Thermal Fault Diagnosis and Localization for Lithium-Ion Batteries

This work presents a framework for estimating thermal fault power and identifying compromised battery cells utilizing sparse temperature measurements. A reduced-order electro-thermal model is constructed for battery packs, requiring only three temperature sensors to characterize normal thermal dynamics, with adaptive parameter tuning to mitigate modeling uncertainties. Based on this model, a proportional-integral observer (PIO) scheme is is deployed to compute thermal anomaly power from critical temperature nodes. The methodology achieves rapid fault detection following fault initiation and subsequent localization through spatiotemporal analysis of sensor activation sequences and response time intervals. Experimental results verify robust diagnostic performance under both steady-state and transient operating conditions, accommodating variable fault onset times. Compared to traditional temperature-residual-based detection approaches, this technique exhibits faster response times and higher precision in isolating faulty cells, demonstrating enhanced diagnostic capabilities for battery pack thermal management systems.