Detection and Classification of Single and Multiple Power Quality Disturbances Based on Independent Component Analysis

Detection of power quality disturbances requires some methods to decompose, identify and then classify waveforms. In this paper, a statistical method called Single Channel Independent Component Analysis (SCICA) is used for this means. In this method, source signals are separated from observed signals by using the characteristics of statistical independence as well as the non-Gaussian distribution of disturbance sources. The proposed method for classification of PQ disturbances consists of three stages: a) data generation, b) feature extraction and c) disturbances classification. The advantages of this method are its lower cost (due to the using of only one sensor), high accuracy in extracting independent components of non-Gaussian, non-linear and non-static signals, excellence of estimated signals, classification of multiple power quality disturbances that include more than two disturbances for any combination (ternary and quaternary disturbances) and estimation the starting time, ending time and duration of voltage swell, voltage sag, flicker and oscillatory transients with great accuracy. The results of the simulations on generated and real waveform show capability and high performance of the proposed method.