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

The integration of Information and Communication Technologies (ICT) into traditional power grids has led to the evolution of smart grids, revolutionizing energy management. However, detecting anomalies within these systems remains challenging due to the complexity of potential events, ranging from cyberattacks to infrastructure faults and equipment malfunctions, compounded by the scarcity of labeled data. Addressing these challenges, this study presents a statistical data-driven framework for explainable anomaly detection in smart grid systems. The framework employs a Gaussian Mixture Model (GMM) to identify anomalous events without reliance on labeled data, followed by machine learning techniques to classify these anomalies into natural events or cyberattacks. Additionally, we utilize SHapley Additive exPlanations (SHAP) to explain the machine learning model’s outputs, thereby enhancing the system’s interpretability and explainability. Experimental results demonstrate the framework’s efficacy, achieving 91% accuracy in anomaly detection and 90% in event classification. This approach not only enhances robustness and transparency in anomaly detection but also holds significant applicability for consumer electronics and cyberphysical systems.