Graph Neural Network- Enhanced Power Flow Adjustment

Abstract

Modern power grids are facing unprecedented operational complexity due to the surge in distributed energy resources (DERs), intermittent renewables, and electric vehicle (EV) charging demands. While traditional methods like Newton Raphson are computationally precise, their iterative nature often fails to meet the sub-second latency requirements of dynamic smart grids. This research proposes a Graph Neural Network (GNN) framework designed to model electrical networks as high dimensional graphs. By capturing the inherent topological relationships among buses (nodes) and transmission lines (edges), the GNN-based approach provides a scalable, data-driven alternative for real-time power-flow estimation. The framework effectively processes non-linear grid behaviours and uncertainties, ensuring stable and efficient grid management, congestion control, and optimal power dispatch in rapidly evolving electrical environments.