基于神经网络的新能源汽车动力电池的串联型故障电弧检测方法

By combining windowed interpolation FFT and MS-RANet model with multi-agent reinforcement learning, the problem of detecting series fault arcs in new energy vehicles has been solved, achieving efficient, real-time, and accurate fault arc identification and ensuring the safety of new energy vehicles.

CN121878502BActive Publication Date: 2026-07-17HEFEI UNIV OF TECH
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HEFEI UNIV OF TECH
Filing Date
2026-02-11
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing technologies are insufficient to effectively detect series fault arcs in high-voltage power batteries in new energy vehicles, especially in complex driving environments where they have weak anti-interference capabilities, poor robustness, and insufficient real-time performance, thus failing to meet safety protection requirements.

Method used

A windowed interpolation FFT algorithm is used for frequency domain feature extraction. Combined with a multi-scale residual attention neural network (MS-RANet) and a multi-agent reinforcement learning optimization model, efficient and real-time series arc detection is achieved through multi-scale feature fusion and residual connection.

Benefits of technology

It achieves accurate identification of series electric arcs, reduces the risk of missed detection, improves the stability and adaptability of the model, meets the real-time safety protection requirements of new energy vehicles, and has high detection accuracy with low false alarm and missed detection rates.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121878502B_ABST
    Figure CN121878502B_ABST
Patent Text Reader

Abstract

为适配新能源汽车高压动力电池串联电弧检测需求，本发明公开基于神经网络的新能源汽车动力电池的串联型故障电弧检测方法。搭建符合UL1699B标准的实验平台，采集多负载工况下电流信号，经高通滤波与加窗插值FFT提取频域特征；构建MS‑RANet模型，通过多尺度卷积融合高低频特征，结合残差连接与CBAM注意力机制强化特征辨识度；MAPPO多智能体协同训练优化模型参数，轻量化部署于车载边缘单元并支持在线微调。实验验证表明，该方法检测准确率≥99.20%，单次推理时间＜14.79ms，误报率≤0.8%、漏报率≤1.2%，兼具高精度、强抗干扰性与实时性，有效保障动力电池安全运行。
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Spectrogram feature enhancement method for arc fault detection and series AC arc fault detection method and system

    CN119986266A

  • Material complete set state early warning model based on spatio-temporal characteristics and attention mechanism

    CN120725465A