Multi-scenario deep learning-based framework to estimate the remaining charge time of lithium-ion power batteries  

作　　者：Jianhui Mou Chengcheng Yu Peiyong Duan Junjie Li Chunjiang Zhang Yuhui Liu Xinhua Liu Akhil GARG Shaosen Su 

机构地区：[1]School of Mechanical,Electrical and Automotive Engineering,Yantai University,Yantai,264005,China [2]Department of Electronics,Electrical and Control,Qilu University of Technology(Shandong Academy of Sciences),Jinan,250353,China [3]The State Key Laboratory of Digital Manufacturing Equipment and Technology,Huazhong University of Science and Technology,Wuhan,430074,China [4]Suzhou Tongyuan Software&Control Technology Co.,Ltd.,Suzhou,215125,China [5]School of Transportation Science and Engineering,Beihang University,Beijing,100191,China

出　　处：《Chain》2024年第3期229-248,共20页链(英文)

基　　金：supported in part by the National Natural Science Foundation of China(Grant No.5217051006);the Shandong Province Natural Science Foundation(Grant No.ZR2021ME223);the Yantai Science and Technology Planning Project(Grant No.2022GCCRC158);the Graduate Innovation Foundation of Yantai University,GIFYTU(Grant No.GGIFYTU2349).

摘　　要：The accurate estimation of the remaining charge time(RCT)is essential in a battery management system(BMS),because it guarantees the safety and dependability of the power battery systems of new energy vehicles.However,the direct estimation of RCT is challenging because of the variability of actual charging scenarios and the complex charging process,which complicates the estimation of RCT in actual scenarios.Hence,this paper proposes an estimation framework based on deep learning for multi-scenario charging data to estimate the remaining charging times.Through an in-depth analysis of multi-scenario charging data,the RCT of the charging process is estimated using the temporal convolutional network(TCN)model,which has a strong generalization ability.Additionally,a dynamic learning rate(DLR)mechanism and an early stopping strategy(ES)are designed in the TCN model(DLR-ES TCN)for the nonlinear characteristics of the battery system to balance the relationship between model convergence speed and accuracy.Finally,compared with the traditional TCN model and four common deep learning models under three different scenarios,the experimental results show the mean absolute percentage error(MAPE)of the proposed method is less than 2%,indicating better accuracy and stability.This research can improve the safety monitoring of power batteries when applied to various target domains.

关 键 词：new energy vehicles lithium-ion power battery multi-scenario temporal convolution network remaining charging time dynamic learning rate and early stopping 

分 类 号：TM9[电气工程—电力电子与电力传动]