摘要
回顾了人工神经网络、支持向量回归、高斯过程回归三种主流数据驱动方法在动力电池健康状态(stateof health,SOH)估算方面的研究进展。人工神经网络适合模拟动力电池,能达到很高的精度;支持向量回归计算量小,理论基础完善,在动力电池SOH估算研究中应用广泛;高斯过程回归精度高并能给出预测结果的置信区间,近年相关文献数量呈现增长趋势。针对现行SOH定义未能反映锂电池额定电压衰退的弊端,提出了利用电池满充能量定义SOH。进而分别建立了BP神经网络、支持向量回归、高斯过程回归模型,利用新能源汽车大数据,对电池充电能量进行了预测,定量对比结果验证了三种方法在计算量和精确度方面的特点。最后展望了数据驱动方法与新能源汽车大数据在动力电池SOH估算研究方面的应用前景。
This paper reviews the research progress of three main data-driven methods,arti ficial neural network,support vector regression and Gaussian process regression,in the estimation of state of health(SOH).Artificial neural network is suitable for simulating power batteries and can achieve high precision.Support vector regression has a small amount of calculation and perfect theoretical foundation.It is widely used in the research of power battery SOH estimation.The Gaussian process has high regression accuracy and can give a confidence interval for the prediction results.In recent years,the number of related literatures shows an increasing trend.In view of the shortcomings of the current SOH de finition that fail to re flect the rated voltage decay of lithium-ion batteries,it is proposed to de fine SOH by using battery full charge energy.In this paper,BP neural network,support vector regression and Gaussian process regression model are established respectively.The new energy vehicle big data is used to predict the battery charging energy.Quantitative comparison results verify the characteristics of the three methods in terms of calculation volume and accuracy.Finally,the application prospects of data-driven methods and new energy vehicle big data in power battery SOH estimation research are prospected.
作者
陈翌
白云飞
何瑛
CHEN Yi;BAI Yunfei;HE Ying(School of Automotive Studies,Tongji University,Shanghai 201804,China)
出处
《储能科学与技术》
CAS
CSCD
2019年第6期1204-1210,共7页
Energy Storage Science and Technology
