Classification of Power Quality Disturbances Using Artificial Neural Networks and a Logarithmically Compressed S-Transform

Accurate and robust classification of power quality events is an important task in smart grid development. In this paper, a novel system for automatic detection and classification of power quality events is presented. The proposed system relies on artificial neural network, as the principal method for classification of power quality signatures extracted from the observed current or voltage waveforms. Power quality signatures are obtained as a set of processed statistical features that describe the result of time-frequency analysis, based on a logarithmically compressed S-transform. The proposed method is evaluated on a large database of simulated power quality disturbances, which include examples of voltage sag, swell, momentary interruption, notch, harmonics, transient oscillation and voltage fluctuation. The results show that the proposed system is able to accurately and robustly detect power quality events in isolation, or in combination, under the noisy and noise-free conditions.