This research presents an innovative approach for the detection and classification of PQD (Power Quality Disturbances) using unsupervised clustering methods, specifically DBSCAN (Density-Based Spatial Clustering of Applications with Noise) and K-Means. The method integrates machine learning models for predicting RMS (Root Mean Square) values and classifying power quality disturbances into distinct categories based on their features. The data is preprocessed using scaling techniques, followed by regression models (Random Forest and XGBoost) for RMS prediction. The classification task is handled using Random Forest and XGBoost models, categorizing PQD events into "Good" or "Poor" labels. Additionally, DBSCAN and K-Means are employed for clustering the data, where DBSCAN's ability to detect noise and identify non-linear clusters and K-Means's effectiveness in partitioning the data into predefined groups offer complementary benefits. The performance of these models is evaluated using standard metrics such as R², RMSE, and accuracy. The research demonstrates high accuracy and precision in PQD classification, with near-perfect regression results and strong clustering performance. This approach offers a promising solution for real-time power quality monitoring, contributing to the development of intelligent grid systems. The study advances machine learning applications in power systems monitoring, fault detection, and anomaly classification.

Keywords: PQD, Unsupervised Clustering, DBSCAN, K-Means, Power Quality Disturbances, RMS Prediction, Machine Learning, Random Forest, XGBoost, Clustering, Real-Time Monitoring.