Comparison of Classifiers for the Recognition of 3-Phase Power Quality Disturbance Events

This paper aims to classify the three-phase (3-\(\phi\)) power Quality (PQ) disturbance events using a set of conventional classifiers. Several 3-\(\phi\) PQ disturbance events such as Transformer Energizing (TE), Line Energizing (LE), fault and non-fault interruption, are synthetically generated for comparing the chosen classifiers in terms of recognition accuracy and computational complexity. It initially explores the multi-resolution ability of the Wavelet Transform (WT) in extracting a suitable feature set of the chosen events. Further, the K-means apriori Feature Selection Algorithm (KAFS) has been applied to the extracted wavelet coefficients to select a reduced feature set for simulating the classifiers. The performance of the classifiers is evaluated in presence of the WT statistical parameters and the extracted WT-KAFS coefficients in determining the recognition accuracy of the chosen PQ disturbance events. The application of KAFS algorithm has indeed enhanced the recognition accuracy as compared to the conventional WT-statistical parameters as revealed from our results. Among the chosen classifiers, the performance of the Random Forest (RF) remains high, while the Discriminant Analyzer (DA) has shown to be the least performer.