Machine Learning Algorithm Analysis for Detecting and Classification Faults in Power Transmission System

Abstract

The importance of Power Transmission System PTS fault detection and classification is increasing day by day as because consumption of electricity is increasing. Short circuit fault in Power Transmission Line Network PTLN can cause severe damage to the power transmission system as well as economic loss. Power Transmission System requires new methods to detect and classify fault behaviour to prevent it from heavy damage. Machine Learning ML algorithms can be very effective to classify and detect various types of faults within the PTLN. There are variety of ML algorithms to recognise and classify the faults but as complexity of PTS is increasing day by day, reliability of these algorithms is decreasing. This study uses various types of ML algorithms to generate predictive models to evaluate what kind of algorithm is more appropriate to recognise and classify faults within the PTLN. Faults investigated in this research work include (L-L) double line fault, (L-L-L) three phase fault, (L-G) line to ground fault, (L-L-G) double line to ground fault, and (L-L-L-G) three phase fault with the involvement of the ground. The data was evaluated using six (06) ML algorithms that are Decision Tree, Support Vector Machine (SVM), K-Nearest Neighbor (Knn), Random Forest, XGBoost (XGB) and Naive Bayes (NB) for recognise of fault and classification within the PTLN. The performance of ML algorithms obtained by comparing the results and determine which algorithm is fast and more accurate. These results can be used to create more effective ML algorithms for PTS. The results indicate that the application of ML algorithms for PTS task could improve the PTLN yield and save time for technical teams.