Series Arc Fault Identification for Photovoltaic System Based on Time-Domain and Time-Frequency-Domain Analysis

This paper aims at providing a reliable algorithm to identify photovoltaic (PV) series arc faults regardless of complex fault-like interferences. Through conducting various arc fault experiments with different PV current levels, arc gap lengths, and load types, PV series arc fault features have been understood comprehensively. To avoid unwanted nuisance tripping, fault-like conditions are analyzed to confirm the unique arc fault features. Based on the loop current signature, a greater unstable fluctuation in the time domain and extra arc noises in the time-frequency domain are chosen as identification features. By quantificational evaluations, optimal detection variables with the Hamming window and the proper time resolution have been established to achieve the best identification results. By building fusion coefficients, two variables are arithmetically fused to achieve the arc fault discovery. The algorithm could also classify fault-like into normal and adjust the threshold value dynamically to fit different normal current levels. Its validity has been verified by experimental results on the simulated platform.