An Artificial Intelligence–Based Approach for Fault Detection and Classification in Electric Power Systems

Abstract

In this study, a machine learning-based approach is developed for the detection and classification of faults occurring in electric power systems. A dataset containing different fault scenarios was obtained from the Kaggle platform, and a comprehensive performance analysis was performed by applying various machine learning algorithms to this data. The models were compared within the scope of fault detection and fault classification tasks. In the analysis, Decision Tree (DT), Random Forest (RF), XGBoost, LightGBM (LGBM), Support Vector Machine (SVM), k-Nearest Neighbors (kNN), Naive Bayes (NB), Logistic Regression (LR), and Multi-Layer Perceptron (MLP) algorithms were evaluated. The results revealed that tree-based and ensemble methods exhibited higher accuracy, reliability, and stability compared to other models. In particular, the RF and XGBoost algorithms stood out as the most successful models in terms of fault detection and classification. The study findings demonstrate that the proposed approach enables rapid, accurate, and reliable fault detection in power systems. Future studies are recommended to test the model with larger, more balanced datasets from different systems and assess its applicability in real-time conditions.