Application d'un code de CFD atmosphrique l'estimation du productible olien en terrain complexe

1
Atmospheric CFD Simulations coupled to Mesoscale
Analyses for Wind Resource assessment in complex
terrain
Laurent Laporte (1), Éric Dupont (2),
Bertrand Carissimo (2), Luc Musson-Genon (2),
Cyril Sécolier (3) (1) CEREA (2)
EDF-RD / CEREA (3) EDF-Energies Nouvelles
Photo GH
Vendredi 12 décembre 2008
2
Introduction

• Building a wind farm requires a preliminary
  estimation of its production taking into account
  
• The wind resource,
• Turbulence levels.
• Current operational methods, based on linear
  models, lack precision in complex terrain.
• The objective of this work has been to develop a
  new methodology to better evaluate wind resource
  in complex terrain.

3
Mercure_Saturne, open source CFD code

• Mercure_Saturne has been used in this work.
• Developed by CEREA laboratory, based on the
  Code_Saturne kernel developed by EDF - RD.
• 3D atmospheric RANS solver for atmospheric flows
• k-e closure has been used.

• Mercure_Saturne has been validated on Askervein
  hill test case.

4
Methodology
correction

• Our methodology is based on 4 steps
• Measurement campaign,
• Meso-scale model outputs,
• Clustering,
• Micro-scale CFD.

Boundary conditions
5
Measurement campaign
m
FP
M80
M
6
Meso-scale model outputs
Mesh Masts
Aladin mesh
7
Meso-scale model outputs
Hourly data
Wind speed m/s

• Under-estimation of the wind speed by the
  meso-scale model Aladin

Direction

• Delay after a direction change by the meso-scale
  model Aladin

8
Clustering

• 1 year 8760 hours
• 8760 different meteorological situations
• 1 CFD simulation takes about 3 hours (on 4
  processors)
• Impossible to run all the 8760 simulations!
• Solution Cluster analysis,
• Chosen method k-means.

Pattern Meteorological situation on the vertical
number 6 defined by Aladin corrected data.
Similarity measure Euclidean distance.

• Variables
• Wind speed,
• Wind direction,
• (100m above ground).

9
Clustering
577 clusters Max radius 2,12m/s Mean radius
1,53m/s Sigma max 0,86m/s Smallest cluster 50
situations Biggest Cluster 216 situations
64 real situations are selected to represent
the year 2007
10
CFD

• Mesh
• 250 632 cells
• Horizontal resolution between 50 et 250m
• First cell height above the ground 10m
• Domain 15,7km x 15,7km
• Altitude min. 104m
• Altitude max. 1079m
• Mesh top 6309m
• Boundary conditions
• Wall functions (detection of the land use to
  evaluate z0)
• In- and outflow conditions functions of the wind
  profile
• Symmetric condition on top.

• Numerical parameters
• Dynamic adaptation
• Constant time step
• 800 time steps
• Cluster IBM EDF RD
• 64 simultaneous simulations
• 8 hours needed

11
Wind potential assessment

• 2 methods to evaluate the wind potential
• center
• weighted

12
Wind potential assessment

• CFD method

• WAsP

• Mercure_Saturne reduces the relative error on 2
  to 3 masts depending on the reference mast chosen
  for WAsP.

13
Wind potential assessment gt Maps
Wind speed 80m above ground
m/s
TKE 80m above ground
m/s
Wind speed 80m above ground
m2/s2
14
Conclusion

• The methodology shows already better results than
  WAsP on complex terrain.
• High-resolution maps of wind speed and turbulence
  are available.
• Long Term wind speeds can be evaluated for the
  same cost.
• Wakes can be simulated directly in the mesh with
  source terms
• There is room for improvement !
• Finer mesh
• Better micro/mesoscale coupling
• Stability as a third clustering parameter.
• Future work European project WAUDIT (Marie Curie
  ITN)