电力信息与通信技术2024，Vol.22Issue(11)：7-15,中插1,10.DOI:10.16543/j.2095-641x.electric.power.ict.2024.11.02
基于CasRel-AttR模型的电力二次作业风险知识图谱构建方法研究
Research on the Construction Method of Risk Knowledge Graph of Power Secondary Operation Based on CasRel-AttR Model
摘要
Abstract
With the continuous accumulation of operation risk data of power system at the present stage,the risk knowledge of power secondary operation represented by relay protection needs to be stored and queried in a more efficient way.The highly structured characteristics of knowledge graph technology provide feasible scheme for solving problems in this field.Firstly,a knowledge graph construction method for the secondary operation of electric power is designed,including data preprocessing,knowledge extraction,knowledge fusion and storage procedure.Then,aiming at the knowledge extraction method,a CasRel-AttR entity relation joint extraction method is proposed.According to the characteristics of the secondary operation corpus,a self-attention mechanism fused with rotary position encoding(RoPE)is added to the subject and object annotation part of the CasRel model.Experiments are set up on the self-constructed secondary operations risk dataset,and the experimental results prove the superiority of the model in recognition effect.Finally,on the basis of the extracted triples,a complete knowledge graph is constructed according to the construction process,which initially achieves the effect of assisting risk classification and operation and maintenance decision-making.关键词
电力二次作业/知识图谱/自然语言处理/知识抽取/自注意力机制Key words
power secondary operation/knowledge graph/natural language processing/knowledge extraction/self-attention mechanism分类
信息技术与安全科学引用本文复制引用
刘丽,白宇昂,李刚,吴智泉,张新,李盈盈..基于CasRel-AttR模型的电力二次作业风险知识图谱构建方法研究[J].电力信息与通信技术,2024,22(11):7-15,中插1,10.基金项目
国家重点研发计划资助(2020YFB0906000,2020YFB0906005) （2020YFB0906000,2020YFB0906005）
河北省在读研究生创新能力培养资助项目(CXZZSS2024163). （CXZZSS2024163）
