Strategic assessment of lithium-ion battery degradation: a data-driven optimization approach

Abstract The performance and lifespan of energy storage systems, including those used for renewable energy storage and electric vehicles, are significantly hampered by the capacity degradation of lithium-ion batteries. For the best battery management, lowest associated maintenance costs, and longest battery life, an accurate estimation of this degradation is required. In this study, in order to improve performance through hyperparameter tuning, the Lichtenberg Optimization Algorithm (LOA) is used to estimate the capacity degradation of lithium-ion batteries using machine learning (ML) techniques. To choose the best, the different ML models were compared using various performance metrics. Among the models, the LOA-XGBoost with the highest values of R 2 (0.9384) and lowest values of MAE (0.0163), MSE (0.0041), RMSE (0.0640), and ME (0.8709) on testing database is the most efficient. The sensitivity analysis also demonstrated that environmental factors, such as the temperature of ambient and battery identification features like battery ID and test ID, contributed the most to degradation, while charge transfer resistance (Rct), traceable database reference (UID), and internal resistance (Re) contributed the least. This study shows how new optimization techniques, such as the Lichtenberg Optimization Algorithm, can be used to improve machine learning models and estimate battery capacity degradation more precisely.