Deep Learning-Based Fault Detection and Classification in Power System Protection: A Comparative Study

Type: MA thesis

Status: running

Date: March 15, 2025 - September 15, 2025

Supervisors: Julian Oelhaf, Andreas Maier, Siming Bayer

Abstract

Faults in power systems can occur during electricity transmission between grids, posing risks to system stability and reliability. Detecting and classifying these faults accurately is essential for effective protection and mitigation strategies. Deep learning models have shown significant promise in automating fault detection and classification, but their performance varies across different scenarios. This thesis presents a comparative study of various deep learning models, including CNNs, RNNs, LSTMs, and GRUs, for fault detection and classification in power system protection. The models are evaluated on a relevant dataset using multiple performance metrics to determine their effectiveness. The study aims to provide a structured performance analysis, offering insights into model suitability for specific fault conditions.