The increasing global demand for sustainable and scalable energy storage solutions has underscored the critical role of lithium-ion batteries (Li-Ion) in applications such as electric vehicles (EVs) and renewable energy systems. To support their widespread deployment, advancements in both manufacturing processes and battery management systems (BMS) are essential. A key challenge in BMS is the accurate estimation of the State of Charge (SoC), which is crucial for ensuring battery reliability, safety, and efficiency. Equivalent Circuit Models (ECMs), though widely used due to their simplicity, rely on static parameters and struggle to adapt to varying operational conditions such as temperature and battery aging. Data-driven methods based on machine learning are promising, but they often require labeled data and retraining for new scenarios. In this work, we propose a hybrid approach that integrates Reinforcement Learning (RL) with Neural Networks (NN) for real-time calibration of ECM parameters, enabling robust and adaptive SoC estimation. The RL agent continuously learns optimal parameter updates through system interaction, eliminating the need for manual recalibration and improving performance across diverse operating conditions. Experimental results demonstrate that the proposed method enhances the accuracy and adaptability of SoC prediction.
Vasilakis et al. (Thu,) studied this question.