基于CEEMDAN-RVM-LSTM模型的锂电池剩余寿命预测  

Remaining life prediction of lithium-ion batteries based onCEEMDAN-RVM-LSTMmodel

作　　者：牛群峰[1] 袁强 王莉[1] 刘江鹏 NIU Qunfeng;YUAN Qiang;WANG Li;LIU Jiangpeng(College of Electrical Engineering,Henan University of Technology,Zhengzhou Henan 450001,China)

机构地区：[1]河南工业大学电气工程学院,河南郑州450001

出　　处：《电源技术》2023年第10期1313-1318,共6页Chinese Journal of Power Sources

基　　金：河南省科技攻关项目(201300210100)。

摘　　要：为了提高锂电池长期使用的可靠性和保证系统的安全运行,提出了一种结合自适应噪声完整集合经验模态分解算法(CEEMDAN)、相关向量机(RVM)和长短期记忆神经网络(LSTM)的剩余使用寿命(RUL)的预测方法。使用CEEMDAN将电池容量数据分解为本征模态分量和残差分量,分别由RVM和LSTM进行预测,最后进行有效集成,得到准确的容量和RUL预测结果,并获得RUL的95%置信区间。采用公共数据集进行实验验证,并对比了其他几种模型。实验结果表明该方法不仅拥有较高的预测精度,而且能够提供不确定性表达,具有良好的工程应用意义。In order to improve the reliability of long-term use of lithium-ion batteries and ensure safe operation of the system,a prediction method for the remaining useful life(RUL)combining the adaptive noise complete ensemble empirical mode decomposition algorithm(CEEMDAN),correlation vector machine(RVM)and long and short-term memory neural network(LSTM)was proposed.The battery capacity data were decomposed into eigenmodal and residual components using by CEEMDAN,predicted by RVM and LSTM respectively,and finally integrated efficiently to obtain accurate capacity and RUL predictions with 95%confidence intervals of RUL.A public dataset was used for experimental validation and several other models were compared.The experimental results show that the method not only possesses high prediction accuracy,but also provides uncertainty expression,which has good engineering application significance.

关 键 词：锂电池 剩余使用寿命 RVM LSTM 不确定性表达 

分 类 号：TM912[电气工程—电力电子与电力传动]