Assignment of AI Detected Buildings from Satellite Images to Registered Meters for Energy Theft Detection

This paper proposes an innovative method to detect energy theft on electrical network. By utilizing satellite imagery to analyze building rooftops and comparing them with registered consumer units, illicit connections in the distribution grid can be identified. The approach solves the assignment problem between AI-detected buildings and registered consumer units, enabling the detection of illegal connections. Suspicious locations without assigned meters are flagged, allowing utility companies to prioritize inspections effectively. Simulated tests with representative data demonstrate promising results in accurately detecting the majority of illegal connections, and an illustrative case of real-world detection is provided. However, further assessment and exploration of alternative methodologies are necessary to enhance accuracy. This research provides utility companies with a valuable tool to combat energy theft, improve system efficiency, and ensure a reliable and profitable electrical energy sector by effectively addressing the assignment problem between detected buildings and registered consumer units using satellite imagery.