| CPC H02J 7/007188 (2020.01) [B60L 53/68 (2019.02); G01R 29/24 (2013.01); G01R 31/392 (2019.01); G06N 20/00 (2019.01); H01M 10/443 (2013.01); H02J 7/0047 (2013.01); G05B 2219/14053 (2013.01)] | 30 Claims |

|
1. A computer-implemented method comprising:
determining, using one or more trained machine learning models, a first likelihood of occurrence of lithium plating of a battery pack;
comparing the first likelihood of occurrence of lithium plating with tear-down data comprising severity levels of actual occurrences of lithium plating of a fleet of battery packs;
based on the comparison, determining whether there is a threshold difference between the first likelihood of occurrence of lithium plating and the tear-down data;
responsive to determining that there is the threshold difference between the first likelihood of occurrence of lithium plating and the tear-down data, retraining the one or more trained machine learning models with the tear-down data comprising the severity levels of actual occurrences of lithium plating of the fleet of battery packs;
determining, using the one or more retrained machine learning models, a second likelihood of occurrence of lithium plating of the battery pack;
based on the second likelihood of occurrence, predicting an anode potential of the battery pack;
determining whether the anode potential satisfies a threshold condition;
responsive to determining that the anode potential satisfies the threshold condition, modifying a charging policy of the battery pack to adjust an anode potential offset; and
controlling, based on the charging policy, charging of the battery pack to adjust the anode potential offset.
|