A dataset consisting of 90 lithium-ion cells obtained from old notebook batteries containing their response to 100 charge–discharge cycles is presented. The resulting degradation patterns are assigned to four clusters and related to possible aging mechanisms. The records in the battery management system (BMS) of each battery are analyzed to understand the influence of first life conditions in the measured degradation patterns. The analysis reveals that a cluster of cells which experienced

more »

... calendar aging in 7–13 years hold ~90% of the rated capacity, and exhibit at 0.4 C discharge a linear capacity degradation throughout cycling comparable to new cells. In contrast, a cluster of cells that experienced extensive calendar and cyclic aging can lose ~50% capacity at 0.4 C discharge in a few cycles after reutilization. A model based on a boosted decision tree is applied to forecast the cluster of each cell, using as features the capacity measured in the first cycle, and the records obtained from the BMS. The highest accuracy (83%) is obtained through capacity, where misclassification arises from two clusters containing highly degraded cells with similar initial capacities, but divergent degradation patterns.