Técnica de detecção de distúrbios para o monitoramento da qualidade da energia

Abstract

This dissertation discusses the disturbance detection problem for power quality monitoring applications when the voltage signal intervals for analysis correspond to submultiples of one cycle component fundamental of the power system. In addition, a technique to detect disturbances in a mono-phase signal if its discrete-time version is constituted by at least 16 samples is introduced. This detection technique makes use of a 2nd order notch filter to decompose the voltage signal into two components. From those components, features based on higher order statistics are extracted. Then, the formed feature vector is used by a detection technique based on Bayes theory when the maximum likelihood criterion is adopted. The main advantage offered by the proposed technique is that if the sampling rate used to acquire the voltagesignal is carefully chosen, then disturbancesin voltagesignals can be detected in a very short interval of time. The performance of the proposed technique is analyzed with synthetic and real data. Also, its performance is compared to the one of similar techniques introduced previously. The numerical results verify that the proposed technique offers satisfactory performance when applied to synthetic and real data. Finally, a comparative analysis with another technique indicates that the proposed presents improved performance.