1 / 15

Dr. Cathérine Mei β ner a Dr. Arne R. Gravdahl a Dr. Xuan Wu b

Short-term Forecasting using Mesoscale Simulations, Neural Networks and CFD Simulations EWEA 2012 Annual Event 16-19 April 2012 Copenhagen. Dr. Cathérine Mei β ner a Dr. Arne R. Gravdahl a Dr. Xuan Wu b a WindSim AS, Fjordgaten 15, N-3125 Tønsberg, Norway

Download Presentation

Dr. Cathérine Mei β ner a Dr. Arne R. Gravdahl a Dr. Xuan Wu b

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Short-term Forecasting using Mesoscale Simulations, Neural Networks and CFD Simulations EWEA 2012 Annual Event 16-19 April 2012Copenhagen Dr. Cathérine Meiβnera Dr. Arne R. Gravdahla Dr. Xuan Wub a WindSim AS, Fjordgaten 15, N-3125 Tønsberg, Norway b WindSim AMERICAS, 470 Atlantic Avenue, Boston, MA 02210, US

  2. Content • The forecasting procedure • Validation on a Chinese wind farm • Conclusion & Outlook

  3. The Procedure: Coupling Mesoscale, ANN and CFD • Understanding how much your wind farm will produce in the next hours is crucial to make the right decisions, either in the energy market or for maintenance planning. • WindSim has developed a system coupling • Mesoscale numerical weather forecasts (Mesoscale) • Artificial Neural Networks (ANN) and • Computational Fluid Dynamics (CFD) • This system is under development together with partners from the energy trading business and will be validated on different sites all over Europe in the next months.

  4. The Procedure: Coupling Mesoscale, ANN and CFD GLOBAL FORECAST Set-up period Forecasting mode MESOSCALE FORECAST Neural Network correction Mesoscale hindcasts Historical wind measurements Neural Network Training WindSim CFD Downscaling CFD look-up tables WindSim wake model WindSim Power Production Forecast

  5. The Procedure: Coupling Mesoscale and CFD Global Models 100 - 16 km e.g. ECMWF, GFS Regional Models 9 - 1 km e.g. WRF Description of the atmospheric conditions Micro Model 100 - 10 m WindSim Accurate description of the local flow field and the wake effects

  6. The Procedure: Coupling Mesoscale and CFD Transfer of mesoscale data into the CFD by using a virtual met mast solution: The forecast of one point in the mesoscale model inside the CFD domain is selected and used to scale the CFD model results at every turbine position Advantage: Very fast as CFD look-up tables can be produced in the set-up phase of the forecasting system and no CFD simulation is necessary during the actual forecast. Micro Model WindSim Regional Model

  7. The Procedure: Why use Artificial Neural Networks? WRF WRF_ANN MEAS wind speed (m/s) 0 5 10 15 20 January February WRF data has phase and model bias errors in wind speed and direction Trained networks can be used to correct each forecasted time series from the mesoscale model before it is used in the CFD simulation

  8. The Procedure: The added value of using CFD 500m 20 m The CFD describes more accurately the local flow field around the turbines and can therefore downscale the mesoscale model results The CFD is able to calculate the wake corrected energy production .

  9. Validation – Chinese wind farm • New legislation in China requires an operational forecasting system for all wind farms • Validation site in China: • Site with 6 measurement masts and 11 turbines • Complexity of the site: steepness up to 45 degrees around the turbine area and absolute height differences of 1000 m • WRF simulations run for 4 months in the winter season on 1 km resolution • WRF results for wind speed and wind direction extracted for every met mast position in the area

  10. Validation – Chinese wind farm 1. The WRF mesoscale model predicts the monthly mean wind speed and direction well but the absolute value is too high WRF MEAS Mast 2

  11. Validation – Chinese wind farm 2. The Neural Networks are able to correct this deviation in wind speed and wind direction

  12. Validation – Chinese wind farm 3. The CFD modelling improves the modelling of the wind profile Mast 2 Mast 3

  13. Validation – Chinese wind farm 4. The CFD is used to calculate the wind speed at the turbines Turbine 2

  14. Validation – Chinese wind farm 5. The CFD is used to calculate the energy output at the turbines – including wake effects Turbine 2

  15. Conclusions & Outlook • A forecasting system has been set-up coupling Mesoscale, ANN and CFD • The system has been validated on a Chinese wind farm in complex terrain • Giving an ANN corrected wind speed delivers improved energy calculation when using CFD compared to using mesoscale data directly • Validation is on going with production data from wind farms in Europe • Searching for partners/users who want to validate the system

More Related