SOC Estimation of Lithium Batteries Based on Deep Learning and Particle Filtering

Abstract

State of Charge (SOC) estimation of Li-ion batteries is important to ensure safe and reliable battery operation. The SOC estimation method based on mechanism knowledge and data fusion can take advantage of the mechanism model and data-driven method to obtain more accurate prediction results. However, it is difficult to determine the accurate degradation mechanism model of Li-ion battery under complex operating conditions. In this paper, we propose a deep learning and particle filter (PF) fusion method for Li-ion battery SOC estimation, solving the dependence of PF method on mechanism model problem by establishing the agent model of Li-ion battery degradation mechanism through deep learning. The specific steps are as follows. Firstly, extract the degradation trend of SOC by Convolutional Neural Networks (CNN) to fit the degradation trajectory of Li-ion battery SOC using the selected health features, constructing the state space model of PF algorithm, and then update and correct the output of CNN model by PF method to achieve the accurate estimation of Li-ion battery SOC. The proposed method is validated by two different lithium battery charge/discharge data sets, and the experimental results show that the proposed method has the features of strong tracking ability of degradation trajectory, high prediction accuracy and stability, and the method does not depend on the mechanism model of the battery, which has strong generalization ability in the practical application scenarios.