High-resolution numerical modelling of weather conditions provides sensitive information of good quality, which is crucial for the development of any wind project and are useful from the early stages of prospecting the wind farm design to long-term adjustments. In particular, the use of meso and micro-scale coupled wind resource products has gained widespread acceptance by the wind industry, offering reliable long term reference data for wind condition characterization and the same has been utilized in this work. Under this study, Meso to micro-scale coupling is solved within the modelling chain by seamless simulations of WRF down to 500m resolution.

The core of the technical modelling approach for this work is the atmospheric model Weather Research and Forecasting System (WRF) developed by NCAR/NCEP. The WRF-system is a community-based, open-source model, where the latest advances in physics and numeric are incorporated in a modular way. It includes a nest domain, allowing zooming atmospheric circulation down to near wind-farm resolution.

WRF model has a long record on usage and it is employed operational in many weather services, cutting-edge research activities and different industry applications. WRF development has engaged a wide community of users, which meant large peer-review validations analysis, and upgrading of advances in the different components of the weather & climate modelling science.

WRF model is allows multi-scale chain modelling that can seamlessly go from regional to wind farm scales. Regarding, micro-scale, WRF incorporates innovation planetary boundary layer sub model (PBL) that can handle effectively turbulence and flow adjustments due to high-resolution orographic effects. Moreover, WRF is a unique solution to provide dynamic representation of wind flow at wind farm resolution including mechanical and thermal turbulence.

Experience acquired from the studies of Vortex in more than 24,000 simulations completed with thousands of them checked against measurements has proven that the WRF model at a 500m resolution produces a realistic representation of flow circulation induced by high-resolution topography effects such as valleys, sea-land transition, hills, etc. By preserving the continuity of the WRF modelling chain, a more consistent site-specific assessment of wind energy parameters can be obtained, minimizing the impact of any artificial interpolation between different atmospheric scales.

Model Run

Under this study, the model has been driven by large-scale conditions prescribed by one of the latest generations of re-analysis projects for the satellite period: NCEP CFS/CFSR. With long-term wind variation into consideration, 10 years of data (2005 – 2014) from NCEP CFS/CFSR was used to initiate the flow modelling. Further, the re-gridded versions of SRTM (no-void) altimetry data were employed to prescribe altimetry conditions. The ESA Globcover (300m) land use database was employed to characterize land-use classes. WRF (Weather &Research Forecast) model is used in order to downscale Reanalysis datasets to the final 500m x 500m resolution through nested domains. Nesting is performed at 27km, 3km, 1km and 500m, each resolution adequate for different scales and characteristic phenomena included in the WRF model. Thus, the WRF model provides output variables at each of the 500m x 500m grid points and therefore no interpolation methods were applied for generating the results. In relation to the complex terrain treatment, no different code is applied. The WRF model is capable of modelling the wind at different resolutions and each of the scales is treated accordingly with the Navier-Stokes equations and the corresponding PBL & Surface Layer schemes parameterizations available in the model. The modelling flow chart is shown in FigureA.1.


Methodology to derive Main Variables

The WRF model provides Wind Speed, Temperature & Pressure among other variables at each of the 500m x 500m grid points and any height between ground level and the troposphere. This allows us to estimate the wind power density by making use of the wind speed and density on an hourly basis -a general gas law for air is used for deriving density values from pressure & temperature.

Joint frequency distribution is obtained by binning the wind speed in 1m/s bins from 0 to 1, 1 to 2 and so on, and binning the wind direction in 30 degrees sectors north centered, that is from -15 to 15, 15 to 45 and so on. The joint frequency distribution can be presented in percentage or number of hours per year. Weibull Parameters are computed by using the WAsP assumption, which emphasizes the most energetic part of the histogram in the Weibull fitting.

Uncertainty of the results is estimated by running some different configurations of WRF model that is different turbulent schemes and slightly different initial conditions, which perturb the model and give an indication on how sensitive the studied region is to some changes in the model. The generated results by the different configurations are then post processed and mixed up in order to have an idea of how wide / uncertain the wind distribution.

Thus for each 500m grid point the following parameters have been derived.

  • Mean Wind Speed(m/s)
  • Weibull Shape factor –k
  • Weibull Scale factor – A(m/s)
  • Mean Wind Power Density(W/m2)
  • Mean Temperature(0C)
  • Mean Atmospheric Pressure(KPa)
  • Mean Air Density(Kg/m3)
  • Wind Direction
  • Joint Frequency Distribution

Methodology adopted for the derivation of CUF and its mapping

  • Wind speed frequency distribution at each grid point were calculated by using the wind speed and Weibull shape parameter ‘k’ for 0 – 25m/s.
  • A normalized power curve was derived from seven modern wind turbines used in the country with approx. 2 MW capacities. The normalized machine was corrected for air density (IEC method) at each grid point.
  • By utilizing both air density corrected power curve and wind speed distribution, gross CUF was estimated for each grid point.
  • Standard correction factors as per practice (95% - Grid availability, 95% - Machine Availability, 3% - Transmission Loss and 10% - Array Loss) was applied to the gross estimates to find out the net values of each grid point at P50 (50% probability of exceedance) confidence level.
  • Thematic map for P50 capacity utilization factor (%CUF) was prepared with classifications of less than 20%, 20-25%, 25-28%, 28-30%, 30-32%, 32-35% and greater than 35% ranges.