A new energy power battery AI-based health check and evaluation method and system

By combining graph neural networks with dual-constraint loss functions, coupling interference in battery modules is eliminated, solving the problem of inaccurate evaluation results in existing technologies, and realizing accurate health status assessment and identification of abnormal cells for new energy power batteries.

CN122133087AInactive Publication Date: 2026-06-02ZHEJIANG GUARDIAN NEW ENERGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ZHEJIANG GUARDIAN NEW ENERGY CO LTD
Filing Date
2026-05-06
Publication Date
2026-06-02
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Existing methods for assessing the health status of new energy power batteries are inaccurate because they ignore the series clamping effect and coupling interference. This makes it impossible to accurately identify abnormal cells and poses a risk of missed detection.

Method used

A graph neural network containing graph convolutional layers and feature decoupling modules is used for training, combined with a dual-constraint loss function. This removes coupling interference and retains electrothermal performance-related information. An electrothermal performance evaluation mapping relationship is established through a local manifold learning method.

Benefits of technology

It significantly improves the evaluation accuracy in complex aging scenarios, accurately identifies abnormal cells and locates their positions, provides a high-quality data foundation, and enhances the credibility and practicality of the evaluation system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122133087A_ABST
    Figure CN122133087A_ABST
Patent Text Reader

Abstract

This invention relates to the field of new energy power battery testing technology, and discloses an AI-based health check and evaluation method and system for new energy power batteries. The invention designs a graph neural network structure including graph convolutional layers and feature decoupling modules, and uses a dual-constraint loss function for training: independence constraints effectively remove coupling interference introduced by series circuits, solving the problem of missed detection of abnormal cells due to neglecting series clamping effects and module-level averaging in traditional methods; correlation constraints ensure that while decoupling, key information strongly correlated with electrothermal performance evaluation is retained. The synergistic effect of these two constraints makes the final intrinsic features both pure and effective, laying a high-quality data foundation for accurate evaluation. Furthermore, the invention introduces a local manifold learning method to establish a mapping relationship between intrinsic features and electrothermal performance scores, solving the evaluation bias problem caused by the mixing of multiple degradation modes in the global model, and significantly improving evaluation accuracy.
Need to check novelty before this filing date? Find Prior Art