Structural surrogate modelling of a floating offshore wind turbine with physics-guided spatial-temporal graph neural network

To address the problems in the current offshore wind industry and the limitation of existing simulation approach, a novel finite element surrogate model of FOWT for DT based on Physics-Guided Graph Neural Network (GNN) is presented. The aero-hydro-servo coupled simulation results are generated from the software QBlade. The semi-submersible OC4 DeepCwind FOWT is considered under the above rated wind speed, rough sea state and water current condition according to metocean of West of Barra, Scotland. In addition, we present the first application of spatial-temporal GNN approach for solving offshore structural dynamics in which this new model can solve the real-time prediction of wind turbine tower and mooring forces under complex wind, wave and current condition for FOWT. Internal forces prediction can allow the remaining useful life calculation in the next stage of fatigue analysis. More importantly, the expressiveness of GNN can provide the interpretability of this new proposed model and highlight the explainability of wind-wave-current dynamics coupling on FOWT, which can overcome the “black box” limitation in traditional artificial intelligence deep learning.