Three-Phase Inverter Dynamics Predictions Using GNN-based Regression Model
Ahmed K. Khamis, Mohammed S. Agamy · 2023
Modern power grids include inverters as a part of the network, and their dynamic behavior is crucial for dependable and sustainable electric systems. Due to its complexity, variability, topology and control variants, it is difficult to forecast systems’ behavior with high degree of accuracy. Utilizing all the circuit to ML framework proposed in [1]–[3] including circuit netlist to graph, feature assignment, dataset generation and GNN model construction, this paper proposes an accurate predictions of the three-phase inverter dynamics by utilizing a regression model based on recently published Graph Neural Network (GNN). Graph convolution networks (GCN) are utilized as subblocks in building such model, resulting in an R2score of 99.56% recorded via experiments and assessments presented in this work.