Electricity theft detection technology is of great importance for identifying theft behaviors
reducing economic losses
and ensuring the safe and stable operation of power grids. To address the issue that existing methods overlook industry differences
leading to difficulties in adapting detection models to the load characteristics of various industries
this paper proposes an electricity theft detection method based on industry characteristics using graph neural networks (GNNs). The method deeply mines user electricity consumption behaviors based on industry characteristics
extracts relevant features
and constructs an electricity theft feature set incorporating industry-specific attributes. It then builds graph-structured data by integrating user correlations and inputs this data into a dual-layer graph neural network with learnable aggregation. Subsequently
electricity theft users are detected through a message-aggregation-update module. Case study analysis demonstrates that the proposed method outperforms comparison models in terms of accuracy
recall
and false positive rate. Ablation experiments further verify that industry characteristic feature extraction and the learnable aggregation mechanism are key to improving electricity theft detection performance.
SHI J,GAO Y,GU D,et al. A novel approach to detect electricity theft based on conv-attentional Transformer Neural Network [J ] . International Journal of Electrical Power & Energy Systems,2023,145:289-303.
MISHRA A K,DAS B. A novel density based clustering approach for electricity theft detection [J ] . IEEE Transactions on Industry Applications,2025,61(4):5537-5548.