State of Health Estimation for Second-Life Lithium-Ion Batteries in Energy Storage System With Partial Charging-Discharging Workloads

Echelon utilization in energy storage systems (ESSs) has emerged as one of the predominant solutions for addressing large-scale retired lithium-ion batteries from electrical vehicles. However, high unit-to-unit health variability and partial charging-discharging workloads render the state of health...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on industrial electronics (1982) 2024-10, Vol.71 (10), p.13178-13188
Main Authors: Jiang, Yiyue, Ke, Yuqi, Yang, Fangfang, Ji, Jinchen, Peng, Weiwen
Format: Article
Language:English
Subjects:
Bayesian neural network
Calibration
Charging
Data mining
Discharge
echelon utilization
Electric vehicles
Energy storage
energy storage system
Error analysis
Estimation
Feature extraction
Integrated circuit modeling
Lithium
Lithium-ion batteries
Neural networks
Rechargeable batteries
second-life lithium-ion batteries
State of charge
state of health
Storage systems
Uncertainty
Workload
Workloads
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Echelon utilization in energy storage systems (ESSs) has emerged as one of the predominant solutions for addressing large-scale retired lithium-ion batteries from electrical vehicles. However, high unit-to-unit health variability and partial charging-discharging workloads render the state of health (SOH) estimation of these second-life lithium-ion batteries (SL-LIBs) in ESSs a crucial and challenging issue. Existing SOH estimation methods are commonly focused on new batteries with consistent health state and complete charging-discharging workloads, while the estimation methods for SL-LIBs have been rarely developed. To fill this gap, this article proposes a novel SOH estimation method with specially designed features and calibrated uncertainty quantification for SL-LIBs. Joint features are first introduced for extracting useful SOH information from the cycling data of SL-LIBs under partial charging-discharging workloads. Then, Bayesian neural network with uncertainty calibration is used to generate SOH estimation results, which can quantify the estimation uncertainty caused by unit-to-unit health variability. To demonstrate the proposed method, a total of 36 retired NCM-18650 power batteries are cycled under 9 different partial charging-discharging workloads. A case study on this SL-LIB lab-test dataset reaches the best results of 2.13% mean absolute percentage error and 0.0178 root-mean-squared error, as well as well-calibrated estimation uncertainty.
ISSN:0278-0046
1557-9948
DOI:10.1109/TIE.2023.3344825