广东电力2025，Vol.38Issue(9)：108-118,11.DOI:10.3969/j.issn.1007-290X.2025.09.010
基于知识图谱的充电站故障智能识别方法
Intelligent Fault Recognition Method of Charging Station Based on Knowledge Graph
摘要
Abstract
Aiming at the problems of low knowledge utilization and insufficient fault localization accuracy of traditional fault identification methods due to the complex structure of the charging station system and various types of faults,this paper proposes an intelligent identification method of charging station faults based on knowledge graph.First,based on the text data of charging station faults,it discusses the process of text preprocessing and knowledge ontology construction.Second,a joint BERT-Seq2Seq extraction model is innovatively designed,which integrates the BERT contextual characterization capability with the Seq2Seq generative framework,and significantly improves the accuracy of entity relationship extraction through the three-tier structure of embedding layer,encoder layer,and decoder layer working together.After extracting the fault entity relations using the model,the Neo4j graph database is used to construct the visualized knowledge graph.Finally,in the knowledge graph,the paper uses the rule-based matching method and string matching algorithm to classify and query problems to complete the construction of charging station fault intelligent recognition.The results show that the BERT-Seq2Seq model has an accuracy rate of 89.22％in the entity-relationship joint extraction task,which can effectively identify the information entities in the fault text and accurately construct the knowledge graph relational edges.At the same time,the response accuracy rate of the intelligent question-answering system constructed on the basis of this model is more than 90％,which can quickly locate the cause of the faults and output the solutions,and has a certain value for engineering applications.关键词
知识图谱/故障识别/问答系统/知识抽取/图谱构建技术Key words
knowledge graph/fault recognition/question-answering system/knowledge extraction/graph construction technology分类
信息技术与安全科学引用本文复制引用
陈国芳,李黄强,向昆,朱添安,马辉..基于知识图谱的充电站故障智能识别方法[J].广东电力,2025,38(9):108-118,11.基金项目
国家自然科学基金项目(52377191) （52377191）
湖北省自然科学基金面上项目(2024AFB584) （2024AFB584）
