Enhancing Power System Reliability with Machine Learning-enabled Fault Management
Rahman Md. Mahidur
Department of Electrical and Electronics Engineering, Faculty of Science and Engineering, Jatiya Kabi Kazi Nazrul Islam University, Trishal, Mymensingh, Bangladesh.
Nayem Zannatun
Department of Electrical and Electronics Engineering, Faculty of Science and Engineering, Jatiya Kabi Kazi Nazrul Islam University, Trishal, Mymensingh, Bangladesh.
Hossain Md. Sajjad
Department of Electrical and Electronics Engineering, Faculty of Science and Engineering, Jatiya Kabi Kazi Nazrul Islam University, Trishal, Mymensingh, Bangladesh.
Shahiduzzaman Kazi Md. *
Department of Electrical and Electronics Engineering, Faculty of Science and Engineering, Jatiya Kabi Kazi Nazrul Islam University, Trishal, Mymensingh, Bangladesh.
*Author to whom correspondence should be addressed.
Abstract
Aims: To develop and evaluate a machine learning-based framework for the accurate detection, classification, and localization of faults in a power transmission system, with the goal of enhancing system reliability and reducing maintenance delays.
Study Design: Simulation-based experimental modeling study.
Place and Duration of Study: Department of Electrical and Electronics Engineering, Jatiya Kabi Kazi Nazrul Islam University, Trishal, Mymensingh, Bangladesh. (Note: The exact duration/dates of the study are not specified in the provided text).
Methodology: A 4-bus power transmission system was simulated using MATLAB Simulink to generate a diverse, realistic dataset of 4,440 samples. This included 4,000 non-fault samples and 440 fault samples representing 11 distinct fault types across four buses. Four machine learning models—Support Vector Machine (SVM), Random Forest, Long Short-Term Memory (LSTM), and K-Nearest Neighbors (KNN)—were trained and evaluated. Their performance was assessed using standard metrics, including accuracy, precision, recall, and F1-score.
Results: The Random Forest model achieved the highest performance, recording an overall accuracy of 96%. It proved to be the most effective model for handling complex fault patterns with minimal misclassifications, consistently outperforming the SVM (95% accuracy), KNN (94% accuracy), and LSTM (93% accuracy) models.
Conclusion: The proposed framework provides a robust, reliable tool for power system engineers to quickly and accurately identify, classify, and localize transmission line faults. Utilizing this approach can improve predictive maintenance, drastically reduce system downtime, and enhance overall power grid safety and efficiency.
Keywords: Fault detection, power system faults, machine learning, random forest, fault classification, predictive maintenance