Low False Positive Rate Electricity-theft Detection Method Based On Line Loss Power Causality Analysis

Electric power industry is an important basic industry of national economy,which is related to the development of national economy and social progress.Maintaining normal power supply and consumption order is a significant guarantee for maintaining social harmony and stability and building a well-off society.In the operation of power grid,electricity-theft and other non-technical factors are the major sources of power loss.Electricity-theft directly results in the loss of operation income of power supply enterprises,so it is a severe obligation of power supply enterprises to carry out electricity theft inspection.Traditionally,without the detailed power consumption data of users,the power supply enterprises mainly identify the high loss area whose line loss is higher than a specific threshold through monthly electricity quantity,then users with high risk are selected to carry out electricity stealing inspection according to the industry characteristics.But there are still some defects in manpower and time efficiency.In recent years,the construction of integrated marketing and distribution system of power supply enterprises and the popularity of smart meters provide abundant available data for identifying abnormal electricity consumption.Researchers have carried out a lot of research on data-driven electricity theft detection based on clustering and classification anomaly identification technology.Most of these methods design and select characteristic index items according to daily load curve and sudden drop of electricity consumption,and then identify abnormal power consumption with adaptive improvement of algorithm.However,the power supply enterprises serve a large number of customers,the type of industry are complex,and the user’s electricity consumption behavior patterns are diverse,which does not necessarily strictly meet the normal user’s electricity consumption behavior characteristics assumed by the selected characteristic index items,and the power consumption of some users will fluctuate greatly under normal conditions,which is easy to cause false alarm.Aiming at the demand of identifying electricity-theft users efficiently and accurately on the basis of targeted screening of high loss distribution area to narrow the detection range,this paper first uses linear regression model to determine the user variables that are significantly related to the line loss of the distribution area.Compared with the simple linear regression model,the multiple linear regression model is more suitable for describing the specific form of quantitative correlation between multi-user electricity and line loss in the distribution area.Considering the abnormal fluctuation of line loss due to the existence of power stealing users,the power loss in the process of line operation presented by the station area has certain similarity with the fluctuation trend of economic variable data which is the same time series when affected by other factors.According to the historical data of line loss and power consumption,the equilibrium relationship and Granger causality between them can be analyzed to identify the users who cause abnormal line loss in the distribution area.The results show that the Granger causality test can accurately reveal the one-way causal relationship between the electricity consumption of electricity stealing users and the power loss in the substation area,and the change of electricity consumption of power stealing users will have a strong impact on the power loss in the substation area.According to the user power consumption and line loss,the power stealing users in the high loss area and under the line can be accurately identified.Aiming at the problem that the data of power users are missing in the process of uploading,a detection method of power stealing in low-voltage substation area based on edge computing is proposed.Granger causality analysis was carried out on the general table of the transformer zone,based on the actual data of a high loss area,the power stealing user detection is carried out,the analysis results of the edge and background are compared,and the effectiveness and accuracy of the proposed method are confirmed through on-site inspection.