基于形态学滤波和OTSU的串联故障电弧识别方法

为有效识别不同负载串联故障电弧,针对不同类型的纯阻性负载,变频机-电机负载,工控机负载等进行故障电弧实验。对采集到的正常工作状态和电弧故障状态下的电流信号使用db4小波基对电流的一阶前向差分信号进行了5层分解,得到电流信号在32个频段的分解波形,作为故障电弧的辨识特征。通过计算同一时刻各个频段的方差,将分解的频段信号重新构成新的信号。利用形态学算法对此重构信号进行滤波,突显出故障情况下的电流特征。通过最大类间方差(OTSU)方法提取波形阈值,并统计阈值与滤波后波形的交点个数。研究结果表明,正常状态和故障电弧状态下滤波后波形与波形阈值的交点个数有明显的区别,可以作为故障电弧的识别特征。

Series fault arc recognition method based on morphological filtering and OTSU

To identify different load series fault arc effectively, this paper has carried out fault arc experiments with different types of load. These loads include pure resistive load, frequency converter-motor load, and industrial computer load. In this experiment, current signals are collected in normal working state and arc fault state. The first-order forward differential current signal is decomposed into five layers by using db4 wavelet basis. The decomposed waveform of current signal in 32 frequency bands are obtained as the identification feature of fault arc. By calculating the variance of each frequency band at the same time, the decomposed frequency band signals are reconstructed into a new signal. Morphological algorithm is used to filter the reconstructed signal to highlight the current characteristics in the case of fault. The number of intersection points is counted between the waveform threshold extracted by OTSU method and the filtered waveform. The research shows that the number of intersection points are evidently different between the normal state and the fault arc state, which can be used as the identification feature of fault arc.