Early-stage distributed thermal fault detection and localization are essential for battery safety but have been relatively underexplored compared to state-of-charge and state-of-health estimation. To address this, a backstepping-based fault localization filter is proposed. Traditional fault localization requires full-state temperature measurements, which are impractical in real applications. Using an interpolation-based approximation, the number of required sensors is reduced to a few, enhancing the fault localization filter’s practical applicability. A structured framework, including fault localization filter design, distributed residual evaluation, and threshold determination, ensures robust and reliable real-time performance. The effectiveness of the approach is demonstrated through theoretical analysis and experimental validation.