IoT-Based Transformer-to-Consumer Comparative Analysis for Real-Time Electricity Theft Detection

Abstract

Electricity theft is a significant source of non-technical losses in electrical distribution networks, particularly when unmetered loads or illegal connections occur between distribution transformers and registered consumers. This paper presents a low-cost IoT-based transformer-to-consumer comparative monitoring system for real-time screening of unexplained electricity losses. The proposed system uses ESP32 microcontrollers and PZEM-004T energy-monitoring modules to measure the output of a three-phase transformer and the consumption of individual single-phase consumers. Measurements are transmitted through Wi-Fi to a Supabase/PostgreSQL cloud backend at approximately five-second intervals, where aggregated registered-consumer power is compared with transformer output power. During the controlled proof-of-concept experiment, unaccounted power was approximately 0.13 kW under nominal operation. After introducing an additional unmetered load, the residual increased to 3.22 kW, exceeding the 0.5 kW experimental threshold. The PostgreSQL trigger consequently changed the theft status from false to true, and the dashboard displayed the alert during the next monitoring cycle. These results demonstrate the functional operation of the sensing, communication, cloud-processing, decision, and visualisation stages. However, structured measurement-error statistics, packet-delivery reliability, precise end-to-end response time, and statistical detection performance were not established during the present prototype experiment. Therefore, the generated alert should be interpreted as a potential theft or abnormal-loss inspection flag rather than conclusive evidence of electricity theft. Repeated calibration, communication-reliability testing, response-time analysis, and field-scale validation are recommended for future work.