摘要
针对锂离子电池储能系统寿命预测不准确问题,提出了一种多时间尺度下的电池剩余寿命混合预测方法。首先,通过带自适应噪声的完全集成经验模态分解(CEEMDAN),将原始数据分解为多个低频和高频分量。接着,利用GS-LM模型对其低频分量进行拟合预测,再将分解后的高频分量作为LSM-STW模型(滑动时间窗口和长短期记忆神经网络)的输入,得到高频分量预测结果。最后,将所有预测结果相加,得到最终的锂离子电池剩余寿命预测。实验结果表明,所提出的预测方法可以准确地预测锂离子电池容量退化趋势,不同预测起点对预测结果影响较小,并能够捕捉电池容量的再生现象。
Aiming at the life forecast inaccurate problem of lithium-ion battery energy storage system,a hybrid prediction method of battery remaining useful life under multiple time scales is proposed.First,the original data is decomposed into multiple lowfrequency and high-frequency components through the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN).The GS-LM model is used to fit and predict its low-frequency components,and then the decomposed highfrequency components are used as the input of the LSM-STW model(sliding time window and long&short-term memory neural network)to obtain the high-frequency component prediction results.Finally,all the prediction results are added together to get the final prediction of the remaining life of the lithium-ion battery.Experimental results show that the proposed prediction method can accurately predict the degradation trend of lithium-ion battery capacity.The different prediction starting points have less effects on the prediction results,and it can capture the regeneration phenomenon of battery capacity.
作者
毛玲
胡慧仲
赵晋斌
徐杰
MAO Ling;HU Huizhong;ZHAO Jinbin;XU Jie(Shanghai University of Electric Power,Shanghai 200090,China)
出处
《供用电》
2021年第6期7-13,共7页
Distribution & Utilization
基金
国家自然科学基金项目(51777120)。
关键词
储能锂离子电池
CEEMDAN
混合预测
预测起点
容量再生
energy storage lithium-ion battery
CEEMDAN
hybrid prediction
prediction starting point
capacity regeneration
作者简介
通讯作者:毛玲(1981-),女,博士,讲师,研究方向为锂离子电池建模及状态估计、移动储能的配置及运行、电动汽车与电网的互动及多机器人的通信及协调;胡慧仲(1997-),男,硕士研究生,研究方向为锂离子电池建模及状态估计;赵晋斌(1972-),男,博士,教授,研究方向为电力电子电路、装置与系统、电力电子电路的智能化及模块化技术、现代电力电子技术在电力系统中的应用、新能源发电技术、无线充电技术。
