Advanced Method for Electricity Theft Detection: A Case Study in Low-Voltage Station

Electricity theft can lead to grid load imbalance, increased grid losses, higher electricity costs, and even safety incidents, severely harming economic interests and social development. Various efficient and intelligent electricity theft detection methods have been researched with the development of the energy internet and advanced metering infrastructure. However, the issue of data imbalance and low quality frequently encountered has not been effectively addressed, leading to the inferior performance of the electricity theft detection model. This paper proposes an intelligent electricity theft detection system as a potential solution to this dilemma. The LOF and KM-IMI are first developed to locate and improve data quality. After that, the AEGRU combined with ACGANs was proposed to address the class imbalance problem. Finally, another deep learning model is employed to execute electricity theft detection. The experimental results indicate that the proposed system has excellent detection performance.