基于递归小脑模型神经网络和卡尔曼滤波器的锂电池荷电状态预测  

State of charge prediction for lithium-ion batteries based on KF-RCMNN

作　　者：徐智帆 李华森 李文院 余凯 XU Zhifan;LI Huasen;LI Wenyuan;YU Kai(State Grid Xiamen Electric Power Supply Company,Xiamen 361001,China;State Grid Zhangzhou Electric Power Supply Company,Zhangzhou 363020,China;State Grid Fujian Procurement Company Limited,Fuzhou 350001,China;State Grid Fujian Construction Company,Fuzhou 350001,China)

机构地区：[1]国网厦门供电公司,福建厦门361001 [2]国网漳州供电公司,福建漳州363020 [3]国网福建物资公司,福州350001 [4]国网福建建设公司,福州350001

出　　处：《综合智慧能源》2024年第7期81-86,共6页Integrated Intelligent Energy

摘　　要：由于储能系统被广泛应用到新能源汽车、分布式发电等领域,其在运行过程中的可靠性是研究的重点之一。荷电状态(SOC)是反映电池续航能力的关键参数。为保证储能系统的正常运行,提出了一种锂电池SOC估计的方法,将递归小脑模型神经网络(RCMNN)和卡尔曼滤波器(KF)都用于荷电状态估计。为了强化RCMNN的捕获动态特征的能力,在联想记忆层和权值记忆层均加入了递归单元。将采集的电压、电流和温度作为模型的输入,用于模拟储能系统的不同充、放电情况。考虑到实际工况下电池放电的复杂性,在不同的放电条件和不同SOC初值的情况下将SOC的实际值与预测值进行对比。试验结果表明,该预测方法在不同条件下都具有较高的精度和鲁棒性。Since energy storage systems(ESSs)are widely used in electric vehicles,distributed power generation and other fields,the reliability of batteries has become an issue of concern for researchers.The state of charge(SOC)is a crucial parameter reflecting the battery endurance.This study proposes a new method for lithium-ion battery SOC estimation to keep the stable working of an ESS.Recurrent cerebellar model neural network(RCMNN)and Kalman filter(KF)are both applied in the estimation method.Integrating recurrent units in associative memory space and weight memory space can improve the RCMNN in dynamic feature capture.The RCMNN and KF with inputs of voltage,current and temperature can simulate the charging and discharging of the ESS.Considering the complexity of discharging process of the battery,the measured and simulated values of the SOC under different charging and discharging conditions with varied SOC initial values are compared.The results show that the proposed method has decent accuracy and robustness under different conditions.

关 键 词：荷电状态 锂电池 储能系统 递归小脑模型神经网络 卡尔曼滤波器 

分 类 号：TM91[电气工程—电力电子与电力传动] TK01[动力工程及工程热物理]