基于神经网络改进PID-AEKF的锂电池SOC估计  

Improved PID-AEKF Based on Neural Network for SOC Estimation of Lithium Battery

作　　者：李洁 赵阳[1] 王鑫 马蒙召 LI Jie;ZHAO Yang;WANG Xin;MA Meng-zhao(College of Electrical and Computer Engineering,Jilin Jianzhu University,Changchun 130118,China)

机构地区：[1]吉林建筑大学电气与计算机学院,吉林长春130118

出　　处：《电力电子技术》2025年第2期43-47,共5页Power Electronics

基　　金：吉林省科技发展计划项目(20200403137SF);吉林省教育厅"十三五"科学技术项目(JJKH20200273KJ)。

摘　　要：针对自适应扩展卡尔曼滤波(AEKF)算法中模型的不稳定以及测量噪声导致的估计误差问题,本文提出了一种基于神经网络改进的自适应反馈补偿的锂离子电池(LIB)荷电状态(SOC)融合估计方法。在二阶RC模型的基础上,基于新息向量的比例积分微分(PID)观测器补偿AEKF中产生的噪声误差,并采用BP神经网络(BPNN)对PID观测器的参数进行自调整,实时估计电池的SOC。开展多工况下算法估计误差对比实验,结果表明,基于神经网络改进PID-AEKF算法可以提高SOC的估计性能,在动态工况下,平均绝对误差相较AEKF降低了82%,达到0.25%。Aiming at the instability of the model and the estimation error caused by measurement noise in the adap-tive extended Kalman filter(AEKF)algorithm,this paper proposes an improved adaptive feedback compensation me-thod for lithium-ion battery(LIB)state of charge(SOC)fusion estimation based on neural network.Based on the sec-ond-order RC model,the innovation vector-based proportional integral differential(PID)observer compensated the no-ise error generated in AEKF and the parameters of the PID observer are self-adjusted by the BP neural network(BPNN)to estimate the SOC of the battery in real time.The experimental results show that the improved PID-AEKF al-gorithm based on neural network can improve the performance of SOC estimation and the average absolute error is reduced by 82%compared with AEKF under dynamic conditions,reaching 0.25%.

关 键 词：锂电池 自适应扩展卡尔曼滤波 神经网络 

分 类 号：TN86[电子电信—信息与通信工程]