广东电力2025，Vol.38Issue(12)：33-43,11.DOI:10.3969/j.issn.1007-290X.2025.12.003
融合双向时序特征与多头注意力机制的电缆故障定位方法
Cable Fault Localization Method Integrating Bidirectional Temporal Features and Multi-head Attention Mechanism
摘要
Abstract
Aiming at the current problems of insufficient accuracy and susceptibility to interference in cable fault location,this paper proposes a cable fault location method that integrates bi-directional timing features and multi-head attention mechanism.Firstly,the fault signal was subjected to phase-mode transformation,and then through multielement variational mode decomposition(MVMD)and K-means clustering,the fault characteristic data was formed.Secondly,the information of fault data was captured in both forward and reverse directions through the bi-directional temporal convolution network(BiTCN).Combining the bi-directional gated recurrent unit(BiGRU)and the multi-head attention mechanism(MHAM)to enhance the model's ability to capture key features,the BiTCN-BiGRU-MHAM model was established,and then the crested porcupine optimizer(CPO)was used to optimize the hyperparameters of the model.Finally,the data was input into the model for training and testing.The results show that the fitting degree of this fault location model reaches 99.992％,and it has a relatively high fault location accuracy.关键词
故障定位/相模变换/K-means聚类/双向时间卷积网络/双向门控循环单元/多头注意力机制/冠豪猪优化算法Key words
fault location/phase mode transformation/K-means clustering/bi-directional temporal convolution network(BiTCN)/bi-directional gated recurrent unit(BiGRU)/multi-head attention mechanism(MHAM)/crested porcupine optimizer(CPO)分类
信息技术与安全科学引用本文复制引用
胡国栋,马宏忠,孙维,田正宏..融合双向时序特征与多头注意力机制的电缆故障定位方法[J].广东电力,2025,38(12):33-43,11.基金项目
国家自然科学基金项目(51577050) （51577050）
江苏省重点研发计划项目(BE2023347) （BE2023347）
