ARCHIVES

As smart grids become more important for reliable and efficient power delivery, the need for strong security measures to prevent cyber and physical attacks has increased. Phasor measurement units (PMUs) with their high-resolution time-synchronized measurements provide valuable opportunities for real-time monitoring and anomaly detection. This project proposes a machine learning-based framework for increasing smart grid security by using PMU data to detect malicious attacks such as false data injection (FDI), replay, and stealth attacks. The proposed methodology involves extensive preprocessing: data cleaning, normalization and feature extraction from voltage, current, frequency deviation, power factors and phases angles differences. Exploratory data analysis (EDA) is performed to look for important features to characterize normal operation vs. attack. Multiple classification algorithms including linear discriminant analysis (LDA), k-nearest neighbor (KNN), gradient boosting, boosting class (AdaBoost) and linear regression for trend estimation are trained and compared in terms of accuracy, precision, recall, F1-score and areas under the receiver operator characteristic (ROC-AUC) evaluation.