Convolutional Neural Network and Data Augmentation Method for Electricity Theft Detection

Abstract

Electricity theft is a severe issue that causes huge revenue loss for utility companies and influences stable operation of power system. With the development of big data analysis, electricity theft detection (ETD) based on data-driven method has received massive attention. However, since available data in low-voltage (LV) network is usually sparse and imbalanced, most of the existing data-driven ETD methods are not applicable to residential customers. In light of this issue, we proposed a convolution neural network (CNN) and data augmentation method for ETD. This method applies kernel density estimator (KDE) and monte carlo method to expand dataset. Then CNN model is implemented on the dataset for classification. Experiment using realistic electricity usage data has been conducted to verify the effectiveness of this method, results show that this method can achieve high performance in terms of different metrics.