版权所有:内蒙古大学图书馆 技术提供:维普资讯• 智图
内蒙古自治区呼和浩特市赛罕区大学西街235号 邮编: 010021
作者机构:AIWARE Limited Company 17 Huynh Man Dat Street Hoa Cuong Bac Ward Hai Chau District Da Nang550000 Viet Nam AIT Austrian Institute of Technology GmbH Giefinggasse 4 Vienna1210 Austria Software Engineering Department FPT University Da Nang550000 Viet Nam Faculty of Information Technology Ton Duc Thang University Ho Chi Minh City700000 Viet Nam Laboratory of Environmental Sciences and Climate Change Institute for Computational Science and Artificial Intelligence Van Lang University Ho Chi Minh City700000 Viet Nam Faculty of Environment School of Technology Van Lang University Ho Chi Minh City700000 Viet Nam
出 版 物:《arXiv》 (arXiv)
年 卷 期:2025年
核心收录:
主 题:Lithium ion batteries
摘 要:Accurate prediction of the Remaining Useful Life (RUL) in Lithium-ion battery (LIB) health management systems is essential for ensuring operational reliability and safety. However, many existing methods assume that training and testing data follow the same distribution, limiting their ability to learn domain-invariant features and generalize effectively to unseen target domains. To approach this opportunity, we propose a novel RUL prediction framework and incorporates an domain adaptation (DA) technique. Our framework integrates a comprehensive signal preprocessing pipeline—including noise-reduction, feature extraction, and normalization—with a robust deep learning model named HybridoNet-Adapt. The model comprises a feature extraction block built from Long Short-Term Memory (LSTM), Multihead Attention, and Neural Ordinary Differential Equation (NODE) layers, followed by two predictor modules implemented using linear layers. These predictors are balanced using trainable trade-off parameters. To enhance generalization to unseen data distributions, we introduce a domain adaptation strategy inspired by the Domain-Adversarial Neural Network (DANN) framework, replacing adversarial loss with Maximum Mean Discrepancy (MMD) to learn domain-invariant features. This approach enables effective transfer learning from source to target domains using cycling data. Experimental results demonstrate that our proposed method significantly outperforms traditional machine learning models such as XGBoost and Elastic Net, as well as deep learning baselines like Dual-input DNN, in predicting RUL. These findings highlight the potential of HybridoNet-Adapt for reliable and scalable Battery Health Management (BHM). Copyright © 2025, The Authors. All rights reserved.