In Uence of Orography On Wind Resource Assessment Programs: January 2007
In Uence of Orography On Wind Resource Assessment Programs: January 2007
In Uence of Orography On Wind Resource Assessment Programs: January 2007
net/publication/254420469
CITATIONS READS
9 60
5 authors, including:
Some of the authors of this publication are also working on these related projects:
GABA is the principal fast-acting excitatory transmitter in the neonatal brain. View project
All content following this page was uploaded by Ana Talayero on 30 April 2019.
Abstract. In order to provide optimal sitting of wind RIX coefficient value is too high, recirculation processes
turbines, a reliable estimate of the wind resource over a given (formation of vortexes) could appear. This comportment is not
area is required. This paper compares the performance of two taken into account by the 2D linear equations which govern
models, WAsP and WindSim predicting the power production WAsP calculations and consequently would justify the use of a
of individual turbines. Wind models are potentially important at CFD model to simulate wind flow.
complex terrain sites in the wind resource assessment since
measurements can only be afforded at selected positions, and This analysis gives a general picture of how well the two
the wind variations may be large over short distances. The models are able to predict wind speed and direction, trying to
purpose of this paper is to compare the accuracy with regard to focus on the causes of the generated differences and
wind resource assessment of the commonly used WASP consequently the eventual necessity of the CFD solver in some
package with the WindSim package which uses a more situations.
complete Navier-Stokes solver based on the PHOENICS code
which should give more accurate results in complex terrain. Due to the constant increscent of wind turbines size and
therefore wind farm dimensions, it is absolutely necessary be
able to predict the error committed by the wind assessment
Key words software extrapolating wind speed in both horizontal and
vertical directions. These analysis will consequently focuses
WAsP, WindSim, RIX coefficient, wind shear, cross checking. first on the way that each software model the wind shear at
three determined sites of the three considered terrains. Then,
1. Introduction considering common series of wind data, using cross-checking
methodology [5], comparing the wind speed obtained at one
met mast site from the data registered at another met mast, it
One of the challenges to effective sitting of wind turbines is the has been evaluated how each model would perform the
ability to make reliable and accurate predictions of the wind horizontal extrapolation of wind data and then its reliability to
power resource. Wind models are potentially important at model wind behavior in each determined terrain configuration.
complex terrain sites in the wind resource assessment since
measurements can only be afforded at selected positions, and Some overall conclusions are then drawn as to the relative
the wind variations may be large over short distances. For accuracy of the two models and what factors affect the degree
micro-scale flow (spatial scales of 1 m to 5 km), the WAsP of accuracy.
model [1] is most commonly used in the wind resource
analysis, but in areas with flow separation the model is not very
well suited for resource assessment. Simple terrain corrections 2. The wind resource evaluation methods
(in function of the so called RIX-coefficient ) have been applied
with success to the WAsP simulation for quite complicated Perhaps the first serious software package that has been
terrain. developed in order to assess wind resource was the WAsP
However, there is increasing interest in applying more complete program . This is generally considered to be the reference wind
Navier Stokes models to flow in complex terrain, given that the resource program at the present time.
required computing power is now more affordable and can
produce results in a reasonable time-frame with an acceptable The program is based on similar boundary layer physics to the
spatial resolution. MS-Micro package using the 3D extension by Mason and
The purpose of this paper is to compare the accuracy with Sykes of the original linear 2D equations by Jackson and Hunt
regard to wind resource assessment of the commonly used for wind flow over low hills [6]. In addition, the WAsP model
WAsP package with the WindSim package [2] which uses a includes a roughness change and shelter model to cater for the
more complete Navier Stokes solver based on the PHOENICS effect of obstacles. The model employs a zooming polar
code which should give more accurate results in complex coordinate grid for optimum resolution at the point of interest.
terrain.
Recent research has been carried out using more complete
To study the limitations of the different software packages have Navier-Stokes solvers. One of the first commercial programs to
in modelling the wind comportment at a site, several European use such a solver is the WindSim model developed by Vector
wind farms sites in Southern-Europe situated presenting AS. This package uses the computational fluid dynamics (CFD)
varying degrees of orography have been studied. In order to model PHOENICS as it core solver. Potentially, the program is
evaluate the real necessity of more complex model, such as capable of more accurate results over steep hills than the WAsP
Computational Fluid Dynamics (CFD) [3], it is necessary to model.
identify zones where the WAsP model could commit important
errors. The RIX coefficient [4], based on a steepness ratio can Both software allows the conversion of time history wind
be used to evaluate the accuracy of an estimation. When the measurements (speed and direction). However, they do not
integrate any temporal verification, considering data as annual The wind rose obtained at met mast A1 and A2 presents two
complete cycles. Thus, it is necessary to analyse the wind data main E-W wind direction as it is shown in the following
measured, clean them and select the study data series. Beside, figure 2.
one of the main characteristics of wind is its variability annual,
diurnal and situational. That means that its mean velocity could
change a lot from one year to another and consequently the
wind potential of a determined site. In order to well evaluate an
elected zone wind potential, it is necessary to get a large set of
annually wind data [7]. Scale
0.56
The first wind farm site studied is located in a relatively flat 0.52
terrain. As can be seen on the following RIX map, met mast A1 0.48
0.44
and A2 are situated in a very flat area, presenting a RIX 0.4
0.28
roughness areas, such as villages, different landscapes, not 0.24
taken into account. This assumption might affect the results 0.2
0.12
height (Higher Height HH: 10m). 0.08
0.48
(close to Raleigh distribution) with few calms. As it is shown in
0.44 following figure 4, the Weibull-fit is very good, presenting a
0.4 high value of shape parameter K.
0.36
0.32
Wind rose obtained at site B presents a dominance of winds
0.28 coming from the NW but not defined as clear as site A. Indeed,
0.24 the wind frequency is generally distributed all over direction-
0.2
binned sectors from W to NNW. This characteristic will be
0.16
0.12
taken into account in the further analysis of wind shear
0.08
The annual average wind speed of this site at met mast A1 at its
higher height (40 m) is 6.6 m/s. However, the high number of Scale
calms cumulated with a low dispersion of high wind speeds
would generate a bad adjustment of the Weibull distribution fit,
which would introduce an increscent of uncertainties. B1 Wind rose B1 Weibull
4. Micro-scale model set-up
4.1 WAsP
3.3 Site C Table 1: Set-up of WAsP simulations for sites A,B and C.
Site_A Site_B Site_C
Site C presents the most complex terrain of the three considered Size of map (Xkm x Ykm) 38.4 x 28.2 4x7 12.2 x 9
Both met masts C1 and C2 are located on the top of north-south Roughness length (m) 0.05 0.03 0.03
high ridge presenting strong slopes and consequently
considerable speed-up for both westerly and easterly flow. As As roughness was not considered as the influent parameter to
shown in figure 5, RIX coefficient can reach values close to study in details in this project, a fixed roughness height was set
60 % in the previously mentioned wind directions. Then, up for each one of the three terrains.
important deflections can be assumed between both met mast However, it is necessary to keep in mind that in the case of
locations. Site A, presenting a very flat terrain with little forest and
villages, little changes in roughness might generate a strong
RIX Coeff. scale affection to the wind speed estimation a low height, such as
Mast A2.
0.6
0.56
0.52
4.2 Windsim
0.48
0.44
Windsim version 4.6.1 was employed in this study. The details
0.4
0.36
of the models set-up for the three wind farms sites considered
0.32 are shown in the following Table 2.
0.28
0.24
0.2
Table 2: Set-up of Windsim simulations for sites A,B and C.
Site_A Site_B Site_C
0.16
0.12
Size of model domain (km2) 15 · 10 4·7 9· 9
C2 Wind rose C2 Weibull To perform the wind shear comparison study, only met mast
presenting various levels of measurements, in order to be able
Figure 6: Site C; Wind rose and Weibull distribution to estimate the vertical evolution of wind speed from theoretical
measured values. Met mast elected for the wind shear study
were the highest met masts with several measurement levels
available for each one of the three terrain configuration.
The following process was realized in each one of the two
model studied to estimate the wind shear at mast A1, B and C1.
Various wind turbines were modeled at meteorological mast in function of the wind direction. Therefore, it has been chosen
location, with different hub height, in order to estimate the to study the vertical evolution of wind for the predominant
vertical evolution of wind direction and speed. direction-binned sectors.
In order to perform a suitable comparison until heights up to Site A1: The following figure 8 presents the wind shear
100 meters, it has been chosen to extrapolate the wind speed calculated theoretically, and computed respectively by WasP
registered at met mast measurement levels applying the so- and Windsim for the predominant sectors W and E.
called potential law, known to be appropriate in flat terrain or .
gentle hills. As the mast permits two level of measurements, it
is possible to determine the alpha coefficient defined by the
120
Power Law following equation, which is commonly used in
most of the cases that require height extrapolation.
α 100
U(h1) ⎛ (h1) ⎞
=⎜ ⎟
U(h2) ⎜⎝ (h2) ⎟⎠ Equation(1)
80
The figure 7 presents the wind shear obtained from met mast
Height [m]
measurements, from the WASP model and finally from the 60
Windsim model.
40
120
20
100
0
80 4 5 6 7 8 9 10 11 12 13 14
Wind Speed [m/s]
Height [m]
20
The closest wind shears to the theoretical measured one is
surprisingly given by Windsim, being however overestimated.
The vertical wind profile given by measurement present for
0 both met mast a more vertical tendency. The inaccuracy of both
5.00 6.00 7.00 8.00
Wind Speed [m/s] numerical models can be due to the absence of roughness
A1 Met Mast A1 WAsP Model A1 Windsim Model
which, in this type of terrain, influence significantly the
B1 Met Mast B1 WAsP Model B1 Windsim Model evolution of wind speed, especially at lower height. Indeed, we
C1 Met Mast C1 WAsP Model C1 Windsim Model
can see that the difference at heights lower than 40m are much
Figure 7: shows the results obtained for mast A1, B1 and C1 higher than at higher heights, underlining the importance of
between the heights of 20 and 100 m: roughness model in very flat terrain where influence of
orography is negligible.
Site A1: Despite of the low changes in orography; both software
overestimate the wind shear at met mast site, underlining the The following table 3, (predominant directions being underlined
importance of an accurate assessment of the roughness at met in yellow), shows the wind shear calculated by both model
mast site, especially in the case of very low measurement height assuming a “power law coefficient” obtained using a potential
and very flat terrain. Moreover, in this case, site A being regression between wind speeds calculated between 20 and 100
located very close to sea, atmospheric stability is another m height for each model in function of deflection and RIX
parameter which could have generated inaccuracy for both coefficient by sectors.
models. It can be easily seen that in this case orography has no influence
of the difference between WAsP and Windsim models.
Site B1: In this case, results obtained with the WASP model is
closer to measurements. The Windsim model tends to produce a Table 3: Sector power Law coefficient estimation for WAsP
too vertical wind shear. The lower influence of roughness and Windsim model in function of defection, Speed-up and
changes, the no very complex orography and the good Weibull RIX. Site A
Sector Frequency alfa Windsim alfa WAsP Speed up [%] Deviation [º] RIX
fit could explain the improvement of WASP model respect to N 0.015 0.174 0.201 3% -4% 0%
NNE 0.015 0.106 0.206 -2% -2% 0%
Site A1. NE 0.016 0.495 0.195 -3% 1% 0%
ENE 0.022 0.657 0.171 0% 3% 0%
E 0.264 0.113 0.121 5% 4% 1%
Site C1: As the logarithmic corrected model of WASP is not ESE 0.068 0.034 0.037 10% 2% 4%
0.023 -0.015 0.084 11% -1% 0%
valid in very complex terrain, as expected, the values obtained SE
SSE 0.036 0.123 0.139 8% -3% 0%
with the CFD model are much closer to real measured wind S 0.051 0.147 0.184 2% -4% 1%
SSW 0.049 0.175 0.217 -2% -2% 2%
speeds. The inverted wind profile modelled by Windsim SW 0.046 0.171 0.206 -3% 1% 3%
0.111 0.157 0.204 0% 4% 1%
represent with much accuracy the comportment of wind in top WSW
W 0.139 0.138 0.144 5% 4% 0%
of a strong hill, as site C1. For this so complex terrain wasp is WNW 0.053 0.096 0.101 10% 2% 0%
NW 0.064 0.122 0.161 11% -1% 0%
not enough to evaluate the site. NNW 0.026 0.141 0.216 8% -3% 0%
Global -- 0.126 0.141 1%
120
120
100
100
80
80
Height [m]
60
Height [m]
60
40
40
20
20
0
5 6 7 8 0
Wind Speed [m/s] 2 3 4 5 6 7 8 9 10 11 12 13 14
C1 Met Mast_SSE C1 WAsP Model_SSE C1 Windsim Model_SSE Wind Speed [m/s]
C1 Met Mast_S C1 WAsP Model_S C1 Windsim Model_S
C1 Met Mast_WNW C1 WAsP Model_WNW C1 Windsim Model_WNW B1 Met Mast_ENE B1 WAsP Model_ENE B1 Windsim Model_ENE
C1 Met Mast_NW C1 WAsP Model_NW C1 Windsim Model_NW
C1 Met Mast_NNW C1 WAsP Model_NNW C1 Windsim Model_NNW B1 Met Mast_SW B1 WAsP Model_SW B1 Windsim Model_SW
B1 Met Mast_WSW B1 WAsP Model_WSW B1 Windsim Model_WSW
B1 Met Mast_NE B1 WAsP Model_NE B1 Windsim Model_NE
Modelled
temporal series of wind data (wind speed and direction) or
registered by the both met mast considered has been set as input B 1_HH 0.2% -1.7% 4.0%
in each software in order to evaluate the ability of each model B 1_LH 1.8% 0.0% 5.8%
to reproduce wind flow at one location (modelled site)., using B 2_LH -3.3% -5.0% 0.2%
the wind data collected at the other location (Reference met
mast site). Afterwards, the average value of the wind speed
Table 9: Site B. Wind speed deviation. Windsim model
obtained at the modelled site is compared with the theoretical
Reference met mast
average value of wind speed calculated from the measured data.
The difference that it can be seen in the WAsP model when the B 1_HH B 1_LH B 2_LH
Modelled
same mast is used as Reference met mast and as Modelled site B 1_HH 0.0% -10.2% -6.8%
is due to the Weibull-fit. B 1_LH 11.3% 0.0% 3.8%
B 2_LH 7.4% -3.7% 0.0%
Site A:
For this type of semi-complex terrain, as in its estimation of
The following tables the deviation of average wind speed wind shear, WASP provides in this case more suitable results.
respect to the theoretical calculated value from measurements The constant roughness of that site avoid the uncertainties met
for each model. in the previous less complex terrain Site A. An error within a
range of 1% to 3,5% is reached with WASP, while an error of
Table 6: Site A. Wind speed deviation. WAsP model +/- 11% is obtained with Windsim between Mast B1 and B2 at
Reference met mast their highest height.
A 1_HH A 1_LH A 2_LH It can seen also that in the case of a series of data presenting a
Modelled
roughness between Reference met mast site and modeled site is C 1_HH 0.0% -3.7% 7.4%
not taken into account and might affect significantly the value C 1_LH 3.8% 0.0% 11.3%
of the average wind speed obtained, especially in the case of C 2_LH -6.8% -10.2% 0.0%
site A2, which measurement height is 10 m. This tendency Table 11: Site C. Wind speed deviation. WAsP model
increases in the case of WAsP model.
Moreover, in the case of site A, located very close to sea, As expected for a such a complex site, important differences are
atmospheric instability is another parameter which could have produced by both models. Results improves when the highest
generated inaccuracy for both models. At site A, the cross- mast is used as reference for the calculations, as it reduces
checking comparison has shown that the linear model would approximations due to vertical extrapolation. It can be seen that,
preferable in this case. in each case, the best results are obtained when the highest
height of site C1 is used as Reference met mast. However, the
Site B: incapacity of WAsP to reproduce the vertical evolution of wind
flows increases the error when Reference met mast site and
modeled met mast presents different heights.
Windsim model is more able to reproduce the changes in wind of wind flows. Nevertheless, in the case of smaller slopes or
directions generated by the orography and consequently elevation changes, this model has a slight tendency to
reproduce the difference of wind rose between met mast C1 and overestimate the wind shear and thus remains conservative.
C2, which influence strongly the value of the average wind
speed. Nonetheless, it is necessary to consider that in the case of
a terrain presenting a very high level of complexity, the CFD 7. Acknowledgement
model results need to be balance by measurements as it is not
completely reliable. This work is being carried out with funding from the Spanish
Ministerio de Ciencia y Tecnologia (projects: DPI2003–09731
The figure 11 resumes the difference between the theoretical and FIT-120000-2004-182) and the Cai-Europa Program.
measured wind speed at one site (met mast) from another met
mast. Results are presented in function of met mast height (LH: 8. References
Lower height and HH: Higher height) from the lowest to
highest complexity of terrain. [1] RISO WAsP web site, . http://www.wasp.dk.
B2_LH 6 3.8% 3.7% [4] A.J. Bowen and N.G Mortensen, WAsP Prediction
1.8% 1.7% 0.2%
Errors Due to Site Orography” , Riso National
B1_LH 5
3.3%
6.8%
4.0%
7.4%
Laboratory, December 2004, pp 28-29, pp 34-35.
10.2%
B1_HH 4 11.3%
0.2% 5.0% 5.8% [5] B. Schaffner and A. N. Gravdhal, “Wind modelling
A2_LH 3 11.9% 0.8% in Mountains: Intercomparison and Validation of
5.0% 3.3%
9.1%
Models”, ECOTECNIA, Barcelona, Spain; VECTOR AS,
A1_LH 2
8.8%
9.9%
3.6% 5.2%
10.4%
Tonsberg, Norway. 2003.
11.4%
A1_HH 1 0.3%
3.3% 16.8% [6] P.S. Jackson, J.C.R. Hunt, Turbulent wind flow over a low
17.7%
hill, Quart. J. R. Met. Soc. 101 (1975) 929-955.
0
A1_H H A1_LH A 2_LH B1_H H B1_LH B2_LH C1_HH C1_LH C2_LH
0 1 2 3 4 5 6 7 8 9 10
Refe rence M e t M ast
[7] AWS TrueWind, LLC “Estimation of the Long-
% Over estimation WAsP
% Under estimation WAsP
% Over estimation Windsim
% Under estimation Windsim
Term Average Wind Speed and Energy Production at the
Proposed. East Mountain Wind Project Site”.
Figure 11: Site A, B and C WAsP and Windsim cross-checking
results [8] I. Troen and E.L. Petersen, “European Wind Atlas”,
Riso National Laboratory, 1989. pp 389.
6. Conclusions [9] R. Derickson, M. McDiarmid, B. Cochran and J.A.
Peterka, “Wind Energy Explained, Theory, Design and
In a very flat terrain, both models seem to be very sensitive to
Application”,. ”, WILEY, University of Massachusetts,
roughness changes, not considered in this project. Moreover, in
2002, pp. 50-78, pp 370-409.
the case of site A, located very close to sea, atmospheric
stability is another parameter which could have generated
inaccuracy for both models. However, the cross-checking
comparison has shown that the linear model would preferable in
this case. The linear WAsP shows an high dependency to
roughness modelling in the case of very flat terrain, with a
almost null influence of orography.