基于语义增强的电网故障处置预案匹配方法  

Matching Method for Power Grid Fault Handling Plan Based on Semantic Enhancement

作　　者：蒙飞 李江鹏 李涛 徐建忠 高海洋 乔咏田 MENG Fei;LI Jiangpeng;LI Tao;XU Jianzhong;GAO Haiyang;QIAO Yongtian(State Grid Ningxia Electric Power Co.,Ltd.,Power Dispatch and Control Center,Yinchuan 750001,China;Guodian Nari Nanjing Control System Co.,Ltd.,Nanjing 211106,China)

机构地区：[1]国网宁夏电力有限公司调度控制中心,宁夏银川750001 [2]国电南瑞南京控制系统有限公司,江苏南京211106

出　　处：《中国电力》2025年第4期237-244,共8页Electric Power

基　　金：国网宁夏电力有限公司科技项目(5229NX220027)。

摘　　要：为提升电网故障处置预案匹配效率和准确率,提出了基于语义增强的电网故障处置预案匹配方法。首先,通过微调基于变换器双向编码器表征(bidirectional encoder representations from transformers,BERT)模型的超参数,将故障处置预案中多调度对象实体表征为可计算词向量,并接入条件随机场(conditional random field,CRF)模型识别调度对象实体类别;然后,基于残差向量-字词嵌入向量-编码向量(residual vector-embedding vector-encoded vector,RE2)计算电网故障信息和调度对象的语义距离,建立基于BERTCRF-RE2的电网故障处置预案匹配模型;最后,通过某地区电网数据进行验证。结果表明,所提模型有效解决了预案匹配准确率低的问题。In order to improve the matching efficiency and accuracy of grid fault handling plan,a semantic enhancement-based grid fault handling plan matching method is proposed.Firstly,the multi-dispatch objects entities in the fault handling plan are characterized as computable word vectors by fine-tuning the hyperparameters of the bidirectional encoder representations from transformers(BERT)model,and integrated into the conditional random field(CRF)model to identify the dispatch objects entity categories.And then,the semantic distance between the grid fault information and dispatch objects are computed based on the residual vector-embedding vector-encoded vector(RE2),and a grid fault handling plan matching model is established based on BERT-CRF-RE2.Finally,through validation of the data of a regional power grid,the proposed model effectively solves the problem of low plan matching accuracy rate.

关 键 词：电网故障处置 预案匹配 语义增强 多调度对象 

分 类 号：TM73[电气工程—电力系统及自动化] TM711