.png){kind=icon}
.png){kind=icon}
Return
A deep learning method combining wavelet transform and transfer learning for lithium-ion battery state of health estimation
.png){kind=icon}
PRE
AI
C
Z
R
Z
P
J
DOI：10.1016/j.jprocont.2026.103706.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Accurate estimation of the battery state of health (SOH) is crucial for ensuring the safety and stable operation of battery systems. However, existing SOH estimation methods focus on local feature extraction, and fail to capture global battery aging features. This paper introduces a novel approach integrating convolutional neural networks (CNN), wavelet transform, attention mechanisms, and transfer learning (TL), which significantly improves the ability of CNN to capture global features of battery aging patterns. The introduced method leverages wavelet transform to extract multi-scale information from charging data while suppressing noise interference. Additionally, an attention mechanism is introduced to optimize feature weighting and focus on critical degradation patterns. To further reduce data dependency, a TL strategy is developed, where the feature extraction layers of a pre-trained model are frozen, and the fully connected layers are fine-tuned, enabling knowledge transfer across battery models and operating conditions. Experimental validation on the XJTU and Toyota-MIT-Stanford datasets demonstrates the superior performance of the proposed model compared to 4 baseline models. On dataset 1, compared to the baseline models, HACNN reduces Mean Absolute Error (MAE) by 38.4%, 21.5%, 36.8%, and 9.1%, respectively. Notably, the TL strategy reduces model errors by over 30% using only 10% of the target dataset. This study provides a high-accuracy framework for SOH estimation in data-scarce scenarios and offers significant practical value for real-world battery management systems.
Keywords:
State of health
Wavelet transform
Convolutional neural network
Transfer learning
Journal
IF:
3.9
Papers:
3.4K
Citations:
7.3K
Organization
Cited Papers