基于多源特征融合去噪网络的配电网故障辨识方法研究

doi:10.3969/j.issn.2097-0706.2026.02.003

摘要/Abstract

摘要：

针对配电网故障辨识过程中故障特征易被噪声掩盖，单一特征难以全面反映故障信息等问题，提出了一种多源特征融合去噪网络的配电网故障辨识方法。利用离散小波变换（DWT）对归一化处理后的零序电压与三相电流信号进行多尺度分解，将原始信号重构为高频与低频分量，减少特征混叠的影响。然后，在高频分支中构建时频残差自适应去噪模块，嵌入傅里叶卷积，提取具有丰富频域信息的高频特征，并结合通道注意力机制，实现信号的精准去噪；在低频分支中，设计轻量级卷积网络提取低频分量的时序特征，增强时频信息互补。引入一种时序-通道注意力机制进行自适应特征融合，增强特征提取能力的同时抑制冗余特征，从而实现复杂工况下故障类型的精确诊断。试验结果表明，所提方法在仿真数据集下取得99.12%的准确率，相较于现有方法显著提高了故障辨识的准确率。

关键词: 配电网故障辨识, 多源特征融合, 傅里叶卷积, 自适应去噪, 时序-通道注意力机制

Abstract:

To address the issues that fault features are easily obscured by noise and single features can hardly capture comprehensive fault information during distribution network fault identification，a distribution network fault identification method based on multi-source feature fusion denoising network was proposed. The discrete wavelet transform （DWT）was used to decompose the normalized zero-sequence voltage and three-phase current signals at multiple scales，and the original signals were reconstructed into high-frequency and low-frequency components to mitigate the influence of feature aliasing. In the high-frequency component，the time-frequency residual adaptive denoising module was constructed，and Fourier convolution was embedded to extract high-frequency features with rich frequency-domain information. Combined with the channel attention mechanism，the accurate denoising of the signal was achieved. In the low-frequency branch，a lightweight convolutional network was designed to extract the temporal features of the low-frequency components to enhance the complementarity of the time-frequency information. A temporal-channel attention mechanism was introduced for the adaptive feature fusion，which enhanced the feature extraction capability while suppressing the redundant features，thus enabling the accurate diagnosis of the fault types under the complex working conditions. The experimental results showed that the proposed method achieved an accuracy of 99.12% in the simulation dataset，which significantly improved the accuracy of fault identification compared with the existing methods.

Key words: distribution network fault identification, multi-source feature fusion, Fourier convolution, adaptive denoising, temporal-channel attention mechanism

中图分类号:

TK01

崔向虎, 徐越飞, 戚佳金, 张静, 陈世喆, 李金诺. 基于多源特征融合去噪网络的配电网故障辨识方法研究[J]. 综合智慧能源, 2026, 48(2): 27-36.

CUI Xianghu, XU Yuefei, QI Jiajin, ZHANG Jing, CHEN Shizhe, LI Jinnuo. Research on distribution network fault identification method based on multi-source feature fusion denoising network[J]. Integrated Intelligent Energy, 2026, 48(2): 27-36.

图/表 16

图1

图2

图3

图4

表1

图5

表2

表3

图6

图7

图8

图9

表4

表5

图10

图11

参考文献 34

[1]	李泽文, 葛俊辰, 陈思旭, 等. 基于VMD-MWPE和PNN的配电网故障分类辨识方法[J]. 江西电力, 2025, 49(3): 28-37.
	LI Zewen, GE Junchen, CHEN Sixu, et al. Fault classification and identification method of distribution network based on VMD-MWPE and PNN[J]. Jiangxi Electric Power, 2025, 49(3): 28-37.
[2]	王建, 吴昊, 张博, 等. 不平衡样本下基于迁移学习- AlexNet的输电线路故障辨识方法[J]. 电力系统自动化, 2022, 46(22): 182-191.
	WANG Jian, WU Hao, ZHANG Bo, et al. Fault identification method for transmission line based on transfer learning-AlexNet with imbalanced samples[J]. Automation of Electric Power Systems, 2022, 46(22): 182-191.
[3]	WANG J, ZHANG B, YIN D, et al. Distribution network fault comprehensive identification method based on voltage-ampere curves and deep ensemble learning[J]. International Journal of Electrical Power & Energy Systems, 2025, 164: 110403. doi: 10.1016/j.ijepes.2024.110403
[4]	RIZEAKOS V, BACHOUMIS A, ANDRIOPOULOS N, et al. Deep learning-based application for fault location identification and type classification in active distribution grids[J]. Applied Energy, 2023, 338: 120932. doi: 10.1016/j.apenergy.2023.120932
[5]	李靖鑫, 李旭锋, 杨金东, 等. 非线性负载配电网异常时频电流行波辨识方法[J]. 电子设计工程, 2024, 32(24): 143-146, 151.
	LI Jingxin, LI Xufeng, YANG Jindong, et al. Nonlinear load distribution network abnormal time-frequency current wave identification method[J]. Electronic Design Engineering, 2024, 32(24): 143-146, 151.
[6]	庄小艇, 孙蒙恩, 赖天德, 等. 基于负荷波动特征量提取和稳健估计的低压配电网回路阻抗辨识方法[J]. 供用电, 2025, 42(2): 29-37.
	ZHUANG Xiaoting, SUN Meng'en, LAI Tiande, et al. A low voltage distribution network loop impedance identification method based on load fluctuation feature extraction and robust estimation[J]. Distribution & Utilization, 2025, 42(2): 29-37.
[7]	GUO M F, YANG N C, YOU L X. Wavelet-transform based early detection method for short-circuit faults in power distribution networks[J]. International Journal of Electrical Power & Energy Systems, 2018, 99: 706-721. doi: 10.1016/j.ijepes.2018.01.013
[8]	JI L P, TIAN X L, WEI Z H, et al. Intelligent fault diagnosis in power distribution networks using LSTM-DenseNet network[J]. Electric Power Systems Research, 2025, 239: 111202. doi: 10.1016/j.epsr.2024.111202
[9]	SAAD M, KIM C H, MUNIR N. Single-phase auto-reclosing scheme using particle filter and convolutional neural network[J]. IEEE Transactions on Power Delivery, 2022, 37(6): 4775-4785. doi: 10.1109/TPWRD.2022.3159256
[10]	叶远波, 王吉文, 邵庆祝, 等. 配电网故障识别Transformer-联邦迁移学习算法设计[J]. 电力系统及其自动化学报, 2025, 37(10): 120-128.
	YE Yuanbo, WANG Jiwen, SHAO Qingzhu, et al. Design of Transformer-based federated transfer learning algorithm for distribution network fault recognition[J]. Proceedings of the CSU-EPSA, 2025, 37(10): 120-128.
[11]	梁栋, 赵月梓, 贺国润, 等. 基于图半监督与多任务学习的配电网故障区段与类型统一辨识[J]. 电力系统保护与控制, 2024, 52(12): 25-32.
	LIANG Dong, ZHAO Yuezi, HE Guorun, et al. Unified identification of fault section and type for distribution networks based on graph semi-supervised and multi-task learning[J]. Power System Protection and Control, 2024, 52(12): 25-32.
[12]	任一鸣, 杜董生, 邓祥帅, 等. 基于GRU和GWO-KELM的电力线路故障诊断[J]. 综合智慧能源, 2024, 46(3): 54-62. doi: 10.3969/j.issn.2097-0706.2024.03.007
	REN Yiming, DU Dongsheng, DENG Xiangshuai, et al. Power line fault diagnosis based on GRU and GWO-KELM[J]. Integrated Intelligent Energy, 2024, 46(3): 54-62. doi: 10.3969/j.issn.2097-0706.2024.03.007
[13]	陈昊蓝, 靳冰莹, 刘亚东, 等. 基于门控循环注意力网络的配电网故障识别方法[J]. 上海交通大学学报, 2024, 58(3): 295-303. doi: 10.16183/j.cnki.jsjtu.2022.091
	CHEN Haolan, JIN Bingying, LIU Yadong, et al. Fault detection in power distribution systems based on gated recurrent attention network[J]. Journal of Shanghai Jiao Tong University, 2024, 58(3): 295-303.
[14]	IBRAHIM A H M, SADANANDAN S K, GHAOUD T, et al. Incipient fault detection in power distribution networks: Review, analysis, challenges and future directions[J]. IEEE Access, 2024, 12: 112822-112838. doi: 10.1109/ACCESS.2024.3443252
[15]	SHALBY E M, ABDELAZIZ A Y, AHMED E S, et al. A comprehensive guide to selecting suitable wavelet decomposition level and functions in discrete wavelet transform for fault detection in distribution networks[J]. Scientific Reports, 2025, 15(1): 1160. doi: 10.1038/s41598-024-82025-2
[16]	FAHIM S R, SARKER S K, MUYEEN S M, et al. A deep learning based intelligent approach in detection and classification of transmission line faults[J]. International Journal of Electrical Power & Energy Systems, 2021, 133: 107102. doi: 10.1016/j.ijepes.2021.107102
[17]	HAO S, FAN S A, MA X, et al. MKDNet: A fault detection method based on multi-scale entropy mode denoising and collaborative knowledge distillation for distribution networks[J]. Electric Power Systems Research, 2025, 249: 112003. doi: 10.1016/j.epsr.2025.112003
[18]	卢晓强, 李刚, 周万竣, 等. 基于最小二乘双支持向量机的配电网短路故障辨识方法研究[J]. 武汉大学学报(工学版), 2022, 55(4): 401-408.
	LU Xiaoqiang, LI Gang, ZHOU Wanjun, et al. Research on identification method of short-circuit fault in distribution network based on least squares twin support vector machine[J]. Engineering Journal of Wuhan University, 2022, 55(4): 401-408.
[19]	李政洋, 曹一家, 陈春, 等. 配电网相电压不对称动态变化下的高阻接地故障辨识[J]. 中国电机工程学报, 2024, 44(22): 8744-8759.
	LI Zhengyang, CAO Yijia, CHEN Chun, et al. Identification of high-impedance ground faults under dynamic variations of voltage unbalance in distribution networks[J]. Proceedings of the CSEE, 2024, 44(22): 8744-8759.
[20]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 7132-7141.
[21]	冯骥, 杨国华, 史磊, 等. 基于并行融合深度残差收缩网络的有源配电网故障诊断[J]. 综合智慧能源, 2024, 46(6): 8-15. doi: 10.3969/j.issn.2097-0706.2024.06.002
	FENG Ji, YANG Guohua, SHI Lei, et al. Research on fault diagnosis of active distribution network based on parallel fusion deep residual shrinkage network[J]. Integrated Intelligent Energy, 2024, 46(6): 8-15. doi: 10.3969/j.issn.2097-0706.2024.06.002
[22]	刘宝稳, 崔露瑶, 曾祥君, 等. 配电网两相接地零序电压空间轨迹区域特征与故障相识别[J/OL]. 电网技术, 2025: 1-15(2025-08-04) [2025-09-22]. https://kns.cnki.net/KCMS/detail/detail.aspx?filename=DWJS20250801007&dbname=CJFD&dbcode=CJFQ.
	LIU Baowen, CUI Luyao, ZENG Xiangjun, et al. Spatial trajectory characteristics of zero-sequence voltage for two-phase grounding fault and fault phase identification in distribution networks[J/OL]. Power System Technology, 2025: 1-15(2025-08-04) [2025-09-22]. https://kns.cnki.net/KCMS/detail/detail.aspx?filename=DWJS20250801007&dbname=CJFD&dbcode=CJFQ.
[23]	缪欣, 黄海悦. 基于Δi-PCA原理的分布式小电流接地选线保护方法研究[J]. 机电信息, 2023(4): 76-81.
	MIAO Xin, HUANG Haiyue. Research on distributed protection method of small current grounding line selection based on Δi-PCA principle[J]. Mechanical and Electrical Information, 2023(4): 76-81.
[24]	甘捷, 姚孝庭, 于景国, 等. 变频器10 kV断路器柜速断保护动作故障分析及处理[J]. 东北电力技术, 2021, 42(1): 41-43.
	GAN Jie, YAO Xiaoting, YU Jingguo, et al. Analysis and treatment of Quick-B reak protection fault for 10 kV switchgear in frequency converter[J]. Northeast Electric Power Technology, 2021, 42(1): 41-43.
[25]	CHI L, JIANG B, MU Y. Fast Fourier convolution[J]. Advances in Neural Information Processing Systems, 2020, 33: 4479-4488.
[26]	ATTAR M S, MIVEH M R. High‐impedance fault detection in distribution networks based on support vector machine and wavelet transform approach (Case study: Markazi Province of Iran)[J]. Energy Science & Engineering, 2025, 13(3): 1171-1183. doi: 10.1002/ese3.v13.3
[27]	张灵芝, 王杨帆, 刘基典, 等. 电网故障行波传感器研究及测试[J]. 电力科学与技术学报, 2020, 35(2): 142-149.
	ZHANG Lingzhi, WANG Yangfan, LIU Jidian, et al. Research and experiments on the transmission characteristics of traveling wave sensor for power fault[J]. Journal of Electric Power Science and Technology, 2020, 35(2): 142-149.
[28]	SIDDIQUE M N I, SHAFIULLAH M, MEKHILEF S, et al. Fault classification and location of a PMU-equipped active distribution network using deep convolution neural network(CNN)[J]. Electric Power Systems Research, 2024, 229: 110178. doi: 10.1016/j.epsr.2024.110178
[29]	ZHAO M H, ZHONG S, FU X, et al. Deep residual shrinkage networks for fault diagnosis[J]. IEEE Transactions on Industrial Informatics, 2019, 16(7): 4681-4690. doi: 10.1109/TII.9424
[30]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017, 30: 03762.
[31]	LI T F, ZHAO Z B, SUN C, et al. WaveletKernelNet: An interpretable deep neural network for industrial intelligent diagnosis[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2022, 52(4): 2302-2312. doi: 10.1109/TSMC.2020.3048950
[32]	WANG Q L, WU B G, ZHU P F, et al. ECA-Net: Efficient channel attention for deep convolutional neural networks[C]// Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020: 11534-11542.
[33]	ZHOU B L, KHOSLA A, LAPEDRIZA A, et al. Learning deep features for discriminative localization[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 2921-2929.
[34]	PRATT H, WILLIAMS B, COENEN F, et al. FCNN: Fourier convolutional neural networks[C]// Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer International Publishing, 2017: 786-798.

网络结构	激活函数	输出形状
线性层 1		128
线型层 2		32
线性层 3	Softmax	10

标签	故障类型	时间步长	故障阻抗/Ω	初始相位角/(°)
AG	A相接地	1 024	0.01，50，100，200，300，500，1 000	0，30，60，90，120，150，180
BG	B相接地
CG	C相接地
AB	AB相短路
AC	AC相短路
BC	BC相短路
ABG	AB相接地
ACG	AC相接地
BCG	BC相接地
ABC	ABC相短路

方案	DWT	注意力机制	准确率/%
1	√	TCA	99.12
2		TCA	96.35
3	√		95.55
4	√	SE	97.20
5	√	ECA	98.15
6	√	CAM	97.36

方法	准确率	召回率
DWT	82.40	81.73
SVM	81.69	81.54
1D-CNN	88.71	88.04
DRSN-CW	96.21	96.10
Trasformer	94.56	94.35
MexHat-WKN	92.48	92.03
本文方法	99.12	98.83