基于FA-BP神经网络的锂离子电池SOH估算  

SOH estimation of Li-ion battery based on FA-BP neural network

作　　者：吴铁洲[1] 刘思哲 张晓星 吴麟章[1] WU Tie-zhou;LIU Si-zhe;ZHANG Xiao-xing;WU Lin-zhang(Key Laboratory of Solar Energy Efficient Utilization and Energy Storage Operation Control in Hubei Province,Hubei University of Technology,Wuhan,Hubei 430068,China)

机构地区：[1]湖北工业大学太阳能高效利用及储能运行控制湖北省重点实验室,湖北武汉430068

出　　处：《电池》2021年第1期21-25,共5页Battery Bimonthly

基　　金：国家自然科学基金(51677058);湖北省技术创新专项重大项(2018AAA056)。

摘　　要：采用萤火虫算法(FA)优化BP神经网络对锂离子电池进行健康状态(SOH)估算,利用FA算法全局寻优的能力和收敛速度快的特点,优化BP神经网络的权值和阈值,解决BP神经网络容易陷入局部最小值和收敛速度慢的问题。对单体磷酸铁锂正极锂离子电池进行充放电实验,选用一阶RC电路模型,利用递推最小二乘法在线辨识模型参数,将电池的欧姆内阻、极化内阻和极化电容作为模型的输入参数。与BP神经网络算法相比,FA-BP神经网络优化算法估算SOH的误差波动范围减小2.50%,最大误差减少3.00%,平均误差减小1.68%,且具备良好的收敛性。Global optimization ability and fast convergence speed of firefly algorithm(FA)were applied to optimize the state of health(SOH)estimation of Li-ion battery by improving BP neural network.The methodology of FA could efficiently optimize the weights and the thresholds of BP neural network and solve the flaws of BP neural network of easily failing into local minimum and slow convergence speed.A single lithium iron phosphate cathode Li-ion battery was subjected to charge-discharge experiments.The first-order RC circuit model was selected,the model parameters were identified online using the recursive least square method.The ohmic internal resistance,polarization internal resistance and polarization capacitance of the battery were used as input parameters of the model.Compared with the traditional BP algorithm,the battery SOH error fluctuation range of the FA-BP optimization algorithm was reduced by 2.50%,the maximum error was reduced by 3.00%,the average error was reduced by 1.68%,which had good convergence.

关 键 词：萤火虫算法(FA) 锂离子电池 BP神经网络 一阶RC电路模型 健康状态(SOH) 

分 类 号：TM912.9[电气工程—电力电子与电力传动]