Selection of wavelets for analysis of power system disturbances

Abstract

The wavelet transform represents a powerful signal processing tool with a variety of applications. The main reason for the growing interest in this field of signal processing is its ability not only to decompose a signal into its frequency components, but also, as opposed to the Fourier transform, to provide a non-uniform division of the frequency domain. The decomposed signals possess a powerful time-frequency localization property, which is one of the major benefits provided by the use of the wavelet transform. This property is extremely important for the analysis of power system disturbances. Large numbers of disturbances occur in a very short time period within the analyzed signal while having the broad frequency spectrum (temporal localization). Some disturbances spread over the entire signal duration while having their significant frequency content very close in frequency to the "normal" signal components (frequency localization). In spite of the great attention given to the application of wavelets in power system analysis, only a small number of papers deal with the details of selecting the right wavelet suitable for a given application. The need for defining a systematic approach for choosing the best wavelet for an application at hand still exists. This thesis focuses on finding the best approach to the wavelet selection for the analysis of typical power system disturbances. The research topics addressed in this document include: o initial theoretical selection of wavelets based on the typical wavelets characteristics such as compact support, orthogonality, symmetry, number of vanishing moments, and regularity. o definition of a criterion for wavelet performance assessment. o implementation of wavelet performance assessment approaches in MATLAB and Simulink. o conclusions regarding the suitability of the specific wavelet for the given application.