Arc faults in the power distribution system of an electric vehicle (EV) can result in damage to cables and associated equipment, as well as threaten the safety of the occupants of the EV. On-board components can create differential background noise, which limits the reliability of current arc detection methods. Within this paper an accurate and real-time arc fault detection method is designed and verified. An arc fault data pipeline is designed, trained and validated by using experimental data of damaged EV cables. Weak-current signals are pre-processed through a two-stage filter, and then the number of continuous over-threshold windows in the wavelet transform result are collected to determine the occurrence of the series arc. Our results, also demonstrate an immunity to background noise with a 150ms detection time and 99% accuracy.

Read the publication here: A Method for DC Arc Fault Detection, Classification and Mitigation in Electric Vehicles