基于QPSO-BP神经网络的锂电池SOC预测  

Estimation for SOC of lithium battery based on QPSO-BP neural network

作　　者：刘征宇[1,2] 杨俊斌 张庆[1] 张利[1,2] 赵爱国[1] 

机构地区：[1]合肥工业大学机械与汽车工程学院,合肥230009 [2]安全关键工业测控技术教育部工程研究中心,合肥230009 [3]安徽富煌建设有限责任公司,合肥238076

出　　处：《电子测量与仪器学报》2013年第3期224-228,共5页Journal of Electronic Measurement and Instrumentation

基　　金：国家电子信息产业发展基金([2010]301);安徽高校省级自然科学研究(KJ2011A220);广东省教育部产学研结合(2009B090300303)项目资助

摘　　要：电池荷电状态(state of charge,SOC)的预测是电动汽车电池管理系统的关键任务之一,为此对锂电池荷电状态的预测进行了研究,提出了一种基于QPSO-BP神经网络的锂电池SOC预测。在分析了磷酸铁锂(LiFePO4)电池充放电机理后,运用MATLAB人工神经网络工具箱建立基于量子微粒群算法(QPSO)的BP(back propagation)神经网络模型,用于预测锂离子电池充放电过程中的任一状态下的SOC。仿真实验验证了方法的准确性。结果表明,与现有的神经网络预测方法相比,基于QPSO-BP神经网络的锂电池SOC预测方法准确度高,且具备很好的实用性。Estimation of the state of charge （SOC） of lithium battery is one of the key missions for battery management system of electric vehicles. Therefore, the estimation for SOC of lithium battery is researched, and an algorithm of the estimation for SOC of lithium battery based on QPSO-BP neural network is proposed in the paper. After analyzing the electrochemical mechanism of lithium battery in charge and discharge process, MATLAB artificial neural network toolbox to establish BP neural network model based on QPSO algorithm is used to estimate the remaining capacity of lithium battery in charge and discharge process. The simulation results show that the proposed algorithm is accurate. The result shows that the algorithm of the estimation for SOC of lithium battery based on QPSO-BP neural network has higher accuracy and proticability than other algorithms.

关 键 词：锂离子电池 荷电状态SOC 神经网络 量子微粒群算法 

分 类 号：U469.72[机械工程—车辆工程]