POWER DISTURBANCES PATTERN RECOGNITION USING SUPPORT VECTOR MACHINE

Abstract

Recently, many signal processing techniques, such as fast Fourier transform, short-time Fourier transform, wavelet transform (WT), and wavelet packet transform (WPT), have been applied to detect, identify, and classify power quality (PQ) disturbances. For research on PQ analysis, it is critical to apply the appropriate signal processing techniques and classifier to solve PQ problems. The aim of this paper is to develop a classification method based on the combination of Hilbert transform (HT) and support vector machine (SVM) for the assessment of power quality events. Recent data mining literature has shown that support vector machine methods generally outperform traditional statistical and neural methods in classification problems involving power disturbance signals. The features obtained from the Hilbert transform are distinct, understandable and immune to noise. Analysis is presented to verify that the merits of HT and SVM combination make it adequate for PQ analysis when compared with the existing techniques in the literature.