基于联合参数辨识的粒子群优化扩展粒子滤波的锂电池荷电状态估计  

State of Charge Estimation of Li-Ion Battery Using Particle Swarm Optimization Extended Kalman Particle Filter Based on Joint Parameter Identification

作　　者：贠祥 张鑫[1] 王超[1] 范兴明[1] Yun Xiang;Zhang Xin;Wang Chao;Fan Xingming(Department of Electrical Engineering&Automation,Guilin University of Electronic and Technology,Guilin 541004 China)

机构地区：[1]桂林电子科技大学电气工程及其自动化系,桂林541004

出　　处：《电工技术学报》2024年第2期595-606,共12页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(61741126);广西自然科学基金(2022GXNSFAA035533)资助项目。

摘　　要：提高参数辨识的精度和SOC算法的精度是提高SOC估计的关键,该文提出了基于联合参数辨识的粒子群优化扩展粒子滤波的荷电状态(SOC)估计方法。在参数辨识阶段,结合遗忘因子递推最小二乘法在线辨识的优势,弥补粒子群辨识精度高但前期缺乏数据无法实时辨识的劣势,联合进行参数辨识;在SOC估计阶段,利用扩展卡尔曼滤波生成重要性密度函数,去克服粒子退化,同时采用粒子群优化算法优化重采样策略改进采样过程缓解粒子贫化。最后在联邦城市运行(FUDS)和US06高速公路运行(US06)工况下将所提算法与F-PF、F-PSO-PF、FPSO-PSO-PF进行了对比,结果表明,在FUDS工况下,方均根误差分别提高了65.4%、56.3%和43.5%;在US06工况下,方均根误差分别提高了45.8%、35.9%和35.1%,验证了所提算法具有较好的适应性和鲁棒性。Improving the accuracy of parameter identification and SOC(state of charge)algorithm is the key to enhancing SOC estimation.Based on joint parameter identification,this paper proposed a SOC estimation method using particle swarm optimization extended Kalman particle filter(EPF).In the early stage of parameter identification,the forgetting factor recursive least squares(FFRLS)is used.However,when errors in the low SOC region become larger,the swarm optimization(PSO)algorithm is used for parameter identification.PSO uses the voltage data collected during the previous FFRLS parameter identification as input,employing the minimum voltage difference as the objective function to calculate the circuit model parameters.The joint parameter identification method can compensate for the accuracy issue of PSO identification but needs more data in the early stage.The SOC of the lithium battery is estimated based on parameter identification.Aimed at the problem of particle degradation and particle shortage in particle filter(PF),an extended Kalman filter algorithm is used to update each particle.The final approximate posterior probability density is used as the importance density function to overcome particle degradation.At the same time,the particle swarm optimization algorithm optimizes the resampling strategy to improve the sampling process and mitigate particle impoverishment.Finally,the proposed method is compared with PF and PSO-PF algorithms under federal urban driving schedule(FUDS)and US06 Highway Driving Schedule(US06)conditions.Under the FUDS condition,regarding the maximum error,PSO-EPF based on joint identification is 14%higher than PSO-PF based on joint identification,32.8%higher than PSO-PF based on FFRLS,and 53.2%higher than PF based on FFRLS.Regarding the mean absolute error,PSO-EPF is 56%higher than PSO-PF based on joint identification,62.5%higher than PSO-PF based on FFRLS,and 67.7%higher than PF based on FFRLS.Regarding the root mean square error,PSO-EPF is 43.5%higher than PSO-PF based on joint identificati

关 键 词：锂电池 粒子群算法 扩展粒子滤波算法 荷电状态 

分 类 号：TM912[电气工程—电力电子与电力传动]