On the Usage of PANS for the CFD Simulation of Wind Turbines

Abstract

With the rapid evolution of o↵shore wind energy, engineering tools are crucial

to catalyze technological developments and increase their maturity, therefore leading to lower
costs. Complex turbine-turbine interactions require a good knowledge of the physics of the flow
on, around and down/upstream of each turbine, which can be provided using high-fidelity CFD
simulations. Turbulence models play a critical role on this matter and an adequate balance
between accuracy and computational e↵ort is necessary. While RANS approaches are quite
e cient, LES should provide the most accurate result. Yet, even nowadays, LES blade-resolved
simulations are still computationally prohibitive for industrial purposes. A middle-ground exists
in SRS formulations, such as hybrid ones as DDES, or bridging ones such as PANS. In the
present work emphasis is placed on PANS, since numerical and modelling errors can be studied
and quantified independently, as opposite to other SRS approaches. Using as a benchmark the
UNAFLOWwindturbine, it is found that traditional RANS and DDES turbulence formulations
are able to predict integral forces, but partially fail in capturing wake mixing. Nevertheless,
PANS, while enabling the user to select the ratio of turbulent quantities modelled, is not able to
properly capture the integral forces due to premature separation in the blades. Several causes are
discussed, including insu cient mesh refinement in the near-wall region and lack of turbulent
content of the numerical inlet, preventing laminar to turbulent flow transition. Future work
should focus on inlet synthetic turbulence generation, in line with existent literature, in order
to improve the shortcomings faced in properly resolving the near-wall flow.