Classification of Power Quality Disturbances using the Unique Combination of Hilbert Transform, Image Processing and K-Nearest Neighbor

Abstract

This paper introduces the unique combination of Hilbert Transform (HT), Image Processing, and K-Nearest Neighbor (KNN) for classifying the Power Quality Disturbances (PQDs). Power Quality (PQ) is a term that is frequently used these days. Everyone is cautious of the power supply they are purchasing from the utility because the end-user sensitive equipment may malfunction or trip as a result ofPQDs. In order to get a clean and disturbance free power supply, the utility needs to identify the type of disturbance, the cause of the disturbance, and mitigate it. This paper presents a novel approach for classifying the commonly occurring PQDs like sag, swell, and interruption. The proposed algorithm is realized by generating voltage signals pertaining to the PQDs using integral mathematical models, Simulink models, and experimentation. The voltage signals related to different PQDs are processed using HT and the processed signals having elliptical shapes are plotted and converted into images. These images are further processed using the image processing technique in order to turn the RGB image into a grayscale image. The statistical parameters namely mean and standard deviation are calculated from the grayscale image input to the algorithm for feature extraction. The KNN classifier is trained and tested using these extracted features. In the KNN classifier, the minimum Euclidean distance is calculated to identify the class of PQDs with high accuracy.