Causal Evaluation of Early-Stage Fast-Charging Protocols Under Cell-to-Cell Variability

초록

Managing the tradeoff between charging rate and degradation in lithium-ion batteries is critical for reliable fast-charging operations. Even in controlled experiments, cells can differ in their early electrochemical and thermal conditions, and these differences may be unevenly distributed across charging-protocol groups. As a result, a protocol may appear favorable partly because it was tested on cells with more favorable initial conditions. In addition, the protocol that performs best on average may not be the best for cells with particular early-cycle states. We formulate the selection of charging protocols as an offline causal decision problem and propose a two-stage framework. First, we use doubly robust estimators to compare protocols after adjusting for cell-to-cell variability. Second, we examine how the preferred protocol changes across early-cycle cell states to identify when faster charging is favorable. After adjustment, protocols that begin conservatively and increase the charging rate later show the lowest expected degradation. This pattern is largely consistent across estimators, alternative degradation metrics, and sensitivity analyses. We further find that acceleration in the earliest charging phase is beneficial only for cells with favorable electrochemical signatures and mild thermal conditions, whereas later early-phase acceleration is mainly favored under mild thermal conditions. These findings provide an offline framework for selecting among predefined fast-charging protocols using early-cycle electrochemical and thermal information.

키워드