Detection of DC arc with statistical feature of high frequency current sample

In order to detect high frequency arc in direct current (DC) system effectively and accurately, this paper proposes a high frequency arc detection method based on statistical features, where features including information entropy, kurtosis, skew and their loss value are extracted from high frequency current signal for identifying arc fault and its level. Furthermore, test data under various arc faults are used to investigate the effectiveness of proposed method. Results show that the arc fault can be distinguished with good quality from current signal variation due to load change, moreover, the arc fault level expressed with arc distance can also be evaluated using proposed method. The findings can be beneficial in practical application for detecting arc fault and corresponding level, such that appropriate mitigations can be taken based on detection results.