Predicting the impact of a new transmission line on grid synchronization: a fuzzy-enhanced graph isomorphism network approach

Abstract As the integration of renewable energy into the power grid increases, enhancing grid capacity and optimizing network topology have become critical strategies to facilitate the energy transition and maintain stable grid operation. However, the addition of a new transmission line to the existing topology significantly affects grid frequency synchronization. This effect results in three distinct outcomes: improved, unchanged, or degraded synchronization, where degradation represents a manifestation of Braess’s paradox. Motivated by this challenge, we propose a prediction approach utilizing graph isomorphism networks (GIN) to evaluate in advance the impact of a newly added transmission line on power grid synchronization. This method uses node degree, closeness centrality, resistance centrality, and power values as graph node features, while edge features are constructed through the nonlinear concatenation of these node features. Moreover, we propose an improvement strategy that incorporates fuzzy layers into the hidden layers to enhance feature representation capability and mitigate overfitting. Experimental results on the IEEE 39, 57, and 118 test systems demonstrate that the proposed GIN+Fuzzy model achieves Macro-F1 scores of 96.51%, 60.28% and 83.52%, respectively. Notably, these results represent improvements of 1.03, 1.68 and 1.18 percentage points over the baseline GIN model, consistently outperforming other mainstream baseline models.