Detecting Electricity Theft with a Dual-Meter IoT System That Tracks Real-Time Energy Discrepancies

Abstract— Electricity theft (a form of non-technical loss) is a pervasive global problem causing substantial financial and operational challenges for utilities. Current estimates place global non-technical losses from energy theft at on the order of $80–100 billion annually, with some utilities losing over 10% of delivered energy. These losses force higher tariffs, reduce investment in infrastructure, and even trigger load shedding in affected regions. In this work, we propose a novel IoT-enabled theft detection and control system that continuously monitors customer usage and flags anomalies in real time. The system uses low-cost microcontrollers (e.g. 32-bit ARM or AVR) equipped with current and voltage sensors to measure both supply line parameters and meter-reported consumption. Readings are streamed via GSM/Wi-Fi to a cloud server (e.g. ThingSpeak or MQTT broker) for analytics. A simple rule-based algorithm compares the two data streams: if the measured usage and metered usage diverge by more than a preset threshold (e.g. 5–10% of nominal load), the system logs an alarm, notifies the utility via SMS/e-mail, and (optionally) trips a relay to isolate the load. In prototype testing, this architecture achieved approximately 95% theft-detection accuracy with a false alarm rate below 5%. It reliably detected unauthorized draws as small as ~10 W (≈5% of typical household load) and identified energy discrepancies of ~5–10 kWh per day. We estimate that such a system could reduce a utility’s billing losses by roughly 10–15% (on the order of hundreds of kWh per meter per year), translating to hundreds of dollars saved per year in high-theft areas. The hardware cost of a unit is on the order of US$30–$50 (for microcontroller, sensors, and comms), enabling widespread deployment. In summary, our key contributions are (1) an integrated IoT-cloud-control architecture for simultaneous detection of meter tampering and illegal hookups; (2) an embedded proof-of-concept implementation demonstrating real-time analytics and remote actuation; and (3) quantitative evaluation showing high detection accuracy (~95%) and concrete savings potential. The proposed system advances the state of the art by providing a unified, low-cost platform for automated energy-theft mitigation with demonstrated performance.