Paper Publications

Hits:

Impact Factor:7.3

DOI number:10.1109/TMECH.2025.3619518

Teaching and Research Group:Tan Y., Wang L., Lin M., & Wu J.

Journal:IEEE/ASME Transactions on Mechatronics

Key Words:Battery pack, Chebyshev Kolmogorov–Arnold networks (CKAN), multilevel feature, residual value, retention value ratio (RVR)

Abstract:Accurate residual value assessment of retired power battery packs is vital for second-life applications. Traditional metrics such as state of health (SOH) tend to underestimate pack value due to their sensitivity to severely aged cells, overlooking the contributions of relatively healthier ones. Moreover, existing methods often require complete charge-discharge data, which is impractical in real-world or postretirement scenarios. To address these limitations, we propose a novel metric, retention value ratio (RVR), which evaluates the overall residual value by incorporating the condition of each individual cell. To enable real-time RVR estimation under practical constraints, real-world electric vehicles charging data is first analyzed to identify the appropriate data segment, from which a multilevel feature set is subsequently constructed to capture battery characteristics from multiple perspectives. The fusion of these features enhances the stability and accuracy of estimation, especially when cells degrade unevenly. Using the extracted features, an improved Kolmogorov–Arnold network, namely, the Chebyshev Kolmogorov–Arnold network, is developed to estimate RVR with high efficiency and stability. Experimental results show that the proposed method supports accurate online estimation and yields residual value estimates around 4% higher than SOH in late-stage batteries.

Indexed by:Journal paper

Discipline:Engineering

Document Type:J

Page Number:Early Access

Translation or Not:no

Date of Publication:2025-10-23

Included Journals:SCI

Links to published journals:https://ieeexplore.ieee.org/document/11215981