基于数据扩展的锂离子电池早期寿命在线预测  

Online prediction of early life of lithium-ion batteries based on data extension

作　　者：李超 汪伟 安斯光 邹国平 LI Chao;WANG Wei;AN Siguang;ZOU Guoping(College of Information Engineering,China Jiliang University,Hangzhou 310018,China)

机构地区：[1]中国计量大学机电工程学院,浙江杭州310018

出　　处：《现代电子技术》2024年第10期171-176,共6页Modern Electronics Technique

基　　金：国家自然科学基金项目(52077203);浙江省属高校基本科研业务费(2021YW06)。

摘　　要：准确的锂离子电池早期剩余使用寿命预测可以确保用户在早期阶段对锂离子电池进行监控,从而为用户提供早期规划。已有的研究方法在训练样本时存在数据利用不充分的问题,因此提出一种充分利用离线电池的历史数据对在线早期数据进行扩展,从而实现对电池使用寿命早期预测的方法。离线建模阶段,使用高斯过程回归(GPR)模型将离线电池所有循环的数据进行训练,建立锂离子电池健康特征随循环次数变化的时序关系模型。在线预测阶段,使用高斯过程回归模型经由前100个周期数据扩展得到后期健康特征,之后将生成的后期特征融合到早期特征中,形成最终的全生命周期特征。最后通过预训练的一维卷积神经网络-长短期记忆模型进行早期预测。在公开的夏威夷NMC-18650电池退化数据集中的测试结果显示,早期预测相对误差小于1.2%,证明了该方法的有效性。Accurate early remaining useful life(RUL)prediction of Lithium-ion batteries can ensure that Lithium-ion batteries are monitored at an early stage,thus providing users with early planning.The existing research methods have the problem of insufficient data utilization when training samples.Therefore,a method is proposed to fully utilize the historical data of offline batteries,expand online early data,and achieve early prediction of battery service life.In the offline modeling stage,data from all cycles of an offline battery are trained by means of Gaussian process regression(GPR)model to construct the time-series relationship model of lithium-ion battery health features with the number of cycles.In the online prediction stage,a GPR model is used to expand the data from the first 100 cycles to obtain later health features,which are then fused into the early features to form the final full lifecycle features.The early prediction is conducted by means of the pretrained 1 dimension convolutional neural network-long short-term memory model.The testing results in the publicly available Hawaii NMC-18650 battery degradation datasets show that the relative error of early prediction is less than 1.2%,proving the effectiveness of the proposed method.

关 键 词：锂离子电池 剩余使用寿命预测 早期预测 数据扩展 高斯过程回归 CNN-LSTM 

分 类 号：TN86-34[电子电信—信息与通信工程]