A Beta-Gaussian deep state-space model for unsupervised multi-domain battery state-of-charge estimation

Yang, Rui, García-Fernández, Ángel F and Lee, Che-Rung (2026) A Beta-Gaussian deep state-space model for unsupervised multi-domain battery state-of-charge estimation Journal of Energy Storage, 152. p. 120737. ISSN 2352-152X, 2352-1538

Text Paper__3_revised__2st_.pdf - Author Accepted Manuscript Available under License Creative Commons Attribution. Download (2MB) | Preview

Abstract

This paper introduces the Beta-Gaussian Multi-Domain Model (BG-MDM), an unsupervised multi-domain approach for accurate, label-free state-of-charge (SOC) estimation in lithium-ion batteries. BG-MDM models the SOC using a beta distribution to capture its natural boundedness, while representing other battery state variables as Gaussians, both during learning and estimation. A key innovation is the Neural-Parameterized Equivalent Circuit Model (NP-ECM), which dynamically predicts ECM parameters via neural networks to capture nonlinear battery behavior. For online inference, the Beta-Gaussian Iterated Posterior Linearization Filter (BG-IPLF) is used. BG-MDM also incorporates a shared-layer architecture for learning domain-invariant features and enabling efficient transfer learning. Experimental results demonstrate that BG-MDM outperforms baseline models across diverse domains. Its ability to adapt to new domains with significantly fewer training samples, along with an ablation study confirming each submodule's contribution, highlights its data efficiency and robustness. BG-MDM offers a unified solution for multi-domain SOC estimation in label-scarce environments.

Item Type:	Article
Uncontrolled Keywords:	State-of-charge (SOC), Multi-domain model, Unsupervised learning, Beta-Gaussian iterated posterior linearization, filter
Divisions:	Faculty of Science & Engineering Faculty of Science & Engineering > School of Engineering Faculty of Science & Engineering > School of Engineering > Electrical Engineering and Electronics
Depositing User:	Symplectic Admin
Date Deposited:	26 Jan 2026 10:11
Last Modified:	23 May 2026 10:56
DOI:	10.1016/j.est.2026.120737
Related Websites:	Author Publisher
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3196772
Disclaimer:	The University of Liverpool is not responsible for content contained on other websites from links within repository metadata. Please contact us if you notice anything that appears incorrect or inappropriate.