Research on Electricity Theft Detection Technology Based on Deep Learning

The global issue of electricity theft has become increasingly severe, posing significant threats to the stability of power systems and public safety. This paper introduces a Multi-Dimensional Perception Network for Electricity Anomaly Detection, which integrates the DSAE module and MSSPAM module to enhance the model's capability in capturing both global and local features and improving temporal dependency modeling. Experimental results demonstrate that the proposed model outperforms existing baseline methods across multiple metrics, including accuracy, precision, and recall, providing an efficient and reliable solution for electricity theft detection.