Modified ResGated Graph Convolutional Network-Based Power Flow Calculation

This is the authors' abstract. We don't add key points for this paper.

With the rapid development of power systems, the stochasticity of renewables, frequent changes of controllable loads, and network failures caused by extreme weather have brought new challenges to power flow calculation. To solve the nonlinear equations of large-scale systems, the traditional power flow calculation methods have difficulties in balancing the requirements of real-time performance and accuracy, and non-convergence cases may occur under extreme conditions. Meanwhile, the existing data-driven based power flow calculation methods also exhibit high dependence on training data. To address these issues, this paper proposes a power flow calculation method based on the modified residual gated graph convolutional network (ResGated GCN). This method integrates the principles of power flow calculation into the entire process of network design, loss function formulation, and model training to enhance the generalization ability of data-driven algorithms. The effectiveness of the proposed method has been systematically verified using the IEEE ll8-bus system.