A Hybrid XGBoost-Augmented Gaussian Process Framework for Accurate Battery State-of-Health Estimation
- 5 March 2026
- journal article
- Published by The Electrochemical Society in Journal of the Electrochemical Society
- Vol. 173 (5) , 050510
Abstract
Accurate estimation of the state of health (SOH) of lithium-ion batteries is crucial for battery management. To address the issues of feature redundancy and noise extracted from charge-discharge profiles that a single model struggles to overcome effectively, the XGBoost feature-augmented Gaussian Process Regression (GPR) cascaded model is proposed for accurate SOH estimation. This method initially extracts electrochemical and statistical features from easily accessible voltage relaxation regions, which are processed though sequential forward selection (SFS) to address feature redundancy. Then, XGBoost model maps the original features into an augmented feature. The augmented feature and the SFS-optimized subset are fed into the GPR model, which outputs SOH estimation with a confidence interval, thus achieving a cooperation of feature enhancement and probabilistic estimation. Furthermore, an improved dung beetle optimization (IDBO) algorithm is proposed for efficient hyperparameter tuning of both XGBoost and GPR models. The proposed framework achieves excellent performance, exhibiting a 19% and 47% lower RMSE compared with the single GPR and XGBoost model, respectively. Furthermore, a transfer learning framework leveraging hyperparameter-tuned GPR is developed, which requires only 127 and 96 target samples from Datasets 2 and 3, achieving an RMSE of 1.55% and 0.83%, respectively.Keywords
Funding Information
- National Natural Science Foundation of China (52402245)
This publication has 42 references indexed in Scilit:
- A deep learning method based on transfer learning and noise-resistant hybrid neural network for lithium-ion battery state of health and state of charge estimationElectrochimica Acta, 2025
- A lithium-ion battery SOH estimation method based on temporal pattern attention mechanism and CNN-LSTM modelComputers and Electrical Engineering, 2024
- An adaptive and interpretable SOH estimation method for lithium-ion batteries based-on relaxation voltage cross-scale features and multi-LSTM-RFR2Energy, 2024
- State of Health (SoH) estimation methods for second life lithium-ion battery—Review and challengesApplied Energy, 2024
- An electrochemical-thermal-aging effects coupled model for lithium-ion batteries performance simulation and state of health estimationApplied Thermal Engineering, 2023
- Stacking integrated learning model via ELM and GRU with mixture correntropy loss for robust state of health estimation of lithium-ion batteriesEnergy, 2023
- A novel battery health indicator and PSO-LSSVR for LiFePO4 battery SOH estimation during constant current chargingEnergy, 2023
- Development of a calibration methodology for fitting the response of a lithium-ion cell P2D model using real driving cyclesEnergy, 2023
- Dung beetle optimizer: a new meta-heuristic algorithm for global optimizationThe Journal of Supercomputing, 2022
- State of health assessment for lithium batteries based on voltage–time relaxation measureElectrochimica Acta, 2016