Fault line selection method for distribution networks based on BMF-GADF and improved Swin Transformer

doi:10.3969/j.issn.2097-0706.2026.02.008

Abstract

Abstract:

Due to the weak fault characteristics when a single-phase-to-ground fault occurs in the small current system of distribution networks，existing fault line selection methods often suffer from low accuracy and weak robustness. Therefore，a fault line selection method for distribution networks based on BMF-GADF and an improved Swin Transformer was proposed. The zero-sequence current was converted into a feature-enhanced Gramian angular difference field image by combining Butterworth mean filtering and Gramian angular difference field. The image samples were then input into the improved Swin Transformer model for feature extraction. Building upon the original architecture，the improved Swin Transformer innovatively introduced a parallel convolution block attention module，enabling more accurate adaptive feature selection and effectively enhancing model accuracy. The line selection results were output through Softmax. The experimental results showed that the fault line selection accuracy of the proposed method reached 98.96%. Compared with other fault line selection methods，the proposed method demonstrated higher line selection accuracy and better noise robustness，offering a new solution for fault line selection in distribution networks.

Key words: fault line selection, Gramian angular difference field, convolution block attention module, Swin Transformer, feature extraction

CLC Number:

TK01

WU Xiaohuan, SHEN Jinggui, ZHANG Xin, HU Yumin, XU Yeling, SHI Mingyu. Fault line selection method for distribution networks based on BMF-GADF and improved Swin Transformer[J]. Integrated Intelligent Energy, 2026, 48(2): 86-95.

Figures/Tables 16

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Table 1

Fig.7

Table 2

Fig.8

Fig.9

Fig.10

Fig.11

Table 3

Table 4

Table 5

References 27

[1]	邓丰, 梅龙军, 唐欣, 等. 基于时频域行波全景波形的配电网故障选线方法[J]. 电工技术学报, 2021, 36(13): 2861-2870.
	DENG Feng, MEI Longjun, TANG Xin, et al. Faulty line selection method of distribution network based on time-frequency traveling wave panoramic waveform[J]. Transactions of China Electrotechnical Society, 2021, 36(13): 2861-2870.
[2]	袁佳歆, 李响, 张哲维. 基于注入信号的有源配电网单相接地故障选线方法[J]. 电测与仪表, 2020, 57(5): 44-49.
	YUAN Jiaxin, LI Xiang, ZHANG Zhewei. Research on single-phase grounding fault selection technology for active distribution network based on injected signal[J]. Electrical Measurement & Instrumentation, 2020, 57(5): 44-49.
[3]	李成钢, 刘亚东, 杨雪凤, 等. 基于变分模态分解与空洞卷积神经网络的配电网故障选线方法[J]. 电网与清洁能源, 2024, 40(2): 110-118, 126.
	LI Chenggang, LIU Yadong, YANG Xuefeng, et al. A method of fault line selection for distribution networks based on variational mode decomposition and dilated convolutional neural network[J]. Advances of Power System & Hydroelectric Engineering, 2024, 40(2): 110-118, 126.
[4]	邓祥力, 赵磊鑫, 熊小伏, 等. 基于多元暂态特征故障度的配电网单相接地选线方法[J]. 电力系统保护与控制, 2024, 52(15): 69-80.
	DENG Xiangli, ZHAO Leixin, XIONG Xiaofu, et al. A single-phase grounding line selection method for a distribution network based on multivariate transient characteristic fault degree[J]. Power System Protection and Control, 2024, 52(15): 69-80.
[5]	雷静, 李晨婧, 郭亮, 等. 基于GAF与卷积神经网络的配电网故障选线新方法[J]. 电力系统及其自动化学报, 2023, 35(2): 53-62.
	LEI Jing, LI Chenjing, GUO Liang, et al. Novel fault feeder selection method for distribution network based on GAF and convolution neural network[J]. Proceedings of the CSU-EPSA, 2023, 35(2): 53-62.
[6]	陈家晖, 周凡博, 陈芳芳. 基于双图转换和ECA-DenseNet的配电网故障选线新方法[J/OL]. 云南民族大学学报(自然科学版),1-10(2025-03-11)[2025-07-29]. https://link.cnki.net/urlid/53.1192.N.20250310.1610.018.
	CHEN Jiahui, ZHOU Fanbo, CHEN Fangfang. Novel fault line selection method for distribution networks based on dual graph transformation and ECA-DenseNet[J]. Journal of Yunnan Minzu University(Natural Sciences Edition),1-10(2025-03-11)[2025-07-29]. https://link.cnki.net/urlid/53.1192.N.20250310.1610.018.
[7]	郝帅, 张旭, 马瑞泽, 等. 基于改进GoogLeNet的小电流接地系统故障选线方法[J]. 电网技术, 2022, 46(1): 361-368.
	HAO Shuai, ZHANG Xu, MA Ruize, et al. Fault line selection method for small current grounding system based on improved GoogLeNet[J]. Power System Technology, 2022, 46(1): 361-368.
[8]	JIN A Q, YANG S, LEI M H, et al. An effective convolutional and transformer cooperation network for underwater acoustic target recognition[J]. Engineering Applications of Artificial Intelligence, 2025, 159: 111791. doi: 10.1016/j.engappai.2025.111791
[9]	CHEN J Y, MA X H, LI S K, et al. A hybrid parallel computing architecture based on CNN and transformer for music genre classification[J]. Electronics, 2024, 13(16): 3313. doi: 10.3390/electronics13163313
[10]	邓飞跃, 郑守禧, 郝如江. 一种轻量化尺度感知调制Swin Transformer模型的轴箱轴承故障诊断方法[J]. 西安交通大学学报, 2024, 58(9): 83-93.
	DENG Feiyue, ZHENG Shouxi, HAO Rujiang. A lightweight scale-aware modulation Swin Transformer network for axlebox bearing fault diagnosis[J]. Journal of Xi'an Jiaotong University, 2024, 58(9): 83-93.
[11]	XIE F Y, WANG G, ZHU H Y, et al. Rolling bearing fault diagnosis based on SVD-GST combined with vision transformer[J]. Electronics, 2023, 12(16): 3515. doi: 10.3390/electronics12163515
[12]	周舟, 陈捷, 吴明明. 基于CWT和优化Swin Transformer的风电齿轮箱故障诊断方法[J]. 振动与冲击, 2024, 43(15): 200-208.
	ZHOU Zhou, CHEN Jie, WU Mingming. Fault diagnosis method for wind power gearbox based on wavelet transform and optimized Swin Transformer[J]. Journal of Vibration and Shock, 2024, 43(15): 200-208.
[13]	杨隽豪, 韦延方, 王鹏, 等. 基于Swin Transformer的柔性直流电网单端量故障诊断[J]. 电力建设, 2023, 44(5): 53-60. doi: 10.12204/j.issn.1000-7229.2023.05.006
	YANG Junhao, WEI Yanfang, WANG Peng, et al. Single-ended fault diagnosis of flexible DC grid based on Swin Transformer[J]. Electric Power Construction, 2023, 44(5): 53-60. doi: 10.12204/j.issn.1000-7229.2023.05.006
[14]	李艳秋, 李胜赵, 孙光灵, 等. 轻量型Swin Transformer与多尺度特征融合相结合的人脸表情识别方法[J]. 光电工程, 2025, 52(1): 240234.
	LI Yanqiu, LI Shengzhao, SUN Guangling, et al. Lightweight Swin Transformer combined with multi-scale feature fusion for face expression recognition[J]. Opto-Electronic Engineering, 2025, 52(1): 240234.
[15]	郑玉婕, 沈兴全, 周进节, 等. 基于混合注意力机制的滚动轴承故障诊断方法[J]. 机床与液压, 2024, 52(22): 218-226.
	ZHENG Yujie, SHEN Xingquan, ZHOU Jinjie, et al. Fault diagnosis method of rolling bearing based on mixed attention mechanism[J]. Machine Tool & Hydraulics, 2024, 52(22): 218-226.
[16]	殷林飞, 蒙雨洁. 基于DenseNet卷积神经网络的短期风电预测方法[J]. 综合智慧能源, 2024, 46(7): 12-20. doi: 10.3969/j.issn.2097-0706.2024.07.002
	YIN Linfei, MENG Yujie. Short-term wind power forecasting based on DenseNet convolutional neural networks[J]. Integrated Intelligent Energy, 2024, 46(7): 12-20. doi: 10.3969/j.issn.2097-0706.2024.07.002
[17]	郭威, 史运涛. 基于空间域图像生成和混合卷积神经网络的配电网故障选线方法[J]. 电网技术, 2024, 48(3): 1311-1321.
	GUO Wei, SHI Yuntao. Fault line selection for distribution network based on spatial domain image generation and hybrid convolutional neural network[J]. Power System Technology, 2024, 48(3): 1311-1321.
[18]	缪欣, 张忠锐, 郭威, 等. 基于信号图像化和CNN-ResNet的配电网单相接地故障选线方法[J]. 中国测试, 2024, 50(6): 157-166.
	MIAO Xin, ZHANG Zhongrui, GUO Wei, et al. Single-phase ground fault line selection method for distribution network based on signal imaging and CNN-ResNet[J]. China Measurement & Testing Technology, 2024, 50(6): 157-166.
[19]	殷浩然, 苗世洪, 郭舒毓, 等. 基于S变换相关度和深度学习的配电网单相接地故障选线新方法[J]. 电力自动化设备, 2021, 41(7): 88-96.
	YIN Haoran, MIAO Shihong, GUO Shuyu, et al. Novel method for single-phase grounding fault line selection in distribution network based on S-transform correlation and deep learning[J]. Electric Power Automation Equipment, 2021, 41(7): 88-96.
[20]	陆汉东, 方明, 刘刚刚, 等. 基于传播模型与神经网络的输变电工程造价分析与预测方法研究[J]. 综合智慧能源, 2025, 47(4): 33-40. doi: 10.3969/j.issn.2097-0706.2025.04.003
	LU Handong, FANG Ming, LIU Ganggang, et al. Research on cost analysis and prediction methods for power transmission and transformation projects based on propagation models and neural networks[J]. Integrated Intelligent Energy, 2025, 47(4): 33-40. doi: 10.3969/j.issn.2097-0706.2025.04.003
[21]	王晓龙, 付锐棋, 金韩微, 等. 融合格拉姆角差场纹理特征的滚动轴承性能退化评估[J/OL]. 振动工程学报,1-8(2024-10-29)[2025-07-29]. https://link.cnki.net/urlid/32.1349.TB.20241028.1719.004.
	WANG Xiaolong, FU Ruiqi, JIN Hanwei, et al. Performance degradation assessment of rolling bearings based on fusion of Gramian angular field texture features[J/OL]. Journal of Vibration Engineering, 1-8(2024-10-29)[2025-07-29]. https://link.cnki.net/urlid/32.1349.TB.20241028.1719.004.
[22]	LIU Z, LIN Y T, CAO Y, et al. Swin Transformer: Hierarchical vision transformer using shifted windows[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021: 10012-10022.
[23]	WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[C]// Proceedings of the European Conference on Computer Vision (ECCV). 2018: 3-19.
[24]	刘伟, 杨东风, 王洪志, 等. 基于格拉姆角场-改进残差网络的小电流接地系统故障选线[J]. 电气技术, 2023, 24(12): 14-19.
	LIU Wei, YANG Dongfeng, WANG Hongzhi, et al. Fault line selection method for small current grounding system based on gramian angular field-improved residual network[J]. Electrical Engineering, 2023, 24(12): 14-19.
[25]	CHENG X R, CUI B J, HOU S Z. Fault line selection of distribution network based on modified CEEMDAN and GoogleNet neural network[J]. IEEE Sensors Journal, 2022, 22(13): 13346-13364. doi: 10.1109/JSEN.2022.3179810
[26]	陈池瑶, 苗世洪, 殷浩然, 等. 基于注意力机制-卷积神经网络的配电网单相接地故障选线方法[J]. 电力建设, 2023, 44(4): 82-93. doi: 10.12204/j.issn.1000-7229.2023.04.010
	CHEN Chiyao, MIAO Shihong, YIN Haoran, et al. Single-phase grounding-fault line selection method based on attention mechanism-convolution neural network for distribution network[J]. Electric Power Construction, 2023, 44(4): 82-93. doi: 10.12204/j.issn.1000-7229.2023.04.010
[27]	何银, 何宇, 聂祥论. 基于CWT-CNN-LSTM的配电网单相接地故障选线方法分析[J]. 集成电路应用, 2024, 41(1): 418-421.
	HE Yin, HE Yu, NIE Xianglun. Analysis of distribution network single-phase grounding fault line selection method based on CWT-CNN-LSTM[J]. Application of IC, 2024, 41(1): 418-421.

故障参数	数值
故障线路	线路1、线路2、线路3、线路4
过渡电阻/Ω	0~2 000
故障初始角/(°)	0,60,120,150
故障位置	沿每条线路按1/8线路长度间隔设置故障点

批尺寸	准确率/%	损失值
128	94.13	0.288
64	96.45	0.153
32	97.73	0.078
16	99.01	0.039
8	98.39	0.066
4	94.71	0.278

网络模型	内存/MB	训练速度/(s·次^-1)	准确率/%
改进Swin Transformer	132	12	98.96
VGG16	508	28	89.91
VGG19	545	31	91.73
ResNet	278	19	93.12
ConvNeXt	336	21	93.44

网络模型	注意力模块嵌入	准确率/%
Swin Transformer	CSP-CBAM	98.96
Vision Transformer		97.18
ResNet		96.88
ConvNeXt		96.67
Swin Transformer	原始CBAM	95.10
Vision Transformer		94.06
ResNet		93.64
ConvNeXt		93.85

注意力模块	准确率/%
CSP-CBAM	98.96
SE	94.89
ECA	94.68
EMA	93.75