| 引用本文: | 富 朕,王 玮,徐丙垠,等.多时间尺度随机性特征融合的低压配电网串联电弧故障检测方法[J].电力系统保护与控制,2026,54(08):47-57. |
| FU Zhen,WANG Wei,XU Bingyin,et al.Multi-timescale stochastic feature fusion method for series arc fault detection in low voltage distribution networks[J].Power System Protection and Control,2026,54(08):47-57 |
|
| 摘要: |
| 串联电弧故障检测易受负载类型影响，因此选取具有高普适性的故障特征对提升检测性能至关重要。电弧故障发生时，电流波形表现出显著的随机特性，利用该随机性特征是实现准确检测的有效途径。通过分析电弧电压的随机特性，揭示了故障电流随机性的形成机理。针对故障电流在不同时间尺度下即半波内、半波间与周波间所呈现的随机特性，分别提出用特征能量熵、奇偶谐波因数比和差分频谱幅值和进行表征，进而提出多时间尺度随机性特征融合的串联电弧故障检测方法。该方法融合了基于河马优化算法的支持向量机与连续电弧累计判断机制，建立了电弧故障综合检测模型。实验结果表明，多时间尺度随机性特征融合的检测模型利用少数特征融合即可实现对多种实验负荷下电弧故障的有效识别，准确率高于 99%，为串联型电弧故障检测提供了有益借鉴。 |
| 关键词: 串联电弧故障 随机性特征量 多时间尺度 河马优化算法 支持向量机 |
| DOI：10.19783/j.cnki.pspc.251034 |
| 分类号: |
| 基金项目:国家自然科学基金项目（52077221）；山东省自然科学基金项目（ZR2022QE100） |
|
| Multi-timescale stochastic feature fusion method for series arc fault detection in low voltage distribution networks |
|
FU Zhen, WANG Wei, XU Bingyin, SUN Zhongyu, ZOU Guofeng, HAN Haoran
|
|
1. Shandong Provincial Key Laboratory of New Power Distribution & Utilization Technology and Equipment, Shandong University of Technology, Zibo 255000, China; 2. School of Electrical and Electronic Engineering, Shandong University of Technology, Zibo 255000, China; 3. School of Electrical Engineering, Shandong University, Jinan 250061, China
|
| Abstract: |
| Series arc fault detection is highly susceptible to load types, therefore, selecting fault features with strong general applicability is crucial for improving detection performance. When a series arc fault occurs, the fault current waveform exhibits significant stochastic characteristics, and leveraging these features is an effective approach for accurate detection. This paper analyzes the stochastic behavior of arc voltage to reveal the formation mechanism of fault current randomness. Considering the stochastic characteristics of fault current across different time scales, namely within a half-cycle, between half-cycles, and between full cycles, feature energy entropy, odd-even harmonic factor ratio, and the sum of differential spectral amplitudes are proposed for characterization, respectively. Based on this, a multi-timescale stochastic feature fusion method for series arc fault detection is developed. The proposed method integrates a support vector machine optimized by the hippopotamus optimization algorithm with a continuous arc cumulative judgment mechanism, thereby establishing a comprehensive arc fault detection model. Experimental results show that the proposed model can effectively identify arc faults under various loads using a small set of fused features, achieving an accuracy exceeding 99%, and providing a valuable reference for series arc fault detection. |
| Key words: series arc fault stochastic feature multi-timescale hippopotamus optimization algorithm support vector machine |
