Ensemble Learning for Enhanced Fault Classification in Smart Distribution System

With the developments in Deep Learning algorithms, researchers have been turning to neural network-based approaches for detection of transmission line faults. However, deep learning methods necessitate a huge data to train properly. Furthermore, they have much less interpretability than traditional machine learning models. Here, some of the most popular ensemble learning techniques are compared for classification of transmission line faults. The study compares Random Forest, XGBoost and AdaBoost. The training data was generated using a 6 bus IEEE system simulated in MATLAB. The data thus obtained was pre-processed and relevant features were extracted from it before training the models. The study shows that AdaBoost performed better than the other classifiers, achieving an accuracy of 99.46%, thus establishing that traditional machine learning models are still relevant.