Series arc fault detection method based on Wavelet energy spectrum entropy

The existing series arc fault detection methods mostly construct the detection criterion by extracting the fault characteristics of current. Due to the diversity of load currents, it is difficult for current detection methods to construct detection criteria with a single characteristic. In contrast, the normal voltage on the load side is not sensitive to the types of loads, thus the arc fault information has a more obvious characterization on the load side voltage. In view of this reason, based on the analysis of basic time-frequency domain characteristics of arc faults, a method for arc fault detection is proposed based on wavelet energy spectrum entropy of the voltage on the load side whose purpose is to overcome the difficulty of selecting fault characteristic frequency band under different loads, and the optimal selection of wavelet decomposition layers and wavelet basis functions are also achieved through analysis and experiments. The experiment and comparison of algorithms show that the fault detection accuracy of the proposed method is more than 98% under different loads. The anti-disturbance ability of the method for load switching is verified by load switching experiment, which verifies the effectiveness of the method.