Enhancing Situational Awareness: Predicting Under Frequency and Under Voltage Load Shedding Relay Operations

Abstract

This paper proposes a machine-learning-based method to enhance online situational awareness in power systems by predicting under frequency load shedding (UFLS) and under voltage load shedding (UVLS) relay operations for several seconds after a disturbance. Voltage magnitudes/angles of electrically closest high voltage buses to the relay locations along with the relay settings are used as the input features to train random forest (RF) classifiers that predict UVLS/UFLS relay operations, respectively. A variety of contingencies considering different operation conditions and topologies of the Western Electricity Coordinating Council (WECC) system data representing the 2018 summer-peak load are studied offline using the GE positive sequence load flow analysis (PSLF) software. The results are used to create a comprehensive dataset for training and testing the classifiers. A comparison between the performances of RF models trained with different periods of input data is conducted in the presence of measurement errors.