Abstract: As arc faults have become a common cause of electrical fire accidents in photovoltaic（PV）systems,it is of great significance to investigate a reliable DC arc fault detection method to ensure the PV system operation and personal safety.In this paper,an experimental platform of DC arc faults is built in the PV system,and typical arc fault current signals are collected. Next,the stability analysis,time-varying analysis,correlation analysis and randomness analysis are carried out.Considering the global effective separation in whole arc fault burning stage, less calculation loads and stronger antiinterference requirements,Euler features are selected as the optimal arc fault characteristic. Then an arc fault detection algorithm is constructed based on the above optimal Euler feature.The algorithm takes the multi-threshold judgment system as the core rule to discover the arc fault characteristic mode different from the system transient processes. Finally,the accuracy of the proposed arc fault detection algorithm is verified under conditions of different arc generation modes and system operating points,and the result shows that the arc faults can be detected successfully in 0.5 s under the conditions of different working points and arc generation modes of the photovoltaic system.