FenceGuard: An IoT and TinyML-Based Smart Unauthorized Electric Fence Monitoring System with GPS-Assisted Fault Localization

Abstract

Unauthorized intrusion, wire damage, and electrical faults in perimeter electric fences remain challenging problems in agricultural and rural environments, particularly when fence boundaries extend over large areas and cannot be inspected continuously. Conventional monitoring approaches primarily depend on manual inspection or predefined electrical thresholds, which provide limited discrimination between different abnormal operating conditions under changing environmental and load conditions. This paper proposes FenceGuard, an Internet of Things (IoT)-enabled smart electric-fence monitoring architecture that combines voltage and current sensing, edge machine learning, GPS-assisted localization, cloud monitoring, cellular alerting, and solar-powered operation. The proposed system uses an ESP32 microcontroller as the field processing node. Electrical measurements obtained from voltage and current sensing circuits are processed locally and supplied to a lightweight machine-learning model deployed using TensorFlow Lite Micro. The model is designed to classify five operating conditions: normal operation, voltage drop, wire cut, short circuit, and tampering. When an abnormal condition is identified, the system activates a local buzzer, obtains geographical information through a NEO-6M GPS module, and transmits the event through Wi-Fi to Firebase for visualization through a Flutter dashboard. GSM communication is additionally proposed as a backup alert mechanism when conventional network connectivity is unavailable. A solar-battery power subsystem is incorporated to support deployment in remote agricultural locations. Unlike cloud-dependent monitoring architectures, the proposed design performs primary fault classification at the edge, reducing dependence on continuous network connectivity. The paper presents the motivation, research gap, system architecture, hardware design, TinyML methodology, communication framework, and proposed evaluation methodology. Experimental accuracy and performance measurements are intentionally reserved for subsequent prototype-validation work.