1、基于 ELM 和 MCSCKF 的锂离子电池 SOC 估计王桥1),叶敏1),魏孟1,2),廉高棨1),武晨光1)1)长安大学公路养护装备国家工程研究中心,西安7100642)新加坡国立大学机械工程系,新加坡117576通信作者,E-mail:摘要为了减少噪声对锂离子电池荷电状态估计的影响,本文提出一种新颖的基于极限学习机和最大相关熵平方根容积卡尔曼滤波的 SOC 估计方法.首先,利用泛化性好、运行速度快的极限学习机作为卡尔曼滤波的测量方程;其次,基于灰狼优化算法,极限学习机的超参数被优化以提高电池荷电状态的估计精度;最后,基于最大相关熵平方根容积卡尔曼滤波,极限学习机的测量噪声被进一步减弱.
2、所提方法可以简化极限学习机繁琐的调参过程,且为闭环的 SOC 估计方法.所提方法在多工况和宽温度范围内被测试以验证其泛化性能.测试结果显示,所提方法明显地提高了锂离子电池的荷电状态估计精度.同时,对比其他算法,所提方法的平均运行时间仅仅为长短时序列和循环门控单元网络的三分之一.当行驶工况复杂、温度变化区间较大时,所提方法的均方根误差小于 1%,最大误差小于 3%.当存在初始误差与环境噪声时,所提方法显示出了优越的鲁棒性.关键词锂离子电池;荷电状态估计;极限学习机;灰狼优化;卡尔曼滤波;鲁棒性估计分类号TM911.3ELM-andMCSCKF-basedstateofchargeestimati
3、onforlithium-ionbatteriesWANG Qiao1),YE Min1),WEI Meng1,2),LIAN Gao-qi1),WU Chen-guang1)1)NationalEngineeringResearchCenterforHighwayMaintenanceEquipment,ChanganUniversity,Xian710064,China2)DepartmentofMechanicalEngineering,NationalUniversityofSingapore,Singapore117576,SingaporeCorrespondingauthor,E
4、-mail:ABSTRACTLithium-ionbatteriesarewidelyusedinelectricvehiclesandenergystoragesystems.Asaprerequisiteforthesafeandefficientapplicationoflithium-ionbatteries,batterymanagementsystemshavereceivedextensiveattentionworldwide.Amongtheseprerequisites,thestateofcharge(SOC),asthebasicparameterofbatteryma
5、nagementsystemonlineapplication,iscrucialforthesafeandefficientoperationofbatterymanagementsystems.However,measurementnoisedecreasestheaccuracyandrobustnessofthestateofchargeestimation.Toreducetheimpactofnoiseonthestateofchargeestimationoflithium-ionbatteries,anovelSOCestimationmethodbasedonanextrem
6、elearningmachineandamaximumcorrelationentropysquarerootvolumetricKalmanfilterisproposedinthis paper.First,the extreme learning machine is used as the measurement equations of the Kalman filter because of its goodgeneralizationandfastrunningspeed,andthevoltageandcurrentareselectedasthemodelinput;second,onthebasisofthegraywolfoptimizationalgorithm,theextremelearningmachinehyperparametersarethoroughlyoptimizedtoimprovetheaccuracyofthestateofchargeestimationforlithium