Active equalization for lithium-ion battery pack via data-driven residual charging capacity estimation

Considering the limitations in existing voltage-based and state-of-charge (SOC)-based active equalization strategies, including the difficulty in threshold value determination for equalization system on/off controlling, repeated estimation for equalization variable and the corresponding tremendous complexity, this paper designs a novel residual capacity-based active equalization strategy via data-driven residual charging capacity (RCC) estimation to fully charge all in-pack cells during constant current (CC) charging stage. Firstly, taking charge accumulation corresponding to specific voltage window as input, we build a data-driven RCC estimation model to online monitor each cell's RCC at specific voltage. Afterwards, considering the initial cell inconsistences, each cell's current RCC before equalization is further computed based on the estimated RCC and the subsequently measured charge accumulation. Finally, a parallel global search algorithm named particle swarm optimization is adopted to search the optimal combination of in-pack cells' equalization current considering equalization time and energy loss simultaneously. The verification results based on the Oxford battery degradation dataset demonstrate that the established data-driven model can realize accurate RCC estimation and all in-pack cells' totally charged capacity can roughly approach their maximum capacity at the end of CC charging stage using the proposed equalization strategy. Meanwhile, the maximum cell voltage difference and maximum cell SOC difference can be limited below 0.01V and 0.02, respectively. Moreover, by extending the charging time by only about 4min, the developed equalization strategy can further increase battery pack capacity by about 10%.