Low-voltage series arc fault detection based on unsupervised multi-scale dynamic collaborative network

NIU Zhili; ZOU Guofeng; WANG Wei; FU Xingle; FU Zhen

doi:10.19783/j.cnki.pspc.251260

English Version Journal Login

您当前的位置：

首页 >

文章列表页 >

Low-voltage series arc fault detection based on unsupervised multi-scale dynamic collaborative network

更新时间：2026-07-19

- Low-voltage series arc fault detection based on unsupervised multi-scale dynamic collaborative network
- Power System Protection and Control Vol. 54, Issue 14, (2026)
- 作者机构：
- 作者简介：
- 基金信息：
- DOI：10.19783/j.cnki.pspc.251260
  CLC：
- Published：2026
- 稿件说明：
移动端阅览
NIU Zhili, ZOU Guofeng, WANG Wei, et al. Low-voltage series arc fault detection based on unsupervised multi-scale dynamic collaborative network[J]. Power System Protection and Control, 2026, 54(14).
DOI：

NIU Zhili, ZOU Guofeng, WANG Wei, et al. Low-voltage series arc fault detection based on unsupervised multi-scale dynamic collaborative network[J]. Power System Protection and Control, 2026, 54(14). DOI： 10.19783/j.cnki.pspc.251260.

摘要

针对低压用电系统电弧故障检测中负载类型多样、背景噪声复杂以及故障样本稀缺等难题，提出一种基于无监督多尺度动态协同网络的串联电弧故障检测方法。该方法采用重构误差驱动的异常检测范式，无需故障样本即可实现稳定检测。首先，提出多尺度特征融合模块提取信号在不同时间尺度下的动态特征。然后，提出上下文感知门控机制，增强专家混合网络的瓶颈层设计，实现多尺度特征与专家网络之间的动态协同调度。最后，在推理阶段提出高斯位置加权的异常分数聚合策略与动态阈值决策机制，有效抑制边界波动干扰，提升检测稳定性。实验结果表明，所提方法在电弧故障检测与异常半波定位任务上的F1分数分别达到90.57%与84.74%，验证了其在复杂工况下的有效性。

Abstract

Aiming at the challenges of low-voltage series arc fault detection including diverse load types

complex background noise and scarce fault samples

an unsupervised multi-scale dynamic synergy network (MDS-Net) is proposed for series arc fault detection. The reconstruction error driven anomaly detection paradigm is adopted

which realizes stable detection without fault samples during training. Firstly

a multi-scale feature fusion module is proposed to extract dynamic features of signals under different time scales. Then

a context-aware gating mechanism is put forward to optimize the bottleneck layer of mixture of experts (MoE) network

realizing dynamic collaborative scheduling between multi-scale features and expert sub-networks. Finally

a Gaussian position-weighted anomaly score aggregation strategy and dynamic threshold decision mechanism are designed in inference stage to effectively suppress boundary fluctuation interference and improve detection stability. Experimental results show that the proposed method achieves F1 scores of 90.57% for arc fault detection and 84.74% for abnormal half-wave localization

which verifies its effectiveness under complex operating conditions.

关键词

Keywords

references

YANG Bo, ZHENG Ruyi, HAN Yiming, et al. Recent advances in fault diagnosis techniques for photovoltaic systems: a critical review[J]. Protection and Control of Modern Power Systems, 2024, 9(3): 36-59.

ZOU Guofeng, FU Guixia, HAN Bangzheng, et al. Series arc fault detection based on dual filtering feature selection and improved hierarchical clustering sensitive component selection[J]. IEEE Sensors Journal, 2023, 23(6): 6050-6060.

WANG Y, HOU L, PAUL K C, et al. ArcNet: series AC arc fault detection based on raw current and convolutional neural network[J]. IEEE Transactions on Industrial Informatics, 2022, 18(1): 77-86.

LIU H, LI J, WANG W, et al. Arc-1DCNN: an enhanced model for series arc fault detection[J]. Measurement, 2024, 234.

SHENTU Q, LI B, ZHAO K, et al. Towards a general time series anomaly detector with adaptive bottlenecks and dual adversarial decoders[C] // The Thirteenth International Conference on Learning Representations, April 24-28, 2025, Singapore, Singapore: 18810-18833.

JACOBS R A, JORDAN M I, NOWLAN S J, et al. Adaptive mixtures of local experts[J]. Neural Computation, 1991, 3(1): 79-87.

SHAZEER N, MIRHOSEINI A, MAZIARZ K, et al. Outrageously large neural networks: the sparsely-gated mixture-of-experts layer[C] // The Fifth International Conference on Learning Representations, April 24-26, 2017, Toulon, France.

NIE Y, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with transformers[C] // The Eleventh International Conference on Learning Representations, May 1-5, 2023, Kigali, Rwanda.

FEDUS W, ZOPH B, SHAZEER N. Switch transformers: scaling to trillion parameter models with simple and efficient sparsity[J]. Journal of Machine Learning Research, 2022, 23(1).

BISHOP C M. Pattern recognition and machine learning[M]. New York: Springer, 2006.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

基于同源录波数据比对的继电保护采样回路异常检测方法

基于电压特征能量的低压交流串联电弧故障检测方法

Abnormal detection of electricity theft using a deep auto-encoder Gaussian mixture model

基于改进PSO-PFCM聚类算法的电力大数据异常检测方法

基于GAN的光伏逆变器数据异常检测技术

Related Author

HAN Xiao

ZHANG Fuze

DAI Zhihui

LIANG Dong

SUN Zhongyu

ZOU Guofeng

Related Institution

LIANG Dong(1. School of Electrical and Electronic Engineering, Shandong University of Technology

School of Electrical Engineering, Shandong University

College of Electrical and Information Engineering, Hunan University

Hangzhou Haixing Electrical Co., Ltd

YU Haining(1. School of Electric Engineering and Electronic Information, Xihua University

AI问答

⁰