锂离子电池的荷电状态(State of charge,SOC)是电池管理系统的重要参数之一.针对开路电压法估计电池SOC需要将电池长时间静置的问题,提出基于松弛特性和主成分分析(Principal component analysis,PCA)的锂离子电池SOC估计模型.首先,提出利用电池电压松弛曲线估计SOC的方法,并据此搭建了门控循环单元神经网络(Gated recurrent unit recurrent neural network,GRU-RNN)模型,使电池静置时间相比开路电压法大幅缩短;然后,针对电压松弛曲线数据维度较高的问题,采用PCA方法对输入数据降维,降低了GRU-RNN模型的复杂度;最后,设计了锂离子电池周期放电试验和动态放电试验,完成了电池电压松弛曲线及 SOC 数据收集,并用于模型训练和测试.试验结果表明,针对各恒流放电或动态放电工况,所提出的锂离子电池SOC估计方法在电池短时间静置的情况下仍具有高精度,PCA方法有效缩短了模型训练时间.
State of Charge Estimation for Lithium-ion Battery Based on Relaxation Effect and Principal Component Analysis
State of charge(SOC)of lithium-ion battery is one of the most important parameters for battery management system.The open-circuit voltage method requires the battery to rest for a long time for to SOC estimation.To solve this problem,a SOC estimation model based on relaxation effect and principal component analysis(PCA)is proposed for lithium-ion batteries.First,a SOC estimation method based on the voltage relaxation curve is proposed,and a gated recurrent unit recurrent neural network(GRU-RNN)model is built,which significantly shorten the resting time compared with the open-circuit voltage method.Then,in order to solve the problem of high dimensionality of voltage relaxation curve data,PCA method is used to reduce the complexity of the GRU-RNN model by reducing the dimensionality of the input data.Finally,the lithium-ion battery periodic discharge experiment and dynamic discharge experiment is designed,and the battery voltage relaxation curve and SOC data are collected and used for model training and testing.The experiment results show that for each constant current discharge or dynamic discharge condition,the proposed SOC estimation method has high accuracy when the battery is rested for a short time,and the PCA method effectively reduces the model training time.
Lithium-ion batterystate of chargerelaxation effectprincipal component analysis
NETL
NSTL
万方数据
