Identification of electricity theft based on the k-means clustering method

Abstract

Electricity theft seriously endangers the safe operation of the power system and damages the interests of the power supply company. Power thieves make the current anti-theft system difficult to identify by constantly researching new methods of electricity theft, so new methods of electricity theft identification need to be studied. In this paper, we propose a method based on layered k-means clustering to identify electricity theft. Firstly, the distribution network line loss data is clustered and analyzed to identify distribution stations with abnormally fluctuating or persistently high line loss rates, and the time dispersion is defined according to the clustering results to measure the suspicion of electricity theft. Then, we analyze the users in the abnormal station area, and study the possible relationship between the change of electricity of a single user and the change of line loss rate in its station area through correlation analysis, determine the suspicion of electricity theft of a single user, and identify the specific electricity theft behavior. The practicality and accuracy of the proposed method is verified by analyzing the line loss rate of two typical station areas, and later by actual data.