Wind Mapping of Sweden

Titel: Wind Mapping of Sweden – Summary of results and methods used, Elforsk rapport 09:04
Författare: Hans Bergström & Stefan Söderberg
Utgivare: Elforsk, Vindforsk
Årtal: 2008
Ämnesord: Årsmedelvind
Sökord/Keywords: MIUU-model, NCEP/NCAR, CFD
Rapport



Sammanfattning/Abstract:
A three-dimensional meso-scale numerical higher-order closure model, the MIUU-model developed at Uppsala University, has been used to investigate the wind climate in Sweden. This type of model takes special notice of the
conditions in the atmospheric boundary layer. Results from a wind resource mapping of Sweden with 1 km2 horizontal resolution is presented here.

A technique to just model some samples of the meteorological conditions governing the wind climate at heights of interest to wind energy has been used. This was shown to give accurate results in good agreement with observations, making it unnecessary to model a long period of ‘true weather’ cases. By choosing the ‘relevant’ samples, the number of model runs could be limited to 192, which correspond to 4608 simulated hours (representing 4 months x 3 speeds x 16 directions x 24 hours each).

Statistics on the horizontal air pressure gradient (geostrophic wind), the major force driving the actual wind, were used to weight the model output together into the final wind climate estimates. Surface pressure data from 3 pressure observing sites in Southern Sweden and 3 sites in Northern Scandinavia were then primarily used. As the geostrophic mean wind varies quite a lot over an area as large as Sweden, it was also needed to take this into account. For this purpose the NCEP/NCAR reanalysis data were used to calculate the geostrophic wind. These results were then used to weight the influence of the geographical variation in geostrophic wind into the results regarding the modelled wind climate.

Comparisons with wind observations at 84 sites showed that the modelled annual average wind speed is in good agreement with measured values. The average difference between model results and observations were found to be -0.03 m/s, and for 87 % of the comparisons the differences were within ±0.4 m/s.

Additional wind statistics, besides the annual average wind speed, have also been determined using the database of modelled atmospheric conditions.

These results include:

• Distributions of wind speed and wind directions, giving the sector wise
Weibull parameters for 12 wind direction sectors.

• Extreme wind speed for 10 min average wind speed and 3 s gust wind
speed with an expected 50-year return period.

• The wind gradient presented as the exponent of the exponential wind
profile.

• Turbulence intensity determined from the modelled turbulent kinetic
energy.

• Wind turbine classes estimated from the modelled average wind speed
and turbulence intensity being translated to reference wind speed and
reference turbulence intensity at 15 m/s.

All wind statistics is available through Internet at:
http://www.geo.uu.se/luva/default.aspx?pageid=13152&lan=0.

It is, however, important to remember that the model does not adequately resolve influence from terrain features smaller than a few times the model grid spacing. With 1 km2 resolution the main characteristics of the geographical variability of the wind climate is captured, but this resolution is often too coarse when planning the details of a wind farm and the exact wind turbine locations within the farm. For that case higher resolution wind resource mapping is needed. This has previously often been done using a model such as WASP, either directly or as a tool within WindPro. This type of model has a simplified physical description of the atmosphere why the results may sometimes be questioned.

Using a higher-order closure meteorological CFD-model such as the MIUU-model has previously in practice been unrealistic for downscaling to the high resolutions needed for local micro scale planning. Due to the hydrostatic approximation in the MIUU-model it was decided also to test another meteorological CDF-model of the same type, but
with a non-hydrostatic closure. The decision fell on COAMPS®, Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS®, a registered trademark of the Naval Research
Laboratory).

A comparison between modelled annual average wind speeds in the Fjällberget/Saxberget area in Sweden showed that both CFD-models give realistic results. Some differences are found, however, and comparisons with observations at the site showed that the MIUU-model seems to perform better in spite of its hydrostatic approximations. The reasons for this may have to do with differences regarding the turbulence closure and also regarding the surface energy balance routine which may in turn result in differences in
thermal stratification.

The conclusion of the comparisons is that downscaling using higher-order meteorological CFD-models is realistic and gives results in agreement with observations. However, observations with much higher horizontal resolution are needed to fully evaluate the differences found between the two models. We may also assume that the results are probably more realistic than results using models with more simplified physics, especially when input data to the simplified models are not available from the site or even only at a specific location with an altitude not representative for the whole area of interest. This
is so because the differences between high and low altitude terrain were found to be much smaller using a simplified model than using a CFD-model.

Note that no local wind observations are included or needed when calculating the wind resources using the MIUU-technique together with the results from a higher-order model. Hence, the model output is totally independent of the observations with which comparisons may be made.