State of health (SOH) estimation in lithium-ion batteries remains challenging in electric vehicles due to the interplay of multiple aging mechanisms. This work presents a DVA-based capacity estimation method for Nickel–Manganese–Cobalt (NMC) cylindrical cells, designed for integration into Battery Management Systems (BMS) to support predictive maintenance. The method estimates cell capacity by tracking the peak position in the high capacity region of the differential voltage (DV) curve during charging. This peak is extracted as a key health indicator and used to construct an aging model. The method is evaluated first using laboratory datasets covering a wide range of cycling and storage conditions including variations in temperature, state of charge, charging rates, and depth of discharge. A comparison with Incremental Capacity Analysis (ICA) shows that DVA yields more stable aging features and achieves lower RMSE values. An optimal charging rate of 0.3C is identified, balancing diagnostic sensitivity with compatibility for standard charging protocols.

The parameterized DVA algorithm is further validated using real world cell data, demonstrating its robustness across diverse operating profiles. Importantly, the results show that this DVA based approach provides a practical pathway for transitioning laboratory diagnostic techniques to on-board BMS implementation. By enabling reliable tracking of the DVA peak in the high capacity region, the method supports predictive maintenance strategies and enhances SOH estimation for electric vehicles.