Series Arc Fault Detection Method for Photovoltaic Systems Based on Spectral Cosine Distance

In case of arc fault in photovoltaic (PV) system, the fault signal is easily affected by such factors as light intensity variation and, moreover, high-frequency fault signals tend to attenuate as they propagate through cables, making series arc fault in PV systems difficult to be detected. The existing arc fault detection methods for PV systems exhibit limited study on the fault detection under high current and long distance cables, and are lack of validation of the effectiveness in real PV systems. For effectively detecting the arc faults occurring in the PV systems, the experimental platforms for arc faults in low-voltage DC (LVDC) system and PV system are firstly set up, and the arc fault data is obtained through series arc drawing experiments. Then, fast Fourier transform is used to analyze the frequency spectrum before and after the occurrence of the arc fault, and the influence of different electrode materials on the fault characteristics in the frequency domain is investigated. Based on the frequency domain characteristics of different system faults, the spectral cosine distance between time windows of arc current within the characteristic frequency band is used as the fault feature parameter, and a series arc fault detection algorithm for PV systems based on spectral cosine distance is proposed. The effectiveness of the proposed detection method is verified in PV systems, with a detection accuracy of 97.6%. This algorithm, compared with the algorithms proposed in existing studies, has a maximum detection current of 25 A, a maximum detection distance of 400 m and a high detection accuracy.