| Title |
Real-Time Detection of DC Series Arc Faults Using On-Device AI |
| Authors |
Wan Kim ; Minseo Jeon ; Hwa-Pyeong Park |
| DOI |
https://doi.org/10.6113/TKPE.2025.30.5.426 |
| Keywords |
DC Series Arc Fault; On-Device AI; Fault Detection; Photovoltaics (PV); DC/DC Converter |
| Abstract |
With the rapid expansion of renewable energy, especially photovoltaic (PV) systems, the occurrence rate of DC series arc faults has considerably increased, necessitating prompt and accurate detection. Although previous studies have proposed AI-based detection methods for DC arc faults, most of these approaches require external hardware to perform learning and inference, resulting in additional costs and detection delay. To overcome these limitations, this study proposes a real-time detection method for DC series arc faults that uses digital signal processor (DSP)-based on-device AI. The proposed system directly processes current signals from sensors, performs real-time feature extraction, and classifies fault conditions by using a random forest model embedded within DSP. It operates independently without reliance on external computing resources. The proposed algorithm is validated through simulation and experiments, and it achieves a classification accuracy of 99.83% and a detection speed that is compliant with the UL 1699B standard. Results demonstrate the practical applicability of the proposed method in enhancing the stability and reliability of PV systems. |