This paper introduces a novel strategy for fault classification and localization in power transmission networks using variational autoencoders (VAEs) to generate synthetic data, which is then used in conjunction with machine learning algorithms. The VAE generates synthetic fault data that capture the statistical properties of real-world data, augmenting the available dataset and improving fault recognition. A CatBoost algorithm, along with SVM, Decision Trees, Random Forest, and K-Nearest Neighbors, is used for classification and localization. The approach achieves 99% accuracy in fault classification and a mean absolute error (MAE) of 0.2 in fault localization, surpassing existing state-of-the-art techniques.

Muhammad Amir Khan, Bilal Asad, Toomas Vaimann, Ants Kallaste, Raimondas Pomarnacki, Van Khang Hyunh