Power Quality Disturbances Classification Using Dimensionality

• Main Author: Prakash a/l Bala
• Language: English
• Published: 2023
• Subjects: Power Quality; Machine Learning; PCA; KPCA

Evolution of the current modern era demands a huge and good power quality supply day by day. Power utility and power trade service suppliers encounters difficult issues in identifying and sorting the Power Quality Disturbances (PQD). My thesis illustrates the technique of PQD classification by utilizing wavelet decomposition, two methods of dimensional reduction, Principal Component Analysis (PCA) and Kernel Principal Component Analysis (Kernel PCA) with a choice of classifier, the multiclass k-Nearest Neighbors (k-NN). A normal wave without disturbance and waves with PQD events of single-type and hybrid-type were generated using Python with Scikit Learn package using the mathematical model as per the definition and parameters outlined by IEEE 1159 and IEC61000 customary. The generated PQD signals was processed with Discrete Wavelet Transform (DWT), to decompose and acquire it’s illustration in time and frequency domain. In this project work, my database consists of 14000 generated signals of a normal wave and the PQDs, which were divided into 80% of train set and 20% of test set for each PQDs. An k-nearest method for the multiclass classifier with a choice of mother wavelet filter function was worked out with the PQD’s feature vector. Main idea of this research was to enhance the output performance of the classifier after applying dimensional reduction method. By training the classifier to observe and analyze the performance, different tests was carried out. The final result outputs the performance variation between training the k-NN classifier versus training the k-NN classifier in dimension reduction. The k-NN classifier with dimensional reduction succeed to classify the PQDs with higher accuracy in both train and test set.