VindKraftNet Mesoscale Workshop 3 March 2010, Vestas Technology HQ, Århus, Denmark Linking mesocale modelling to site conditions Jake Badger, Andrea Hahmann, Xiaoli Guo Larsen, Claire Vincent, Caroline Draxl, Joakim Refslund Nielsen, Ebba Dellwick, Jesper Nissen, Alfredo Peña Diaz, Niels Gylling Mortensen
Introduction Risø DTU mesoscale modelling group 3 PhD student-staff 5 scientist-staff immersed in the Meteorology Programme 30 scientists working in areas of boundary layer meteorology; resource estimation, modelling, extremes turbulence, measurements, remote sensing immersed in the Wind Energy Division 100+ scientists and engineers working in wide of disciplines, aerodynamics, CFD, structural engineering, electrical engineering Carry out research projects and commercial tasks Links within the themes, and disciplines is reflected in application of the modelling activities, and will be ever more so in the future
Introduction Running themes: Mesoscale modelling methods Application to site conditions Climates mean wind climate, wind resources Extreme wind climate Forecasting Understanding model behaviour Data assimilation Variability Microscale modelling Aggregation Future and conclusions
Wind resource map Wind resource map for Mali 50 m a.g.l. using KAMM/WAsP Classic picture of a wind resource map. What can I do with this? What more would I like to do with this? Maps are not the final product
Generalizing wind climates Post-processing Example: simulated wind wind corrected to standard conditions flat terrain with homogeneous roughness higher roughness + orographic speed-up low roughness higher roughness ~30km
Generalizing wind climates Post-processing Figures below show WRF domain generalized wind for 50 m a.g.l. where hypothetical mesoscale model wind is uniform 10 m/s everywhere. wind direction 90 deg wind direction 270 deg 500 450 400 350 300 250 200 150 100 50 0 0 50 100 150 200 250 300 350 400 450 500 15 14.5 14 13.5 13 12.5 12 11.5 11 10.5 10 9.5 9 8.5 8 7.5 7 500 450 400 350 300 250 200 150 100 50 0 0 50 100 150 200 250 300 350 400 450 500
Two main approaches statistical-dynamical downscaling Wind Classes Generalized wind at 50 m and 3cm roughness length
Two main approaches dynamical downscaling Example: WRF mean wind speed [m/s] at 80 m a.g.l. 1999-2009 (11 years) x = 45 km x = 15 km
Application of mesoscale modelling results In 2009 generated over 1,000,000 geo-referenced WAsP generalize wind atlases: WAsP Indian Wind Atlas Finnish Wind Atlas (Risø downscaling method applied to FMI s mesoscale simulations) Development towards universal interface between mesoscale models to microscale models.
Extreme wind atlases selective episode dynamical downscaling Storm 1 Storm 2 +,,+ storm n Annual 50-year wind Maximum Method Apply post-processing to results to get generalized extreme wind climate data
Extreme wind atlases statistical-dynamical downscaling and application Extreme wind classes WAsP Engineering
Forecasting Developed a real-time wind (weather) forecast system for Denmark. System operational since March 2009 running twice daily on Risø linux clusters. Serves to important roles WRF topography and configuration Δx = 18, 6 and 2km provide source of forecast data for measurement campaigns and for retrospective comparisons learning about the model characteristics Forecast potential wind energy at 80 m a.g.l.
Forecasting Similar systems running for South Africa, Brazil, and India India South Africa Brazil
Verification against Høvsøre data Averaged diurnal cycle of various quantities U 10 U 100 u * heat flux Comparison of profiles from the flat and homogeneous sector (minimize site effects) at Høvsøre
Learning about the model behaviour day night Verification of WRF real-time wind forecast system over Denmark Profiles too neutral at night? June October 10-meter wind speed mean error (WRF-OBS) for the months of June and October 2009
Data assimilation: initial conditions PhD Caroline Draxl Other causes of forecast error: Initial conditions influence the forecasts physics Initial conditions forecast Observation network and data assimilation give initial conditions for forecast model
Data assimilation: initial conditions PhD Caroline Draxl How best to incorperate new observational data into data assimilation? physics Initial conditions forecast Data Assimilation Project: include additional observations to improve the wind forecasts in the short range
Variability PhD Claire Vincent Fluctuations in wind speed over the North Sea 9 m/s Prolonged episodes of fluctuating wind speeds are sometimes observed at Horns Rev 10 m/s Large and sudden changes in wind speed mean sudden changes in power output: Deficit needs to be filled with reserve power. 10 m/s 9 m/s
Variability PhD Claire Vincent Open cellular convection over the North Sea United Kindom Denmark Open cellular convection is an example of a mesoscale feature that contributes to wind fluctuations on time scales of tens of minutes to hours. It is mostly an offshore phenomena cell walls horizontal wind fluctuations
Variability PhD Claire Vincent Can mesoscale models capture mesoscale wind fluctuations? WRF produces open cells with the right dimensions. Total outgoing longwave radiation (like what the satellite would see) Vertical velocity Cell walls
Aggregation Microscale to mesoscale PhD Joakim Refslund Nielsen Micro-scale flow effects 1. Internal boundary layers 2. Edge effects How can these be upscaled into parameterizations within mesoscale models? A number of microscale models will be employed to aggregate the effects of sub-grid scale roughness change. different land use different roughness length z 0, skog 2 z 0,skov z 0, sø z 0,mark z 0, gras
Application and aggregation of generalized wind climates
Application and aggregation of generalized wind climates
Conclusions Mesoscale modelling plays a wide and exciting role in wind energy Mesoscale modelling results applicability, usefulness, and accuracy can be greatly enhanced by post-processing to allow for extension of modelling to the microscale this post-processing step provides the link to actual site conditions Forecasting has dual role providing useful data (i.e. short term wind power prediction) test driving and learning the characteristics of the model itself Outlook Role will expand as dialogue opens between those parties with requirements/demands and the mesoscale modellers Microscale models should exploit more of the information from the mesoscale models Microscale modelling based on mesoscale results will provide important aggregated parameters for Integrated Assessment Models for long term planning for renewable energy
Thank you for your attention jaba@risoe.dtu.dk