Venkatesh Duraisamy Jothiprakasam

1
Downscaling wind energy resource from mesoscale
to CFD model with data assimilation of field
measurements

• Venkatesh Duraisamy Jothiprakasam
• Eric Dupont
• Bretrand Carissimo
• EDF RD, France
• EWEA, Vienna, 4th-7th Feb 2013

2
Objective Introduction

• Annual Energy Prediction (AEP)
• Accurate simulations of wind speed, direction and
  turbulence for wind farm.
• Turbine siting.

• Operational mesoscale data - Météo-France
• k-means clustering method - reduce number of CFD
  simulation needed to reproduce 1 year of flow
  condition.
• Coupling method - apply mesoscale data on CFD
  grid
• 4DDA (nudging) - incorporate field measurements
  into CFD model
• Measurement data - nudging and validation

3
Complex terrain Measurement campaign

• Site
• Southern France
• Very complex, strong slopes (30), valleys and
  forest (80)
• SE Plain, NW Valley and further NW plateau
  at 1100 m
• IGN - Digital Elevation Map Land Cover Use
• Measurement campaign
• 1-year field measurement by EDF RD and EDF EN
• 80m mast with cup and sonic anemometers, vanes,
  temperature and humidity sensors
• 2- 50m masts with cup anemometers and vanes
• 2 sodars

Institut Géographique National France
4
Coupling Mesoscale and CFD models

• Purpose of coupling
• Problem how do we apply mesoscale data on CFD
  grid?
• Mesoscale grid (?xy ? a few km and ?z ? some tens
  of meters) and CFD grid (?xy ? some tens of m and
  ?z ? some meters)
• Large differences at domain boundary
• Some unavailable values
• To solve it Translation, Extrapolation and
  Cressman interpolation

CFD Model Code_Saturne
Mesoscale Model - ALADIN
5
Coupling Methods-Translation, Extrapolation,
Cressman interpolation

• Translated along the absolute co-ordinates of the
  microscale grid
• Extrapolated between the lowest level of the
  mesoscale profile and the microscale grid
• Cressman interpolation NWP, linear combination
  of the values provided by the nearby mesoscale
  data

6
Cressman interpolation-NW

• Velocities calculated using Cressman
  interpolation for all inlet microscale grid faces
• Sensitivity study for radius of influence
• Lowest radius uses the closest ALADIN profiles
• Highest radius includes all the ALADIN profiles.

7
Data assimilation-Nudging

• NWP technique, not common in CFD
• Combining observations of variables into
  numerical models
• Adding in prognostic equations a term that nudges
  the solution towards observation.

• Radial drop-off rate is used
• for
• We have sonic, cup anemometer and sodar data
  available for nudging and validation

8
Case description

• Structured grid with 4.7 Million cells
• Steady state simulation
• Standard k-? turbulence model
• Inlet boundary profiles computed using
  translation, extrapolation and Cressman
  interpolation
• Simulations
• CFD No nudging.
• CFDnudging CFD simulation with nudging
  at M80 Mast location.
• Case 1 CFD predicted far from measurement and
  how CFDnudging performed?
• (7th December 2007 at 1600 CET from NW
  305 direction)
• Case 2 CFD predicted well and how CFDnudging
  performed?
• (17th November 2007 at 2000 CET from NW
  335 direction)

9
Results-CFD with and without nudging Case - 1
10
Results-CFD with and without nudging Case - 1
Without nudging With
nudging Without - With nudging

• Figures show the velocity magnitude at 795 m for
  CFD without nudging, CFD with nudging and
  difference between the 2 simulations.

11
Results-CFD with and without nudging Case - 1
Without nudging
With nudging
12
Results-CFD with and without nudging Case - 2
13
Results-CFD with and without nudging Case - 2
Without nudging With
nudging
14
Conclusion future work

• Conclusion
• Implemented the operational mesoscale data into
  the CFD grid.
• Cressman interpolation better represented inlet
  condition.
• Incorporated field measurement into the CFD model
  using nudging.
• CFD simulation - cannot correct the mesoscale
  wind direction, nudging helps correct the
  directional change
• Nudging - good improvement is noticed at the
  nearby mast locations
• Future work
• Calculate Annual Energy Prediction (AEP) using
  data clustering with assimilation.
• Include thermal stratification and forest model.

15
Thanks for your attention