FYP: Final Year Projects (in National University of Singapore)
优质论文阅读
综述
【Nature 子刊】Predicting the state of charge and health of batteries
文章信息 | 背景、目的及结论 | 结果与讨论 | 文章好在哪里 |
---|---|---|---|
题目:Predicting the state of charge and health of batteries using data-driven machine learning | 背景: “Rechargeable lithium-ion (Li-ion) batteries are currently the best choice for EVs due to their reasonable energy density and cycle life1.” (Ng 等, 2020, p. 161) “An advanced battery management system (BMS) that can monitor and optimize battery behaviour and safety is thus essential for the entire electrification system2.” (Ng 等, 2020, p. 161) “accurate prediction of the current and future state of batteries will open up vast opportunities in battery manufacturing, usage and optimization” (Ng 等, 2020, p. 161) | 目前,研究最多的两种电池状态预测模型是 equivalent circuit models(ECM) 和 physics-based models(PBM)。尽管它们很受欢迎并且不断发展,但在使用这些模型进行在线电池状态预测时,计算效率和准确性之间仍然存在明显的权衡。 具有机器学习功能的Data-driven model (DDM) 是一种很有前景的电池建模方法,有可能解决使用 ECM 或 PBM 的传统建模所面临的困境。 目前,大多数机器学习模型都给出黑匣子电池状态预测,这使得很难推广到其他电池化学成分。领域知识的结合为可解释的“白盒”预测铺平了道路。 此外,高通量实验(也许以初步机器学习结果为指导)是为机器学习提供真实且高质量的电池性能数据集的关键。 随着计算技术和数学算法的进步,加上数据存储设备和高通量实验成本的降低,我们预计数据驱动的机器学习将成为未来实时电池建模的一项有前途的技术。 |
|
期刊:Nature Machine Intelligence-计算机TOP | |||
单位:(A*STAR), Singapore | 目的:
| ||
作者:Man-Fai Ng | |||
文章类型:综述 | 不足: | ||
阅读程度:全文精读 |
【ANN】Applications of artificial neural network based battery management systems: Aliterature review
文章信息 | 背景、目的及结论 | 文章好在哪里 |
---|---|---|
题目:Applications of artificial neural network based battery management systems: A literature review | 背景: Greenhouse 近年来,基于机器学习的研究的流行,特别是基于Meachine Learning的算法提供了比基于模型的算法更高的可用性,使得它们更受研究人员的青睐,即使这两种方法各有优缺点[ 7 ]。 | 主要研究ANN,并且包含了SOTA methods |
期刊:Renewable and Sustainable Energy Reviews-工程技术TOP | ||
单位:Ardahan University, Department of Computer Engineering, Ardahan, Türkiye | 目的:
| |
作者:Mehmet Kurucan | ||
文章类型:综述 | 不足: | |
阅读程度:泛读 |
小论文
SOH 方向:
Driving behavior-guided battery health monitoring for electric vehicles using extreme learning machine
RUL 方向
【代码开源·已复现】A Hybrid Ensemble Deep Learning Approach for Early Prediction of Battery Remaining Useful Life
文章信息 | 背景、目的及结论 | 结果与讨论 | 文章好在哪里 |
---|---|---|---|
题目:A Hybrid Ensemble Deep Learning Approach for Early Prediction of Battery Remaining Useful Life | 背景:
| proposed hybrid model outperforms SOTA approaches in terms of two evaluation criteria in the primary test set, and also has better generalization ability to the secondary test set.![]() | |
在先前的工作[5]中缺少从这些基于领域知识的特征中正确选择关键特征的过程 为了解决这一局限性,作者提出了一种基于斯皮尔曼相关系数的非线性相关特征选择方法,以从领域知识特征中识别最相关的特征。
| |||
单位:Institute for Infocomm Research, Singapore | |||
期刊:IEEE/CAA Journal of Automatica Sinica-Q1 | 目的:有效地将手工特征与深度网络学习到的领域知识和潜在特征相结合,以提高 RUL 早期预测的性能 | ||
**作者:**Qing Xu | |||
不足&展望: 我们打算将基于物理的模型与深度学习模型相结合,因为基于物理的模型以其物理可解释性和准确建模LIB退化过程的出色能力而闻名[37],[38]。这两类模型的结合将产生新的混合模型,这些模型在物理上更加一致、可解释和准确。此外,我们还打算在未来的工作中研究准确预测电池 RUL 所需的最小循环次数,因为它可以显着降低实际应用中电池制造的实验成本。 | |||
代码运行结果
pytorch_hybrid_model_snapshot_train.py
项目进度
-
配置环境
见博客:
-
寻找领域开源数据集
- NASA
- CALCLE
- Oxford
- MIT
- Aging
-
学会读取数据集(特指 .mat 类)的方法
查看数据集内部结构见博客:
MATLAB的.mat格式数据集如何查看内部内容_怎么看mat文件内容-CSDN博客
读取数据集参考NASA数据集处理的开源项目
-
阅读领域综述
-
下载顶刊论文的开源代码并在本地运行复现(hybrid model)
-
简单提取NASA数据集的features, 即Health Indicators
见博客:
-
使用简单的深度模型算法处理提取的HIs来预测电池的State of Health
极限学习机 (ELM) 和 全连接 Dense layer 模型两个模型已经能处理HIs,具体将在后续的博客中展示,目前还并不完善
下一阶段安排:
-
弄懂 SOC (state of health)该怎么使用数据集的数据计算,积分?
-
尝试更多的模型来处理HIs
- WOA-ELM ?
- RNN如LSTM等:之前尝试过,但是开源模型代码无论如何调用都显示报错
-
在同学的合作下加入更多的HIs
下面是梳理的五大最常用的开源数据集地址:
开源数据集
最常见五大的数据集:
NASA dataset
https://c3.ndc.nasa.gov/dashlink/resources/133/
- Batteries 电池 (此数据集在论文中比后者使用更频繁)
数据集格式:.mat
Experiments on Li-Ion batteries. Charging and discharging at different temperatures. Records the impedance as the damage criterion. The data set was provided by the NASA Prognostics Center of Excellence (PCoE).
锂离子电池实验。在不同温度下充电和放电。记录阻抗作为损坏标准。该数据集由 NASA 卓越预测中心 (PCoE) 提供。
- Download Mirror: https://phm-datasets.s3.amazonaws.com/NASA/5.+Battery+Data+Set.zip
- 下载镜像:https://phm-datasets.s3.amazonaws.com/NASA/5.+Battery+Data+Set.zip
- Data Set Citation: B. Saha and K. Goebel (2007). “Battery Data Set”, NASA Prognostics Data Repository, NASA Ames Research Center, Moffett Field, CA
- 数据集引用:B. Saha 和 K. Goebel (2007)。 “电池数据集”,NASA 预测数据存储库,NASA 艾姆斯研究中心,莫菲特菲尔德,加利福尼亚州
数据集结构
- Accelerated Battery Life Testing 加速电池寿命测试
This data set presents accelerated-Li-ion battery lifecycle data focused on a large range of load levels and the characterization of the lifecycle of a battery pack composed of two 18650 battery cells. The lifecycle study is conducted with a total of 26 battery packs that are grouped by constant and random loading conditions, loading levels, and number of load-level changes. The data also includes load cycling on second-life batteries, where surviving cells from previously aged battery packs were assembled for second-life packs.
该数据集提供了加速锂离子电池生命周期数据,重点关注大范围的负载水平以及由两个 18650 电池组成的电池组的生命周期特征。生命周期研究总共对 26 个电池组进行,这些电池组按恒定和随机负载条件、负载水平和负载水平变化次数进行分组。该数据还包括二次电池的负载循环,其中将先前老化的电池组中的剩余电池组装为二次电池组。
Download: https://data.nasa.gov/download/xg3n-ngei/application%2Fzip
下载:https: //data.nasa.gov/download/xg3n-ngei/application%2Fzip
Data Set Citation: Fricke, K., Nascimento, R., Corbetta, M., Kulkarni, C., & Viana, F. “Accelerated Battery Life Testing Dataset”, NASA Prognostics Data Repository, Probabilistic Mechanics Lab, University of Central Florida, and NASA Ames Research Center, Moffett Field, CA
数据集引用:Fricke, K.、Nascimento, R.、Corbetta, M.、Kulkarni, C. 和 Viana, F.“加速电池寿命测试数据集”,NASA 预测数据存储库,概率力学实验室,中佛罗里达大学和美国宇航局艾姆斯研究中心,莫菲特菲尔德,加利福尼亚州
Publication Citation: Fricke, K., Nascimento, R., Corbetta, M., Kulkarni, C., & Viana, F. (2023). Prognosis of Li-ion Batteries Under Large Load Variations Using Hybrid Physics-Informed Neural Networks. Annual Conference of the PHM Society, 15(1). https://doi.org/10.36001/phmconf.2023.v15i1.3463
出版物引用:Fricke, K.、Nascimento, R.、Corbetta, M.、Kulkarni, C. 和 Viana, F. (2023)。使用混合物理信息神经网络对大负载变化下的锂离子电池进行预测。 PHM 协会年会, 15 (1)。 https://doi.org/10.36001/phmconf.2023.v15i1.3463
This data set was generated from a custom-made testbed to cycle battery packs designed and developed by Kajetan Fricke, Renato Nascimento, and Professor Felipe Viana from the Probabilistic Mechanics Laboratory at the University of Central Florida (UCF). This work is the result of a collaboration between the Probabilistic Mechanics Lab at the University of Central Florida, and the Intelligent Systems Division Diagnostics & Prognostics Group at NASA Ames Research Center.
该数据集是由中佛罗里达大学 (UCF) 概率力学实验室的 Kajetan Fricke、Renato Nascimento 和 Felipe Viana 教授设计和开发的循环电池组定制测试台生成的。这项工作是中佛罗里达大学概率力学实验室和美国宇航局艾姆斯研究中心智能系统部诊断与预测小组合作的结果。
CACLE dataset
数据文件与种类太多,暂时不知道如何利用
https://calce.umd.edu/battery-data#CS2
数据集格式:.txt
Oxford dataset
两种数据集不知道用哪个
Oxford Battery Degradation Dataset 1
Readme.txt:
Oxford energy trading battery degradation dataset
MIT dataset (A123)
A Hybrid Ensemble Deep Learning Approach for Early Prediction of Battery Remaining Useful Life
此论文使用的就是这个数据集
https://data.matr.io/1/projects/5c48dd2bc625d700019f3204
The following repository contains some starter code to load the datasets in either MATLAB or python:
以下存储库包含一些用于在 MATLAB 或 python 中加载数据集的起始代码:
https://github.com/rdbraatz/data-driven-prediction-of-battery-cycle-life-before-capacity-degradation
Aging dataset
与数据集相关的开源代码:
https://github.com/xtangai/Ageing-Data-Set
与数据集相关的论文:
Xiaopeng Tang, Liu Kailong, Li Kang, Widanage Widanalage, Kendrick Emma, Gao Furong. Recovering large-scale battery aging dataset with machine learning. Patterns 2021;2. http://dx.doi.org/10.1016/j.patter.2021.100302.
参考资料
[1] Ng, MF., Zhao, J., Yan, Q. et al. Predicting the state of charge and health of batteries using data-driven machine learning. Nat Mach Intell 2, 161–170 (2020).
[2] Mehmet Kurucan, Mete Özbaltan, Zeki Yetgin, Alkan Alkaya, Applications of artificial neural network based battery management systems: A literature review, Renewable and Sustainable Energy Reviews, Volume 192, 2024, 114262, ISSN 1364-0321.
[3] Nanhua Jiang, Jiawei Zhang, Weiran Jiang, Yao Ren, Jing Lin, Edwin Khoo, Ziyou Song, Driving behavior-guided battery health monitoring for electric vehicles using extreme learning machine, Applied Energy, Volume 364, 2024, 123122, ISSN 0306-2619.
[4] Q. Xu, M. Wu, E. Khoo, Z. Chen and X. Li, “A Hybrid Ensemble Deep Learning Approach for Early Prediction of Battery Remaining Useful Life,” in IEEE/CAA Journal of Automatica Sinica, vol. 10, no. 1, pp. 177-187, January 2023, doi: 10.1109/JAS.2023.123024.
[5] DASHlink - Li-ion Battery Aging Datasets. https://c3.ndc.nasa.gov/dashlink/resources/133/.
[6] Battery Data | Center for Advanced Life Cycle Engineering. https://calce.umd.edu/battery-data#CS2.
[7] Birkl, C. Oxford Battery Degradation Dataset 1. 2017年. ora.ox.ac.uk, https://ora.ox.ac.uk/objects/uuid:03ba4b01-cfed-46d3-9b1a-7d4a7bdf6fac.
[8] Reniers, J. M., 等. Oxford Energy Trading Battery Degradation Dataset. 2020年. ora.ox.ac.uk, https://ora.ox.ac.uk/objects/uuid:9aae61af-2949-49f1-8ad5-6aea448979e5.
[9] Experimental Data Platform. https://data.matr.io/1/#projects/5c48dd2bc625d700019f3204.
[10] Xtangai. xtangai/Ageing-Data-Set: Data and Code related to 《Recovering large-scale battery ageing data-set with machine learning》. Zenodo, 2021年5月30日. DOI.org * (Datacite)*, https://doi.org/10.5281/ZENODO.4867041.