Lithium-Ion Battery SOC Estimation Method Based on LSSVM-UKF

Abstract

<div class="section abstract"><div class="htmlview paragraph">Since entering the 21st century, the world has faced extremely serious environmental pollution and energy crises. In this context, new energy vehicles have been vigorously developed. Lithium-ion batteries have gradually become one of the most important energy sources for electric vehicles due to their excellent performance. State of charge (SOC) is one of the important indicators of the battery. Accurate estimation of SOC is of great significance to establishing a safe and accurate Battery management system (BMS).</div><div class="htmlview paragraph">Among related methods for SOC estimation, observation-based methods have been widely used. However, this type of method has the disadvantages of being susceptible to disturbance and requiring high accuracy of the battery model. The traditional equivalent circuit model cannot meet the needs. This paper proposes a battery SOC estimation method based on LSSVM-UKF through research on lithium-ion battery modeling and SOC estimation methods. First, the least squares support vector machine (LSSVM) is used to model the battery. During the training process of LSSVM, Northern Goshawk optimization (NGO) is used for parameter optimization to improve the fitting performance of the model. In addition, a time window mechanism is introduced to allow the current sampling point data and historical data to be integrated during training to more accurately capture the timing characteristics of the battery system. After building the battery model, the unscented Kalman filter (UKF) is used for battery SOC estimation.</div><div class="htmlview paragraph">In the verification experiment, this paper used public data sets to verify the battery model accuracy and SOC estimation performance under different ambient temperatures. The results show that in most cases, the error of the battery model does not exceed 0.05V, with only slight fluctuations under medium and high temperatures. UKF has a certain error in the initial stage, but it can adjust quickly. The error does not exceed 0.07 under low temperatures and the error does not exceed 0.03 under medium and high temperatures, showing high accuracy and robustness.</div></div>