Researchers at the Massachusetts Institute of Technology (MIT) and TU Darmstadt have developed an advanced method using machine learning for early detection of faults in electric vehicle (EV) battery systems, significantly improving safety and reliability.

The research team’s approach enables real-time monitoring and analysis of battery cell performance, addressing a critical issue in lithium-ion battery technology: preventing potentially catastrophic failures.

“A single faulty battery cell can lead to serious issues, such as fire,” explain researchers from TU Darmstadt's Institute of Automation and Mechatronics. This new method allows for continuous online monitoring, marking a revolutionary improvement in battery safety for EVs and energy storage systems.

The study utilized a unique dataset from TU Darmstadt, featuring 28 battery systems recalled by manufacturers due to defects. This dataset, containing over 133 million data points from 224 battery cells, offers invaluable insights into battery degradation and failure mechanisms.

The research findings revealed that often just one faulty cell can compromise the entire system's operation and safety. The machine learning model accurately detects temporal and operational changes in battery cells, enabling timely intervention.