基于改进K-medoids聚类和SVM的异常用电模式在线检测方法

Online detection method for abnormal electricity model behavior based on improved K-medoids clustering and SVM

窃电已成为电网电能非技术损失的主要问题,快速捕捉用户用电异常行为数据,精准定位窃电用户位置成为研究关键。针对现有异常用电辨识精度低的问题,提出一种基于改进K-medoids聚类和支持向量机(support vector machines,SVM)的用电异常行为在线检测方法。首先提取并分析电网电量及其相关表征数据,在此基础上构建基于用电量、线损等用电异常参数特性表征指标;然后对用户用电相关数据进行清洗、缺省值补全,并采用基于密度权重Canopy的改进K-medoids算法将异常数据依据表征指标进行标签化聚类;最后将已获得标签的数据采用SVM分类器进行训练进一步完成异常参数的检测和评估,并通过电网实际运行数据验证了所提异常数据辨识方法的检测精度,研究可为电网异常用电的辨识提供理论基础。

Stealing electricity has become the main problem of non-technical loss of power grid.It is the key to quickly capture the abnormal behavior data of users′electricity and accurately determine the location of users.In order to solve the problem of low identification accuracy of abnormal power consumption data,an online detection method of abnormal electricity consumption behavior based on improved K-medoids clustering and support vector machine is proposed.Firstly,the power grid power and its related characterization data are extracted and analyzed.On the basis,the characteristic index of power consumption abnormal parameters is constructed based on the power consumption,line loss and so on.Then,the user electricity related data is cleaned,and the improved K-medoids algorithm based on density weight Canopy is used to cluster the abnormal data according to the characterization index.Finally,the acquired tag data is trained by SVM classifier to further complete the detection and evaluation of abnormal parameters.The detection accuracy of the proposed method is verified by the actual operation data of power grid.This study can provide a theoretical basis for the identification of abnormal power consumption in grid.