Simulation of Wind over complex terrain
comparison of CFD codes
Rémi Gandoin, DONG Energy
European Wind Energy Conference & Exhibition 2010, Tuesday 20 - Friday 23 April 2010, Warsaw, Poland
normalized speed-up at 80m agl for (left to right) the flat, moderately and steep hills as a
function of the distance from the hill center
turbulence intensity levels at 80m agl for (left to right) the flat, and moderately steep hills as
a function of the distance from the hill center
The objective of this study is to estimate the uncertainties of the CFD codes
EllipSys3D (Risø-DTU) and MeteodynWT (Meteodyn) in complex terrain.
A mountainous site in Norway has been selected for this purpose. A
comparison with the linear model WAsP has been also conducted.
The models results have been compared with results using standardized site
calibration procedures based on measurements.
It was decided at ﬁrst to compare both codes, EllipSys3D and MeteodynWT,
on ﬂat terrain in order to discuss the meshing process and the boundary
Secondly, WAsP and CFD simulations were run on 2D hills, whose
topography was extracted from a real terrain, in order to quantify the
uncertainties between the models.
Finally, WAsP and CFD codes were run on the real terrain, results were
compared between each other, and with met mast measurements.
This project has identiﬁed two main differences between MeteodynWT and
EllipSys3D, namely the mesh resolution and the turbulence models.
The CFD codes were found to capture non linearities, showing a different
behaviour than WAsP, and identifying severe wind conditions. They follow
similar trends in simple and moderately steep terrain. However, EllipSys3D
did predict recirculation after a very steep theoretical proﬁle, while
MeteodynWT did not, presumably because of the limitation in the mesh
accuracy due to the necessary limitations of the 32 bits version of the
At hub height on real terrain, discrepancies were found to be limited, and
within the measurement uncertainties.
To conclude, ﬂat terrain and 2D hill simulations were found to be useful to
explain the real terrain simulations. Even if MeteodynWT on a32bits machine is
a more pre-conﬁgured code, and contains limitations, it showed comparable
results with EllipSys3D.
Three topics require further investigations: only one site and two wind directions
were simulated, steep theoretical proﬁles led to discrepancies between the
codes, measurements were not conducted especially for this project.
Finally the two codes agree fairly well on the identiﬁcation of unsuitable
Flat Mid Steep
180 degrees slices
wind data analysis
relative speed difference [%] and
turbulence intensity [%] for each
mast and wind direction
maximum speed-up at the top of the
hill as a function of the slope, in