Post-earthquake rapid assessment of interconnected electrical equipment based on hybrid modelling

Abstract. Electrical substations play an indispensable role in the power grid for adjusting power voltage and controlling power flow, ensuring normal power transmission and distribution. However, substation equipment is particularly vulnerable to seismic events due to the use of brittle porcelain materials and highly slender structures, increasing the risk of structural failure and cascading effects. While significant research has focused on seismic performance and innovative damping techniques for different standalone equipment, critical gaps remain, particularly in accounting for the interconnected nature of substation equipment and post-earthquake emergency response. This study addresses these gaps by proposing a hybrid post-earthquake rapid assessment model that utilizes monitored ground motion signals. Hybrid modeling combines physics-based models with data-driven approaches, leveraging the strengths of both to overcome their individual limitations. Specifically, the study integrates Graph Neural Networks (GNN) with mechanistic models of standalone equipment to capture the dynamic behavior and interaction forces of interconnected equipment under seismic loads. The spatial relationships of the interconnected equipment are represented through a graph structure, and temporal dependencies are learned through recurrent computations. Thus, this approach circumvents the need for costly and impractical sensor installations on every piece of equipment, offering a cost-effective and accurate assessment method. The efficacy of the proposed hybrid modelling technique is demonstrated with a case study on combinations of 800 kV post insulators interconnected by busbars. By providing a rapid assessment of each equipment's post-earthquake condition, the proposed model can be applied to inform emergency repair actions and enhance the resilience of power infrastructure.