Back to the programme printer.gif Print

Delegates are invited to meet and discuss with the poster presenters in this topic directly after the session 'How does the wind blow behind wind turbines and in wind farms?' taking place on Tuesday, 11 March 2014 at 16:30-18:00. The meet-the-authors will take place in the poster area.

Oxana Agafonova Lappeenranta University of Technology, Finland
Oxana Agafonova (1) F P Aapo Koivuniemi (1) Ashvinkumar Chaudhari (1) Boris Conan (2) Jari Hämäläinen (1)
(1) Lappeenranta University of Technology, Lappeenranta, Finland (2) Orleans University, Orleans, France

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2014 and are published here unedited

Oxana Agafonova was born in St.Petersburg, Russia in 1988, completed her bachelor degree in St. Petersburg State Polytechnical University in 2009, and then came in Lappeenranta, Fionland for double master degree in CFD. Now, she is a doctoral student of LUT, her research is related to wind park simulations.


Limits of WASP modelling in comparison with CFD for wind flow over two-dimensional hills


Simulation of atmospheric flows around a complex topography is important for wind energy applications, such as it helps to locate and construct as well as to control the wind farms. The goal of this work is to evaluate different numerical modelling tools for wind flow prediction over complex terrain. Thus, numerical simulations of wind flow over two-dimensional (2D) hills are carried out here using two different modelling aspects such as the Wind Atlas Analysis and Application Program (WAsP) and Computational Fluid Dynamics (CFD) models.


WAsP has been employed in wind power meteorology and wind power industry for the last 25 years and has become the standard PC software in wind idustry for wind resource estimation and wind turbines location. WAsP solves linearized Navier-Stokes equations and uses very simple roughness, obstacle and stability models so that performing WAsP simulations is very fast. But linear model is not able to predict accurately highly complex flow, especially conditions with flow separation or backflow. That is why we need to examine and identify the limits when the Wind Atlas Analysis and Application Program WAsP gives accurate wind prediction.
Numerical simulation using the Reynolds-Averaged Navier-Stokes (RANS) models has been the traditional technique for mean-wind prediction over complex terrain ([3]; [6]; [7]). Due to inherent flow variability and large-scale unsteadiness especially over complex terrains, it can be difficult to be predicted by the RANS model, as it is based on the time-averaged equations. Therefore, Large Eddy Simulation (LES) is expected to be more suitable for this kind of flow simulation although it is computationally more expensive than the RANS approach.
Thus, in this paper, three WAsP, RANS and LES computations are carried out to reproduce the turbulent flows over the 2D hills, and the results are compared with the wind tunnel measurement of the RUSHIL experiment by Khurshudyan et al. (1981).
The generic hill geometries used in this study are the same as those used in the RUSHIL experiment. The hill height H is fixed to 0.117 m in both cases but the hill half-length a varies from 3H to 5H. The corresponding average slope n=H/a of the hills are 0.33 and 0.2, and they are named here as Hill3 and Hill5, respectively. However, series of hills represent more complex terrains than the isolated hill, further simulations have been also carried out for series of three hills using RANS and WAsP models.

Main body of abstract

Numerical methods and Results.
The finite volume meshes for RANS simulations consist of 300800 and 354000 quadrilaterals type of structured cells in case of isolated hill and three periodic hills, respectively. The SST k-w turbulence model with so-called “Low Reynolds Corrections” for near-wall treatment has been employed for the RANS simulations (see in [2]). The inflow boundary conditions for the velocity and turbulent quantities have been taken from the flat terrain simulations. The RANS simulations have been done for steady state condition.
Apart from RANS simulations, some LES calculations are also carried out for both Hill3 and Hill5 hills. For this, the 2D hill geometry is extruded in the span-wise direction in order to use 3D geometry for LES. In LES, larger eddies are always grid-filtered and resolved directly, but the smaller eddies are modelled. The dynamic Smagorinky sub-grid scale model is used to model the smaller eddies. The time dependent turbulent inflow profile is generated by adding the artificial turbulence to the logarithmic law. The LES computations for both hills are run until t = 40 s with all quantities time averaged over the last 30 s, which is approximately 20 advection times along the whole domain. In addition to time averaging, the results are also space averaged over span-wise direction. In this study, the finite-volume method based commercial software ANSYS FLUENT 13.0 is employed for both LES and RANS calculations.
WAsP maps that describe similar hills or series of round hills are created for comparison with CFD. WAsP simulations are performed at real scale. The height of the hills is taken as 117m and computational domain is extended up to 1km. Surface roughness is similar to RANS modelling. The flow model parameters were configured in effort to model a neutrally stratified situation. The actual flow was examined with reference sites along the streamwise axis of the hill.
In this work, RANS results for Hill3 and Hill5 are compared with LES and experimental (Khurshudyan et al., 1980) results. Figs. 1 (a, b) compare the simulated mean velocity with the measurements. Hill3 has a massive flow separation in the lee side of the hill, which is well predicted by both RANS and LES. The reattachment point is predicted at 5.75H in both RANS and LES, which is somewhat upstream than the measured value 6.5H.

(a) Hill3, n=0.33

(b) Hill5, n=0.2
Fig. 1: Vertical profiles of mean stream-wise velocity (U/Uinf) compared with measurements for flow over (a) Hill3 and (b) Hill5 at some longitudinal locations.
Both CFD simulations predict the mean flow properties reasonably good, however the prediction of the turbulence property (see Fig. 2) is not as good as that of mean velocity.

(a) Hill3, n =0.33

(b) Hill5, n=0.2
Fig. 2: Vertical profiles of stream-wise turbulence intensity (u’/Uinf) compared with measurements for flow over (a) Hill3 and (b) Hill5 at some longitudinal locations.
According to Fig. 3, LES overestimates the Reynolds shear stress in the separated region. Similar results have been reported in [6] and [1].

(a) Hill3, n =0.33

(b) Hill5, n =0.2
Fig. 3: Vertical profiles of Reynolds shear stress (-/U2inf) compared with measurements for flow over (a) Hill3 and (b) Hill5 at some longitudinal locations.
As mentioned earlier, WAsP approach has been also utilized here in order to evaluate its capability toward wind flow prediction over complex terrain. From the Fig. 4, it can be noticed that for the stepper hill shape WAsP does not predict the flow separation at all, although Hill3 has strong reverse flow in the lee side. Outside the reverse flow region, the agreement with RANS results is satisfactory. Also, series of three hills (Hill5) with different spaces (sp=0, a) between the hills have been simulated and compared with RANS (in Fig. 5). In this case, WAsP rather well represents flow behaviour over hills in comparison with RANS and differs in a quite small range especially near ground surface. Further results concerning the turbulent properties will be shown later in the conference.

Fig. 4: Vertical profiles of mean stream-wise velocity (U) compared with RANS for flow over Hill3 at some longitudinal locations.

(a) series of three Hill5, sp = 0

(b) series of three Hill5, sp = a
Fig. 5: Vertical profiles of mean stream-wise velocity (U) compared with RANS for flow over Hill5 at some longitudinal locations.


In current study, we carried out RANS, LES and WAsP to investigate the turbulent boundary layer flows over isolated two dimensional hills with different slopes and series of three hills with the fixed slope. That was shown, our CFD predictions underestimate the length of recirculation in Hill3. On the other hand, our reattachment-length prediction is closer to the measurements than those of reported in [3] and [1]. LES prediction is better in Hill5 case than that of Hill3. LES predicts small flow separation on the downwind slope of Hill5; whereas flow is found to be attached in the measurements. Also, LES overestimates the turbulence intensity in the lee side of both hills. RANS also overestimates the turbulence intensity in the separated region of Hill3. But for Hill5, it underestimates the measured values behind the hill. Measurements show almost similar profiles for shear stresses in both cases in spite of different flow behaviour. In [7] was suggested that hot-wire anemometry is not convenient instrument for measuring flow in the region having the reverse flow and high turbulence intensity. The current RANS agrees well enough with the experiment, LES and other similar numerical study, and might be used for later validation of WAsP results.
WAsP results were in line with the expectations from previous works, especially the ones explored in [8]. WAsP seems to predict wind speeds with reasonable accuracy when slopes of 20 % or less, after which the results diverge. Introducing multiple hills in series does not seem to have a meaningful effect.
However, WAsP produces reasonably realistic results for flow over low-pitched hills which are commonly found in nature. This research work proves that WAsP results might be used as an inflow for performing later careful and time-consuming CFD over rather long and high complex terrain with slopping hills, forest and turbines.
Also, the performing cases with Hill2 and Hill4 and comparison with the experiment conducted in the VKI-L2 wind tunnel of the von Karman Institute for Fluid Dynamics using the PIV methodology are planned in near future ([4]).

Learning objectives
The main idea of this paper is showing when it is possible to use WAsP wind prediction for time consuming CFD simulations. Also, thee different numerical methods to observe wind flow over two-dimensional hills are described and compared with the experiment.

[1] Allen, T., Brown, A., 2002. Large-eddy simulation of turbulent separated flow over rough hills. Boundary-Layer Meteorol. 102, 177-198.
[2] ANSYS FLUENT, 2011. ANSYS FLUENT Theory guide.
[3] Castro I.P., Apsley D.D., 1997. Flow and dispersion over topography: a comparison between numerical and laboratory data for two-dimensional flows. Atmospheric Environment, 31, 839-850.
[4] Conan. B., 2012. Wind Resource Assessment in Complex Terrain by Wind Tunnel Modelling.
[5] Khurshudyan L. H., Snyder W. H., and Nekrasov L. V., 1981. Flow and Dispersion of Pollutants Over Two-Dimensional Hills. United States Environmental Protection Agency Report EPA-600/4-8 I-067.
[6] Loureiro, J. B., Alho, A. T., Freire, A. P. S., 2008. The numerical computation of near-wall turbulent flow over a steep hill. J. Wind Eng. Ind. Aerodyn. 96, 540 - 561.
[7] Ying, R., Canuto, V., 1997. Numerical simulation of flow over two-dimensional hills using a second-order turbulence closure model. Boundary-Layer Meteorol. 85, 447 - 474.
[8] Bowen A. J., Mortensen N. G., 1996. Exploring the limits of WAsP. European union wind energy conference, Göteborg, 20-24.5.1996