Electricity Theft Detection Of High-loss Line With Regression Analysis

The power distribution system has always been the difficulty and focus of line loss management in the power system.Statistics show that the line loss of 10kV and below in the distribution link accounts for 65%to 70%of the bus loss,and electricity theft is an important reason for the high line loss rate,so accurately identifying the user of power theft has important research value.In the era of extreme lack of available data,engineering and technical personnel have explored the idea of abnormal power inspection work for high-loss lines.Under normal conditions of online,variable,household and table relationship,abnormal line loss is mainly caused by electricity theft by special change users,and there are generally abnormal electricity users among the special variable users who access non-technical high-loss lines(such as line loss rates higher than 5%).This method can effectively narrow the scope of electricity inspection,significantly improve the targeting of electricity theft detection,and is a practical method that has withstood the test of practice.After continuous data governance,the integrated line loss management system of power supply enterprises in the same period has been able to provide basically accurate daily line loss.Electricity management personnel can select highloss lines in a targeted manner,and identify electricity theft users according to the correspondence between the access user’s electricity consumption data and the daily line loss change of the line.At present,such methods rely on the individual experience of the staff and lack systematic data mining and analysis,and it is still difficult to accurately identify abnormal users with hidden electricity theft performance,and it is urgent to study the user identification algorithm for high-loss lines.In order to solve the problem that most of the existing studies are based on the electricity load data to construct the input characteristic quantity,the correlation between the electricity consumption data and the line loss data is not considered,and the engineering application is less.Combined with the experience of engineering electricity theft audit,the paper uses the correlation between the electricity consumption of the user and the line loss to detect the electricity theft,considering that the user’s electricity theft will directly lead to abnormal fluctuations in the line loss,so there is a long-term dynamic interaction relationship between the user’s electricity consumption and the loss of the line.Establish a vector autoregressive model,analyze the internal relationship and interaction mechanism between user electricity consumption and line loss according to impulse response analysis and variance decomposition,and designate the users who have a significant impact on line loss and the greatest contribution to the fluctuation of line loss as electricity theft users;accurate and efficient electricity theft audit not only needs to accurately identify the power users,but also refine the detailed information of abnormal users.The existing high-loss line electricity theft detection research focuses on identifying and locating users who steal electricity,and less considers the time period information of users stealing electricity,which is difficult to facilitate on-site audit.Therefore,a high-loss line stealing user detection method based on forward stepwise regression and dummy variables is proposed.Finally,a test data set including long and short lines,the number of access users and various means of electricity theft is constructed,and the applicability of the two algorithms is verified from the engineering application requirements to better serve the production practice.