etc. Subsequently, knowledge fusion is realised by means of pairing entities and stored in the Neo4j graph database to construct a knowledge graph oriented to the grading of power equipment defects for better knowledge querying and assisted decision making.