際際滷

際際滷Share a Scribd company logo

Comparison and Terrain Influence on Predictions with Linear and CFD Models

Michel Carreau
Michel Carreau

Comparison between linear (WAsP) and CFD model (Meteodyn) on a potential project RANS equation with one-equation closure scheme (k-L turbulence model) Project covers an area of 11km x 8km Equipped with 12 meteorological masts (recording from 6 months to 6 years of data) Relatively complex (deep valleys, ridges, rolling mountains) Mix of coastal and inland areas

Convert to study materialsBETA

Transform any presentation into ready-made study materialselect from outputs like summaries, definitions, and practice questions.

1 of 20
Downloaded 30 times
Comparison and Terrain Influence on Predictions with Linear and CFD Models CANWEA Annual Conference, Vancouver, BC October 04, 2011 GILLES BOESCH, M.Eng, Wind Project Analyst Hatch (Montreal), Canada
Overview Introduction Presentation of a test case Model comparison, terrain influence Conclusions and investigations 2
Introduction CFD is now well established in the wind industry Need to quantify the uncertainty associated to these models Compare the errors with linear models Influence of the errors with topography complexity And how to deal with it 3
Test case Comparison between linear (WAsP) and CFD model (Meteodyn) on a potential project RANS equation with one-equation closure scheme (k-L turbulence model) Project covers an area of 11km x 8km Equipped with 12 meteorological masts (recording from 6 months to 6 years of data) Relatively complex (deep valleys, ridges, rolling mountains) Mix of coastal and inland areas 4
Test case Altitude RIX Masts (m) (%) Forest diversity: M1 540 10.1 Logged area M2 560 11.0 M3 421 22.4 15m high trees M4 420 17.9 Regrowth M5 448 15.1 M6 521 16.6 RIX (Ruggedness Index) M7 560 8.0 % of slopes >30% in a 3500m radius M8 433 22.1 RIX Variations: M9 440 11.8 2 to 25 over the entire project M10 665 14.3 M11 567 2.7 2.7 to 22.4 at the meteorological masts M12 540 12.1 Variety of conditions to evaluate the behavior of the models 5
Test Case Meteodyn settings : Structured Mesh (30m cell size within the project area) Use of a forest model (windflow over canopy) Neutral stability class assumed (can induce errors for sea shore sites) Resulting shear verified for some masts Data : Measured and Quality controlled At 50m or 60m high (to avoid extrapolation errors) Adjusted to long term with standard MCP method (to have the same reference) 6
Results Methodology Cross-Prediction Matrix Predictors : Mast that predicts the others Predicted : Wind Speed at the 束 Predicted Mast 損 Predicted M1 M2 M3 M12 M1 M1 measured M1 predicts M2 M2 M2 predicts M2 measured Predictor M1 M3 M3 measured M12 M12 measured 7
Results - Methodology Cross-Prediction Matrix 12 x 12 matrix 132 cross predictions For both WAsP and Meteodyn No correction is applied to both models output Correction often applied with WAsP because of wind speed inconsistencies in complex terrain Converted into a Relative Error Matrix : V predicted Vmeasured %E Vmeasured Resulting in 132 relative error values for each cross-prediction 8
Altitude Masts RIX (%) (m) M1 540 10.1 M2 560 11.0 Results - Comparison M3 M4 M5 M6 421 420 448 521 22.4 17.9 15.1 16.6 M7 560 8.0 M8 433 22.1 M9 440 11.8 M10 665 14.3 Mean absolute errors M11 567 2.7 M12 540 12.1 Prediction Errors 25.0% 20.0% 15.0% Error (%) WAsP Error 10.0% Meteodyn Error 5.0% 0.0% M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 < 2km from water 9
Results - Comparison Absolute errors (direct output from models) WAsP Meteodyn Min Error 0.0% 0.0% Max Error 34.0% 14.1% Average 7.7% 4.6% On average, reduction of the error by 40%. Some exceptions : 33 cases out of 132 show better results with WAsP 10
Results - Comparison Generally, errors from both models have the same sign (positive/negative) 40.0% 30.0% 20.0% Relative Error (%) WAsP 10.0% Meteodyn 0.0% -10.0% -20.0% The difference is in the magnitude 11
Results RIX Analysis RIX dependency: WAsP : Error increase sharply when RIX > 15% Meteodyn : Error is more constant RIX influence on cross-prediction errors 25.0% 20.0% Average Error (%) 15.0% Wasp Meteodyn 10.0% 5.0% 0.0% 0.0 5.0 10.0 15.0 20.0 25.0 RIX (%) 12
Results RIX Analysis RIX dependency: Possibility to correct WAsP with RIX (between 2 masts) Correction based on a correlation between logarithmic error and RIX for each cross- prediction : E(%) = A* RIX + B Can we correct Meteodyn based on the RIX ? Error vs dRIX - Meteodyn Error vs dRIX - Wasp 40.0% 40.0% 30.0% y = 0.5552x 30.0% y = 1.0632x R族 = 0.6345 R族 = 0.7025 20.0% 20.0% Error (%) Error (%) 10.0% 10.0% 0.0% 0.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -10.0% -10.0% -20.0% -20.0% -30.0% -30.0% RIX (%) RIX (%) 13
Results RIX Analysis CFD RIX dependency: Error increases when RIX increases Error and RIX seem to be correlating (not as good than Wasp however) The slope is lower for Meteodyn Influence of site topography differences is lower Error vs dRIX - Meteodyn Error vs dRIX - Wasp 40.0% 40.0% 30.0% y = 0.5552x 30.0% y = 1.0632x R族 = 0.6345 R族 = 0.7025 20.0% 20.0% Error (%) Error (%) 10.0% 10.0% 0.0% 0.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -10.0% -10.0% -20.0% -20.0% -30.0% -30.0% RIX (%) RIX (%) 14
Results RIX Analysis Wasp RIX Correction: 12 towers available Equation based on 11 towers and evaluate how it corrects the 12th tower Prediction Errors 25.0% 20.0% 15.0% Error (%) WAsP Error Meteodyn Error 10.0% WAsP RIX Corrected Error 5.0% 0.0% M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 15
Results RIX Analysis Meteodyn RIX Correction: Same methodology with updated correction equation Prediction Errors 25.0% 20.0% 15.0% WAsP Error Error (%) WAsP RIX Corrected Error Meteodyn Error 10.0% Meteodyn RIX Corrected Error 5.0% 0.0% M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 16
Results RIX Analysis Summary of average error: Wasp 7.7 % Wasp RIX Corrected 4.3 % Meteodyn 4.6 % Meteodyn RIX Corrected 3.1 % RIX correction with Meteodyn produces promising results Reduction by 44% of the error after correcting Wasp with the RIX. Reduction by 33% of the error after correcting Meteodyn with the RIX. RIX correction with Wasp compared to Meteodyn direct output shows similar errors. 17
Conclusions In general, a project in complex terrain requires lots of masts An alternative is the use of a CFD model but linear corrected models can give good results too Only few litterature over relation between RIX and CFD models But quantification of CFD errors is more complex (topography / volume discretisation, forest model etc.) In some cases error is bigger 18
Conclusions To go further : Try with concurrent data (when possible) to avoid MCP related errors How does RIX correction with CFD performs for other sites ? Introduction of new complexity index (takes into account RIX, distance, vegetation, stability) 19
Thank you for your attention Gilles Boesch, M.Eng Wind Project Analyst Hatch Ltd GBoesch@hatch.ca 20

More Related Content

Comparison and Terrain Influence on Predictions with Linear and CFD Models

  • 1. Comparison and Terrain Influence on Predictions with Linear and CFD Models CANWEA Annual Conference, Vancouver, BC October 04, 2011 GILLES BOESCH, M.Eng, Wind Project Analyst Hatch (Montreal), Canada
  • 2. Overview Introduction Presentation of a test case Model comparison, terrain influence Conclusions and investigations 2
  • 3. Introduction CFD is now well established in the wind industry Need to quantify the uncertainty associated to these models Compare the errors with linear models Influence of the errors with topography complexity And how to deal with it 3
  • 4. Test case Comparison between linear (WAsP) and CFD model (Meteodyn) on a potential project RANS equation with one-equation closure scheme (k-L turbulence model) Project covers an area of 11km x 8km Equipped with 12 meteorological masts (recording from 6 months to 6 years of data) Relatively complex (deep valleys, ridges, rolling mountains) Mix of coastal and inland areas 4
  • 5. Test case Altitude RIX Masts (m) (%) Forest diversity: M1 540 10.1 Logged area M2 560 11.0 M3 421 22.4 15m high trees M4 420 17.9 Regrowth M5 448 15.1 M6 521 16.6 RIX (Ruggedness Index) M7 560 8.0 % of slopes >30% in a 3500m radius M8 433 22.1 RIX Variations: M9 440 11.8 2 to 25 over the entire project M10 665 14.3 M11 567 2.7 2.7 to 22.4 at the meteorological masts M12 540 12.1 Variety of conditions to evaluate the behavior of the models 5
  • 6. Test Case Meteodyn settings : Structured Mesh (30m cell size within the project area) Use of a forest model (windflow over canopy) Neutral stability class assumed (can induce errors for sea shore sites) Resulting shear verified for some masts Data : Measured and Quality controlled At 50m or 60m high (to avoid extrapolation errors) Adjusted to long term with standard MCP method (to have the same reference) 6
  • 7. Results Methodology Cross-Prediction Matrix Predictors : Mast that predicts the others Predicted : Wind Speed at the 束 Predicted Mast 損 Predicted M1 M2 M3 M12 M1 M1 measured M1 predicts M2 M2 M2 predicts M2 measured Predictor M1 M3 M3 measured M12 M12 measured 7
  • 8. Results - Methodology Cross-Prediction Matrix 12 x 12 matrix 132 cross predictions For both WAsP and Meteodyn No correction is applied to both models output Correction often applied with WAsP because of wind speed inconsistencies in complex terrain Converted into a Relative Error Matrix : V predicted Vmeasured %E Vmeasured Resulting in 132 relative error values for each cross-prediction 8
  • 9. Altitude Masts RIX (%) (m) M1 540 10.1 M2 560 11.0 Results - Comparison M3 M4 M5 M6 421 420 448 521 22.4 17.9 15.1 16.6 M7 560 8.0 M8 433 22.1 M9 440 11.8 M10 665 14.3 Mean absolute errors M11 567 2.7 M12 540 12.1 Prediction Errors 25.0% 20.0% 15.0% Error (%) WAsP Error 10.0% Meteodyn Error 5.0% 0.0% M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 < 2km from water 9
  • 10. Results - Comparison Absolute errors (direct output from models) WAsP Meteodyn Min Error 0.0% 0.0% Max Error 34.0% 14.1% Average 7.7% 4.6% On average, reduction of the error by 40%. Some exceptions : 33 cases out of 132 show better results with WAsP 10
  • 11. Results - Comparison Generally, errors from both models have the same sign (positive/negative) 40.0% 30.0% 20.0% Relative Error (%) WAsP 10.0% Meteodyn 0.0% -10.0% -20.0% The difference is in the magnitude 11
  • 12. Results RIX Analysis RIX dependency: WAsP : Error increase sharply when RIX > 15% Meteodyn : Error is more constant RIX influence on cross-prediction errors 25.0% 20.0% Average Error (%) 15.0% Wasp Meteodyn 10.0% 5.0% 0.0% 0.0 5.0 10.0 15.0 20.0 25.0 RIX (%) 12
  • 13. Results RIX Analysis RIX dependency: Possibility to correct WAsP with RIX (between 2 masts) Correction based on a correlation between logarithmic error and RIX for each cross- prediction : E(%) = A* RIX + B Can we correct Meteodyn based on the RIX ? Error vs dRIX - Meteodyn Error vs dRIX - Wasp 40.0% 40.0% 30.0% y = 0.5552x 30.0% y = 1.0632x R族 = 0.6345 R族 = 0.7025 20.0% 20.0% Error (%) Error (%) 10.0% 10.0% 0.0% 0.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -10.0% -10.0% -20.0% -20.0% -30.0% -30.0% RIX (%) RIX (%) 13
  • 14. Results RIX Analysis CFD RIX dependency: Error increases when RIX increases Error and RIX seem to be correlating (not as good than Wasp however) The slope is lower for Meteodyn Influence of site topography differences is lower Error vs dRIX - Meteodyn Error vs dRIX - Wasp 40.0% 40.0% 30.0% y = 0.5552x 30.0% y = 1.0632x R族 = 0.6345 R族 = 0.7025 20.0% 20.0% Error (%) Error (%) 10.0% 10.0% 0.0% 0.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -30.0% -20.0% -10.0% 0.0% 10.0% 20.0% 30.0% -10.0% -10.0% -20.0% -20.0% -30.0% -30.0% RIX (%) RIX (%) 14
  • 15. Results RIX Analysis Wasp RIX Correction: 12 towers available Equation based on 11 towers and evaluate how it corrects the 12th tower Prediction Errors 25.0% 20.0% 15.0% Error (%) WAsP Error Meteodyn Error 10.0% WAsP RIX Corrected Error 5.0% 0.0% M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 15
  • 16. Results RIX Analysis Meteodyn RIX Correction: Same methodology with updated correction equation Prediction Errors 25.0% 20.0% 15.0% WAsP Error Error (%) WAsP RIX Corrected Error Meteodyn Error 10.0% Meteodyn RIX Corrected Error 5.0% 0.0% M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M11 M12 16
  • 17. Results RIX Analysis Summary of average error: Wasp 7.7 % Wasp RIX Corrected 4.3 % Meteodyn 4.6 % Meteodyn RIX Corrected 3.1 % RIX correction with Meteodyn produces promising results Reduction by 44% of the error after correcting Wasp with the RIX. Reduction by 33% of the error after correcting Meteodyn with the RIX. RIX correction with Wasp compared to Meteodyn direct output shows similar errors. 17
  • 18. Conclusions In general, a project in complex terrain requires lots of masts An alternative is the use of a CFD model but linear corrected models can give good results too Only few litterature over relation between RIX and CFD models But quantification of CFD errors is more complex (topography / volume discretisation, forest model etc.) In some cases error is bigger 18
  • 19. Conclusions To go further : Try with concurrent data (when possible) to avoid MCP related errors How does RIX correction with CFD performs for other sites ? Introduction of new complexity index (takes into account RIX, distance, vegetation, stability) 19
  • 20. Thank you for your attention Gilles Boesch, M.Eng Wind Project Analyst Hatch Ltd GBoesch@hatch.ca 20