Online Analytical Characterization of Outliers in Synchrophasor Measurements: A Singular Value Perturbation Viewpoint
Online Analytical Characterization of Outliers in Synchrophasor Measurements: A Singular Value Perturbation Viewpoint
复制标题
DOI：
10.1109/tpwrs.2017.2771782
复制
发表时间：
2018-07
影响因子：
6.6
通讯作者：
Kaveri Mahapatra;N. Chaudhuri;R. Kavasseri;S. Brahma
中科院分区：
文献类型：
--
作者：
Kaveri Mahapatra;N. Chaudhuri;R. Kavasseri;S. Brahma
提出了一种基于主元分析的同步相量测量异常值在线表征方法。为此，建立了一个线性化框架来分析系统在标称和非标称（例如，故障）条件，其包含在同步相量数据的相同窗口中。受奇异值摄动理论的启发，提出了PC分数作为系统状态矩阵的函数的范数变化的界。结果表明，在坏的数据离群值的存在下，这些边界为高维PC分数将显着大于较低的维度。建立了数据窗口中样本数量对分析结果的影响。一个模拟的测试系统和从美国公用事业收集的现场数据的案例研究，以支持分析结果。最后，一个在线分类器的特征离群值的开发，以说明机器学习为基础的方法所提出的框架的有用性。
This paper presents a principal component (PC) analysis based method for online characterization of outliers in synchrophasor measurements. To that end, a linearized framework is established to analyze dynamical response from a system under nominal and off-nominal (e.g., faulted) conditions, which are contained in the same window of synchrophasor data. Inspired by the singular value perturbation theory, a bound on the change in the norm of the PC scores as a function of system state matrices is presented. It is shown that in the presence of bad data outliers these bounds for higher dimensional PC scores will be significantly larger compared to lower dimensions. The effect of the number of samples in the data window on the results of the analysis is established. Case studies on a simulated test system and on field data collected from a US utility are presented to support the analytical results. Finally, an online classifier for the characterization of outliers is developed to illustrate the usefulness of the proposed framework for machine learning based methods.