Spacecraft Anomaly Detection: Machine Learning Based Detection of Lithium-Ion Battery Degradation in Space Conditions

As space travel becomes increasingly complex and sought after with the prospects brought about by international and national space missions such as NASA’s Artemis II and Europa missions, monitoring battery safety and health in spacecraft has become even more critical. Under space conditions and stresses, electrical systems and components such as batteries face exposure to high energy particle radiation, thermal fluctuations, and operational autonomy in remote environments. The industry standard for satellite, probe, and rover batteries has been favorable in regards to Lithium-ion batteries (LiBs), which, despite their high energy density, long life cycle, and wide operating temperature range, are still vulnerable to solid electrolyte interphase (SEI) degradation, capacity fade, thermal runaway, and impedance shifts caused by these harsh conditions, significantly impacting mission success. Current spacecraft battery monitoring methods rely heavily on human oversight and telemetry data, resulting in delays or inaccuracies. This study aims to address this limitation by employing machine learning (ML) methods, such as linear (LR) and random forest (RF) regression. Utilizing the nascent PyBaMM library to artificially synthesize LiB radiation and thermal data, the ML model will be trained on labeled data to improve anomaly detection accuracy and reduce false positives in battery systems monitoring, offering future potential for real-time autonomous responses to battery health deterioration in space without human intervention.