Identification of power quality disturbances based on singular value decomposition of S-transform time-frequency matrixes

This paper presents a novel identification method for power quality disturbance (PQD) based on singular value decomposition (SVD) of S-transform (ST) time-frequency matrixes (TFM). The ST has excellent time-frequency resolution characteristics and makes it simple to observe and analysis the PQD signals. The SVD is an effective algebra feature extraction method, so SVs can reflect the inherent characteristics of TFM nicely. Firstly, the feature vectors are constructed with helps of the first two SVs and the first column of V matrix. Thus, using Euclid distance quantifies the difference between different feature vectors of disturbance signals to classify feature vectors in terms of a minimum distance. The disturbances are then identified. Simulation results show that the proposed method can achieve a satisfied accuracy of identification with different noise SNR, so it is an effective method to identify the PQD. (4 pages)