基于BP神经网络的梯次利用电池健康状态诊断

State of health estimation for echelon-used batteries based on BP neural network

在大规模储能产业迅猛发展及退役车用动力电池数量逐年增长的背景下,阐述了梯次利用电池及其储能应用场景,以及梯次利用电池健康状态估算的重要性。介绍了影响电池健康状态的几种因素,将电池直流内阻、放电倍率及表面温度作为输入构建了3层反向传播(BP)神经网络。试验表明:在30块梯次利用电池的样本训练下,网络能够有效收敛且对梯次利用电池健康状态的计算误差在3%内,根据BP神经网络估算电池健康状态具有一定的可行性,该方法对梯次利用电池的分选以及储能应用具有重大意义。

In the context of rapid developing energy storage industry and the gradually increasing decommissioned power batteries for vehicles,the echelon-used batteries and their application scenarios in energy storage are introduced,and the necessity of their state of health is expounded.Influence factors for battery state of health are discussed.A three-layer BP neural network is constructed by taking battery DC resistance,discharge rate and surface temperature as inputs.Experiment results show that trained by 30 echelon-used batteries,the network can effectively converge and keep the health state estimation errors of echelon-used batteries within 3%.Estimating battery state of health of batteries with BP neural network is feasibility and of great significance in sorting as well as energy storage for echelon-used batteries.