A Lithium-Ion Battery Degradation Model Agnostic to Cell Chemistry with Integrated State-of-Charge and Temperature Dependence

Abstract

Accurately forecasting lithium-ion battery degradation is essential for safe and cost-effective electrification. This work presents a cycle-wise degradation model that estimates capacity loss based on usage conditions, using only data from a single reference degradation campaign. The model characterizes each equivalent cycle by features extracted from the battery’s State of Charge (SoC) profile; specifically, the SoC Range (SR) and Average Swing Range (ASR), and the average ambient temperature. A Similarity-Based Model maps SR and ASR to a normalized expected cycle life, which is further adjusted using a temperature correction factor derived from empirical studies. Unlike approaches requiring chemistryspecific testing, this method assumes, and validates, that cells under similar conditions degrade similarly, allowing generalization across battery types. The degradation rate also incorporates uncertainty through Kernel Density Estimation of observed cycle-to-cycle variations in supervised datasets. Validation was performed using a public lithium-ion degradation dataset, where the model predicted the State of Health (SoH) trajectory of a test cell with a Mean Absolute Error (MAE) of 0.27% of SoH percentage. Because the model uses only operational features readily measured in battery systems, it is practical for integration into battery management systems for real-time SoH tracking, predictive maintenance, and usage optimization. Future work will expand the feature set and refine uncertainty quantification to further improve predictive robustness.