Power System Transient Stability Assessment Using Dictionary Learning on PMU Data With Noise

This paper presents a transient stability assessment (TSA) approach that leverages dictionary learning technique applied to data acquired from phasor measurement units (PMUs). The signal data encompasses measured voltages, active and reactive powers, along with other pertinent network quantities. The dictionary learning process employs matrix factorization to derive a limited set of fundamental components, referred to as atoms, which can sparsely represent the observed data. By analyzing the reconstruction errors resulting from the application of the dictionary to the PMU data, we propose a binary classification model capable of effectively identifying unstable TSA events, specifically those arising from short-circuit incidents. Furthermore, we investigate the impact of noise present in the PMU data and its subsequent effect on the performance of the classification model. The proposed methodology is evaluated using the IEEE New England 39-bus power system as a test case.