数据缺失约束下电力通信网全链路异常二分递归分割检测  

Binary Recursive Segmentation Detection of Full Link Anomaly in Power Communication Network under Data Absence Constraint

作　　者：甘莹 萧展辉 梁置铭 柯婷 GAN Ying;XIAO Zhan-hui;LIANG Zhi-ming;KE Ting(Digital Grid Group,CSG,Guangzhou 510700 China;China Southern Power Grid Digital Power Grid Research Institute Co.,Ltd.,Guangzhou 510700 China)

机构地区：[1]南方电网数字电网集团有限公司,广东广州510700 [2]南方电网数字电网研究院有限公司,广东广州510700

出　　处：《自动化技术与应用》2025年第3期105-109,共5页Techniques of Automation and Applications

基　　金：数据运维工具(全链路监控)研发项目(0002200000076144)。

摘　　要：针对在缺失数据环境下电力通信网全链路异常数据检测存在检出异常数据误差大的问题,提出一种数据缺失约束下电力通信网全链路异常二分递归分割检测方法。采用滑动平均法和非线性回归分析算法平滑处理与缺失值补全处理数据,设计对应链路的单独子模型,获取数据特征向量,引入二分递归分割技术,划分全链路数据为正常数据子集与异常数据子集,实现全链路异常检测。实验数据显示:所提出方法获得的检出异常数据占比数值与实际数值最大误差仅为2.53%,全部异常数据检出耗时最小值为2 s,充分证实了提出方法异常检测效果较佳。Aiming at the problem of large error in detecting abnormal data in the whole link of power communication network under the environment of missing data,a dichotomy recursive segmentation method for detecting abnormal data in the whole link of power communication network under the constraint of missing data is proposed.The moving average method and nonlinear regression analysis algorithm are used to smooth and complete the data,design a single sub model of the corresponding link,obtain the data feature vector,introduce the dichotomy recursive segmentation technology,divide the whole link data into normal data subset and abnormal data subset,and realize the whole link anomaly detection.The experimental data shows that the maximum error between the percentage of abnormal data detected by the proposed method and the actual value is only 2.53%,and the minimum time spent for detecting all abnormal data is 2 s,which fully confirms that the proposed method has a better anomaly detection effect.

关 键 词：全链路数据 电力通信网 异常数据 决策树算法 异常检测 滑动平均法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]