In one aspect, a computer-implemented method may include receiving, at a cloud-based computing system, historical data comprising (i) battery charging profiles associated with a fleet of vehicles comprising battery packs, (ii) geographical climate information associated with the fleet of vehicles, (iii) lithium plating prediction results associated with the fleet, or (iv) some combination thereof. The method includes training, using the historical data, lithium plating prediction models to predict a likelihood of occurrence of lithium plating for a battery pack, receiving, from an edge processor communicatively coupled to a battery pack, battery pack charging data, determining, using the models based on the battery pack charging data, the likelihood of occurrence, based on the likelihood of occurrence, modifying a battery charging policy for the battery pack, and transmitting the battery charging policy to cause the edge processor to control charging of the battery pack.