Estimação de harmônicos/interharmônicos: uma abordagem multitaxa

Abstract

This dissertation discusses parameter estimation of harmonics/inter-harmonics of electrical signals under time-varying conditions. Two new suitable approaches for processing this kind of signals are proposed. The first one is based on phase error of DFT (Discrete Fourier Transform) result, occurring due to the frequency deviation of input signal. This method makes use of digital filters to prevent oscillations of DFT results, estimating the frequency with a simple linear equation. Magnitude and phase response of DFT filter are then used to provide correct estimations of amplitude and phase. The second one, the central focus of this work, is obtained from the association of a digital analysis filter bank and multirate processing with an estimation tool: the Phase-Locked Loop (PLL). The filter bank is responsible for decomposition of the analyzed signal, separating it in its harmonic components. Down-sampler devices follow the filter stage, resulting in an estimation stage working with reduced sampling rate. For high-order harmonics, parameters are estimated from signals obtained by undersampling operation. Simulations results are presented, comparing performance of proposed methods with performances of traditional STFT (ShortTime Fourier Transform) algorithm, DFT Recursive algorithm and PLL-based single rate structure found in literature. With the aim in a DSP-based platform implementation, approximated counts show the number of arithmetic operations of each method, pointing out the methods more suitable for this purpose.