Application of WRF and Fluent for Wind
Energy Forecasting
College of Engineering
College of Engineering
Alan Russell, Kevin Nuss, Paul Dawson and Todd Haynes, Boise State University
Project Objectives
Forecasting for Wind Energy
Grid Integration
• Develop forecast methodology
• Reliably forecast power production in MW
• Produce 5 minute, 1 hour and 24 hour
forecasts
• Weather forecast using WRF, Weather
Research and Forecasting, a mesoscale
meteorological model
• Couple WRF with FLUENT computational
fluid dynamics (CFD) model
o CFD model for individual wind farms
o Complex terrain effects
o Buoyancy effects
o Wind fields and turbulence
• Develop met tower network and SODAR
field data to improve accuracy and forecast
ramp events
• Evaluate forecast model accuracy
Mesh Inputs
o 30 meter USGS DEM (digital elevation map) data
o DEM turned into Gambit (ANSYS mesh software)
face
o Mesh domain built above surface
o Hexagonal mesh with surface boundary layer
o 2 meter resolution in surface layer
o 2 million cells model a 2.4 km by 2.6 km by 600
meter volume domain.
CFD Model
o FLUENT: ANSYS commercial CFD code for
science and engineering
o k-ε turbulence closure
o Good compromise between accuracy and speed
o Second order upwind scheme
o Numerical solution convergence requires up to 12
hours run time on a 2 processor Linux workstation
FLUENT Inputs
Wind Farm
Results
CFD Model
o WRF results for vertical wind speed profiles, wind
direction, k and ε along 1 km inner nest boundary
o WRF output converted into FLUENT inlet velocity
profile data grids
o Set domain boundaries for wind direction
o Non-standard k-ε model constants for atmospheric
flow
o Surface roughness height
o Surface heat flux for buoyancy model
FLUENT neutral model results for wind speed
contours at 80 m, 2/10/09 at 6 AM. WRF input
wind was 318°(blue arrow) at 12.51 m/s. At 80
m, SODAR recorded 13.2 m/s wind at 310°
CFD Output
• CFD model outputs velocity at hub height
• Velocity and turbulence across rotor plane of
each turbine
• CFD process is too slow for real time
forecasting
• Use CFD to model
o Neutral and buoyancy driven flow
o Diurnal and seasonal effects
o Wind ramp events
• Project goal: Generate look-up tables using
CFD simulations to match the wind rose
Wind farm contour
map showing
turbine and
SODAR locations.
The yellow square
is the CFD domain
area. The wind
rose shows the
predominant NWSE wind pattern
CFD & SODAR
FLUENT 10-Feb
FLUENT 4-Jan
SODAR 10-Feb
SODAR 4-Jan
14
Wind Speed (m/s)
13
Forecast Network
12
11
10
FLUENT buoyancy model results for wind speed
contours at 80 m, 1/4/09 at 1 PM. WRF input wind was
100°(blue arrow) at 11.4 m/s. SODAR recorded 11.2
m/s wind at 100°at 80 m during the same time period
9
40
60
80
100
Height AGL (m)
Next Steps
Top instrument height, distance, direction and
ramp forecast time from wind farm
• 80 m tower 69 km NW ~ 3 hours upwind
• 30 m tower 51 km NW ~ 2 hours upwind
• 50 m tower 5 km NW
• SODAR at the wind farm
• 50 m tower 56 km SE ~ 2.5 hours upwind
Thank you for assistance and advice:
Dr. Inanc Senocak and Marty Lukes, Boise State
University College of Engineering
Dr. John Gardner, Boise State University Dept. of
Energy Research, Policy and Campus Sustainability
Kurt Myers, Idaho National Lab
Doug Taylor, John Deere Wind Energy
Second Wind Triton SODAR
Bonneville Power Administration (BPA)
Research Funding: BPA Contract 00039902
• Complete data network
o Install 30 m tower
o Update existing towers
• Refine FLUENT model
• Generate CFD lookup tables
• Compare model power forecast to
wind farm SCADA
• Run actual forecasts
• Complete final BPA grant report
120