부산대학교 도서관

This paper evaluates and compares commonly used methods for analysis of waveform distortions in power supply systems, including one new wavelet-based method. The evaluation is performed using a set of selected test waveforms (TWs) that have known waveform distortion characteristics, representing various types of deviations from an ideal sinusoidal waveform with a constant magnitude and frequency. The presented analysis uses both stationary and nonstationary TWs, featuring low and high frequency harmonics, interharmonics and subharmonics, as well as the frequency and magnitude variations of the fundamental and nonfundamental (including dc) components. The capabilities of the considered methods to correctly identify and accurately quantify specific waveform distortion features are evaluated by three general types of error metrics: a) statistical error indices, b) energy-based quantities, and c) waveform-related indicators. The combining of these metrics allows for a more detailed comparison of the original time-domain TWs with the waveforms reconstructed by the different methods. The presented results are part of the efforts to formulate a general methodology for a physically relevant evaluation of accuracy of different methods and specification of standardised test waveforms for a direct and comprehensive comparison and benchmarking of different methods.