Recently, the advances in the technology causes the power systems switch to smart grids, ever-increase of the distributed power generation, and expansion of microgrid causes power quality problems to the consumers. Power Quality Disturbances causes serious damages to the electrical appliances and hence their prior detection classification gained significant interest. Towards such prospect, this paper introduced a new method for PQDs detection and classification based on composite features and ensemble learning. At features extraction, each PQD is signified through two different set of features extracted from S-transform and Statistical methods. S-transform reveals time-frequency characteristics and statistical features reveals the feature independency between inter PQDs. Further at classification, two machine learning algorithms namely Support Vector Machine (SVM) and Kernalized Extreme Learning Machine (KELM) are employed. KELM explores the perfect discrimination between PQDs and is formulated as a combination of polynomial and Radial Basis Function Kernels. Extensive simulations of synthetic PQDs shows the effectiveness of proposed method, especially at the mixed PQDs. The proposed method gained only noticeable improvement at single PQDs classification, but it gained significant improvement at mixed PQDs classifications.

power Quality Disturbances, S-transform, Statistical features, Mixed PQDs, KELM, SVM, and Accuracy.