Comparative Evaluation of Machine Learning Algorithms for Fault Classification in Smart Grid Systems

Citation Count

Abstract Views

35

Downloads

3

Abstract

Frequent power outages and inefficient fault detection mechanisms remain major challenges in Nigeria’s electricity distribution systems. This study develops a machine learning-based model for real-time fault detection in smart grid networks. The aim is to improve fault prediction accuracy and reduce system downtime. We adopted a computational experimental design using historical power system data, training and validating supervised learning algorithms (Random Forest and Support Vector Machine). We evaluated model performance using accuracy metrics, regression analysis, and hypothesis testing. Findings revealed a significant improvement in fault detection accuracy (93.6%) compared to traditional methods (t = 6.48, p < 0.001). A strong correlation (r = 0.85) was observed between predictive model outputs and actual fault occurrences. System downtime was reduced by 58.9%, while response time improved by 46.3%. The study recommends integrating machine learning-based fault detection systems into national grid infrastructure and training technical personnel. In conclusion, intelligent fault detection systems offer a scalable solution to improve power reliability and efficiency.

Keywords: Fault detection, Machine learning, Power systems, Smart grid, Reliability



Journal Title: Journal of Technical and Vocational Education and Training

Category: Electrical and Electronics Engineering

ISSN: 3141-2742

Year of Establishment: 2025

Section: Faculty of vocational and Technical Education University of Benin in collaboration with School of Technology Education, FCET, Asaba

Volume: 1

Issue: 5

Abstract Views:

Total Download: 3

Oguntunwase Akinola Ojo, Aimayo Vincent Joshua. 2026 Comparative Evaluation of Machine Learning Algorithms for Fault Classification in Smart Grid Systems. Journal of Technical and Vocational Education and Training. 1 (5). 102-114. https://doi.org/10.61955/IPZHOT

Our Partners