摘要
锂离子电池剩余使用寿命(RUL)可以评估电池的可靠性,是电池健康管理的重要参数。准确地预测电池的RUL可以有效提高设备的安全性并降低工作风险。该文提出一种自适应数据预处理结合长短期记忆神经网络(LSTM)的RUL预测框架。选取容量作为健康因子,数据预处理阶段,首先使用自适应双指数模型平滑方法减少容量回升现象产生的负面影响,然后通过自适应白噪声完整集成经验模态分解(CEEMDAN)对数据进行降噪;模型构建阶段,利用预处理后的数据训练得到用于RUL预测的LSTM模型。以NASA和CALCE公开数据集为研究对象进行算法性能测试,实验结果表明,所提方法鲁棒性好,能够提供精确的RUL预测结果。
The remaining useful life(RUL) of lithium-ion battery can evaluate the reliability of battery, which is an important parameter of battery health management. Accurate prediction of RUL of battery can effectively improve the safety of equipment and reduce the working risk. In this paper, a RUL prediction framework combined with the adaptive data preprocessing method and long-term and short-term memory neural network(LSTM) was proposed. Selecting capacity as the health factor, in the data preprocessing stage, the adaptive double exponential model smoothing method was used to reduce the negative effect of capacity recovery and the adaptive white noise integrated empirical mode decomposition(CEEMDAN) is used to suppress the noise. In the model constructing stage, the LSTM model was built for RUL prediction by training the preprocessed data. The NASA and CALCE open source data were selected to verify the performance of the proposed method. The experimental results show that it has good robustness and can provide RUL prediction results with high precision.
作者
黄凯
丁恒
郭永芳
田海建
Huang Kai;Ding Heng;Guo Yongfang;Tian Haijian(State Key Laboratory of Reliability and Intelligence of Electrical Equipment Hebei University of Technology, Tianjin 300130 China;School of Artificial Intelligence Hebei University of Technology, Tianjin 300130 China)
出处
《电工技术学报》
EI
CSCD
北大核心
2022年第15期3753-3766,共14页
Transactions of China Electrotechnical Society
基金
河北省自然科学基金资助项目(E2019202328)。
关键词
锂电池
剩余使用寿命
自适应双指数模型平滑方法
自适应白噪声完整集成经验模态分解
长短期记忆神经网络
Lithium-ion battery
remaining useful life
adaptive bi-exponential model smooth method
the complete ensemble empirical mode decomposition with adaptive noise
long short-term memory network
