Multicharacteristics Arc Model and Autocorrelation-Algorithm Based Arc Fault Detector for DC Microgrid

Abstract

DC microgrids contribute to the increasing penetration of renewable energies. However, a DC arc fault will lead to malfunctions and even fire hazards in a DC microgrid. To ensure the safe and efficient operation of a DC microgrid, the characteristics of DC arc faults should be determined, and a reliable arc-fault detection technique is required. Nevertheless, it is difficult to investigate the DC arc faults in a practical system. Thus, an accurate and reliable arc model for simulation research of arc faults is significant. Since a DC microgrid has numerous branches, the existing arc models cannot express the complex and random characteristics of potential arc faults. Besides, the detection points need to be determined for the appropriate installation of arc fault detectors. Furthermore, the arc features may be ambiguous and affected by the environmental noise and power electronics noise leading to nuisance tripping to arc detectors. In this paper, a multi-characteristics arc model is established based on the volt-ampere, current sag, and power spectral characteristics of arc faults. According to the frequency domain features of arc faults and interaction effects between different branches, the arc-detection-point selection principle is formed. A non-invasive arc fault detector based on magnetic-field sensing and autocorrelation algorithm is developed. It can avoid the effects of periodic environmental noise and power electronics noise by comparing the correlation between arc features in periods. The experimental results verify that the arc faults can be detected with high accuracy by installing the autocorrelation-algorithm-based arc detector at detection points selected by the arc-detection-point selection principle in a DC microgrid.