Enhancing Fraud Detection in Renewable Energy Grids through Behind-the-Meter PV Disaggregation

Detecting non-technical losses (NTL) in electrical grids with high renewable energy penetration presents unique challenges. This study proposes a methodology that combines a two-step behind-the-meter (BTM) photovoltaic (PV) disaggregation algorithm with an unsupervised fraud detection model to improve the detection rate in renewable energy grids. The methodology is validated using real-world data from the Ausgrid dataset, comprising several prosumers with BTM PV systems. The disaggregation algorithm accurately estimates the PV capacity with a Mean Average Percentual Error (MAPE) of 7.85%. Applying the disaggregation model improves fraud detection performance, increasing the Matthews Correlation Coefficient (MCC) from 0.654 to 0.747. These services have been developed and tested within the OMEGA-X European project.