LSTM-Based SOC Estimation Leveraging Coulomb Counting: A Machine Learning Approach*

Estimating the state of charge (SOC) of lithiumion batteries is an essential part of electric vehicle battery management strategies. SOC actually provides real-time details about the remaining percentage of charge in the battery. So that it helps the driver to understand how much energy is remaining for driving. However, fluctuating driving conditions—such as rapid accelerations, decelerations, and the battery’s temperature sensitivity—pose significant challenges to accurate SOC estimation. To overcome these challenges, this study presents the Long Short-Term Memory (LSTM)-based machine learning approach. Here SOC is calculated using the Coulomb counting method, which determines charge levels by integrating current over time. LSTM model is trained on a dataset that includes current and time, making it to learn temporal dependencies According to the experimental evaluation, the suggested model achieves remarkable accuracy, with a mean absolute error (MAE) of 0.2598% and a root mean squared error (RMSE) of 0.1877% throughout the test. In comparison to more straightforward approaches, the model exhibits improved prediction performance by incorporating more battery metrics, offering accurate and up-to-date SOC data. These results demonstrate the dependability and effectiveness of the LSTM-based approach for battery management systems, which helps to make electric vehicle operations safer and more effective.