Relevant lenght scales for turbulence modelling

Multi-scale Turbulence Modelling for RANS Wake Simulations

Wakes from wind turbines are important in wind farms, since they can account for losses in the total power output of up to 20% and also contribute to fatigue loads. It is both difficult and costly to perform 1:1 scale experiments on such large systems, while down-scaled wind tunnel experiments inevitably have blockage, scaling and other undesired side effects. Instead, computer modelling has traditionally been used as a ”numerical wind tunnel” to investigate wind turbine wakes.

Background

Many approaches to modelling wakes have been considered during the last decades ranging infidelities from simple engineering models to fully-resolved Computational Fluid Dynamics (CFD) simulations. In this project the medium-fidelity Reynolds-Averaged Navier-Stokes (RANS) CFD method implemented in the in-house EllipSys3D CFD-solver will be used to investigate the wake in an ensemble-averaged manner (equivalent to time-averaged for stationary conditions). This type of model relies heavily on the turbulence model used and models like the k-ε-fP -model have previously been used successfully for ideal atmospheric conditions.

This project seeks to extend the ”area of validity or application” of this turbulence model to unstable and stable atmospheric boundary layers (ABL) with inversion caps caused by the geostrophic wind and Coriolis force. These ABL’s define some natural length scales, while complex terrain also have some characteristics length scales.

Objectives
  • Tuning of the current k-ε-fP turbulence model or development of a new model to include the effect of unstable and stable inversion-capped ABL’s. This is achieved by comparison with experiments and LES.
  • Including complex terrain in the developed model.
  • Investigating the performance of the model in wind farm scenarios.
  • Implementation of the developed models in EllipSys3D and the dependent software (PyEllipSys and PyWake RANS).
Impact

Increasing computer power makes medium-fidelity CFD models like RANS more attractive to use for some of the cases that was previously reserved for engineering models. An obstacle for industry use has previously been the steep learning curve of setting up and run-ning simulations, but also here a movement has happened with the newly introduced Python interfaces for EllipSys3D: PyEllipSys and PyWake RANS. It is expected that the models and tools developed in this PhD project will push this development even further and help making RANS a common tool in the industry for calculation of wind turbine wakes and their interaction in wind farms.

Contact

Mads Christian Baungaard
PhD student
DTU Wind Energy