With the increasing number of retired lithium-ion batteries from electric vehicles (EVs), effective screening and classification are necessary to enable their secondary utilization. The existing retired lithium-ion battery screening methods have limitations in accuracy and high-energy consumption. This article proposes a method based on local discharge curves and an improved CrossFormer model to reduce energy consumption and improve screening accuracy. Considering the actual conditions of retired lithium-ion batteries, features are extracted from the discharging stage. Select the segments of the capacity sequence that exhibit the most significant variation with the voltage sequence and construct a partial capacity-voltage sequence. The standard deviation (Std) and Shannon entropy (ShanEn) of the capacity sequence, as well as the variance, energy, and the maximum amplitude of the low-frequency part of the Fourier-transformed voltage sequence, are used as features. Then, the CrossFormer utilizes a flexible embedding mechanism and a hierarchical structure for handling the dependencies of the feature. The CrossFormer is improved for classification by adding a cross-attention layer at the final stage of the CrossFormer decoder, which enhances its ability to capture the dependencies between features and improves the model’s performance in battery screening classification. Finally, the validation of 252 retired batteries shows that the method achieves a classification accuracy of 94.74%, outperforming other advanced models, including the original CrossFormer, Transformer, ResNet, and multilayer perceptron (MLP) models.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Retired Lithium-Ion Batteries Screening Based on Partial Discharge Curves and an Improved CrossFormer


    Contributors:
    Lin, Mingqiang (author) / Xu, Zijun (author) / Zheng, Gengfeng (author) / Meng, Jinhao (author) / Wang, Wei (author) / Wu, Ji (author)

    Published in:

    Publication date :

    2025-08-01


    Size :

    1514359 byte




    Type of media :

    Article (Journal)


    Type of material :

    Electronic Resource


    Language :

    English