基于多源信息熵融合的局部放电模式识别方法

打开文本图片集

关键词： 局部放电； 气体绝缘开关设备； 图谱数据； 信息熵； 深度残差网络； 决策融合中图分类号：TN86⁃34；TM595 文献标识码： A 文章编号：1004⁃373X（2025）12⁃0096⁃09

Abstract：Partial discharge （PD） is closely related to the insulation degradation of gas insulated switchgear （GIS）， and the accurate identification of PD types is of great significance for ensuring the safe operation of the power grid. In allusion to the limitations of a single spectrum recognition mode in PD pattern recognition tasks， a method of PD identification based on multi⁃ source information entropy fusion is proposed. Five typical PD defect models in GIS are built， and the phase resolved partial discharge （PRPD） graph data and phase resolved pulse sequence （PRPS） graph data are collected by the experiments. A deep residual network （ResNet） is employed to extract the two graph features， so as to train the optimal network model. The testing set is input into the ResNet to obtain the Softmax output probabilities to calculate the information entropy， assigning different weights to each classifier. The final classification result is obtained by means of the information entropy decision fusion. The experimental results show that， in comparison with single graph recognition mode， the proposed method can significantly improve accuracy， reaching 98.4% . For real field data， the accuracy of the proposed method can exceed 90% ， indicating promising application prospects.

Keywords：partial discharge; gas insulated switchgear; graph data; information entropy; deep residual network; decision fusion

0 引 言

因为气体绝缘开关设备（Gas Insulated Switchgear，GIS）具有更小的占地面积、更高的运行可靠性以及更低的检修频率等优点，被广泛应用于电力系统中的输配电领域[1⁃3]。（剩余9870字）