DC Series Arc Detection Algorithm Using FFT-Based Frequency Band Analysis and RMS Current Variations

With the increasing penetration of renewable energy, the detection of DC series arcs has become increasingly critical in DC power systems. This paper proposes an arc detection algorithm that uses frequency-band variations of current signals. Current data sampled at 100 [kHz] are processed using a sliding FFT (Fast Fourier Transform) and divided into seven bands spanning 0–50 [kHz]. The RMS current for each band is computed per frame, the absolute frame-to-frame difference is calculated, and the result is normalized using the Z-score method. An arc is declared if the metric exceeds a threshold within a detection window. DC series arc tests conducted under four load conditions and eight electrode opening speeds show pronounced changes in the 0.2–1 [kHz] and 1–5 [kHz] bands. The 1–5 [kHz] band, being less dependent on load yet sensitive to arcs, is selected for the algorithm. The proposed method detected arcs within 0.15 seconds across all tested conditions.