边缘计算节点异常恢复方法、系统、电子设备及存储介质

CN121636236BActive Publication Date: 2026-07-17HUAYAN INTELLIGENT TECH (GRP) CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HUAYAN INTELLIGENT TECH (GRP) CO LTD
Filing Date
2025-12-01
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing edge computing nodes struggle to effectively handle anomalies in complex and dynamic environments. Traditional fixed thresholds or simple watchdog mechanisms cannot adapt to scenarios with frequent voltage and current fluctuations and devices susceptible to high temperatures, resulting in untimely or inaccurate anomaly recovery.

Method used

By acquiring multiple operating parameter sequences of edge computing nodes, performing sliding window smoothing and classification model analysis, and combining reinforcement learning algorithms, the target recovery action instructions are determined, thereby achieving intelligent anomaly recovery of edge computing nodes.

Benefits of technology

It improves the accuracy of anomaly identification, enables early prevention and control of anomalies, quickly and efficiently restores node stability, reduces business interference, and improves the operational reliability and intelligent management level of edge computing nodes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121636236B_ABST
    Figure CN121636236B_ABST
Patent Text Reader

Abstract

本申请提供一种边缘计算节点异常恢复方法、系统、电子设备及存储介质,涉及边缘计算与智能控制技术领域,所述方法包括:获取边缘计算节点的多个不同运行参数序列;分别对各运行参数序列进行平滑处理,得到多个平滑参数序列;将当前时刻下,各平滑参数序列中分别对应的平滑运行参数值,输入预设分类模型,得到当前运行状态;将各平滑参数序列输入预设时间序列模型,得到多个预测运行参数值,将各预测运行参数值输入预设分类模型,得到预测运行状态;若当前运行状态或预测运行状态为异常,则获取增强状态向量;根据增强状态向量确定目标恢复动作指令并发送给边缘计算节点。本申请提升了异常识别精度。
Need to check novelty before this filing date? Find Prior Art