The Study of Knowledge Graph Framework for Faults Disposal of Power Transformer

Currently, the maintenance strategy decision making mainly relied on the preventive tests and artificial participation, and this cannot meet the increasing scale of power grid and the number of equipment. So, it is necessary to study the fault maintenance strategy of power transformer under the background of new power system characterized by big data. In this paper, a knowledge graph is designed for transformer fault disposal in the field of power grid regulation, composed of knowledge extraction, knowledge fusion and knowledge update. For the text data in the OMS and the PMS system, using the natural language processing (NLP) technology, knowledge extraction includes establishing the power proper noun dictionary and specifying the regularization template, such as “station +voltage level +transformer”. When a transformer fault occurs, the type identification of the transformer fault information is carried out according to the knowledge graph: on the basis of the grid dispatching operation and fault disposal rules stored as knowledge elements, the similarity between this fault event and historical fault case events is calculated to give disposal plans for fault events with a similarity greater than 70% and sorted by similarity priority. The result is imported into the Neo4j graph database to generate multiple discrete data nodes, so as to connect the discrete nodes through relationships. The knowledge graph for transformer fault disposal can provide traversal and high performance retrieval of the data network, the similarity of fault description reaches more than 90%, and the traversal speed almost not affected by the amount of data, suitable for providing the corresponding intelligent decision making plan in the field of regulation.