A Machine Learning-Based Robust Classifier for Power Quality Disturbances in Smart Grids

The integration of renewable energy sources and advanced monitoring technologies in smart grids has improved power supply efficiency and reliability, but also introduced challenges in managing power quality (PQ) disturbances. This study proposes a machine learning-based approach combining Deep Learning algorithms and advanced signal processing techniques to identify and classify PQ disturbances, such as voltage sags, swells, flickers, and harmonics. Utilizing Deep Auto-encoder networks for feature extraction and dimensionality reduction, the method enhances classification accuracy through an ensemble of Support Vector Machines, Random Forest, and Extreme Learning Machines. Simulation results based on comprehensive datasets demonstrate the framework's superior accuracy, robustness, and computational efficiency. The findings suggest the methodology's potential as a reliable tool for PQ management in smart grids, enabling early detection and classification of disturbances to mitigate their impacts, thereby supporting the advancement of smart grid technology and maintaining high PQ standards amidst increased renewable energy integration.