基于改进动态时间弯曲的PMU频率异常检测及类型识别  

Frequency Anomaly Detection and Type Identification of PMU Based on Improved Dynamic Time Warping

作　　者：黄子蒙 余娟[1] 向明旭 张江南 李文沅[1] HUANG Zimeng;YU Juan;XIANG Mingxu;ZHANG Jiangnan;LI Wenyuan(State Key Laboratory of Power Transmission Equipment&System Security and New Technology(Chongqing University),Chongqing 400044,China;Electric Power Research Institute of State Grid Henan Electric Power Company,Zhengzhou 450052,China)

机构地区：[1]输配电装备及系统安全与新技术国家重点实验室(重庆大学),重庆市400044 [2]国网河南省电力公司电力科学研究院,河南省郑州市450052

出　　处：《电力系统自动化》2022年第24期104-112,共9页Automation of Electric Power Systems

基　　金：国家自然科学基金资助项目(52077016);四川省科技计划资助项目(2021YFH0029);重庆市院士牵头科技创新引导专项(cstc2021yszx-jscxX0002)。

摘　　要：高比例可再生能源电力系统中不确定性、复杂性和脆弱性问题日益凸显,亟须利用同步相量测量单元(PMU)的高质量量测数据支撑系统动态安全监控。然而,PMU数据受到各类因素影响,存在不同程度的质量问题,影响着数据的各类高级应用。对此,以频率量测数据为切入点,提出了一种数据驱动的PMU频率数据异常检测及类型识别方法。首先,归纳了频率数据的典型异常类型,并构造各类频率数据异常特征。进一步,提出了一种动态时间弯曲改进策略,通过动态调整弯曲窗口来有效量化各类异常特征。最后,基于局部离群因子法实现频率数据的异常检测及类型识别。以实际电网PMU频率数据为例,验证了所提方法的有效性。The problems of uncertainty, complexity and vulnerability in power systems with high proportion of renewable energy are increasingly prominent. It is urgent to use high-quality measurement data of synchronous phasor measurement units(PMUs) to support dynamic security monitoring of the system. However, PMU data is affected by various factors, and there are quality problems of varying degrees, which affect various advanced applications of data. In this paper, a data-driven PMU frequency data anomaly detection and type identification method is proposed based on the frequency measurement data. First, the typical types of frequency data anomalies are summarized, and the characteristics of various types of frequency data anomalies are constructed.Furthermore, an improved dynamic time warping strategy is proposed, which can effectively quantify various abnormal characteristic by dynamically adjusting the warping window. Finally, anomaly detection and type identification of frequency data are realized based on the local outlier factor method. The effectiveness of the proposed method is verified by taking the PMU frequency data of an actual power grid as an example.

关 键 词：相量测量单元 数据质量 异常检测 类型识别 动态时间弯曲 局部离群因子 

分 类 号：TM933.313[电气工程—电力电子与电力传动]