Comparative Analysis on Deep Learning Methods to Estimate the State of Charge in Li-ion Battery

Estimating the State of Charge (SOC) is a component of a Battery Management System (BMS). Calculating the State of Charge using traditional methodologies is difficult, resulting in significant inaccuracies or a loss of SOC accuracy and complexity in the hardware. A data-driven model is one that calculates the SOC using machine learning and deep learning, which are currently popular methods and can be utilized to solve this challenge. This paper focused on the performance comparison of SOC estimation using Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM) and simple Recurrent Neural Network (RNN) compared using a publicly available dataset and the data, which consists of charging and discharging batteries at various temperatures, are utilized to train the models. The Gated Recurrent Unit has achieved a Root Mean Square Error (RMSE) in a range of 1.11 - 1.38 % and a Mean Absolute Error (MAE) in a range of 0.8 -1.11 % at a different temperature to estimate the SOC.