Power transmission networks are inherently susceptible to faults arising from uncontrollable environmental factors such as lightning and storms. Rapid identification and localisation of these faults are critical to maintaining system reliability and enabling protective actions. While conventional methods rely on measurements from current and voltage transformers, this study proposes an artificial intelligence-driven approach for enhanced fault detection and classification in secondary power distribution networks. A custom-designed sensing prototype captured voltage and current data under simulated fault conditions, including short-circuit and open-circuit faults. Fundamental variables, such as fault type, sensor placement topology, and line distance, were rigorously controlled during data acquisition. The acquired dataset was analysed using two machine learning classifiers: an Artificial Neural Network (ANN) and an Adaptive Neuro-Fuzzy Inference System (ANFIS). Performance was validated under variable load conditions across line distances of 200–800 meters. Simulation results demonstrate that the ANFIS classifier achieved superior accuracy in fault classification (99.7%) and minimal distance estimation error (0.5%). Both ANN and ANFIS delivered high precision in fault detection, localisation, and classification, with ANFIS exhibiting significantly faster training convergence (1 ms). The framework presents a robust, computationally efficient solution for real-time fault management, recommending the integration of dedicated sensors by distribution utilities to enable targeted grid interventions.
Udofia et al. (Tue,) studied this question.