LAUSR

Abstract This paper presents wavelet transform, independent component based detection of the power quality disturbances in a microgrid. Wavelet transform (WT) is well known for its better time-frequency multi-resolution analysis because of selection of adaptive window size depending inversely with the frequency. This makes it suitable for identification of both stationary and non-stationary characteristics of the power quality disturbances. But, it is analyzed in the study that the detection capability of wavelet is degraded when the input voltage signal contains noise and completely fails when the noise level is increased to 25 dB. Of course, a wavelet-denoising technique is tested to improve the ability of wavelet transform under noisy conditions, but observed to increase the detection time. Hence, independent component analysis (ICA) is proposed for detection based on its property of finding the significant components which are statistically independent and non-Gaussian. In addition, the statistical features are extracted to formulate the input dataset for support vector machine (SVM) to classify different power quality disturbances like sag, swell, interruptions, harmonics, transients, and flickering. A comparative analysis is presented and it is observed from the study that ICA-SVM hybrid combination provides better classification accuracy as comparison to WT-SVM under different operating scenarios.