A Systematic Review of AI-Driven DC Arc Fault Detection Methods for High-Voltage Electric Vehicle Systems: Techniques, Challenges, and Future Directions

Direct current (DC) arc faults are serious safety hazards in high-voltage electric vehicle (EV) systems. Sustained high-energy discharges can cause thermal runaway and fires. Conventional detection methods often underperform in the dynamic environments of EVs. This paper reviews artificial intelligence (AI)-based detection techniques for EVs, assesses methods from photovoltaic (PV) systems, and defines deployability criteria such as inference time and hardware needs. We analyzed 72 peer-reviewed studies published between 2018 and 2025, sourced from IEEE Xplore, ScienceDirect, Web of Science, SpringerLink, and Wiley Online Library after a strict quality assessment. Hybrid AI models achieve high accuracy (97-99.99%) but face real-time deployment challenges, with inference times from 4 ms to 200 ms depending on hardware. Deep learning needs large, labeled datasets. Variable-frequency traction inverters produce electromagnetic interference, creating unique EV challenges. Key deployment barriers include sensor integration costs, limited automotive ECU computation, and a lack of standardized validation protocols. Future research should focus on explainable AI for safety certification and federated learning to address data scarcity, offering practical guidance for robust detection systems.