首页|基于贝叶斯优化LSTM的锂电池健康状态评估方法

基于贝叶斯优化LSTM的锂电池健康状态评估方法

扫码查看
锂电池的健康状态(State of Health,SoH)是电池管理系统的主要指标之一,为了提升锂电池SoH评估的精确性,将能够解决人为经验调参困难的贝叶斯优化(Bayesian Optimization,BO)算法和具有长期记忆能力的长短期记忆(Long Short-Term Memory,LSTM)神经网络组合起来,形成一种新的BO-LSTM神经网络.选取美国国家航空航天局(NASA)的公开锂离子电池数据集进行验证,结果显示,相比于单个的BP神经网络模型和单个的LSTM神经网络模型,BO-LSTM神经网络对评估锂电池SoH的精确率更高,其中在B0005 电池上分别提高了 14.3%和 15.3%,在B0006 电池上分别提高了 23.8%和 20.5%.这表明在评估锂电池SoH时,基于贝叶斯优化的LSTM神经网络具有更好的效果,在实际应用中有着更高的价值.
Lithium Battery State of Health Assessment Method Based on Bayesian Optimization LSTM
The State of Health(SoH)of lithium-ion batteries is one of the main indicators of battery management systems.In order to improve the accuracy of SoH assessment of lithium batteries,a new BO-LSTM neural network is formed by combining Bayesian Optimization(BO)algorithm,which can solve the difficulty of human experience parame-ter tuning,and Long Short Term Memory(LSTM)neural network with long-term memory ability.Selecting the publicly available lithium battery dataset from NASA for validation,the results showed that compared to a single BP neural net-work model and a single LSTM neural network model,the BO-LSTM neural network has higher accuracy in assessing the SoH of lithium batteries,with improvements of 14.3%and 15.3%on battery B0005 and 23.8%and 20.5%on battery B0006,respectively.This indicates that the LSTM neural network based on Bayesian optimization has better performance in assessing the SoH of lithium batteries and has higher value in practical applications.

lithium batteryBayesian optimization algorithmlong short-term memory neural networkstate of healthassessment

潘子良、朱成杰、余梦书

展开 >

安徽理工大学 电气与信息工程学院,安徽 淮南 232001

锂电池 贝叶斯优化算法 长短期记忆神经网络 健康状态 评估

国家自然科学基金

62003001

2024

仪表技术
上海市仪器仪表学会,上海仪器仪表研究所等

仪表技术

影响因子:0.217
ISSN:1006-2394
年,卷(期):2024.(3)
  • 10