基于随机矩阵理论的用电高峰场景中电力网络异常检测模型构建研究

Research on Constructing an Anomaly Detection Model for Power Networks in Peak Electricity Consumption Scenarios Based on Random Matrix Theory

随着新型电力系统的发展,电力网络规模日益扩大,传统的电力网络异常检测模型已经不能满足当前的需求。此次研究提出了一种基于随机矩阵理论的电力网络异常检测模型。实验结果表明,在低信噪比的情况下,数据集尺寸为900时,平均谱半径方法、最大特征值方法和样本协方差矩阵的准确率分别是0.746、0.764和0.788。在高信噪比的情况下,三种方法的准确率分别是0.921、0.934和0.947。在高信噪比情况下,异常节点个数为5时,平均谱半径方法、最大特征值方法和样本协方差矩阵方法的响应时间分别是3.1、2.6和2.1。在低信噪比情况下,三种方法的响应时间分别是4.0 s、3.7 s和2.8 s。研究结果表明所提出的方法能够给电力网络的稳定运行提供更好的保障。

With the development of new power systems,the scale of power networks is expanding day by day,and traditional power network anomaly detection models can no longer meet the current needs.This study proposes a power network anomaly detection model based on random matrix theory.The experimental results show that under low signal-to-noise ratio conditions,when the dataset size is 900,the accuracy of the average spectral radius method,maximum eigenvalue method,and sample covariance matrix are 0.746,0.764,and 0.788,respectively.In the case of high signal-to-noise ratio,the accuracy of the three methods is 0.921,0.934,and 0.947,respectively.Under high signal-to-noise ratio,when the number of abnormal nodes is 5,the response times of the average spectral radius method,maximum eigenvalue method,and sample covariance matrix method are 3.1,2.6,and 2.1,respectively.In the case of low signal-to-noise ratio,the response times of the three methods are 4.0,3.7,and 2.8,respectively.