Detection of Series Arc Faults Based on Time-Frequency Domain Disturbances

In contemporary low-voltage AC distribution systems, series arc fault currents are significantly lower than overload currents and short-circuit currents, rendering circuit breakers ineffective in their detection. Consequently, such faults can persist in the circuit for extended periods and potentially leading to electrical fires. This paper extracts the characteristic time-frequency domain disturbances in line current induced by series arc faults through various load experiments conducted on an experimental platform. Utilizing an improved CUSUM algorithm, disturbances in the current data are identified, and the time points of these perturbations are precisely located based on the standard deviation analysis. Subsequently, series arc fault detection is achieved by examining changes in even-order harmonics before and after the identified time points. The proposed method exhibits lower requirements for data sampling frequency and ensures a response time of within 50ms for arc fault detection, thereby effectively addressing series arc fault detection in low-voltage AC distribution systems.