Usage of Harmonic Synchrophasors for High-Impedance Fault Classification in Microgrids

The growing presence of monitoring systems based on phasor measurement units (PMU) in distribution systems has strengthened the development of data-based solutions. The correct identification of events in distribution networks is strictly benefited from large data availability, especially those challenging to detect, such as high-impedance faults (HIFs). In this sense, this work presents the usage of harmonic synchrophasors on HIF classification in active distribution networks. A total of 2800 simulated events from a microgrid are considered, and six classic machine learning classification models are evaluated. The performed analyses focus on establishing how the amount and quality of the PMU data can influence the distinction of common microgrid events. Results show that low-resolution PMU data, like one phasor estimated per cycle, is sufficient to discriminate the events accurately. Regarding synchrophasors, the classification corroborates that odd harmonic contents present better discriminant potential when compared to even, and the usage of angle information subtly enhances outcomes. When dealing with noisy data, the classification stays stable for mild scenarios of signal-to-noise ratio. By validating the classification task with real HIF data, 99