基于时频域瞬时特征的配电网弧光接地故障检测

Arc Grounding Fault Detection in Distribution Network Based on Transient Features in Time-frequency Domain

摘要:

配电网弧光接地故障对电力系统安全和人身安全带来很大的威胁，快速准确的弧光检测是实现故障隔离与处置、预防电气安全风险的重要技术手段之一。为实现在不同电压等级和系统配置下均具有高水平检测速度和准确度的弧光检测，提出了一种基于时频域特征的弧光接地故障检测方法。首先，通过奇异值分解与变分模态分解将零序电流分解为不同频带的单分量子信号，得到的子信号有不同的频带与中心频率，且避免了模态混叠带来的误差，有利于对不同谐波分量进行细化分析；然后，利用希尔伯特变化获得子信号的瞬时幅值、相位与频率，并将瞬时特征作为检测的主要依据，为保证在线计算时对时序信号分类的准确性，设计了基于长短期记忆（LSTM）网络的故障分类器以区分弧光接地与其他情况；最后，通过不同系统配置的实际配电网弧光接地的故障数据验证了该方法的性能。

Abstract:

Arc grounding faults in distribution network are a great threat to system safety and personal safety, and fast and accurate detection means are one of the important technologies to prevent the risk. In order to achieve a high level of detection speed and accuracy for different voltage levels and system configurations, an arc grounding fault detection method based on time-frequency domain characteristics is proposed. The zero-sequence current is decomposed into single component sub-signals of different frequency bands by singular value decomposition and variational mode decomposition, and the obtained sub-signals have different band and center frequency, and it avoids the errors caused by mode mixing, which ensures the refine analysis of different harmonic components. And then, the instantaneous amplitude, phase and frequency of sub-signals are obtained by using Hilbert transform, and the instantaneous characteristics are taken as the main basis for detection. In order to ensure the classification accuracy of sequence signals during online calculation, a fault classifier based on long short-term memory (LSTM) network is designed to distinguish arc grounding from other cases. Finally, the performance of the proposed method is verified by the fault data of arc grounding in an actual distribution network.