Knowledge Graph Model of Power Grid for Human-machine Mutual Understanding

In power grid research, the high complexity of the power generation and distribution requires the power grid system to be robust and resilient. To assure the stability of the power grid, massive human intervention is necessary currently, which is labor intensive and inflexible. However, considering the vital importance of the power grid, it is impossible to leave human being out of the power system at present. To increase the degree of automation in power grid and keep the system robust meanwhile, a smart grid with human in the loop is required. A novel modeling method of the power grid is developed in this study to enable efficient human-machine operation. Specifically, a knowledge graph model of power grid is developed which incorporates the operation rules and knowledge from corresponding documents and literatures. According to the property of the documents, two task oriented methods have been designed to extract the entities and relations from texts. For regularized document with items well organized, TextRank algorithm is adopted to extract the keyword entities and grammatical rule analysis is used to extract the logical relation entities and related event entities. For general literatures, semantic role labeling based on dependency parsing has been employed to extract event triplets to simplify the text analysis. The original sentence of the event triplet is used to extract the logical relation and co-occurrence relation. Based on the text analysis results, knowledge fusion and knowledge processing are carried out and the results are imported into Neo4j to form a visual knowledge graph which can be queried and used. The constructed model could facilitate mutual understanding of power grid for both human and machine.