基于LSTM-AEKF联合算法的锂电池SOC估计研究  

Research on Lithium Battery SOC Estimation Based on LSTM-AEKFHybrid Algorithm

作　　者：孙冠宇 沈金荣[1] SUN Guanyu;SHEN Jinrong(College of Information Science and Engineering,Hohai University,Changzhou,Jiangsu 213000,China)

机构地区：[1]河海大学信息科学与工程学院,江苏常州213000

出　　处：《自动化与仪器仪表》2024年第6期231-235,240,共6页Automation & Instrumentation

基　　金：江苏省重点研发计划项目(BE2022100)。

摘　　要：针对如何准确且精密的估算出锂电池荷电状态(SOC)问题,提出了一种自适应扩展卡尔曼滤波(AEKF)和长短时记忆神经网络(LSTM)相结合算法对锂电池SOC优化估计。采用混合脉冲功率特性(HPPC)测试方法,利用带遗忘因子的递推最小二乘法(FFRLS)对三阶等效电路模型(ECM)等效电路模型进行在线参数辨识,然后根据方程使用LSTM-AEKF算法对电池进行SOC实时估计实验,得出算法对SOC估计误差控制在1%以内。最后由对比实验结果证明:该算法与EKF和LSTM两种算法相比在均方根误差上分别提高1.25%和0.81%,具有更高的准确性和精密度。To address the problem of how to accurately and precisely estimate the state of charge(SOC)of lithium-ion batteries,a combined algorithm of adaptive extended Kalman filter(AEKF)and long short-term memory neural network(LSTM)is proposed for the optimal estimation of lithium-ion battery SOC.Using the hybrid pulse power characterization(HPPC)test method,the online parameter identification of the third-order equivalent circuit model(ECM)is performed by the forgetting factor recursive least squares method(FFRLS),and then the LSTM-AEKF algorithm is applied to the battery SOC real-time estimation experiment according to the equation,obtaining the algorithm to control the SOC estimation error within 1%.Finally,the comparative experimental results show that the algorithm improves the root mean square error by 1.25%and 0.81%compared with the EKF and LSTM algorithms,respectively,and has higher accuracy and precision.

关 键 词：锂电池 荷电状态 最小二乘法 自适应扩展卡尔曼算法 长短期记忆网络 均方根误差 

分 类 号：TP202[自动化与计算机技术—检测技术与自动化装置]