Electricity Theft Detection Based on SMOTE Oversampling and Logistic Regression Classifier

Abstract

Many countries are starting to construct the smart grid (SG) due to it creates a reliable, clean, and efficient power system compared to the traditional power grid. However, electricity theft can be harmful to grid operators, smart meter data from the Advanced Metering Infrastructure (AMI) can be tampered by thieves by using advanced digital instruments or cyber attacks to reduce electricity bills, which can have devastating financial consequences for utilities. The performance of the existing detection algorithm is influenced by the serious imbalance of data categories. In this paper, a Synthetic Minority Oversampling Technique (SMOTE) oversampling is employed to solve the imbalance problem, so the processed data and normal data can achieve a relative balance, and select logistic regression algorithm for power theft detection.