Ensemble learning for event detection and disturbance classification in power quality data from solar energy systems

Integrating renewable energy sources into power grids introduces complex challenges, particularly in accurately detecting and classifying power quality (PQ) disturbances due to the variability and intermittency of renewable energy generation. This study proposes a classification framework that employs data balancing techniques and ensemble learning models to classify key PQ events, such as voltage sags, waveform distortions, and over-under frequency disturbances. A comprehensive dataset was collected from a solar farm in Norfolk, England, covering January to December 2023, to perform this analysis. By investigating this high-resolution, high-fidelity big PQ data, the research explored real disturbances and provided insights into the solar site’s operational behavior, contributing to improved grid reliability. This work offers valuable insights into solar farm operations, helping utility-scale owners and operators implement more effective and cost-efficient condition monitoring strategies.