Design of robust battery capacity model for electric vehicle by incorporation of uncertainties
Design of robust battery capacity model for electric vehicle by incorporation of uncertainties
复制标题
考虑不确定性的电动汽车鲁棒电池容量模型设计
DOI：
10.1002/er.3723
复制
发表时间：
2017-08-01
影响因子：
4.6
通讯作者：
Liang, Xinyu
中科院分区：
文献类型：
--
作者：
Garg, Akhil;Vijayaraghavan, V.;Liang, Xinyu
电动车辆的运行范围的改善可以通过对电池组系统的能量存储容量的设计和优化的鲁棒建模来实现。在这项工作中，作者对电池建模方法进行了全面调查，并确定了对估计电池容量至关重要的关键改进领域。这项工作提出了人工智能方法的自动神经网络搜索（ANS）在开发的强大的电池容量模型的基础上的输入（温度和放电率）的锂离子电池。模型的鲁棒性是通过引入输入（温度和放电速率，算法和模型的架构）的不确定性来引入的。模型的统计分析和验证表明，使用ANS方法制定的模型优于响应面回归模型，相关系数高达0.97。正态分布的输入的基础上的不确定性分析表明，从ANS制定的模型是最不敏感的输入条件的变化相比，响应面回归模型。全局敏感性分析表明，温度是准确的电池容量估计的主导因素。版权所有（c）2017约翰威利父子有限公司
The improvement in the operating range of electric vehicles can be accomplished by robust modelling of the design and optimization of the energy storage capacity of the battery pack system. In this work, the authors have conducted a comprehensive survey on battery modelling methods and identified critical areas of improvement vital for estimating the battery capacity. This work proposes the artificial intelligence approach of automated neural networks search (ANS) in development of the robust battery capacity models for the lithium ion batteries based on the inputs (temperature and discharge rates). The robustness in the models is introduced by incorporating uncertainties in the inputs (the temperature and discharge rates, the architecture of algorithm and the models). The statistical analysis and validation of the models reveal that the models formulated using an ANS approach outperform the response surface regression models with correlation coefficient achieved as high as 0.97. The uncertainty analysis based on normal distribution of the inputs suggests that the models formulated from ANS are least sensitive to change in the input conditions when compared to response surface regression models. The global sensitivity analysis reveals that the temperature is a dominant factor for accurate battery capacity estimation. Copyright (c) 2017 John Wiley & Sons, Ltd.