MSc defence Joaquin Villanueva Figueroa

Supervisors: Senior Researcher Alfredo Pena, DTU Wind Energy - Researcher Rogier Ralph Floors, DTU Wind Energy

Examiner: Lars Landberg, DNV GL

Title: Evaluation of the wind atlas method to predict wind resources from numerical wind simulations
Understanding the methodologies behind the production of numerical wind atlases (NWA) is essential to exploit their potential as planning tools. Their ability to estimate wind resources in a fast and robust manner is limited by a number of reasons, including their resolution and the variable error of their predictions. Such issues are related to the numerical simulations used in the production of a NWA, which can be outputs from reanalyses or mesoscale models. These coarse simulations do not capture microscale features of wind flows, which are essential to predict wind resources at specific sites.

Numerical simulations are corrected with the wind atlas method embedded in the Wind Atlas Analysis and Application Program (WAsP). This is called the numerical wind atlas method (NWAM). To account for the microscale effects caused by the terrain, WAsP requires topographic maps. Maps can be derived from multiple sources, but some lead to more reliable predictions than others. An evaluation of the NWAM with multiple inputs was carried out, aiming to identify the requirements that lead to the most accurate and consistent predictions.

Six sites in Northern Europe with at least 2 years of observational records were used: 2 offshore and 4 onshore. For each site, a total of 8 numerical simulations from ERA5 reanalysis and the WRF mesoscale model were used. Multiple maps were available at each site. Predicted and observed wind climates were compared through their overall mean wind speeds and power densities. Vertical wind speed profiles were compared visually and using the rotor-equivalent wind speed (REWS).

The REWS-based power density was introduced as another metric. For each metric, all the prediction errors were calculated and the mean average error (MAE) was used to quantify uncertainty.

For onshore sites, it was found that selecting the best map reduces the error in the NWAM by more than 50%. The maps that consistently led to the smallest errors were derived from GlobCover2009. At 3 sites, NWAM predictions using simulations from ERA5 at 100 m height result in the lowest errors and show the best resemblance to the observed wind profiles. However, at the fourth site it caused 67% power density overpredictions. Such variability in the errors do not occur when using WRF simulations. The NWAM with WRF simulations at 150 m height leads to the lowest MAE: 2.5%.

For offshore sites, interpolations from ERA5 and WRF simulations at different heights resemble wind profiles better than the NWAM. The NWAM errors are minimized when using simulations from WRF at heights below 100 m. Considering all results from this study, the use of WRF simulations at 100 m height is recommended for the production of a NWA including offshore and onshore sites.



Mon 29 Jul 19
9:00 - 11:00



DTU Risø Campus
Frederiksborgvej 399, bld. 118 - Tycho Brahe meeting room
4000 Roskilde