摘要
电网通过监控全链路电网数据可以发现电网异常,但是数据监控方法存在杰卡德系数低、监控准确率低的问题。在此背景下,研究一种基于数据中台的电网数据全链路监控方法。该方法以数据中台三部分功能作为支撑,实施电网数据全链路采集与预处理,提取电网数据样本中提取特征,包括角度离散程度、平均角的正弦和余弦、圆角标准差,以特征为输入,利用改进神经网络算法构建异常识别模型,实现电网数据全链路监控。结果表明:所研究方法的杰卡德系数更大,该值均在0.9以上,数据异常监控准确率在97.5%以上,证明了所研究监控方法的准确性。
The power grid can find the grid anomaly by monitoring the full link power grid data,but the data monitoring method has the problems of low Jackard coefficient and low monitoring accuracy.Under this background,a full link monitoring method of power grid data based on data center is studied.This method is supported by the three functions of the data center,implements the full link collection and preprocessing of the power grid data,extracts the features extracted from the power grid data samples,includ-ing the degree of angle dispersion,the sine and cosine of the average angle,and the standard deviation of the rounded corners.With the features as the input,the improved neural network algorithm is used to build an anomaly identification model to realize the full link monitoring of the power grid data.The results show that the jackard coefficients of the methods studied are larger,the value is above 0.9,and the accuracy of data anomaly monitoring is above 97.5%,which proves the accuracy of the monitoring methods studied.
作者
萧展辉
甘莹
柯婷
梁置铭
XIAO Zhan-hui;GAN Ying;KE Ting;LIANG Zhi-ming(China Southern Power Grid Digital Power Grid Research Institute Co.,Ltd.,Guangzhou 510700 China;Digital Grid Group,CSG,Guangzhou 510700 China)
出处
《自动化技术与应用》
2025年第2期100-104,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)。
关键词
数据中台
电网数据
数据采集
特征提取
改进神经网络算法
全链路监控
data center
grid data
data acquisition
feature extraction
improved neural network algorithm
full link monitoring
