学习迁移
计算机科学
领域(数学分析)
锂(药物)
电池(电)
估计
锂离子电池
传输(计算)
电池容量
人工智能
机器学习
工程类
数学
物理
系统工程
心理学
功率(物理)
热力学
并行计算
数学分析
精神科
作者
Yubo Zhang,Youyuan Wang,Dongning Huang
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2024-05-10
卷期号:: 2201-2206
被引量:2
标识
DOI:10.1109/cieec60922.2024.10583610
摘要
The state-of-health (SOH) estimation algorithm for lithium-ion batteries is crucial for ensuring the safety and cost-effectiveness of batteries, garnering widespread research and attention. However, data-driven models face challenges in generalizing models established on one type of battery to others, mainly due to the assumption of independent and identically distribution. Therefore, this paper proposes a semi-supervised transfer learning approach to address this issue. By leveraging trace amount of labeled data and unlabeled data from the target domain, the proposed method designs a semi-supervised transfer process and loss function for the source model through fine-tune and a domain adaptation method based on Maximum Mean Discrepancy (MMD). This facilitates the development of a capacity estimation model that can be universally applied to batteries with different chemistries and operating conditions. Cross-validation results indicate that, compared to original models and other benchmarks, the proposed method achieves the generalization of the capacity estimation model to lithium-ion batteries with different data distributions. These efforts contribute to enhancing the practical applicability of SOH algorithms in real-world scenarios.
科研通智能强力驱动
Strongly Powered by AbleSci AI