Enhancing Smart Grid Security: A Data-Driven Anomaly Detection Framework

Abstract

The rapid evolution of smart grid systems has significantly improved the efficiency, reliability, and automation of modern power infrastructure; however, it has also introduced critical cybersecurity challenges due to increased connectivity and data exchange. This project presents a data-driven anomaly detection framework designed to enhance smart grid security by identifying malicious activities and abnormal patterns in real-time operational data. The proposed system leverages a combination of machine learning and deep learning techniques, including Random Forest, Naive Bayes, Decision Tree, and Long Short-Term Memory (LSTM) networks, to effectively classify normal and anomalous grid behavior. The dataset is preprocessed using label encoding and feature scaling to ensure optimal model performance, and missing values are handled to maintain data integrity. The framework evaluates multiple algorithms using performance metrics such as accuracy, precision, recall, and F1-score, enabling a comprehensive comparison of model effectiveness. Additionally, visualization techniques such as confusion matrices and comparative graphs are used to interpret results and support decision-making. A user-friendly web interface built using Django allows users to upload datasets, train models, and perform real-time anomaly prediction. Experimental results demonstrate that the proposed approach can accurately detect anomalies, thereby strengthening the resilience and security of smart grid systems against potential cyber threats and operational failures.