点击次数：

DOI码：10.1109/IPEMC-ECCEAsia60879.2024.10567835

发表刊物：The 10th International Power Electronics and Motion Control Conference-ECCE Asia

关键字：State of health; Back-Propagation Neural Network; Gaussian Process Regression; Support Vector Machine; Compensation strategy

摘要：The widespread adoption of lithium-ion batteries in electric vehicles, renewable energy systems, and portable electronic devices has positioned them as the primary energy supply solution. The State of Health (SOH) of a battery, representing its current performance relative to the brand-new state, significantly impacts capacity, voltage stability, charging and discharging efficiency, battery lifespan, and safety. Within this context, SOH estimation plays a pivotal role in addressing safety concerns through early issue detection, mitigating environmental implications, and achieving cost efficiency by managing battery lifecycles and optimizing maintenance. This study contributes to this field by conducting a comparative analysis of three commonly used SOH estimation methods: Back-Propagation Neural Network (BPNN), Gaussian Process Regression (GPR), and Support Vector Machine (SVM). The results reveal that BPNN has the highest accuracy with an RMSE (Root Mean Square Error) of 0.25%, followed by GPR with an RMSE of 0.83%. SVM exhibits the lowest accuracy, having the highest RMSE at 1.29%. Based on the above three methods, a rate compensation model was constructed to achieve State of Health (SOH) estimation for batteries at different discharge rates using only a single model. The Root Mean Square Error (RMSE) of SOH estimation for batteries at discharge rates of 0.5C, 2C, and 3C using the BPNN-based compensation model were 0.238%, 0.517%, and 1.099%, respectively. The GPR-based compensation model yielded RMSE values of 0.7688%, 1.113%, and 2.383% for the same discharge rates, while the SVM-based compensation model resulted in RMSE values of 0.5978%, 2.630%, and 2.560%, respectively.

合写作者：Changpen Tan

第一作者：Li Wang

论文类型：论文集

通讯作者：Muyao Wu

学科门类：工学

文献类型：C

是否译文：否

发表时间：2024-05-18

收录刊物：EI

发布期刊链接：https://ieeexplore.ieee.org/document/10567835