Comparison of transformations and feature extraction techniques to characterize fault-induced voltage sags

Abstract

This paper presents a comparative study of domain transformations and feature extraction techniques to characterize fault-induced voltage sags. For this purpose, synthetic signals of fault-induced voltage sags are generated through extensive simulations in MATLAB/Simulink. Next, some relevant transformations are applied to the synthetic signals, namely, the space phasor model, discrete Fourier transform, and short-time Fourier transform. A set of statistical, time series, and spectral features are extracted from transformation outputs to obtain signal characterization useful, for instance, for classification of voltage sags employing artificial intelligence techniques. The comparison of the applied domain transformations and feature extraction techniques covers quantitative and qualitative aspects including computation time, storage requirement, linear separability and physical interpretation of features, and suitability for characterizing voltage sags. Finally, the main findings of the work are discussed, and conclusions are remarked.