Description
During the last decades, the wind energy produced in Europe has reached up to 11% and is expected to keep growing, reducing the CO2 yearly produced and the usage of non-renewable energy. One crucial factor for wind energy to reach this relevance is the improvement ofsimulation tools in the early stages of the wind farm design. Simulation has allowed the enhancement of the emplacements of the farms and the prediction of better configuration of the turbines. The mesh is one of the key ingredients of wind farm simulations: it must be adapted to the topography and to the turbine models. Standard simulation pipelines model the wind farm in a preprocessing step, orientating the turbines with an estimation of the flow. In onshore locations, the effect of the topography alters the wind direction, causing a turbine misalignment with the wind, which is commonly ignored when designing the geometry. Even in offshore locations, upwind turbines affect the direction of the wind downstream,and this effect is not known until the simulation is performed.
REWIND's main objective is to develop a modeling and adaptation strategy toenable a simulation framework that can perform a turbine realignment with the actual flow. We aim to impact the accuracy of the simulations and produce a relevant step forward in wind power forecasting: being able to orientate the turbines according to the actual flow. Simulating wind farms with high fidelity is of significant importance not only to accurately compute the production of a wind farm for a given configuration and inflow direction, but also to properly locate the turbines in the early stages of the wind farm design. When simulating wind farms, the computational domains are usually extensive, spanning kilometers, while simultaneously necessitating the accurate representation of local wind dynamics around turbines at the scale of meters. This juxtaposition of expansive domains and fine spatial scales may lead to space discretizations that require a substantial number of nodes. As a result, this can impose significant computational solver demands, especially when utilizing structured meshes.
