期刊
ENERGY STORAGE MATERIALS
卷 42, 期 -, 页码 794-805出版社
ELSEVIER
DOI: 10.1016/j.ensm.2021.08.025
关键词
Lithium-ion battery (LiB); Battery ageing; Li4Ti5O12 (LTO); LiNiMnCoO2 (NMC); Incremental capacity analysis (ICA); Machine learning
资金
- German Federal Ministry for Trans-port and Digital Infrastructure (BMVI) [03B10502B, 03B10502B2]
This study investigates the cyclic and calendar ageing of 43 same-typed LTO cells under 16 different operation conditions. Results show a two-stage ageing mechanism in which the anode and cathode gradually limit the extractable capacity. Additionally, a capacity gain of up to 2.42% is observed for cells operated and stored below 50% SOC.
Lithium-titanate-oxide (LTO) batteries are one of the most promising technologies for various types of future applications in electric mobility, stationary storage systems and hybrid applications with high-power demands due to their long cyclic stability and superior safety. This paper investigates the cyclic and calendar ageing of 43 same-typed LTO cells considering 16 different operation conditions under variation of state of charge (SOC), temperature, depth of discharge, cycle SOC range and current rate. The ageing results are presented and the relative shift in incremental capacity is analysed in order to detect degradation mechanisms, separate the influence of degradation enhancing parameters and attribute them to their origin source. Our results show that the cells exhibit a two-stage ageing mechanism with stagewise increasing degradation gradient. In the first ageing stage the anode is limiting the amount of extractable capacity while the capacity fade mainly results from cathode degradation. After a certain level of degradation is reached the cathode starts limiting the amount of extractable capacity, initiating the second ageing stage with stronger occurring capacity fading gradient. A capacity gain of up to 2.42% becomes visible for cells operated and stored in a range below 50% SOC. For these cells an extended three-stage ageing mechanism is shown to be more applicable. The degradation behaviour is then estimated using a machine learning approach based on a recurrent neural network with long short-term memory, for which the presented incremental capacity data is used as training input.
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