Multi-Class Fault Type Classification Using Sequence Components and Wavelet Features with Comparative Study of Machine Learning Models and Neural Networks

The growing complexity of modern power networks requires intelligent fault classification techniques capable of handling multiple fault conditions with high accuracy. Classifying faults in transmission lines with higher accuracy is very important to provide reliability to the power system along with a shorter fault clearance time. The presented work provides a reliable fault classification system in transmission lines using a hybrid approach in feature extraction, taking into consideration both the symmetrical components and time-frequency analysis using wavelet transform. The hybrid technique will be able to analyze imbalance in both transients. Some machine learning algorithms along with an artificial neural network are presented for their implementation in a unified platform. A comparative study will be performed based on standard parameters for fault classification accuracy. The simulation results show improved accuracy with lower misclassifications among related faults using ensemble machine learning algorithms.