SIA OpenIR  > 工业控制网络与系统研究室
基于粒子滤波算法的锂离子电池剩余寿命预测方法研究
Alternative TitleResearch on Remaining Useful Life Prediction of Lithium-ion Battery with Particle Filter
张凝1,2
Department工业控制网络与系统研究室
Thesis Advisor徐皑冬
Keyword锂离子电池 剩余使用寿命 粒子滤波 自回归模型 扩展卡尔曼粒子滤波
Pages81页
Degree Discipline检测技术与自动化装置
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract锂离子电池作为众多设备的能量来源,其安全性和可靠性对系统的正常运行起着至关重要的作用,因此提前预测其剩余使用寿命是系统故障预测和健康管理(Progonostics and Management, PHM)至关重要的环节。粒子滤波算法不受系统模型的限制,针对非高斯非线性系统依然有很好的参数估计能力,且能给出估计结果的置信程度分布函数,对于电池这样的复杂系统有很好的适应性,因此本文将基于粒子滤波算法预测电池剩余使用寿命。粒子滤波方法需要构建系统模型来逼近真实系统的特性。本文以锂电池容量的衰退为电池寿命衰退的特征,构建了基于电池容量的双指数经验退化模型。该模型将作为粒子滤波算法系统模型中的状态转移方程,为粒子滤波算法产生的大量粒子指导分布。本文对该模型进行了变形,这样做不仅使模型的形式顺应了粒子滤波的状态转移方程,而且减少了参数个数,降低了参数训练的难度。随后针对粒子滤波算法对系统建模的观测方程由状态值加观测噪声得出,导致整个粒子滤波算法对系统模型的构建完全依赖于电池容量衰退模型,而电池容量衰退模型对电池寿命的表达能力也极其有限,为减少粒子滤波算法对该模型的过度依赖,引入了自回归模型(本文叫做AR时间序列模型)修正观测值,以提高预测的准确性;为了使指导粒子分布的函数更接近真实的分布,本文使用扩展卡尔曼粒子滤波算法对粒子滤波算法加以改进。该算法由扩展卡尔曼滤波算法进行粒子滤波算法中的重要性采样环节,其优势在于使用扩展卡尔曼滤波算法为每个粒子产生均值和方差时,吸纳了最新的观测信息,因此产生的建议密度分布函数更接近真实的分布。这样做可以提升粒子滤波算法的参数估计性能,进一步提高锂电池剩余使用寿命预测的准确性。本文对模型的训练数据和验证预测效果的试验数据均来自NASA PCoE电池数据集。实验结果表明,本文基于粒子滤波算法的锂电池剩余使用寿命预测方法能够给出预测结果和不确定性分布。基于标准粒子滤波算法的改进方法能够极大地提升预测准确性。
Other AbstractThe safety and reliability of lithium-ion batteries play an important role in systems to provide energy to many devices. Therefore, remaining useful life prediction of lithium-ion batteries is very important in progonostics and management (PHM). Particle filter algorithm has good parameter estimation ability in non-Gaussian and non-linear systems, it can also analyze uncertainty of prediction results. Above all, particle filter algorithm is very suit for complex systems such as lithium-ion batteries. It will be the key point in this paper. System model needs to be built to approximate the characteristics of real system in particle filter algorithm. In this paper, a double exponential empirical degradation model based on the capacity loss of lithium battery is constructed. The model will be used as the state transition equation in particle filter algorithm. The distribution of particles generated by the particle filter algorithm is guided by the model. In this paper, the model is deformed to adapt to the state transition equation of particle filter and reduce the number of parameters and the difficulty of parameter training. Because the observation equation of the system model is composed of the state value and the observation noise, it will lead to a complete dependence on the model for particle distribution. The problem is the degradation model has very limited expression ability for battery life. Overdependence on the model can cause a decline in prediction accuracy. To solve this problem, an autoregressive model is introduced to correct observations and improve the prediction accuracy. The extended Kalman particle filter algorithm is introduced to guide the distribution of particles. It will make the distribution closer to the real. The extended Kalman filter is uesd to generate the importance function in the extended Kalman particle filter. Compared to the standard particle filter algorithm, the latest observation information is considered when the mean and variance of each particle is generated by extended Kalman filter in extended Kalman particle filter algorithm, so the proposed density distribution is closer to the real distribution. This method can improve the performance of particle filter algorithm and accuracy of remaining useful life prediction of lithium batteries. Training data of the model and test data of the method are all from NASA PCoE battery set. The experimental results show that particle filter algorithm is suit for remaining useful life prediction of lithium-ion batteries. The improved methods based on particle filter algorithm greatly improve the accuracy of prediction.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25194
Collection工业控制网络与系统研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
张凝. 基于粒子滤波算法的锂离子电池剩余寿命预测方法研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
基于粒子滤波的锂离子电池剩余寿命预测方法(2804KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张凝]'s Articles
Baidu academic
Similar articles in Baidu academic
[张凝]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张凝]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.