Cell-level versus module-level degradation of Li-ion batteries for grid-scale storage: A four-year parallel aging experiment and empirical capacity-fade analysis

Lithium-ion batteries dominate grid-scale storage, yet the link between single-cell and module-level performance remains unvalidated. We present a four-year experimental comparison of commercial NMC811/graphite 18650 cells and a 7.6 kWh (48P14S) module aged in parallel under standardized U.S. Department of Energy duty cycles: thirteen cycles spanning peak shaving, frequency regulation, electric vehicle, and calendar aging for cells, with peak shaving applied to the module. Among three empirical capacity-fade models, cell power-law exponents of 0.25–0.46 indicated diffusion-limited SEI growth. A single power law fit the full module dataset with a markedly higher exponent (0.73), but this reflects late-stage acceleration beyond ≈1400 days. Truncated to that window, the module exponent falls to 0.47, and a composite power-law-plus-knee model—decisively preferred by information criteria—recovers cell-like kinetics (n = 0.44) with an incipient knee, implicating current redistribution and thermal non-uniformity. Differential voltage analysis confirmed loss of lithium inventory as the dominant mechanism at both scales. Cell-calibrated expressions track early module degradation but underestimate it as a module-level knee develops, requiring scale-aware lifetime corrections.