基于BP神经网络的锂电池SOC在线精确估算

Accurate estimation of charge state of lithium battery based on BP neural network

文中以4节12 V的串联锂离子电池组模块为研究对象,通过实验采集动力电池充放电时的电压、电流、温度、内阻和放电量数据来估算电池的荷电状态(State Of Charge,SOC),重点考虑内阻对动力电池SOC预测结果的影响。以动力电池的电压、电流、温度和内阻作为输入,SOC作为输出,建立四输入一输出的神经网络仿真模型。实验结果表明SOC的预测精度为1.6%,比未考虑电池内阻的预测精度提高45%左右。本文提出的预测方法,其运行时间为0.27 s左右,比不考虑电池内阻时稍有延长,但完全能满足不同工况动力电池充放电时SOC在线估算的速度要求,从而能实现SOC的在线准确预测。

In this paper,the 4 section 12 V series lithium ion battery module is taken as the research object.The voltage,current,temperature,internal resistance and discharge data are collected to estimate the charge state of the battery(State Of Charge,SOC)by experiment.The influence of internal resistance on the SOC prediction results of the power cell is mainly considered.Taking the voltage,current,temperature and internal resistance of the power battery as input and SOC as output,a neural network simulation model with four inputs and one output is established.The results show that the prediction accuracy of SOC is 1.6%,which is 45%higher than that without considering the internal resistance of battery.The prediction method proposed in this paper has a running time of about 0.27s,which is a little longer than that without the internal resistance of the battery,but it can fully meet the speed requirements of the SOC on-line estimation of the battery charging and discharging in different working conditions,thus the accurate online prediction of SOC can be realized.