Power Primary Equipment Defect Knowledge Graph And Question Answering System Based On Deep Learning

According to statistics,in 2022,China’s electricity consumption in the whole society will reach 8,637.2 billion kilowatt hours,and the huge power pressure will lead to the longterm high-load working state of power equipment,which needs to be frequently repaired and maintained,and the maintenance records of these equipment will form a large amount of power equipment defect data.At present,the State Grid equipment is large in scale and varied,and the equipment defect information lacks management,which makes it difficult to quickly analyze the cause of defects and accurately give defect disposal plans.At the same time,the fault maintenance of power equipment mainly relies on manual analysis,which has high requirements for the knowledge reserve,maintenance experience and adaptability of operation and maintenance personnel.In recent years,the continuous development of knowledge graph and natural language processing technology has provided a better solution for knowledge graph question and answer.In view of this,this paper uses named entity recognition and text classification technology to design and implement a knowledge graph and question answering system based on deep learning,and the main work content is as follows:(1)Collect and sort out relevant knowledge in the field of electric power,and construct a knowledge map of power primary equipment defects.Firstly,the relevant literature in the field of electric power,web materials,some materials of the State Grid Hubei Provincial Company and "QGDW1904.1-2013 Transmission and Transformation Equipment Defect Specification Part 1: Primary Partial Specification for Power Transformation" were extracted.Then,the extracted data is subject to knowledge fusion and knowledge processing,and then the knowledge graph is constructed.Finally,the knowledge graph is stored in the Neo4 j database,with a total of 3046 nodes and 3045 relationships.(2)The power primary equipment defect question dataset is constructed,and a power primary equipment defect knowledge graph question answering algorithm model is proposed.Firstly,the defect question dataset is constructed,which is divided into two parts:named entity recognition and text classification,the former using the BIO system for labeling,and the latter using five question types for labeling.Secondly,a named entity recognition based on BERT-wwm-ext+Bi LSTM+CRF network is designed,and the text classification of questions is performed in combination with the Text CNN network.Then,the extracted entity is inserted into the corresponding category of Cypher template,and the conversion of Chinese question to Cypher query statement is completed,so as to find the solution in the knowledge graph.Experimental results show that the design of each module of the model is reasonable,and it has good effects on named entity recognition and text classification of Chinese questions.(3)Design and implement the power primary equipment defect knowledge graph question and answer and visualization system.Based on Spring Boot,Vue,Pycharm,My SQL database and other technical frameworks,the system is implemented,which can be divided into visual question answering subsystem and background management subsystem.The former implements the login registration module,knowledge graph question answering module,defect analysis module,question answering history analysis module and knowledge graph visualization module;The latter implements the background login module,user data management module,defect device management module,question history management module and knowledge graph data management module.(4)System testing was completed.The functional and non-functional tests of the power primary equipment defect knowledge graph question and answer and visualization system are carried out,and the test results show that the system has perfect functions,and the question answering algorithm module is reasonable and meets the expected goals.