Comparative Analysis of Supervised Machine Learning Models for Fault Detection and Classification in Smart Grid Terminal Systems

Abstract

The increasing complexity and digitalization of smart grid terminal systems have heightened the need for reliable and intelligent fault detection and classification mechanisms. Traditional rule-based and threshold-driven protection schemes often struggle to adapt to dynamic operating conditions and diverse fault scenarios. This study presents a comparative analysis of supervised machine learning models for fault detection and classification in smart grid terminal systems. Using a comprehensive dataset comprising electrical, operational, and device-level features, four widely used supervised learning algorithms Random Forest, Support Vector Machine, Decision Tree, and Logistic Regression were implemented and evaluated. The models were trained using standardized preprocessing techniques, stratified data partitioning, cross-validation, and hyperparameter optimization to ensure fair and reproducible performance assessment. Experimental results demonstrate that ensemble-based models, particularly Random Forest, achieve superior performance in terms of accuracy, precision, recall, F1-score, and generalization stability, while simpler models offer faster training times and improved interpretability. The findings highlight critical trade-offs between predictive accuracy and computational efficiency and provide practical insights for selecting suitable machine learning models for real-time smart grid fault detection. Overall, this research contributes a robust evaluation framework that supports intelligent decision-making and enhances the reliability and resilience of modern smart grid infrastructures.