Comparative study of machine and deep learning frameworks for fault classification in HVAC transmission lines

The probability of faults increases in the ever-expanding power system. Rapidly rebuilding the system after classifying and clearing faults enhances power system quality and reduces the likelihood of blackouts. Therefore, it is crucial to assess faults quickly and accurately. Transmission lines are particularly susceptible to faults because they are significant components of power systems. Consequently, it is necessary to enhance reasonable procedures for identifying and classifying fault sites in transmission systems. An exhaustive literature review was conducted to understand the existing methods and emerging trends in fault classification. This study compared various machine learning and deep learning algorithms for fault classification using the open-source Python tool. An evaluation of the algorithms was performed to demonstrate their efficiency based on accuracy, precision, F1-score, and recall. The prediction and training times for each model were also considered. Ensemble techniques, such as decision trees, random forests, and extra trees, achieved the best performance, followed by deep learning methods, such as artificial neural networks and long short-term memory. However, the training and prediction times for the deep learning methods were the highest. Furthermore, hyperparameter tuning suggestions were offered to improve performance.