基于多视图稀疏特征选择的架空输电线路故障原因判别  

Fault Cause Identification of Overhead Transmission Line Based on Multi-view Sparse Feature Selection

作　　者：苏超 杨强[1] SU Chao;YANG Qiang(College of Electrical Engineering,Zhejiang University,Hangzhou 310001,China)

机构地区：[1]浙江大学电气工程学院,浙江杭州310001

出　　处：《智慧电力》2023年第3期96-103,共8页Smart Power

基　　金：国家自然科学基金资助项目(52177119)。

摘　　要：日渐增加的多源异构数据为输电线路故障原因判别带来了信息融合的机遇和挑战。为解决故障录波和多源关联信息的特征融合问题,引入多视图学习概念,提出了基于多视图稀疏特征选择的架空输电线路故障原因判别方法。根据故障录波和关联信息区分并提取双视图故障特征,随后基于稀疏表示提出了层次多视图特征选择算法(HMVFS)。该算法引入ε-dragging扩大分类类别的标签间距,并通过Frobenius范数和l2,1范数的正则化项分别从故障视图和故障特征的高低维度实现特征选择。最后采用某地区输电线路故障数据进行对比实验,结果验证了该方法在输电线路故障原因判别的有效性和优越性。The increasing multi-source heterogeneous data brings the opportunities and challenges about multisource information fusion to the fault cause identification of transmission line.In order to solve the feature fusion of fault recording data and multi-source contextual information,the paper introduces the conception of multi-view learning,and proposes a fault cause identification method for overhead transmission line based on multi-view sparse feature selection.Firstly,dual-view fault features are distinguished and extracted according to the fault recording data and contextual information.Then hierarchical multiview feature selection(HMVFS)is proposed based on the sparsity representation,integratingε-dragging technique into the algorithm to enlarge the boundary distance between class labels,and realizing the feature selection from the high and low dimensions of fault view and fault feature with the regularization item of Frobenius norm and l2,1 norm.Finally,the comparative experiments are done using the fault data from the transmission line in a certain region,and the results show the effectiveness and superiority of the proposed method in the fault cause identification.

关 键 词：输电线路 故障原因判别 多视图学习 稀疏表示 特征选择 

分 类 号：TM726[电气工程—电力系统及自动化]