%0 Journal Article
%T 基于二维灰度图与CNN-GRU的单相接地故障选线方法<br>Single-Phase Ground Fault Line Selection Method Based on Two-Dimensional Grayscale Image and CNN-GRU
%A 胡祥谢
%A 聂祥论
%A 谢宪源
%J Modeling and Simulation
%P 1-10
%@ 2324-870X
%D 2025
%I Hans Publishing
%R 10.12677/mos.2025.141001
%X 针对配电网单相接地故障特征提取困难和现有检测方法精度低的问题&#65292;本文提出一种基于原始数据生成的二维灰度图和CNN-GRU的故障选线方法。首先&#65292;获取原始的电流波形进行截取并生成二维灰度图&#65292;以有效地保留明显而全面的故障特征原始信息&#65307;然后&#65292;CNN自适应提取时频灰度图像的局部特征&#65292;GRU从CNN层学习到的局部特征中学习上下文依赖关系。最后&#65292;通过SoftMax层实现故障选线。仿真结果表明&#65292;本文所提方法的选线准确率为99.41%&#65292;与现有方法相比准确率更高。<br>Aiming at the difficulty of single-phase grounding fault feature extraction in distribution network and the low accuracy of existing detection methods, this paper proposes a fault line selection method based on two-dimensional grayscale generated by original data and CNN-GRU. Firstly, the original current waveform is obtained and intercepted to generate a two-dimensional grayscale image, so as to effectively retain the obvious and comprehensive original information of fault characteristics. Then, CNN adaptively extracts the local features of the time-frequency grayscale image, and GRU learns the context dependency from the local features learned by the CNN layer. Finally, the fault line selection is realized through the SoftMax layer. The simulation results show that the accuracy of the proposed method is 99.41%, which is higher than that of the existing methods.
%K 故障选线&#65292
%K 卷积神经网络&#65292
%K 循环神经网络<br>Fault Line Selection
%K Convolutional Neural Network
%K Recurrent Neural Network
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=104374