IoT and GPS-Based Fault Detection in Electrical Distribution Networks Using Machine Learning
Faults on electrical distribution networks are usually found through manual patrols, which delays the restoration of supply. This paper presents an Internet of Things (IoT) and GPS-based system that detects, classifies and locates faults on an 11 kV distribution feeder using machine learning. The feeder was modelled in MATLAB/Simulink, and five fault types were simulated at selected feeder locations with different fault resistances. Statistical and difference features were taken from the three-phase voltage and current signals, and five models were trained and compared: Decision Tree, Bagged Trees, Support Vector Machine, Artificial Neural Network and Physics-Informed Neural Network. Bagged Trees gave the best balanced result, with 90.74% accuracy for fault type, 96.00% for fault location and an average of 93.37%. This model was built into a MATLAB graphical user interface, which maps the predicted feeder section to GPS coordinates, uploads the result to the Ubidots cloud platform and sends an email alert with a Google Maps link. The results show that machine learning and IoT monitoring can be combined to find and locate faults on distribution lines. The system was tested on simulated data only, so validation with field measurements is recommended.
Authors
- Joseph Tetteh
- David Akorli
- Abdul Ghafaru
- John Ransford Arthur
- Isaac M.K Prah
Publication Details
- Published
- 2026-11-18
- DOI
- https://doi.org/10.68258/distribution_network_design.65995
- Primary Topic
- Power Systems Fault Detection
- Type
- article
- Field-Weighted Citation Impact
- 0.00