A Hybrid Ensemble Deep Learning Approach for Early Prediction of Battery Remaining Useful Life

Accurate estimation of the remaining useful life(RUL)of lithium-ion batteries is critical for their large-scale deployment as energy storage devices in electric vehicles and stationary storage.A fundamental understanding of the factors affecting RUL is crucial for accelerating battery technology development.However,it is very challenging to predict RUL accurately because of complex degradation mechanisms occurring within the batteries,as well as dynamic operating conditions in practical applications.Moreover,due to insignificant capacity degradation in early stages,early prediction of battery life with early cycle data can be more difficult.In this paper,we propose a hybrid deep learning model for early prediction of battery RUL.The proposed method can effectively combine handcrafted features with domain knowledge and latent features learned by deep networks to boost the performance of RUL early prediction.We also design a non-linear correlation-based method to select effective domain knowledge-based features.Moreover,a novel snapshot ensemble learning strategy is proposed to further enhance model generalization ability without increasing any additional training cost.Our experimental results show that the proposed method not only outperforms other approaches in the primary test set having a similar distribution as the training set,but also generalizes well to the secondary test set having a clearly different distribution with the training set.The PyTorch implementation of our proposed approach is available at https://github.com/batteryrul/battery_rul_early_prediction.

A machine learning framework for remaining useful lifetime prediction of li-ion batteries using diverse neural networks

Accurate prediction of the remaining useful life(RUL)of lithium-ion batteries(LIBs)is pivotal for enhancing their operational efficiency and safety in diverse applications.Beyond operational advantages,precise RUL predictions can also expedite advancements in cell design and fast-charging methodologies,thereby reducing cycle testing durations.Despite artificial neural networks(ANNs)showing promise in this domain,determining the best-fit architecture across varied datasets and optimization approaches remains challenging.This study introduces a machine learning framework for systematically evaluating multiple ANN architectures.Using only 30%of a training dataset derived from 124 LIBs subjected to various charging regimes,an extensive evaluation is conducted across 7 ANN architectures.Each architecture is optimized in terms of hyperparameters using this framework,a process that spans 145 days on an NVIDIA GeForce RTX 4090 GPU.By optimizing each model to its best configuration,a fair and standardized basis for comparing their RUL predictions is established.The research also examines the impact of different cycling windows on predictive accuracy.Using a stratified partitioning technique underscores the significance of consistent dataset representation across subsets.Significantly,using only the features derived from individual charge–discharge cycles,our top-performing model,based on data from just 40 cycles,achieves a mean absolute percentage error of 10.7%.

Life prediction of Ni-Cd battery based on linearWiener process

Predicting the life of Ni-Cd battery for electric multiple units(EMU)can not only improve the safety and reliability of battery,but also reduce the operating costs of EMU.For this reason,a life prediction method based on linear Wiener process is proposed,which is suitable for both monotonic and non-monotonic degraded systems with accurate results.Firstly,a unary linear Wiener degradation model is established,and the parameters of the model are estimated by using the expectation-maximization algorithm(EM).With the established model,the remaining useful life(RUL)of Ni Cd battery and its distribution are obtained.Then based on the unary Wiener process degradation model,the correlation between capacity and energy is analyzed through Copula function to build a binary linear Wiener degradation model,where its parameters are estimated using Markov Chain Monte Carlo(MCMC)method.Finally,according to the binary Wiener process model,the battery RUL and its distribution are acquired.The experimental results show that the binary linear Wiener degradation model based on capacity and energy possesses higher accuracy than the unary linear wiener process degradation model.

Remaining useful life prediction of aircraft lithium-ion batteries based on F-distribution particle filter and kernel smoothing algorithm

As an emergency and auxiliary power source for aircraft,lithium(Li)-ion batteries are important components of aerospace power systems.The Remaining Useful Life(RUL)prediction of Li-ion batteries is a key technology to ensure the reliable operation of aviation power systems.Particle Filter(PF)is an effective method to predict the RUL of Li-ion batteries because of its uncertainty representation and management ability.However,there are problems that particle weights cannot be updated in the prediction stage and particles degradation.To settle these issues,an innovative technique of F-distribution PF and Kernel Smoothing(FPFKS)algorithm is proposed.In the prediction stage,the weights of the particles are dynamically updated by the F kernel instead of being fixed all the time.Meanwhile,a first-order independent Markov capacity degradation model is established.Moreover,the kernel smoothing algorithm is integrated into PF,so that the variance of the parameters of capacity degradation model keeps invariant.Experiments based on NASA battery data sets show that FPFKS can be excellently applied to RUL prediction of Liion batteries.

Remaining useful life prediction of lithium-ion battery based on auto-regression and particle filter

Purpose-With the rapid development and stable operated application of lithium-ion batteries used in uninterruptible power supply(UPS),the prediction of remaining useful life(RUL)for lithium-ion battery played an important role.More and more researchers paid more attentions on the reliability and safety for lithium-ion batteries based on prediction of RUL.The purpose of this paper is to predict the life of lithium-ion battery based on auto regression and particle filter method.Design/methodology/approach-In this paper,a simple and effective RUL prediction method based on the combination method of auto-regression(AR)time-series model and particle filter(PF)was proposed for lithiumion battery.The proposed method deformed the double-exponential empirical degradation model and reduced the number of parameters for such model to improve the efficiency of training.By using the PF algorithm to track the process of lithium-ion battery capacity decline and modified observations of the state space equations,the proposed PF t AR model fully considered the declined process of batteries to meet more accurate prediction of RUL.Findings-Experiments on CALCE dataset have fully compared the conventional PF algorithm and the AR t PF algorithm both on original exponential empirical degradation model and the deformed doubleexponential one.Experimental results have shown that the proposed PFtAR method improved the prediction accuracy,decreases the error rate and reduces the uncertainty ranges of RUL,which was more suitable for the deformed double-exponential empirical degradation model.Originality/value-In the running of UPS device based on lithium-ion battery,the proposed AR t PF combination algorithm will quickly,accurately and robustly predict the RUL of lithium-ion batteries,which had a strong application value in the stable operation of laboratory and other application scenarios.

基于SCSSA-CNN-BiLSTM的行驶工况下锂电池寿命预测

随着锂离子电池广泛应用,电池寿命预测的重要性日益突显。锂离子电池剩余寿命(RUL)的准确预测是其健康管理的关键组成部分。基于此提出了一种名为SCSSA-CNN-BiLSTM的算法,旨在实现应用于整车的锂离子电池RUL预测。采用卷积神经网络(CNN)和双向长短时记忆神经网络(BiLSTM),并结合了正余弦和柯西变异的麻雀优化算法(sine-cosine and Cauchy mutation sparrow search algorithm, SCSSA),形成了一种新型的混合神经网络,以提高锂离子电池RUL预测的准确性和稳定性。CNN用于电池健康状态(SOH)深度特征的全面提取,而BiLSTM以双向方式研究这些深度特征,并通过密集层生成锂离子电池的RUL预测。为验证所提出方法的有效性,首先使用NASA的电池数据,将多个常用模型与所提出的混合神经网络模型进行比较。研究结果显示,混合模型的决定系数(R2)提高了4%~23%,RUL绝对误差降至1,这表明模型具备更高的预测准确性。随后,在整车层面进行了CLTC动态工况下的循环试验,并对寿命衰减数据进行了预测。最终的结果显示,SCSSA-CNN-BiLSTM模型对应的均方根误差(RMSE)、R2分别为1.64、0.98 Ah,展现出了良好的预测和泛化性能。

基于改进GWO–SVR算法的锂电池剩余寿命预测

锂离子电池性能优越,已在B787等机型上得到应用.锂离子电池性能随着使用次数增加而衰退,准确预测锂电池剩余使用寿命从而及时维护/更换,对航班安全飞行具有重要意义.面向锂离子电池剩余寿命预测问题,本文采用容量增量分析等方法提取特征,基于灰色关联分析计算特征与电池容量的关联程度并筛选特征,提出一种基于改进灰狼优化算法(Improved grey wolf optimization, IGWO)和支持向量回归(Support vector regression, SVR)的锂离子电池剩余寿命预测方法.作为近年研究热点的灰狼优化(Grey wolf optimization, GWO)算法寻优性能出色,但是在应用中容易陷入局部最优.针对此问题,IGWO对GWO算法中的位置更新方程进行优化,对狼群中的个体添加了记忆与飞行功能,增强了算法全局搜索和收敛能力;同时基于Skew Tent映射产生混沌序列,优化狼群初始位置分布.基于标准测试函数对比GWO和IGWO算法的寻优能力,结果表明IGWO算法的收敛速度和寻优效果更好,能够避开GWO陷入的局部最优,在部分测试函数上将寻优精度提升了几十个数量级;基于NASA锂离子电池数据集开展IGWO–SVR、GWO–SVR和SVR的剩余寿命预测能力对比实验,结果证明IGWO–SVR能够有效提高预测精度,与GWO–SVR相比预测均方根误差值降低了10%以上.

电动汽车动力锂离子电池可靠性关键技术综述

随着动力锂离子电池在电动汽车行业发挥关键性作用,保证其工作可靠性已成为当下研究热点。文中综述锂离子电池材料和制造工艺;详细归纳总结电池状态估算方法以及剩余寿命预测方法,并讨论各种方法的优缺点;从电池管理系统层面,先后梳理均衡管理系统和热管理系统相关知识;从电动汽车混合储能系统层面阐述实际工况下性能退化机理及相关技术。最后总结电动汽车动力锂离子电池与可靠性相关的四个方面关键技术的现状,并展望未来发展可能。

融合K-means聚类和序列分解的实车锂电池剩余使用寿命预测

电动汽车锂离子动力电池健康状态(SOH)衰退过程受使用工况影响存在较多波动,导致模型预测精度下降,在锂电池剩余使用寿命(RUL)短期预测时,SOH波动情况不可忽略,为了准确预测SOH短期内波动情况,须从实车上传的锂电池运行数据中提取有效的健康因子。本文建立一种联合分布特征输入和序列分解融合的锂电池RUL预测方法,使用K-means聚类方法构建车辆锂电池运行过程的联合分布特征,并通过S-G滤波器对SOH衰退曲线进行序列分解,分别使用长短时记忆神经网络(LSTM)和多层感知机(MLP)对趋势部分和波动部分进行预测,融合得到最终预测结果。理论分析和实车采集数据验证表明,融合模型可以在预测车辆锂电池RUL短期衰退趋势的同时预测SOH的波动情况,有较高的短期预测精度。

基于PCA-GWO-GRU的锂离子电池剩余使用寿命预测

为了提高GRU神经网络模型预测锂离子电池剩余使用寿命时的准确性,提出基于PCA-GWO优化的GRU模型,并应用于锂离子电池剩余寿命预测。结果表明,与传统GRU模型相比,经PCA-GWO算法优化的GRU模型具有更高的预测精度。预测起始点为原始数据90%时,预测精度达到最大,对应的均方根误差RMSE为0.0049、平均绝对误差MAE为0.0036、决定系数R^(2)为0.9863。

基于多尺度分解的LSTM-ARIMA锂电池寿命预测

锂电池剩余使用寿命(Remaining useful life,RUL)预测是锂电池研究的一个重要方向,通过对RUL的准确预测,可以更好地管理和维护电池,延长电池使用寿命。为了能够准确预测锂电池的RUL,提出了一种集合变分模态分解(Variational mode decomposition,VMD)、长短时记忆网络(Long short-term memory,LSTM)和自回归移动平均模型(Autoregressive integrated moving average,ARIMA)相结合的锂电池RUL预测模型。该模型首先采用VMD算法将NASA锂电池数据集中的容量数据分解为多个高频分量和低频分量,以此减少容量数据中的噪声干扰,然后针对各个分量的特点,分别利用LSTM和ARIMA对分解所得的高频分量和低频分量建立预测子模型,最后将各个子模型的预测值进行叠加重构得到锂电池的RUL结果。实验结果表明VMD-LSTM-ARIMA预测模型相比于其他预测模型,该模型具有较好的锂电池RUL预测能力。并在CALCE锂电池数据集上进行了泛化性实验,结果表明该模型适用于不同电池RUL预测任务。

基于机器学习方法的锂电池剩余寿命预测研究进展

随着技术的不断进步和成本的逐步降低,锂电池在电动汽车、储能系统、便携式电子设备等多个领域实现了广泛应用,有效促进了清洁能源的普及和能源结构的优化。掌握锂电池衰变和剩余使用寿命(RUL)对于确保设备稳定运行、提高能源利用效率以及保障用户安全至关重要。通过优化电池设计和使用策略,可以延长锂电池的使用寿命,降低更换成本,进一步推动锂电池的规模化应用。锂电池的性能衰变是一个涉及多尺度化学、电化学反应的复杂过程,涉及其内部从材料、界面到多孔电极、器件等诸多因素影响。各种机器学习(ML)的方法正是建模处理复杂数据、寻找规律、反馈应用的重要手段。本文针对锂电池RUL建模研究的科学问题,综述了ML算法在预测电池RUL领域的最新进展,重点介绍数据驱动的电池管理、预测建模以及利用ML方法来提高电池性能和寿命方面的突破。最后,对当前领域内面临的关键问题进行了归纳总结,以期提供一个基于ML算法的电池RUL预测技术的全面视角,并展望其未来的发展趋势。

混合模糊信息粒化和时间序列密集编码器的锂离子电池剩余寿命区间预测方法

锂离子电池的剩余使用寿命(remaining useful life,RUL)是电池健康状态的关键指标,对其进行预测具有重要的现实意义。该工作将模糊信息粒化(fuzzy information granulation,FIG)技术与时间序列密集编码器模型(timeseries dense encoder,TiDE)相结合,提出了一种对锂离子电池的RUL进行区间预测的模型。首先将锂离子电池容量退化时间序列通过FIG技术转化为粒子序列信息,以此得到模糊信息粒子的上下界序列。其次,分别对上下界序列使用TiDE模型进行训练预测,从而得到区间预测的结果。实验结果表明,与基于支持向量回归(support vector regression,SVR)和长短期记忆网络(long short term memory network,LSTM)的区间预测模型以及不使用狐狸优化算法(fox-inspired optimization algorithm,FOA)优化的TiDE模型相比,该工作提出的基于FIG技术结合TiDE模型与FOA的区间预测方法在锂离子电池RUL预测性能上具有更高的可靠性。

基于弛豫电压模型的锂离子电池RUL预测

锂离子电池在长期使用过程中呈非线性退化趋势,预测非线性退化对延长电池寿命和确保安全意义重大。提出一种利用弛豫电压作为特征序列的非线性退化拐点预测方法,进行拐点和剩余使用寿命(RUL)的联合预测,建立结合拐点退化特征的RUL预测框架,提高预测精度。通过迁移学习,在不同的电池数据集上验证所提联合预测方法的性能,拐点和RUL预测的平均绝对误差在26次循环内,均方根误差低于28次循环。该方法利用弛豫电压来预测拐点和RUL,从而间接预测电池健康状态(SOH),具有预测精度好、应用范围广等特点。

基于VMD和Bat-KELM的仿真变电站蓄电池剩余寿命预测

仿真变电站蓄电池的工作模式呈现间歇非连续性,导致电池性能在退化过程中存在容量再生现象,退化规律具有非平稳性和随机性,增大了蓄电池精确剩余寿命RUL(remaining useful life)的难度。针对存在容量再生现象的蓄电池剩余寿命预测问题,提出了变分模态分解VMD(variational mode decomposition)和蝙蝠(Bat)优化核极限学习机KELM(kernel extreme learning machine)组合的预测方法。基于VMD将蓄电池健康状态SOH(state of health)时间序列分解为整体退化分量和容量再生分量;利用Bat优化KELM构建各分量预测模型,以提高分量趋势预测精度;通过各分量独立预测结果的叠加,得到精确的蓄电池健康状态及剩余寿命预测值。将该方法应用于蓄电池退化数据实例分析中,结果表明该方法相较于KELM模型及VMD-KELM模型,预测精度更高,验证了该方法的优越性。

Application of Digital Twin in Smart Battery Management Systems

Lithium-ion batteries have always been a focus of research on new energy vehicles,however,their internal reactions are complex,and problems such as battery aging and safety have not been fully understood.In view of the research and preliminary application of the digital twin in complex systems such as aerospace,we will have the opportunity to use the digital twin to solve the bottleneck of current battery research.Firstly,this paper arranges the development history,basic concepts and key technologies of the digital twin,and summarizes current research methods and challenges in battery modeling,state estimation,remaining useful life prediction,battery safety and control.Furthermore,based on digital twin we describe the solutions for battery digital modeling,real-time state estimation,dynamic charging control,dynamic thermal management,and dynamic equalization control in the intelligent battery management system.We also give development opportunities for digital twin in the battery field.Finally we summarize the development trends and challenges of smart battery management.

实车数据驱动的锂电池剩余使用寿命预测方法研究

锂离子动力电池剩余使用寿命(RUL)预测对于认识全生命周期电动汽车的安全和可靠性、改善电池管理系统的设计具有重要意义。通常基于深度学习的时序预测方法,本质上是一个递推的过程,每一次预测的误差会随预测次数增加而累积,难以保证预测精度和预测效率。基于深度学习序列预测和误差分析理论,建立一种ARIMA-EDLSTM融合模型的锂电池RUL预测方法,使用编码器-解码器(ED)框架改进长短时记忆神经网络模型(LSTM)构建从序列到序列预测的EDLSTM模型,并融合ARIMA模型预测误差趋势,进而修正最终预测结果。理论分析和实车采集数据验证表明,该方法在预测比例超过历史数据总量35%的情况下,仍然能较好地拟合实车SOH衰退曲线,有效提高锂电池剩余使用寿命的预测精度。

融合经验老化模型和机理模型的电动汽车锂离子电池寿命预测方法研究

为提升实际应用中锂离子动力电池寿命预测精度,本文中提出一种融合经验老化模型和电池机理模型的电池寿命预测方法。该方法以基于经验老化模型SOH预测值作为卡尔曼算法的先验估计,以基于机理模型估计电池未来容量衰减量进而预测得到的SOH作为卡尔曼算法的后验修正,从而实现对锂离子电池寿命的准确预测。基于电芯试验数据的动力电池寿命预测算法验证结果表明,锂离子动力电池剩余寿命预测误差≤5.83%、基于实车数据的锂离子动力电池的剩余寿命预测误差≤8.12%,取得了良好的预测效果,丰富了锂离子动力电池寿命预测的方法。

基于VMD-PSO-GRU模型的锂离子电池剩余寿命预测

准确预测锂电池的剩余使用寿命(remaining useful life,RUL)对降低电池使用风险和保证系统的安全运行起着非常重要的作用。为了消除电池容量序列受容量再生等影响,提高预测结果的准确性和稳定性,提出了一种基于变分模态分解(variational modal decomposition,VMD)与参数优化的门控循环神经网络(gate recurrent unit,GRU)相结合的RUL预测模型。首先采用VMD算法将锂电池的容量序列分解为一系列平稳分量;然后采用多层GRU网络对各分量进行预测,针对预测结果不稳定的问题,在模型训练前利用粒子群算法(particle swarm optimization,PSO)对GRU模型的参数进行优化;最后叠加各分量的预测值作为最终预测结果。在NASA数据集上对本模型进行了验证,当采用20个已知电池序列数据预测时,预测结果的最大平均绝对百分比误差和均方根误差控制在0.88%和0.0148以内,RUL预测的最大误差不超过2个充电周期,具有较高的鲁棒性和预测精度。

基于SSA-VMD-GRU的锂电池剩余寿命预测方法研究

锂离子电池性能在衰退过程中呈现非平稳性和非线性,寿命预测往往被再生容量所干扰,衰退趋势难以捕捉,进而影响寿命预测。针对该问题,以容量为特征,构建一种基于麻雀搜索算法优化变分模态分解和门控循环单元的锂离子电池寿命预测方法。首先,利用麻雀搜索算法优化了变分模态分解的分解层数和惩罚因子,再通过优化了的变分模态分解算法将锂电池容量分解为若干分量,最后引入门控循环单元预测所分解的若干分量,将若干预测结果进行集成。通过NASA电池数据集对所提方法进行验证,并与两种模型相比较,结果表明,该方法相较于另两种方法预测精度平均提升了60%。