Comparative Study of SOC Estimation Methods Based on Artificial Intelligence for Lithium Batteries

Abstract

Battery technology plays a key factor in the advancement of electric cars and mobile robots. Efficient battery management, particularly accurate estimation of the state of charge (SOC), is of paramount importance in both theoretical research and practical applications. However, SOC estimation for batteries is a challenging task due to their inherent time-varying and non-linear characteristics. In this study, we delve into the complex nature of lithium cells and explore the potential of neural networks for SOC prediction. Specifically, we compare three different neural network architectures using the LG HG2 dataset, which provides valuable insights into lithium cell behavior. By analyzing the theory of neural networks and conducting comprehensive experiments, we aim to improve the accuracy of SOC estimation. Our focus is on understanding the intricate relationship between input variables and SOC, and harnessing the power of neural networks to capture and model this relationship effectively. The results of our study not only contribute to the field of battery management systems but also shed light on the broader understanding of SOC estimation in the context of lithium cells. By evaluating the performance of different neural network architectures and analyzing the obtained results, we gain valuable insights into optimizing battery performance and enhancing the overall efficiency of electric vehicles and mobile robots.