Variance Based Offline Power Disturbance Signal Classification Using Support Vector Machine and Random Kitchen Sink

In this paper, five classes of different power quality disturbances such as swell, sag, harmonics, sag with harmonics and swell with harmonics are synthesized using MATLAB/SIMULINK software which is further decomposed into 8 intrinsic mode functions using variational mode decomposition (VMD). VMD is an adaptive signal processing method that decomposes the signal in to several intrinsic mode functions (IMF) or components. The variance calculated from each of the mode is taken as feature representation. It is found that sines and cosines of variance vector of eight different IMF candidates of a signal acts as feature vector that can accurately extract salient and unique nature of the power disturbances. The classification is performed using Support Vector Machines (SVM) and Random Kitchen Sink (RKS) algorithm. The classification results in a highest accuracy of 94.44% for RKS method when compared to SVM.