摘要
随着新能源汽车迅速发展普及,对动力电池健康状态进行准确评估成为了一个亟待解决的问题。针对这一问题,本文以轻量梯度提升机(light gradientboostingmachine,LightGBM)为基础,采集数据处理特征并进行训练,最终构建了一个动力电池健康状态评估平台。首先,依据相关工作,在原始数据中提取了6个特征工程--放电电压的最小值的时间、负载电压的75百分位、放电电压平均值、放电负载电压的25百分位、放电电压的25百分位和放电电压的标准差。其次通过引入直方图算法、单边梯度采样算法和斥特征捆绑算法,对数据进行进一步处理,以减少平台的内存消耗与计算代价。最后采用NASA艾姆斯卓越预测中心的数据,对平台功能进行验证并与相似工作进行对比,结果表明该平台可以提供良好的预测精度(均方根误差仅0.0103)。该平台对动力电池SOH预测方法的发展具有积极的影响,具有巨大的实际运用潜力。
With the rapid development and popularization of electric vehicles(EVs),accurate evaluation of the state of health(SOH)of vehicle power batteries has become a pressing issue.To address the problem,this study adopts a machine learning approach based on the Light Gradient Boosting Machine(LightGBM)framework.This approach involves collecting data,processing characteristics,carrying out training,and finally constructing a data-driven battery SOH analytic system.As a first step,six characteristics are extracted from the original data:the time of minimum discharge voltage;the 75th percentile of load voltage;the average discharge voltage;the 25th percentile of discharge load voltage;the 25th percentile of discharge voltage;and the standard deviation of discharge voltage.Secondly,the data are further processed to reduce memory consumption and computing cost.This processing utilizes LightGBM's key algorithms:histogram-based decision tree learning,Gradient-Based One-Side Sampling,and Exclusive Feature Bundling.Finally,system function is verified and compared with similar work,using data from the NASA Ames Prognostics Center of Excellence.Results demonstrate that the SOH platform can deliver high predictive accuracy(a root-mean-square error of 0.0103).The system represents significant progress in vehicle battery SOH prediction methodology and has high potential for practical application.
作者
常泽宇
张之琦
张晓东
李丽
郁亚娟
CHANG Zeyu;ZHANG Zhiqi;ZHANG Xiaodong;LI Li;YU Yajuan(Department of Energy and Environmental Materials,School of Material Science&Engineering,Beijing Institute of Technology,Beijing 100081,China)
出处
《储能科学与技术》
CAS
CSCD
北大核心
2022年第6期1847-1853,共7页
Energy Storage Science and Technology
基金
国家自然科学基金项目(52074037)
内蒙古自治区科技计划项目(2020ZD0018)。
