基于等效电路模型RBF网络的电池容量快速预测  

Fast Prediction of Battery Capacity Based on Equivalent Circuit Model and RBF Network

作　　者：刘虹灵 别传玉 宋华伟 张宇平 李晨威 高标[3] Liu Hongling;Bie Chuanyu;Song Huawei;Zhang Yuping;Li Chenwei;Gao Biao(GEM Co.,Ltd.,Shenzhen 3518101,China;Wuhan Power Battery Regeneration Technology Co.,Ltd.,Wuhan 430014,China;The State Key Laboratory of Refractories and Metallurgy and the Institute of Advanced Materials and Nanotechnology,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]格林美股份有限公司,广东深圳518101 [2]武汉动力电池再生技术有限公司,湖北武汉430014 [3]武汉科技大学耐火材料与冶金国家重点实验室,湖北武汉430081

出　　处：《稀有金属》2023年第6期915-922,共8页Chinese Journal of Rare Metals

基　　金：深圳市创新创业计划技术攻关面上项目(重20200206)资助。

摘　　要：针对退役动力电池容量一致性分选问题,提出了一种对退役电池放电容量快速预测的方法,该方法主要是通过在选取合适的等效电路模型以及电池荷电状态(SOC)下,将退役电池测试的交流阻抗谱(EIS)结合等效电路模型用ZSimpWin软件进行拟合,由此获得表征电池内部状态的等效电路模型中的各个特征参数,并且利用径向基(RBF)神经网络建立了各个特征参数与退役电池放电容量的神经网络模型,并对退役电池的放电容量进行了预测。结果表明:将电池放电至80%SOC状态下利用LR(CR)(C(WR))(L为电感,R为电阻,C为电容,W为Warburg阻抗元件)等效电路模型对EIS进行拟合的结果最好,是获取电池特征参数的有效途径,并且通过RBF神经网络建立了各个特征参数与电池放电容量关系的神经网络模型,该模型对电池放电容量预测的最大误差不超过0.6503 Ah。Faced with the increasingly serious environmental pollution and energy crisis,countries around the world are gradually reducing the output of fuel vehicles and forming a consensus to vigorously develop electric vehicles(EVs).In the past 10 years,the number of global EVs have grown exponentially,with a growth rate of 4.2% in 2020.Lithium-ion batteries(LIBs)are used as the mainstream energy source for electric vehicles due to their high energy density,light weight,long cycle life and large power capacity.The global installed capacity of LIBs was 137 GWh in 2020 and is expected to reach 1500 GWh in 2030.To ensure safety and performance,LIBs in EVs must be decommissioned when their capacity decays to 70%~80% of their initial capacity.In order to facilitate the cascade utilization of the battery,a method for predicting the capacity and internal resistance of the battery with low energy consumption and high accuracy and speed was developed to sort the batteries reasonably.The battery was charged and discharged in capacity test to obtain the discharge capacity of the battery,and then the battery was discharged to a different state of charge(SOC)state for electrochemical impedance spectroscopy(EIS)test.A suitable equivalent circuit model was selected to fit EIS,and EIS of the battery in the SOC state with the best fitting result was selected as a way to obtain multiple characteristic parameter values.The neural network was used to establish a model of the relationship between multiple characteristic parameters and the discharge capacity,and a method that could realize the rapid prediction of the discharge capacity of the decommissioned battery was proposed.The discharge capacity distribution of the same batch of decommissioned ternary lithium batteries was relatively concentrated on,mainly between 44~47 Ah.Nyquist curves of EIS under 20%,50% and 80%SOC states showed that the changing trends of EIS curves under different SOC states were the same.Nyquist curve of EIS could be divided into three parts:high,medium and low fr

关 键 词：交流阻抗谱(EIS) 等效电路模型 径向基(RBF)神经网络 快速预测 

分 类 号：TM912[电气工程—电力电子与电力传动]