基于波形时域特征聚类法的谐振接地系统故障选线

Faulty line selection based on waveform feature clustering in time domain for resonance grounding system

利用谐振接地系统的零序网络分析单相接地后各线路零序电流波形相似性,基于此提出一种基于暂态零序电流波形时域特征聚类的谐振接地系统单相接地故障选线新方法。根据对各线路故障后首半个周期的零序电流波形的直方图分解结果,采用反映状态差异的相对熵矩阵表示零序电流波形的极性信息,结合幅值信息构造表征暂态零序电流波形时域特征的综合相对熵特征矩阵,由免阈值设定的模糊核聚类算法判别故障线路。在不同工况下对所提方法进行仿真验证,结果表明,所提方法不受电弧故障、噪声干扰、采样不同步和补偿度的影响。

The waveform similarity of line zero-sequence currents after the single-phase grounding fault of resonance grounding system is analyzed based on its zero-sequence network and a method of faulty line selection based on the waveform feature clustering in the time domain is proposed for it. The line zerosequence current waveform of the first post-fauh half-period is decomposed in histograms and the relative entropy matrix is adopted to reflect the state difference among the zero-sequence currents,as well as their polarity information. Combined with the amplitude information,an integrated relative entropy matrix is established to represent the time-domain waveform feature of transient zero-sequence currents. The fuzzy KFCM（Kernel Fuzzy C-Means） without the threshold setting is used to detect the faulty line. The proposed method is verified by the simulations under different working conditions and the results show that,it is immune to the arc fault,noise disturbance,non-synchronized sampling and compensation degree.