Improved hyperparameter Bayesian optimization-bidirectional long short-term memory optimization for high-precision battery state of charge estimation.

Wang, Shunli; Ma, Chao; Gao, Haiying; Deng, Dan; Fernandez, Carlos; Blaabjerg, Frede

doi:10.1016/j.energy.2025.136598

Improved hyperparameter Bayesian optimization-bidirectional long short-term memory optimization for high-precision battery state of charge estimation.

Wang, Shunli; Ma, Chao; Gao, Haiying; Deng, Dan; Fernandez, Carlos; Blaabjerg, Frede

Authors

Shunli Wang

Chao Ma

Haiying Gao

Dan Deng

Dr Carlos Fernandez c.fernandez@rgu.ac.uk
Associate Professor

Frede Blaabjerg

Abstract

At a time when new energy sources are constantly developing, mitigating the safety hazards of lithium batteries and prolonging their lifespan. In this paper, we take a ternary lithium-ion battery as an experimental object and carry out research based on the fusion method of deep learning and modeling for its high-precision state of charge (SOC) estimation requirements. This paper explores the construction of a battery dynamic model and hyperparameter optimization method based on a neural network. It also incorporates Kalman filter to investigate the noise correction strategy of a neural network model. Experimentally verified that the BO-BiLSTM-UKF fusion algorithm in this paper has a maximum error of only 0.113 %, which verifies the accuracy and strong robustness of the model. Its MAE and RMSE are reduced by 96.13 % and 95.73 % compared with the LSTM network model, which has better adaptability and estimation ability. In this paper, a network dynamic prediction fusion method based on the equivalent model is constructed and experimentally verified by different temperatures, complex working conditions and step-by-step simulation.

Citation

WANG, S., MA, C., GAO, H., DENG, D., FERNANDEZ, C. and BLAABJERG, F. 2025. Improved hyperparameter Bayesian optimization-bidirectional long short-term memory optimization for high-precision battery state of charge estimation. Energy [online], 328, article number 136598. Available from: https://doi.org/10.1016/j.energy.2025.136598

Journal Article Type	Article
Acceptance Date	May 13, 2025
Online Publication Date	May 18, 2025
Publication Date	Aug 1, 2025
Deposit Date	May 23, 2025
Publicly Available Date	May 19, 2026
Journal	Energy
Print ISSN	0360-5442
Electronic ISSN	1873-6785
Publisher	Elsevier
Peer Reviewed	Peer Reviewed
Volume	328
Article Number	136598
DOI	https://doi.org/10.1016/j.energy.2025.136598
Keywords	New energy lithium-ion battery; SOC; Noise reduction
Public URL	https://rgu-repository.worktribe.com/output/2842912