电工技术学报2025，Vol.40Issue(9)：2996-3012,17.DOI:10.19595/j.cnki.1000-6753.tces.241243
基于融合特征和OOA-BiGRU的锂离子电池剩余使用寿命预测方法
A Novel Remaining Useful Life Prediction Method Based on Fusion Feature and OOA-BiGRU for Lithium-Ion Batteries
摘要
Abstract
With the continuous development of the new energy vehicle industry,lithium-ion batteries are used in large quantities as on-board power batteries.The battery management system(BMS)is responsible for monitoring,evaluating,maintaining,and optimizing the performance and life of Li-ion batteries,and the prediction of c is an important part of the BMS.Accurate prediction of a battery's RUL helps identify batteries that are nearing the end of their life to prevent potential safety risks such as overheating,combustion,or explosion,and allows O&M personnel to schedule battery maintenance and replacements based on the battery's actual state of health,rather than on a pre-determined schedule,thereby reducing unnecessary O&M costs.However,lithium-ion batteries exhibit nonlinear aging trends due to their complex internal chemical reactions during use,and the aging process of batteries usually exhibits multi-stage degradation,which increases the difficulty of RUL prediction.In view of this,this paper proposes a RUL prediction method for lithium-ion batteries based on public battery data from the University of Maryland and lithium iron phosphate battery data collected by the group's own laboratory,and the main research contributions are as follows: Aiming at the problem that battery capacity is difficult to be measured directly,and the poor ability of traditional network models to capture multi-feature input information,a method is proposed to optimize the bidirectional gated recurrent unit(BiGRU)network based on the fusion feature and the osprey optimization algorithm(OOA)for RUL prediction of lithium-ion batteries.Simple and easy-to-measure current,voltage and time data during battery aging are collected,from which the health factors that can reflect the aging trend of the battery are extracted.The Savitzky-Golay filtering method is selected to reduce the influence of noise on the prediction accuracy.A fusion feature screening strategy combining filter and wrapper is proposed to reduce the complexity of the model and prevent model overfitting.Considering the insufficient ability of the traditional model to capture battery aging information when dealing with multi-feature inputs,the GRU network,which predicts only based on historical information,is upgraded to the BiGRU network,which is capable of handling both forward and backward information of the sequences.The BiGRU network is able to understand the overall structure and dynamic properties of the sequences in a more in-depth manner,better integrate the multi-dimensional features,and adapt to dependencies in different time scales.OOA is used to effectively optimize the hyper parameters inside the BiGRU model,which improves the prediction accuracy of the model and also realizes the automatic configuration of the parameters.Different types of battery data are used to compare the proposed method with traditional network models to verify the reliability of the proposed OOA-BiGRU model.In addition,the effect of the proposed fusion feature prediction is compared with all feature prediction and filtered feature prediction,which proves that the fusion feature better represents the aging degree of the battery and improves the accuracy of the model prediction. The research results of this paper provide a new method and idea for RUL prediction of lithium-ion power batteries,which can be applied to the BMS system of new energy vehicles and is of practical significance.关键词
锂离子电池/剩余使用寿命/双向门控循环单元/健康因子/融合特征Key words
Lithium-ion batteries/remaining useful life(RUL)/bidirectional gated recurrent unit(BiGRU)/health factor(HF)/fusion feature分类
信息技术与安全科学引用本文复制引用
孙静,翟千淳..基于融合特征和OOA-BiGRU的锂离子电池剩余使用寿命预测方法[J].电工技术学报,2025,40(9):2996-3012,17.基金项目
烟台市科技创新发展计划基础研究类项目(2023JCYJ043)、山东省自然科学基金项目(ZR2021ME236)和山东省高校青年创新团队科技支撑计划(2020KJN005)资助. （2023JCYJ043）
