AI-driven fault location in medium-voltage underground cables using wavelet packet transform

Fault diagnosis and fault location are challenging in hazardous environments such as oil fields and refineries due to explosion and fire risks. Conventional high-current fault location methods for fault diagnosis and fault location in medium-voltage (MV) underground cables are impractical and unsafe in these sites. This paper introduces a novel AI-based approach for fault location that employs the Bergeron transmission line model to simulate transient behavior, while integrating wavelet packet transform (WPT), inception angle and Shannon entropy for feature extraction and intelligent fault analysis. The method safely and accurately locates faults in real time using only three-phase current signals and without additional hardware. The method extracts key waveform features, improving efficiency and accuracy, and is robust to changes in fault resistance, inception angle, and distance. A neural network adapts to system variations, and streamlined feature extraction reduces complexity. Simulation results confirm the algorithm effectively locates faults under dynamic conditions, enhancing safety and reliability in high-risk environments through the integration of the Bergeron model, WPT, Shannon entropy, and AI. The novel Bergeron-WPT-ANN framework achieves a highly accurate and rapid fault location, making it an efficient solution for real-time industrial power-distribution protection.