Electricity Anomaly Detection Research of Flue-Cured Tobacco Users Considering the Characteristics of Industry Electricity Consumption Behavior

Abstract

Power theft inspection is an effective means to improve the operating income of power supply enterprises. However, in practice, there has been a problem that the accuracy rate of power theft detection is not high. Cluster analysis of power consumption behaviors by subdividing power users by industry will not only make it possible to extract characteristic index items that accurately describe user behaviors in combination with industry power consumption behavior characteristics, but also reduce user classification dimensions and false positives for power theft detection. Rate is the development direction of promoting the practical application of data-driven power theft detection. This article takes flue-cured tobacco users as the object. Firstly, it analyzes the industry electricity consumption patterns of flue-cured tobacco users, and then based on the statistical analysis of industry characteristics, establishes power consumption index characteristic items that accurately describe the industry characteristics of flue-cured tobacco users. The propagation clustering algorithm performs cluster analysis of users in flue-cured tobacco industry. Numerical simulation results show that the proposed method can not only accurately identify misclassified non-flue-cured tobacco users such as public properties in rural areas, reduce the scope of power theft detection, but also effectively detect specific types of flue-cured tobacco specific power stealing behavior.