Hierarchical dynamic wake modeling of wind turbine based on physics-informed generative deep learning

With the increasing demand for electric power, the size of wind farms is becoming much larger than ever before. Power and load prediction are two of the most essential topics in wind farm layout optimization. Traditional wake modeling methods, such as analytic models and CFD simulations, struggle to handle such large-scale problems accurately and efficiently. In this study, a novel hierarchical dynamic wake modeling approach for wind turbines using generative deep learning, PHOENIX (PHysics-infOrmed gEnerative deep learniNg for hIerarchical dynamic wake modeling eXploration), is proposed to capture the spatial–temporal features of the unsteady wake field in wind turbine clusters. The dynamic wake meandering (DWM) model is employed to generate the corresponding datasets for training, testing, and validating the deep learning-based wake prediction framework. This research is expected to accelerate the prediction process and improve accuracy, and it can be further applied to wind turbine design and wind farm layout optimization.