A Knowledge Graph Completion Method for the Power Transformer Based on Graph Neural Network and TransR

With a high proportion of new energy access in the power grid, the risks and the latent dangers of the operation of the power equipment increase. At present, the prediction of the latent failure of the power equipment largely depends on the experience and the expertise of the maintenance personnel. A completed construction of the knowledge graph for the power equipment provides a more objective way for the maintenance personnel to make a more accurate judgment and decision. In order to further mine the hidden knowledge in the knowledge graph of the power equipment and improve the integrity of the knowledge graph of the power equipment, a dynamic completing method for the knowledge graph of the power transformer maintenance based on Learning Entity and Relation Embeddings for Knowledge Graph Completion (TransR) and graph neural network is presented. A graph neural network is designed for the graph embedding of the triples and obtaining the representation ability of the out-graph entities from the triples. The external knowledge triples were scored by the TransR scoring algorithm, which determines whether a new triplet is added to the knowledge map or not. The testing results show that the proposed method has a better effect on the completion and the extension of the knowledge graph.