Machine Learning Based Capacity Prediction of Lithiumion Batteries Using Charging Characteristics

This study explores the critical aspect of ensuring safety and reliability in lithium-ion batteries by accurately monitoring their State of Health (SOH). Because of the intricate nature of battery degradation, this research delves into data-driven methods as viable alternatives to traditional electrochemical modelling techniques. This study compares the efficacy of ARD Regressor, GPR, and ANN methods in predicting battery capacity, with a specific focus on charging time and temperature dynamics. Among the tested methods, Artificial Neural Networks (ANN) outperformed others, demonstrating better capability in capturing battery degradation trends, achieving a root mean square error (RMSE) of less or equal to 0.01 Ahr for three batteries.