The paper presents WindSeer, a neural network trained with synthetic data from computational fluid dynamics simulations. It predicts low-altitude wind in real-time on limited-compute devices using sparse measurement data. The model can predict real wind fields over terrain with known topography from just a few noisy and spatially clustered wind measurements. It generates accurate predictions at different resolutions and domain sizes on previously unseen topography without retraining. The model successfully predicts historical wind data collected by weather stations and wind measured onboard drones.
Publication date: 19 Jan 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2401.09944
