A lean statistical cell-to-pack framework for sustainable battery pack capacity estimation

Accurate battery pack capacity estimation is vital for electric vehicle and energy storage safety, but conventional pack-level aging tests are costly, energy intensive, and wasteful. This study proposes a lean statistical cell-to-pack (LS-CTP) framework that eliminates pack-level tests. Using aging data from only a few independent test cells, it combines lightweight machine learning and decision algorithms to achieve multi-feature statistical transformation and fusion from routine fragmented charge data. It requires few hyperparameters and low computational overhead and applies to parallel, series, and mixed configurations. Monte Carlo analyses show a mean absolute percentage error below 1.24% for parallel packs and 1.92% for series packs using as few as two test cells. A sustainability assessment indicates over 90% reduction in battery waste and testing energy. By decoupling algorithm development from pack-level cycling, LS-CTP advances cost-effective, energy-efficient, and sustainable battery health diagnostics, supporting the clean energy transition.