An Experimental Study of Low-Current DC Series Arc Faults for Condition Monitoring Purpose

Abstract

DC arc faults present a challenging protection problem in DC systems, such as photovoltaic and electric vehicle supply systems. Incidents of DC arc faults in DC systems are becoming more common especially for series arc faults, and if undetected, would finally cause severe damage to the systems. This paper studies DC series arc fault characteristics including quasi-stationary V-I characteristic and arc current spectrum characteristic. Based on experimental tests, the impacts of different load current levels, DC operating voltage levels, air gap lengths on arc current and its spectrum are investigated and evaluated. A new method based on wavelet packet decomposition combined with entropy theory has been developed to extract the common features of arc signals under different conditions. It is found that arc fault currents share some common wavelet-packet entropy in different bandwidths even if the fault conditions are different. The results provide better characterization of DC arcing phenomena and help to develop more effective detection algorithms.