Machine Learning (ML)-Based Fault Detection Strategy for Power Switches in Electric Mobility
Document Type
Conference Proceeding
Date of Original Version
1-1-2025
Abstract
With the rapid proliferation of electric mobility, especially electric vehicles (EVs), ensuring the reliability and safety of various power electronics components, such as inverters, electrical machines, and the on-board charger (OBC), has become a critical challenge. This paper proposes a deep learning-based fault detection framework tailored for power electronic systems in EVs. By leveraging a commercial circuit simulation tool (Altair PowerSIM), fault data are generated through artificial fault injections under various load and input conditions to mimic both normal and fault states. Two deep learning models, namely a simple recurrent neural network (RNN) and a long short-term memory (LSTM) network, are trained on these time-series datasets to analyze their classification accuracy and execution speed trade-offs. To enhance diagnostic accuracy while minimizing computational load, an algorithm that incorporates instance normalization into the model design is proposed. The experimental results demonstrate that the proposed RNN-based approaches achieve significant improvements in diagnostic accuracy and response time compared to traditional rule-based and model-based methods, thereby providing a promising basis for real-time fault detection in actual electric mobility.
Publication Title, e.g., Journal
Proceedings of the 15th International 2025 IEEE Symposium on Diagnostics for Electrical Machines Power Electronics and Drives Sdemped 2025
Citation/Publisher Attribution
Kim, Youngkeun, Han Shin Youn, and Yeonho Jeong. "Machine Learning (ML)-Based Fault Detection Strategy for Power Switches in Electric Mobility." Proceedings of the 15th International 2025 IEEE Symposium on Diagnostics for Electrical Machines Power Electronics and Drives Sdemped 2025 (2025). doi: 10.1109/SDEMPED53223.2025.11154287.