Lithium-Ion Battery State of Charge Estimation Using Gated Recurrent Unit Neural Network under Dynamic HPPC Profiles

Abstract

Precise assessment of the state of charge (SOC) for lithium-ion batteries is crucial for enhancing performance and energy storage systems. Conventional methods, such as ampere-hour counting and extended Kalman filters (EKF) are prevalent but frequently fail to accommodate the highly nonlinear and dynamic characteristics of contemporary cells. This study presents a robust deep learning framework employing the Gated Recurrent Unit (GRU) network to improve State of Charge (SOC) estimation accuracy. The proposed model was evaluated using experimental data from Samsung 50E batteries under Hybrid Pulse Power Characterization (HPPC) profiles, in contrast to the traditional Ah method. The findings indicate that the GRU method markedly surpasses traditional methods, attaining a Mean Absolute Error (MAE) of $0,004 \%$ and a Root Mean Square Error (RMSE) of $0,006 \%$. The results demonstrate that the GRU based method provides a high-accuracy solution for real-time BMS applications.