摘要
低压交流串联电弧故障是目前研究的热点,非典型负载的特殊性更增加了电弧故障辨识的难度。针对常用负载的低压交流串联故障电弧,以及点接触电弧试验的并联交流故障电弧和调光灯调光状态等非典型负载情况,提出一种基于BP神经网络与电流特征提取组合的故障电弧辨识方法,采用基于小波变换细节分量的BP神经网络辨识串联电弧故障。对于非典型负载情况,在BP神经网络辨识的基础上,通过引入负载电流时域信号的半周期积分变化量,辅助解决调光灯调光状态易产生的误动作;结合负载电流时域信号的变化率与"平肩部"特性,辨识点接触引起的并联电弧故障。实测结果表明,所提出的辨识方法在常用负载的串联电弧故障与非典型负载波形测试中均辨识准确。
Low-voltage AC series arc fault is a hot research topic and particularities of atypical load makes it even more difficult to identify arc faults.Aiming at atypical loads such as low-voltage AC series fault arcs of common load,parallel AC fault arcs in the point contact arc test as well as dimming state of dimming lights,a fault arc identification method was proposed,based on the combination of BP neural network and current feature extraction.According to this scheme,series arc faults were identified by the BP neural network based on wavelet transform detail components.In the case of atypical load,on the basis of identification through BP neural network,the half-cycle integral variation of the time domain signal of the load current was introduced to help solve the common problem of wrong operation of dimming state.Parallel arc faults due to point contact were identified through change rate and“flat shoulder”feature of the load current time domain signal.Measured results showed that the proposed approach could accurately recognize series arc faults of common load and achieve accurate identification in the waveform test of atypical loads.
作者
陆凯峰
张峰
张士文
汪洋堃
Lu Kaifeng;Zhang Feng;Zhang Shiwen;Wang Yangkun(College of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《电气自动化》
2020年第3期45-48,共4页
Electrical Automation
关键词
故障电弧
小波变换
细节信号
神经网络
时域特征量
fault arc
wavelet transformation
detail signal
neural network
time domain characteristic quantity
