Deep learning estimation of state of health for lithium-ion batteries using multi-level fusion features of discharge curves

Jiaming Zhou, Jianfei Rong, Jinming Zhang, Chunrui Liu, Fengyan Yi, Zhipeng Jiao, Caizhi Zhang,
Deep learning estimation of state of health for lithium-ion batteries using multi-level fusion features of discharge curves,
Journal of Power Sources,
Volume 653,
2025,
237781,
ISSN 0378-7753,
https://doi.org/10.1016/j.jpowsour.2025.237781.
(https://www.sciencedirect.com/science/article/pii/S0378775325016179)
Abstract: The complex operating environments of electric vehicles significantly affect battery state of health (SOH) estimation, often resulting in large estimation errors. To address this issue, we propose an innovative deep learning framework that utilizes multi-level fusion of battery features for more accurate SOH estimation under real-world operating conditions. First, we employ short-time working conditions (STWC) to collect raw data and develop a multi-level feature extraction method. This method extracts multiple features from discharge curves and utilizes a Multiple Regression–Multilayer Perceptron (MR-MLP) model to reduce dimensionality and enhance the correlation among features. Then, we propose an advanced deep learning technique based on a Long Short-Term Memory (LSTM) and Transformer hybrid model, which enhances temporal feature integration and overall estimation performance. The model's robustness is validated on batteries with varying capacities and types, achieving high estimation accuracy with an error margin as low as 1 %. This result represents a significant step forward in overcoming the limitations of traditional estimation models.
Keywords: Lithium-ion battery; Multiple regression; Multilayer perceptron; LSTM-Transformer; State of health