Battery state of charge estimation using machine learning and electrochemical impedance spectroscopy measurements

Abstract

<h2>Abstract</h2> Efficient energy management in battery-powered devices requires reliable estimation of the battery state of charge. We developed a data-driven state-of-charge estimation method based on machine learning and electrochemical impedance spectroscopy. Several states-of-charge models were trained and tested using an original measurement dataset from a set of commercial Samsung ICR18650-26 J lithium-Ion batteries. The implications of the curse of dimensionality for this task have been analyzed, and the effectiveness of different feature reduction techniques to avoid classification model overfitting was investigated.