Using Machine Learning for Classifying PMUs Observation in Smart Grids

Phasor Measurement Units (PMUs) are vital in smart grids, providing real-time electrical waveform measurements such as voltage, current, and frequency across the smart grids (SGs). PMUs can work together with GPS to enhance the location and time. Moreover, PMUs enable quicker responses to disturbances, improve fault detection and localization, and support the integration of renewable energy sources. In this regard, classifying the PMUs’ measured quantities is desired to enhance the monitoring, control, and stability of the SGs. This paper presents a machine learning (ML) classifier based on extreme gradient boost (XGB) to efficiently classify the PMUs’ observations in an IEEE Bus system. More particularly, the proposed model is developed to classify the normal operation, faults, line outages, generation changes, and load changes or drops. The proposed model is examined by six evaluation metrics namely, accuracy, F1-score, kappa score, elapsed time, learning curve, and confusion matrix. The proposed XGB classifier outperformed the benchmarks achieving an accuracy of 99.9%.