Classification of complex power quality disturbances under noisy environment using Root-Music and least square techniques

Abstract

In this paper, a hybrid classification approach for single and complex power quality disturbances is proposed. This approach combines the Root-Music and least square techniques. Firstly, the frequencies are extracted by the Root-Music technique, including those of harmonic components. Afterward, the amplitude components are extracted using the least square technique. Once frequency and amplitude components are extracted, the waveform distortion of voltage is analyzed using the known criteria of total harmonic distortion. The value of this criterion allows for pre-classifying the disturbance into one of two classes. Then, the value of the extracted amplitudes of signals allows the identification of the corresponding disturbance. The classifier performances are validated by several tests using real power and synthetic data. The achieved results prove the effectiveness of the classifier compared to recent state-of-the-art techniques, especially in highly noisy environments. These results show that the proposed classifier leads to the highest classification accuracy of 100% at 20 dB, compared to those of other classifiers. Furthermore, this classifier demonstrates high classification accuracy even under challenging conditions with a high noise level of 10 dB.