Health status estimation for lithium-ion batteries with partial charging information using mixed inputs LSTM

Determining the health status of batteries on a personalized level poses challenges given the variety in usage patterns, dynamic charging protocols, and the scarcity of historical data. This study introduces a mixed-input LSTM network that synergistically combines partial charging history with operational conditions. Our experiments span diverse working profiles, temperatures, and charging protocols to evaluate the methodology based on a few RPT results rigorously. For NMC532/graphite batteries, utilizing features from the voltage range of 3.65V to 4.1V, we achieve an RMSE and MAE of 1.54% and 1.18%, respectively. This research highlights the potential of data-driven approaches for monitoring battery health throughout its entire life cycle.