Застосування графових нейронних мереж для аналізу та прогнозування стану енергетичних мереж

Abstract

Modern energy systems are complex infrastructure networks that generate large amounts of operational data. Traditional analysis methods are often insufficient to account for spatial dependencies between network elements. This paper considers an approach using graph neural networks for modelling and analysing energy infrastructure. The power supply system is represented as a graph, where nodes correspond to substations and edges correspond to transmission lines. A graph convolutional neural network is used to account for spatial relationships between network elements and predict system states. The proposed model allows identifying critical nodes and detecting potential overloads in the network. Such approaches contribute to improving the reliability and efficiency of energy infrastructure