BibTeX
@ARTICLE{
         CTP2025Api,
       author = "Gafoor CTP, Azhar and Kumar Boya, Sumanth and Jinka, Rishi and Gupta, Abhineet and
         Tyagi, Ankit and Sarkar, Suranjan and Subramani, Deepak N.",
       title = "A physics-informed neural network for turbulent wake simulations behind wind
         turbines",
       journal = "Physics of Fluids",
       volume = "37",
       number = "1",
       pages = "015110",
       year = "2025",
       month = "01",
       abstract = "Fast simulations of wind turbine wakes are crucial during the design phase of
         optimal wind farm layouts. Wind turbine wakes affect the performance of downstream turbines.
         Physics-informed neural networks (PINNs), a deep learning approach to simulate dynamical systems
         governed by partial differential equations, are gaining traction in computational fluid dynamics due
         to their fast inference capability. We developed a PINN model using the 2-equation
         $k-\varepsilon$ model and the actuator disk method to simulate the wakes behind the wind
         turbines. Crucially, training of the developed PINN model does not rely on high-fidelity simulation
         data, thus reducing the end-to-end training time by saving simulation data generation time. We
         tested the model against traditional solvers and field data to simulate the turbulent wake behind
         the HOLEC WPS 30/3 Wind Turbine from Sexbierum and a three-blade 630-kW Nibe-B wind turbine.
         Detailed computational studies are completed to establish convergence properties with increasing
         sampling collocation points and the number of graphical processing units. A transfer learning
         strategy is introduced to accelerate training for new scenarios resulting in a 5× speedup. Our
         results establish the efficacy of the PINN model in simulating turbulent flows. Compared to field
         data, our PINN model and traditional Reynolds-averaged Navier–Stokes (RANS) numerical solvers,
         such as the shear stress transport $k - \omega$ and Reynolds stress model have similar errors,
         suggesting its utility as a replacement to these RANS solvers. The model architecture, trained
         weights, and code are available in
         https://github.com/quest-lab-iisc/PINN\_WakeTurbulenceModel.",
       issn = "1070-6631",
       doi = "10.1063/5.0245113",
       url = "https://doi.org/10.1063/5.0245113",
       eprint =
         "https://pubs.aip.org/aip/pof/article-pdf/doi/10.1063/5.0245113/20328050/015110_1_5.0245113.pdf",
       ad_area = "Computational Fluid Dynamics",
       ad_tools = "TensorFlow"
}