Non-Technical Loss Detection using Missing Values' Pattern

Non-Technical loss (NTL) has gradually become a threat of the continuous stability of power supply, and is of great significance to social production and people's lives. Electricity theft not only brings losses to the power supply company, but also reduces the quality of power supply. The wide application of advanced metering infrastructure (AMI) makes data-based electricity theft detection algorithms possible. The current data-based methods mainly focus on the feature of electricity consumption. This paper proposes a new feature data, the location information of missing values, and explores its relationship with electricity theft from the perspective of electricity theft means. Then, a convolutional neural network model is built and missing value location data was fitted. The good performance of this model confirms the close relationship between missing values and electricity theft. Finally, the specific missing value pattern is analyzed through the k-means clustering algorithm.