基于在线支持向量回归的锂离子电池SOC估计  

Lithium-ion battery SOC estimation based on online support vector regression

作　　者：朱江[1] 张伟[2] 马嵩 ZHU Jiang;ZHANG Wei;MA Song(Tianjin Lishen Battery Joint-Stock Co.,Ltd.,Tianjin 300384,China;Qingdao University of Science & Technology,Qingdao Shandong 266061,China;Qingdao MKL Technologies Corporation,Qingdao Shandong 266199,China)

机构地区：[1]天津力神电池股份有限公司,天津300384 [2]青岛科技大学自动化与电子工程学院,山东青岛266061 [3]青岛美凯麟科技股份有限公司,山东青岛266199

出　　处：《电源技术》2019年第10期1611-1614,共4页Chinese Journal of Power Sources

摘　　要：对于锂离子电池来说,其化学特性是动态非线性的,并具有较强的耦合性,但是现在常用的电池模型并不能准确表达其上述特性。训练样本数量定量时,在线支持向量回归机可以在线实时更新模型,且具有全局最优、良好的泛化能力。训练模型时,采用输入变量为工作电压和温度,输出变量为荷电状态。仿真结果表明,与BP神经网络相比,在线支持向量回归可以准确预测电池的充电状态,具有较高的SOC预测精度和稳定性。For lithium-ion batteries,the chemical properties are dynamically nonlinear and have strong coupling properties,but the commonly used battery models can not accurately describe the above characteristics.The online support vector regression machine was adopted in this paper,which could update the model online in real time under the limited sample with the global optimal and good generalization ability.When training the model,the input variables were the operating voltage and temperature,and the output variables were the state of charge.The simulation results show that the online support vector regression can accurately predict the state of charge of the battery compared with the BP neural network,and has higher SOC prediction accuracy and stability.

关 键 词：在线支持向量回归 SOC 核函数 神经网络 

分 类 号：TM912[电气工程—电力电子与电力传动]