基于权值选择粒子滤波算法的锂离子电池SOC估计  

SOC Estimation of Lithium-ion Battery based on Weight Selection Particle Filter Algorithm

作　　者：彭方想 南金瑞[1] 孙立清[1] PENG Fangxiang;NAN Jinrui;SUN Liqing(School of Machinery and Vehicles, Beijing Insitute of Technology, Beijing 100081, China)

机构地区：[1]北京理工大学机械与车辆学院,北京100081

出　　处：《太原理工大学学报》2020年第5期750-755,共6页Journal of Taiyuan University of Technology

基　　金：国家重点R&D计划项目(2017YFB0103801);上海汽车工业技术发展基金会项目(1620)。

摘　　要：针对锂离子动力电池的荷电状态(SOC)估计问题,以三元锂(MNC)电池为研究对象,选用Thevenin等效电路模型,建立电池模型的状态方程和观测方程,完成了带遗忘因子的递推最小二乘法(FFRLS)的理论推导。进行电池单体混合动力脉冲功率特性测试(HPPC测试),基于测试数据和FFRLS算法完成电池模型的在线参数辨识,并通过锂离子电池的端电压精度来验证算法的可行性;在此基础上,提出一种权值选择粒子滤波(WSPF)算法来实现锂离子电池SOC估计,该算法中全部粒子都参与粒子滤波过程,但只选择较优权重粒子用于电池状态估计,从而解决粒子滤波的粒子退化问题,提高粒子的多样性。通过HPPC测试和动态工况测试(DST)结果验证,WSPF算法的估计精度能控制在2%以内。与重采样粒子滤波(SIR-PF)算法相比,WSPF算法的估计精度高,鲁棒性好。Aiming at the estimation of the state of charge(SOC)of lithium-ion power batteries,this paper took ternary lithium(MNC)batteries as the research object,selected Thevenin equivalent circuit model to establish the state equation and observation equation of the battery model and completed the theoretical derivation of recursive least squares method(FFRLS).Hybrid pulse power characteristic test(HPPC test)on battery cells was performed,online parameter identification of battery model was achieved by using test data and FFRLS algorithm,and the feasibility of the algorithm was verified by the battery terminal voltage.On this basis,a weighted selection particle filter(WSPF)algorithm was proposed to realize the SOC estimation of lithium-ion batteries.All particles in the algorithm participate in the particle filter process,but only the particles of which weight are better are used for battery state estimation,thereby solving the problem of particle degradation of particle filtering and improving the diversity of particles.Through HPPC test and dynamic working condition test(DST)result verification,the estimation accuracy of WSPF algorithm can be controlled within 2%.Compared with that of the resampling particle filter(SIR-PF)algorithm,the estimation accuracy of the WSPF algorithm is high and the robustness is good.

关 键 词：Thevenin模型 在线参数辨识 SOC估计 权值选择粒子滤波算法 

分 类 号：TQ306[化学工程] TM911[电气工程—电力电子与电力传动]