基于时间序列提取和维诺图的电力数据异常检测方法

电力网络中信息系统与物理系统的深度融合,导致现代电力系统易受异常数据的影响。现有的电力数据异常检测方法未能充分挖掘数据特征,存在计算复杂、灵活性差、精度较低等缺点。提出一种基于时间序列提取和维诺图的异常数据检测方法,利用重要点分段的时间序列提取方法,将高维数据进行降维处理,并将其映射到二维平面上,构造维诺图分区,进而检测出异常数据。该方法可降低数据维度和算法复杂度,能根据序列特征灵活设定异常阈值,实现异常数据的准确检测,仿真实验证明所提方法的有效性。

Outlier Detection Method Based on Compressed Time Series and Voronoi Diagram for Power Data

The deep integration of information system and physical system made power system easily affected by outlier data, while the existing outlier detection methods for power system didn`t take the advantages of data features, and had problems such as heavy computation, bad flexibility and low precision, etc.This paper proposes an outlier detection method based on compressed time series and Voronoi diagram, which adopts the time series extraction method of important points section to reduce the dimension of data in power system, map it to a two-dimensional plane, construct the Voronoi diagram partition, and then detect the abnormal data.This method can reduce the data dimension and algorithm complexity, set anomaly threshold according to the sequence features flexible, and realize the accurate detection of abnormal data.The simulation results have verified the effectiveness of the proposed method.