基于小波神经的动力电池SOC估计的研究  

Research on Estimating for Power Battery State of Charge Based on Wavelet Neural

作　　者：于洋[1] 纪世忠[1] 魏克新[1] 

机构地区：[1]天津理工大学,天津市复杂工业系统控制理论及应用重点实验室,天津300384

出　　处：《电力电子技术》2012年第6期90-92,共3页Power Electronics

基　　金：国家自然科学基金(50977063);国家高技术研究发展计划项目(863项目)(2008AA11A145);天津市科技支撑计划重点项目(09ZCKFGX01800)~~

摘　　要：提高电池荷电状态(SOC)估计的准确性对电动汽车电池管理系统的研究至关重要。根据历史实验数据,在分析了电池电压、电流、电池表面温度及放电总电量等因素对电池SOC的影响后,提出一种利用小波神经网络的SOC在线实时估计方法。通过与BP神经网络的预测结果进行比较,结果表明,小波神经网络能更准确地逼近实际SOC值,并克服了神经网络容易陷入局部次优点的缺点,且将神经网络隐含层的传输函数用小波函数代替,算法易实现且易于应用和推广。Improving the accurcy of the state of charge （SOC） evaluation is critical for the research of battery management system of electric vehicle.After analyzing the historical data on battery voltage,current,cell surface temperature and power factors of total discharge battery influence, a novel way to estimate SOC on line by the wavelet neural network is presented.By comparing with the BP neural network algorithm,verify that the proposed method has more exactly results in approaching actual results and the method overcomes the disadvantages of local optimum which the neural network is easy to fall into,becauce of using the wavelet function replace the neural network hidden layer transfer function algorithm, the wavelet neural network is easy to realize, applicate and promote.

关 键 词：动力电池 电动汽车 荷电状态 小波神经网络 

分 类 号：TM912[电气工程—电力电子与电力传动]