Serial-arc detection by use of Spectral Dispersion Index (SDI) analysis in a low-voltage network (270V HVDC)

•This series arc fault detection algorithm is performed by the current frequency analysis based on the calculation of the “Spectral Dispersion Index”.•The method is tested with resistive, inductive and switching appliances.•We propose a process of dynamic thresholding when calculating the spectrogram.•The decision logic is based on a counting up and counting down of the anomaly indicator.•We study and detect series Arc faults in DC - 270 Volts power networks.•The good results of detection are presented throughout a matrix of confusion. In this paper we will present a method for detecting series-arc faults, based on a frequency analysis that uses a decision block. We will define the parameters of a Spectral Dispersion Index (SDI) that has been calculated using data taken from a binarized spectrogram of the line current (provided by employing a dynamic-thresholding process). The decision block uses a counter mechanism that increased or decreased in value at a quicker or slower rate, depending on whether or not a fault was present in the power line. The status of its operating mode was recorded using a finite-state machine. In this method of detection, a trip indicator is activated whenever a fault is present. The method was tested on a large variety of loads, including resistive, inductive and switching loads. The confusion matrix we obtained, showed that series-arc faults can be successfully detected in this way. We will therefore be able to minimize the number of false detections that often occur in the case of switching loads or voltage generators.