Comparative Analysis of Classification and Location of Faults in a Long Transmission Line Using Random Forest Technique

This paper represents random forest based (decision tree) location estimation and classification of fault in 400 KV, 300 km long transmission line (TL). The technique extracts single cycle post fault current and voltage signal and fed to signal processing method wavelet packet transform for useful features. Optimal features are extracted by feature selection method and then normalized. Taking a dissimilar simulation state like (10 types of fault, fault resistance, inception angle of the fault and distance of the fault) train and test matrix is generated. Then decision tree based (random forest) technique is used for location and classification of fault under study model. Simulation results are presented the proposed technique gives (99.5%) accuracy of location of faults and (less than 0.05%) error for classification of fault. The projected method is compared with another researcher approach.