Online Fine-Grained Root Cause Classification for Power Quality Disturbances

Power quality disturbances (PQDs) account for an important share of power quality problems that hurt the resilience of electric power systems. Timely identification of the root causes for such disturbances will enable transmission system operators take necessary countermeasures to decrease their prevalence, and eventually help improve power system resilience. This paper presents comparative evaluation results of learning-based experiments of fine-grained root cause classification for PQDs in the Turkish electricity transmission grid. Various machine learning, ensemble, and deep learning based models are considered. The best performing ensemble learning model is successfully integrated into an operational power quality monitoring system covering the whole transmission grid, and hence the model is currently performing online root cause classification for PQDs almost in real-time. Online classification of disturbances detected in the transmission grid and monitoring these classification results will lead to better power system operation, management and planning, and eventually, to improved power system resilience.