Wavelet packet and support vector machine analysis of series DC ARC fault detection in photovoltaic system

Abstract

AbstractIn a photovoltaic (PV) system, the serial arc is mainly due to the discontinuity in the current‐carrying conductor. Different from the AC arc, the DC arc does not have a periodic zero‐crossing and more easily blossoms into sustained arc, which is more likely to cause accidents. When a serial arc occurs in the DC system, it would lead to a steep drop in current or some unpredictable, irregular change of the current wave. The occurrence of the serial arc can be detected by analyzing the change of amplitude at different frequencies. An arc‐fault detection method based on wavelet packet (WP) and support vector machine (SVM) analysis is proposed in this paper. The threshold to distinguish arcing from normal operation is determined by analyzing the characteristics of the WP coefficients of the DC arc current collected from a real system. Compared to that of the fast Fourier transform (FFT), the effectiveness of the proposed algorithms has been validated with experiments in a 5‐kW grid‐connected PV inverter. © 2018 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.