Research On The Knowledge Graph Pattern Layer Construction Of Anhui Power System Dispatching Regulation

With the expansion of the power system,the information increases greatly.In order to respond and process information quickly,and improve the electric field of knowledge management ability,we need do research into the knowledge graph of power system.besides,the power dispatching regulation is the rule of manufacturing and management,it is practically significance to do research into the knowledge graph.Specifically,at first this thesis describes the research background and previous research results,and introduces the concept of domain knowledge graph and power system dispatching regulation,then introduces knowledge graph construction method and Neo4j graph database storage.Knowledge graph construction needs to complete information extraction,mainly complete entity information extraction and relationship information extraction,since the entity information of power system dispatching regulation is mainly composed of concepts and terms,the knowledge graph pattern layer construction is mainly to extract the concepts and relationships from the regulation,then complete Neo4j graph database storage.Secondly,the technology of information extraction is studied.For the statistics-based method of information extraction,this thesis introduces hidden Markov model,maximum entropy Markov model and conditional random field model,and introduces the advantages of conditional random field from the perspective of directed graph and undirected graph,generative model and discriminant model.Then introduces corpus tagging and the selection of feature templates,and introduces conditional random field implementation software CRF++ and the use of feature templates.For the rule-based method of information extraction,this thesis introduces how to write rules and use regular expressions to realize information matching,and the use of python compiler Py Charm.Finally,this thesis uses the statistics-based method and the rule-based method to realize the information extraction of dispatching regulation,uses the statistics-based method to complete the entity information extraction of dispatching regulation,and realizes the entity extraction and evaluation by CRF++ software of conditional random field.This thesis uses the rule-based method to complete the triplet information extraction of entities and relationships,and uses regular expressions to write rules,mainly completes regex match of entities and dispatching regulation,and gets texts to be extracted,then through given verb list to generate rules,and completes regex match of texts to be extracted and rules,then gets the triplet information of entities and relationships.Then,Cypher statement and py2neo method are used to realize the Neo4j database storage of extracted information respectively.Compared with Cypher statement,py2neo method is more convenient and efficient,which can create nodes and relationships in batches,and realize the storage and display of extracted information in batches.The research shows that the selected information extraction methods and techniques are feasible to construct the knowledge graph pattern layer of dispatching regulation.