Simulating tidal turbines with multi-scale mesh optimisation techniques

Embedding tidal turbines within simulations of realistic large-scale tidal flows is a highly multi-scale problem that poses significant computational challenges. Herein this problem is tackled using actuator disc momentum (ADM) theory and Reynolds-averaged Navier-Stokes (RANS) with, for the first time, dynamically adaptive mesh optimisation techniques. This enables the mesh to be refined dynamically in time and only in the locations required, thus making optimal use of limited computational resources. Ambient turbulence intensity has a significant effect on the structure of the turbine wake and its recovery. Therefore, both $k-\omega$ and $k-\omega$ SST RANS models have been implemented within the Fluidity framework in order to account for the effects of turbulence. The model is validated against three sets of experiments and a comparison against a similar OpenFOAM model is also presented to portray the benefits of the finite element discretisation scheme employed in the Fluidity ADM-RANS model. With the aid of the ADM-RANS model, the eddy viscosity value used in depth-averaged models is tuned to improve the wake structures predicted. Thereafter, OpenTidalFarm (OTF) is used to optimise turbine positions in order to maximise the total extracted power from an array. The depth-averaged results are then compared against 3D simulations of the arrays of tidal turbines using the ADM-RANS model, thus allowing for an investigation into the accuracy of the adjoint-based optimisation used in OTF for the first time. Finally, a methodology for embedding the ADM-RANS model within large scale simulations with real tidal forcing and bathymetry is presented. This allows for an accurate prediction of the power output from an array at a realistic tidal site and it can also provide valuable information on how the local and global flows are affected.Open Acces