期刊
ENERGY
卷 223, 期 -, 页码 -出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.energy.2021.120114
关键词
Feature extraction; State of health; Remaining useful life; Battery aging; Gaussian process regression
资金
- National Natural Science Foundation of China [51975355, 72071127]
- National Major Science and Technology Projects of China [J2019IV0018]
- National Key R&D Plan Key Special Project [2017YFE0102000]
- General Research Fund [CityU 11204419]
This paper proposes a voltage-temperature health feature extraction method to improve PHM of lithium-ion batteries. This method can accurately estimate and predict the health conditions and remaining useful life of lithium-ion batteries. Results show that this method provides higher accuracies than existing methods.
Prognostics and health management (PHM) of lithium-ion batteries are important to ensure the safety of electric vehicles. To date, there has not been an adequate method to accurately estimate battery health conditions and predict battery lifetime under fast charging. A voltage-temperature health feature extraction method is proposed to improve PHM of lithium-ion batteries in this paper. Since voltage change can reflect battery degradation process, a difference model is firstly proposed to extract voltage dependent health features from partial voltage profiles, which does not need to fully discharge a battery. Simultaneously, as battery aging is affected by temperature, battery surface temperature is selected as a thermal-dependent health feature. Subsequently, the extracted voltage-temperature health features are fed into a developed battery degradation model. Using the proposed method, state of health and remaining useful life (RUL) of lithium-ion batteries can be estimated and predicted with uncertainty measurements. Battery degradation data collected from accelerated battery tests under two different charging policies are utilized to validate the accuracy of the proposed method. Results show that root mean square error (RMSE) is smaller than 1% in all capacity estimation and relative RMSE is around 5% for RUL prediction, which provide higher accuracies than the existing methods. (c) 2021 Elsevier Ltd. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据