The benchmark is open to participants of the Wakebench project using flow models over topography in neutral conditions. This initial benchmark will carry out an assessment of the sensitivity to modeling criteria that is adopted when approaching the simulation of a realistic site for wind farms in complex terrain. In order to simplify the analysis, only the North and South wind direction sectors will be analyzed.
Carry out sensitivity tests on different elements of the modeling chain that require user-dependent decisions: domain dimensions, mesh type and resolution, roughness definition, and wind direction binning (1).
The benchmark is offered to participants of the IEA Task 31 Wakebench who has signed the NDA attached to the test case guide.
The conditions for simulating the Alaiz flow field in neutral conditions are:
- Digitized elevation map based on 1:10000 maps (elevation lines every 5 m) and roughness maps based on orthophotos from the Navarra Geoportal (www.sitna.navarra.es)
- Coordinates of the mast positions and sensor levels
Inlet conditions are not measured. It is up to the user to select the most appropriate profiles. The results will be made independent of this choice by normalizing with the MP5 118 m velocity (U0) and turbulent kinetic energy values (TKE0). As guideline, A target U0 of 9 m/s and turbulence intensity of [6, 10] % for the North and South sectors respectively can be used since these will be wind conditions for the validation case (Alaiz_BlindNeutral).
This benchmark does not include validation data. The evaluation will consist on model intercomparison for the following parameters:
- Velocity ratio with respect to MP5 (U/U0)
- Ratio of turbulent kinetic energy with respect to MP5 (tke/tket0)
The study will focus on the prevailing wind direction sectors from North and South. East and West boundaries are placed at a distance of 7 and 10 km from MP5 to make sure they do not have any influence on the results. The following simulation runs are requested:
- Domain dimensions: Do large upstream hills influence the results? How far North and South should the domain be extended in order to have results independent of the inlet profile? In order to answer these questions North and South wind directions will be run for two domains (Figure 1): D1) a long one including the most important topographic obstacles upstream (D1: Xmin = 609000 m, Xmax = 623000 m, Ymin = 4715000 m, Ymax = 4737000 m) and D2) a short one not including them (D2: Xmin = 609000 m, Xmax = 623000 m, Ymin = 4719000 m, Ymax = 4732000 m). The far field terrain, beyond D1/D2 to extend to the boundaries, can be smoothed out to flat terrain (2). This is a up to the user. A uniform roughness of 0.05 m in the entire domain should be used.
- Run 1: North, D1, M1, z01 = 0.05 m
- Run 2: South, D1, M1, z01 = 0.05 m
- Run 3: North, D2, M1, z01 = 0.05 m
- Run 4: South, D2, M1, z01 = 0.05 m
Figure 4: Domain limits D1 and D2. The far field terrain, beyond D1/D2 to extend to the boundaries, can be smooth out to flat terrain (2).
- Grid sensitivity: Based on D1 domain, and using M1 as a reference, increase the grid resolution to assess grid dependencies in the horizontal and vertical directions. A uniform roughness of 0.05 m should be used.
- Run 5: North, D1, M2z (refine in the vertical direction keeping the same horizontal levels of M1), z01 = 0.05 m
- Run 6: South, D1, M2z, z01 = 0.05 m
- Run 7: North, D1, M2x (refine in the horizontal directions keeping the same vertical levels of M1), z01 = 0.05 m
- Run 8: South, D1, M2x, z01 = 0.05 m
- Run 9: North, D1, M3 (optimum mesh by further refinement in any direction), z01 = 0.05 m
The grid sensitivity can be also done by coarsening, i.e. starting from the finest mesh and systematically reducing the number of cells in the horizontal and vertical directions. In any case, please keep the relationship between run number and mesh size as indicated.
Since the mesh size depends on the available computational resources it is up to the user to adopt the strategy that best fit the constraints.
- Roughness: Orography is the main flow driver but which positions and wind direction sectors are more influenced by the roughness of the terrain?
- Run 10: North, D1, M3, z02 = 0.4m
- Run 11: South, D1, M3, z02 = 0.4m
- Run 12: North, D1, M3, z03 = map
- Run 13: South, D1, M3, z03 = map
- Wind direction variability: What shall be the wind direction bin width that should be selected to integrate the directional distribution of the wind climate? The variability of the results for different incidence angles will be evaluated:
- Run 14: North - 10º, D1, M3, z03 = map
- Run 15: North - 22.5º , D1, M3, z03 = map
- Run 16: South - 10º, D1, M3, z03 = map
- Run 17: South - 22.5º , D1, M3, z03 = map
Real topographic coordinates will be used with X pointing East, Y pointing North and Z pointing up.
The requested output data consist of:
- North-South transect, inlet to outlet boundary, passing by the MP5 position at 40 and 118 m levels (profID = "profNS#").
- Vertical profiles from ground level to the top of the domain at MP* (profID = "profMP#") and A* (profID ="profA#").
For each output point you should provide: velocity components (U, V, W), turbulence kinetic energy (tke) and eddy viscosity (nu_t). Hence, the required outputs are, in this order: X(m), Y(m), Z(m), U(m/s), V(m/s), W(m/s), tke(m2/s2), nu_t(m2/s).
Use the file naming and format convention described in the Windbench user's guide with profID = prof#, where # = [NS40, NS118, MP1, MP3, MP5, MP6, A1, A2, A3, A4, A5, A6], i.e. 12 output files per user and model run (12x17 = 204 files).
(1) The benchmark constitutes a step-by-step sensitivity analysis on the elements of the modeling chain that require some expert judgment from the user. The systematic approach imposes constrains in order to limit user dependencies and allow easier model intercomparison studies.
(2) For easier comparison between the two domains, you shall build a mesh for the domain D1 (M1) by extending the mesh of domain D2 with two additional blocks in the North and South sectors.
You can use buffer zones to make a transition to flat terrain conditions (wind tunnel approach). This has the following advantages:
- Horizontally homogeneous inlet boundary conditions can be used more consistently
- Better convergence close to boundaries avoiding reverse flow
- Possibility to use periodic boundary conditions
In principle, all the topographic features that affect the flow in the area of interest shall be included in the inner domain based on the real topography. Then, the outer buffer zone is just a numerical artifact to allow easier boundary conditions and improve convergence.
(3) The grid sensitivity is not fully constrained in order to accommodate different grid sensitivity strategies and computational resources. Two approaches will be typically adopted:
- Grid sensitivity by refinement: The standard approach in consulting, where a cost-effective sensitivity analysis is sought, consist on starting from a mesh of reasonable size and then refine in areas of large gradients typically increasing the mesh size by at least a factor 2. By monitoring the relative change of some target variables, for instance the mast(s) velocity and tke, it is possible to judge if further refinement is necessary: If the deviations are small compared to the bias from the validation data then you can assume that the grid dependency is under control.
- Grid sensitivity by coarsening: A systematic coarsening of an initial mesh is performed by removing cells in each direction. Starting from the finest mesh you can reasonably afford, in each coarsening step you leave out every second node in each direction at a time.
Either way, both methods are user and case dependent. Therefore it is important that the user convincingly justifies the rationale behind the grid sensitivity and how this is monitored in order to reach solutions that are as close as possible to being grid independent.
(4) Based on the results of this benchmark it should be more comprehensive to perform an optimized simulation for the validation phase that will be the object of the next Alaiz benchmark in blind conditions ("Alaiz_BlindNeutral"). Then, runs 12 and 13 could be directly used for the blind test or, if the results show large grid dependencies, the user could decide to perform further refinements.