Enhanced Locational FDIA Detection in Smart Grids: A Scalable Distributed Framework

Locational detection of the false data injection attack (FDIA) is essential for smart grid cyber-security. However, the FDIA detection techniques often falter in scalability as power network complexity increases. To address the research gap, this paper introduces an innovative distributed framework for locational FDIA detection that optimizes both performance and scalability. The proposed framework initially partitions the power grid using the improved Louvain community detection algorithm. The proposed solution utilizes the Electrical Functional Strength (EFS) matrix and power supply modularity. Subsequently, a dedicated multi-label one-dimensional convolutional neural network model (1D CNN) locational detector is designed for each derived cluster. The proposed methodology is designed to increase detection accuracy and enhance the scalability of the model. This is achieved by reducing training and detection times, as well as lowering memory requirements, compared to traditional centralized approaches. The effectiveness of the proposed framework is validated through simulations on the IEEE 39-bus system. These simulations demonstrate the framework’s capability to enhance detection accuracy by simplifying the locational FDIA detection challenge, achieved through strategic grid partitioning.