Life Prediction under Charging Process of Lithium-Ion Batteries Based on AutoML

Accurate online capacity estimation and life prediction of lithium-ion batteries (LIBs) are crucial to large-scale commercial use for electric vehicles. The data-driven method lately has drawn great attention in this field due to efficient machine learning, but it remains an ongoing challenge in the...

Full description

Saved in:
Bibliographic Details
Published in:Energies (Basel) 2022-07, Vol.15 (13), p.4594
Main Authors: Luo, Chenqiang, Zhang, Zhendong, Qiao, Dongdong, Lai, Xin, Li, Yongying, Wang, Shunli
Format: Article
Language:English
Subjects:
Aging
Algorithms
automated machine learning
Automation
Charging
Computer applications
COVID-19
Datasets
Deep learning
Electric vehicles
incremental capacity
Kalman filters
Learning algorithms
Life prediction
Life span
Lithium
Lithium-ion batteries
lithium-ion battery
Machine learning
Methods
Neural networks
Prediction models
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate online capacity estimation and life prediction of lithium-ion batteries (LIBs) are crucial to large-scale commercial use for electric vehicles. The data-driven method lately has drawn great attention in this field due to efficient machine learning, but it remains an ongoing challenge in the feature extraction related to battery lifespan. Some studies focus on the features only in the battery constant current (CC) charging phase, regardless of the joint impact including the constant voltage (CV) charging phase on the battery aging, which can lead to estimation deviation. In this study, we analyze the features of the CC and CV phases using the optimized incremental capacity (IC) curve, showing the strong relevance between the IC curve in the CC phase as well as charging capacity in the CV phase and battery lifespan. Then, the life prediction model based on automated machine learning (AutoML) is established, which can automatically generate a suitable pipeline with less human intervention, overcoming the problem of redundant model information and high computational cost. The proposed method is verified on NASA’s LIBs cycle life datasets, with the MAE increased by 52.8% and RMSE increased by 48.3% compared to other methods using the same datasets and training method, accomplishing an obvious enhancement in online life prediction with small-scale datasets.
ISSN:1996-1073
1996-1073
DOI:10.3390/en15134594