Mean-Shift Based Clustering Research On The Screening Method Of Abnormal Electricity Users

The non-technical loss(NTL)on the power consumption side is an important evaluation index reflecting the economic benefits of the power system.The NTL problem caused by electricity theft is the main reason for the loss of net profit of power companies,and it is also a problem faced by global power companies.A persistent historical conundrum.With the continuous development of electric energy metering technology and the construction of smart grid,user power consumption data shows a trend of high-dimensional and exponential growth,which provides effective support for the development of data-driven abnormal power consumption detection methods.It has become a research hotspot to study how to mine effective information from massive electricity consumption data and identify electricity stealing behavior.In recent years,universities and scientific research institutions at home and abroad have carried out a lot of research on how to accurately identify abnormal electricity consumption of users from the perspectives of abnormality principle,feature extraction,model construction,etc.,laying an important foundation for data-driven electricity theft detection methods.It should be pointed out that the main problem of electricity stealing detection is how to improve the recognition rate and reduce the false alarm rate.The existing classification-based method can effectively improve the accuracy rate in the optimization process,but it is difficult to reduce the false alarm rate.It is necessary to conduct on-site investigations one by one.The high false alarm rate will increase the difficulty of on-site inspection by operation and maintenance personnel,and consume a lot of manpower and material resources.Electricity theft detection is always a difficult problem.Aiming at the high false alarm rate of existing detection methods,this paper proposes a screening method for abnormal electricity users based on mean-shift clustering.The method first uses electricity consumption and its statistical characteristics as the feature vector to characterize users;then uses the classification-based method to train the sample set to obtain a classification model,and inputs the newly collected samples into the model to identify whether it is a suspected electricity stealing user;The method based on MeanShift clustering is used for secondary screening of suspected electricity users,that is,through cluster analysis of historical electricity consumption data of suspected electricity stealing users,the clusters of historical electricity consumption patterns of users are obtained;When the characteristics and the normal low-power production and operation state are clustered into the same category,it is considered to be an abnormality caused by the normal transition of the user state,and the suspicion of abnormal power consumption can be excluded.Finally,an experiment is carried out using the real electricity stealing data in a certain place.The simulation results show that the method can effectively screen out the falsely reported electricity stealing users and improve the economic benefits of the power grid.