基于随机矩阵理论的电力扰动事件时空关联

发布时间：2018-04-27 00:21

本文选题：电能质量 + 数据清洗　；参考：《高电压技术》2017年07期

【摘要】：为全面挖掘扰动事件的内在关系、分析影响扰动源关联特性的外界因素,从扰动源的特征提取、时空关联分析、关联异常辨识方面开展研究,通过数据分析的方法验证了基本扰动事件之间的时空关联性。为克服传统数据清洗方法重复检出率低、计算成本高的缺点,提出自适应近邻排序数据清洗方法,并且实现了扰动源特征信息的快速提取。在此基础上,利用随机矩阵理论从宏观角度分析扰动源的时空关联特性。针对可能存在的局部扰动源关联异常,提出一种基于相关矩阵信息增益的动态辨识方法。最后,依据实际电能质量监测系统记录的暂态事件数据,验证了该方法的可行性和有效性。研究结果表明,所提出的数据清洗方法适用于大数据分析框架,扰动源之间的关联程度与电网的运行方式和天气状况有关。信息增益方法利用移动窗口和动态阈值技术有效跟踪电网运行状态变化,相较于PCA方法具备更高的异常事件识别率。
[Abstract]:In order to fully excavate the intrinsic relationship of disturbance events and analyze the external factors that affect the correlation characteristics of disturbance sources, the research is carried out from the aspects of feature extraction of disturbance sources, temporal and spatial correlation analysis, association anomaly identification, etc. The temporal and spatial correlation between the basic disturbance events is verified by the method of data analysis. In order to overcome the disadvantages of low repetition rate and high computational cost in traditional data cleaning methods, an adaptive nearest neighbor sorting data cleaning method is proposed, and the feature information of disturbance source can be quickly extracted. On this basis, the temporal and spatial correlation characteristics of the disturbance source are analyzed from the macroscopic point of view by using the stochastic matrix theory. A dynamic identification method based on the information gain of correlation matrix is proposed for the possible local disturbance source correlation anomalies. Finally, the feasibility and effectiveness of the proposed method are verified by the transient event data recorded by the real power quality monitoring system. The results show that the proposed data cleaning method is suitable for big data analysis framework, and the correlation between disturbance sources is related to the operation mode and weather conditions of the power grid. The information gain method makes use of moving window and dynamic threshold technique to track the change of power grid operation state effectively. Compared with the PCA method, the information gain method has a higher detection rate of abnormal events.
【作者单位】：广州供电局有限公司;上海交通大学电气工程系;
【基金】：国家自然科学基金(51577115) 中国南方电网公司科技项目(GZHKJ00000002)~~
【分类号】：TM732

【参考文献】