Implementing Real-Time Edge AI for Anomaly Detection in Smart Grids: A Pilot Study on Power Distribution Networks

In this pilot study, we aim to establish the changes needed to investigate whether edge AI can lead to real-time anomaly detection in the grid. In designing anomaly detection techniques, numerous other AI techniques such as artificial neural networks, among many others, have been investigated in previous works. However, only a few investigate their use for real-time anomaly detection in power systems. This pilot research looks into the possibility of developing a real-time anomaly detection methodology for smart grids. A part of this is casting the anomaly detection algorithm in a way it can be deployed on the edge. This section aims to review a literature survey that contains all the methods and algorithms used in the anomaly detection process. The section starts by addressing the need for the intervention of anomaly detection systems to mitigate the risk of attacks. Then the survey presents the well-established methods of anomaly detection. The last part of this section will review the previous attempt of moving anomaly detection to the edge of networks. In-depth details of each algorithm will be presented in the next section.