1

Project Title: Development of a High-Resolution Virtual Wind Simulator for Optimal Design of Wind Energy Projects Contract Number: RD3-42 Milestone Number: 7 Report Date: 12/17/12

Principal Investigator: Fotis Sotiropoulos Contract Contact: Bridget Foss

fotis@umn.edu 612-624-5571

Congressional District: (Corporate office) Minnesota 5th

Congressional District: (Project location) Minnesota 5th

MILESTONE REPORT Executive Summary: This project aims at developing a ‘Virtual Wind Simulator’ (VWS) for the prediction of atmospheric boundary layer flow and its interactions with wind turbines and wind farms. The use of the simulator will assist in the improved design of potential wind energy projects by providing more accurate predictions of local and wind turbulence at site and turbine levels. Additionally, the VWS will help increase the level of wind energy utilization and reduce the cost of energy production. Computational Fluid Dynamics (CFD) methods are used in this project to develop a computational framework for conducting high-resolution simulations of wind turbulence at the meso and micro scales. In particular, the Large-Eddy Simulation (LES) technique will allow for accurate simulations of the turbulent flow at spatial resolutions as small as one to ten meters, and temporal resolutions of just a few seconds. Parameterizations for wind turbine forces will also be developed in the LES framework. In addition, three-dimensional, time-evolving flow fields obtained from LES at any location within a potential wind farm site could then be used as the inflow condition for even more detailed simulations of the turbulent flow around the blades of specific wind turbines using a hybrid Reynolds-Averaged Navier-Stokes (RANS)/LES technique. The SAFL computational models will be coupled to macro-scale regional models to develop a powerful multi-scale computational tool, the VWS. The VWS will integrate the latest advancements in computational fluid dynamics research and provide reliable, high-resolution descriptions of wind turbulence across the entire range of scales that are relevant to wind power production. This information will provide objective, scientifically

Twin Cities Campus Saint Anthony Falls Laboratory Engineering, Environmental, Biological and Geophysical Fluid Dynamics College of Science and Engineering

Mississippi River at 3rd Avenue S.E. Minneapolis, MN 55414 Dept. Main Office: 612-624-4363 Fax: 612-624-4398

2

based criteria that can be used by wind energy project developers for the site-specific, optimal selection and placement (micro-siting) of wind turbines. As planned, during this reporting period activities have been carried out that address the following objectives: (a) The computational fluid dynamics (CFD) framework for studying wind-turbine interaction is completed. (b) We have performed the (one-way) coupling and integration of two-level models. (c) Application of the VWS has been performed including the meteorological modeling. (d) We have started the on-site assessment of the VWS and the comparative study. Project funding provided by customers of Xcel Energy through a grant from the Renewable Development Fund.

Task 1. Development of the Virtual Wind Simulator for high-resolution simulations

of wind turbulence and their effect on energy production Progress has continued on the development and testing of the Virtual Wind Simulator. Our current efforts are focused on: 1) turbine wake interactions for staggered wind farms; 2) one-way coupling of WRF model (Weather Research and Forecasting Model) with LES-CURVIB model (Large-eddy simulation-curvilinear immersed boundary model); and 3) wind field simulation at the Prairie Island accounting for the effects of complex topography. We continue the investigation of turbine-turbine interactions. The spacing effects in staggered wind farms are studied and compared with the corresponding aligned wind farms. Different from aligned wind farms, the wake from one turbine has complex interactions with its downstream wind turbine wakes. Figure 1 shows the contours of the mean streamwise velocity at hub height. Figures 1(a) to 1(b) show the contours for staggered wind farms. Figures 1(e) to 1(f) show the contours for the aligned wind farms, which has the same turbine occupied area with case shown in figure 1(c).

3

Figure 1: Contours of mean streamwise velocity at hub height for staggered wind farms with comparions with aligned wind farms.

Following analysis of its high resolution modeling of the Mower County wind farm, WindLogics provided model output to the University of Minnesota St. Anthony Falls Laboratory (SAFL) for input to the higher resolution, turbulence-resolving Virtual Wind Simulator (VWS). An example of the data provided is displayed graphically in the colored fields Figs. 2 and 3. The simulations serve two purposes. The first is to model the larger scale weather patterns that affect the wind farm as laid out in Figure 3. It is necessary to account for these patterns because they are the main factor determining wind speed and direction at the smaller scales. Once the wind speeds and direction are known, finer, turbulence-resolving simulations can be performed to understand how atmospheric boundary layer turbulence and turbine wakes interact as a result of the larger scale patterns.

4

Figure 2: Comparison between wind speed at 80 meters above ground level in a high resolution modeling run (colored field—see color bar to the left for scale) and SCADA observations of wind speed (colored squares) for July 2, 2010 and 2220 UTC. The smaller, dark points are turbines from neighboring wind farms, which have been built since the Xcel RDF grant funding this research was provided. The arrows from the colored squares denote the wind direction from the SCADA observations.

Figure 3: The large scale weather patterns associated with the selected period of modeling at Mower County. The colored fields show temperature (K) and the contours show sea level pressure (hPa).

5

The data from the larger scale modeling is used for initial conditions (start time) of the VWS simulation (see Fig. 4 for an example). A time series of lateral boundary conditions is needed, and vertical slices of the WindLogics high resolution modeling, provided at 10-minute intervals, and taken at the edges of the VWS domain, provide the inflow and lateral boundary conditions for the flow in the VWS. With this sort of configuration, the flow of information is from the exterior (large scale) domain (modeled by WindLogics) to the finer resolution domain (VWS) and completely one-way—that is, from the larger scales of motion to the smaller scales. There is no possibility for the smaller-scale information from the VWS to feed back to the larger scale, exterior domain. The best configurations of lateral boundary conditions to employ for turbulence-resolving simulations are still an open academic question. In particular, allowing resolved turbulence to develop from lateral boundary conditions, which are coarser and do not resolve this turbulence, is a particularly difficult problem. Additional runs are being conducted to provide the lateral conditions for all variables (including turbulence) that allow this transition to occur most smoothly.

Figure 4: Impact of large scale weather patterns on the wind farms showing colored fields of potential temperature (Kelvin) and wind direction (arrow). Colored squares denote wind speed measured at the turbine locations within the Mower County wind farm.

6

As pointed, the present VWS for multi-scale modeling uses one-way coupling. As shown in Figure 5, microscale simulations using LES-CURVIB model are carried out in the region of the wind farm and the mesoscale meteorological simulations using WRF model are carried out in the whole domain. The boundary conditions for the microscale simulations are from the mesoscale simulation, while the microscale simulations have no effects on the mesoscale simulations.

Figure 5: Schematic of the multi-scale modeling of VWS.

The velocity and temperature boundary conditions at the inlet of LES region are interpolated spatially and temporally from the WRF solution. The turbulence scales explicitly resolved by WRF are much larger that in LES because the grid spacing employed in WRF is usually ten or twenty times larger than that used in LES. To reconstruct the small turbulence scales for inflow velocities of the LES, a synthetic turbulence generation technique developed by Mann (Prob. Engng. Mech., Vol. 13, No. 4, pp. 269-282, 1998) is used. In Figure 6, we show the synthetic turbulent flow field generated from this technique. From top to below, we show the contours of the streamwise, spanwise and wall-normal velocity fluctuations at an x-z plane (where x and z represent the streamwise and wall-normal directions, respectively).

7

Figure 6: Synthetic velocity fluctuations.

Using this multi-scale modeling for our VWS, we simulate the Mower County wind farm and performed comparison with field measurements for the period ranging from 6 am to 8 am on July 2, 2010. In Figure 7, we show the computational domains for this case. The left image shows the computational domain for WRF with the size of approximate 50 kilometers. The right one shows the computational domain for LES with the size of approximate 2 kilometers. The WRF simulation is carried out by Robert J Conzemius at WindLogics Inc. The grid spacings for WRF simulations are approximately 170 and 20 meters in the horizontal and wall-normal directions, respectively. The grid spacings employed in LES are 20 and 9 meters in the horizontal and wall-normal directions, respectively. There are five turbines within this area as shown in the LES-domain of Figure 4. The field measurements were taken at two locations s1 and s2. Actuator disk parameterizations are used for the wind turbine modeling. The atmospheric boundary layer was stably stratified during the period of analysis. Those stratification effects are considered in the simulations, where a wall model for temperature (Phys. Fluids, 23, 126603 (2011)) is used.

8

 

Figure 7: Computational domains for multi-scale simulation.

Figure 8 illustrates the contours of the instantaneous streamwise velocity at hub height, in which the left and right figures show the computed results without and with adding synthetic turbulence, respectively. Similarly, Figure 9 illustrates contours of the mean streamwise velocity at hub height, where the left and right figures show the computed results without and with adding synthetic turbulence, respectively. It is seen that the wakes from turbines recover faster for the case with added synthetic turbulence.

Figure 8: Instantaneous flow field at hub height for left figure without added synthetic turbulence and right figure with added synthetic turbulence.

9

Figure 9: Mean flow field at hub height with (right) and without synthetic turbulence.

Figure 10 shows the comparison of the computed mean streamwise velocity profiles with the field measurements. Very good agreements are obtained for the measurements at location S1, which is not located within the wind turbines wakes. For location S2 which is located within the wake of turbine T41, some deviations are observed. This is probably because the actuator disk model for wind turbine.

Figure 10: Comparison of the computed mean streamwise velocity profiles with field measurements.

The simulated power is also compared with that from field measurements as shown in Table 1. Overall the computed power output is within the same level with the measurements. The largest discrepancy happens for turbine T42, which is in the wake of turbine 41. Using inflow with added turbulence does improve the predictions. However, the prediction still deviates from the measurements significantly.

10

Table 1: Comparison of the mean power output

T39 (MW) T40 (MW) T41 (MW) T42 (MW) T43 (MW)Measurements 2.15 2.14 1.86 1.85 LES (With added turbulence)

2.42 2.22 2.30 0.95 2.07

LES (Without added turbulence)

2.38 2.25 2.07 0.79 1.94

The computation using actuator line model is underway. Two-way coupling, which requires more complex coupling strategies and programming efforts, will be considered in future work. Task 4. Virtual wind simulator Application (Wind Assessment at Undeveloped Site with Complex Terrain) The present VWS is also applied to simulate the effects of the complex topography for the wind field at the Prairie Island. Figure 11 shows the contours of elevation at the Prairie Island and the domain for LES computation. For the selected time period, the wind is from southeast. Figure 12 shows the contours of the instantaneous streamwise velocity at different downstream locations. Further comparison with field measurements is underway.

Figure 11: Contours of elevation at the Prairie Island site. The two points highlighted in the figure show the locations of the Sodars. The two arrows show the directions of the mean wind velocity during the selected time period. The region enclosed by the white rectangle is the domain for LES computation.

11

Figure 12: Contours of instantaneous streamwise velocity at different downstream locations.

Field data provided by WindLogics (SODAR and SCADA) is being used for comparison between the VWS and the observations (an example of the data is plotted on Figs. 3 and 4). Additionally, the numerical modeling output provided by WindLogics will be used for direct comparisons between the VWS and more traditional, larger scale modeling methods. WindLogics is also beginning modeling for comparisons between VWS runs performed at Prairie Island and existing, larger-scale numerical weather models. WindLogics earlier provided SODAR and meteorological tower data from Prairie Island to SAFL for initializing runs of the VWS. During the next two months, WindLogics will be providing model output to SAFL to compare with VWS results.

Additional Milestones: Work is in progress towards Milestone 8.

12

Project Status: The project is ahead of schedule due to the fact that work on the project started before the contract was finalized.

LEGAL NOTICE THIS REPORT WAS PREPARED AS A RESULT OF WORK SPONSORED BY NSP. IT DOES NOT NECESSARILY REPRESENT THE VIEWS OF NSP, ITS EMPLOYEES, OR THE RENEWABLE DEVELOPMENT FUND BOARD. NSP, ITS EMPLOYEES, CONTRACTORS, AND SUBCONTRACTORS MAKE NO WARRANTY, EXPRESS OR IMPLIED, AND ASSUME NO LEGAL LIABILITY FOR THE INFORMATION IN THIS REPORT; NOR DOES ANY PARTY REPRESENT THAT THE USE OF THIS INFORMATION WILL NOT INFRINGE UPON PRIVATELY OWNED RIGHTS. THIS REPORT HAS NOT BEEN APPROVED OR DISAPPROVED BY NSP NOR HAS NSP PASSED UPON THE ACCURACY OF ADEQUACY OF THE INFORMATION IN THIS REPORT. Appendix: List publications Papers in Refereed Journals 10. Chamorro, L.P., Arndt, REA and Sotiropoulos F. 2013. ‘Drag reduction in large

wind turbines through riblets: Evaluation of riblet geometry and application strategies’. Renewable Energy. 50, 1095-1105.

9. Yang X., Kang S. and Sotiropoulos F. 2012. “Computational study and modeling of turbine spacing effects in infinite aligned wind farms”. Phys. Fluids. 24, 115107.

8. Chamorro, L.P., Guala, M., Arndt, REA and Sotiropoulos F. 2012. ‘On the evolution of turbulent scales in the wake of a wind turbine model’. J. of Turbulence. 13, No.27, 1-13.

7. Chamorro, L.P., Arndt, REA and Sotiropoulos F. 2012. ‘Reynolds number dependence of turbulence statistics in the wake of wind turbines’. Wind Energy. 15, 733-742.

6. Chamorro, L.P., Arndt, REA and Sotiropoulos F. 2011. ‘Turbulent flow properties around a staggered wind farm’. Bound. Layer Meteorol. 141, 349-367.

5. Chamorro, L.P. and Arndt, REA. 2012 ‘Non-uniform velocity distribution effect on the Betz limit’ Wind Energy. DOI: 10.1002/we.549.

4. Wu, Y.-T., and Porté-Agel F. 2011. ‘Large-eddy simulation of wind-turbine wakes: Evaluation of turbine parameterizations’. Boundary-Layer Meteorology, 138:345–366.

13

3. Chamorro, L.P., and Porté-Agel F. 2010. ‘Thermal stability and boundary-layer effects on wind turbine wakes. A wind tunnel study’. Boundary-Layer Meteorology, 136, 489-513.

2. Lu, H., and Porté-Agel F. 2010. ‘A modulated subgrid-scale model for large-eddy simulation: Application to a neutral atmospheric boundary layer’. Physics of Fluids, 22(1), 015109.

1. Chamorro, L.P., and Porté-Agel F. 2009. ‘A wind tunnel investigation of wind turbines wakes: Boundary-layer turbulence and surface roughness effects’. Boundary-Layer Meteorology, 132, 129-149.

Conference Presentations 13. Chamorro L.P., Arndt R.E.A., Yang X. and Sotiropoulos F. 2012 Characterization of

the turbulent structure and transport processes in the wake of a wind turbine under different canonic configurations. Euromech Colloquium 528, Oldenburg, Germany.

12. Yang X., Sotiropoulos F. 2011“LES investigation of turbine spacing effects in wind farms”. The 64th Annual Meeting of the American Physical Society, Baltimore, Maryland.

11. Chamorro L.P., Arndt R.E.A. and Sotiropoulos F. (2011) Turbine layout effects on the flow structure inside an above large wind farms. American Physical Society, Baltimore, Maryland.

10. Chamorro, L.P., R.E.A. Arndt and F. Sotiropoulos, 2011. ‘Turbulence patterns around a large wind farm. 49th AIAA Aerospace Sciences Meeting, Orlando, Florida.

9. Chamorro, L.P., R.E.A. Arndt and F. Sotiropoulos, 2010. ‘Turbulence characteristics around a staggered wind farm configuration. A wind tunnel study’. American Physical Society Meeting. Minneapolis, Long Beach, CA.

8. Porté-Agel, F., Y.-T. Wu, and H. Lu, 2009. ‘Large-eddy simulation of wind-turbine wakes’. American Physical Society Meeting. Minneapolis, MN.

7. Chamorro, L.P., and F. Porté-Agel, 2009. ‘A wind tunnel investigation of wind turbine wakes’. American Physical Society Meeting. Minneapolis, MN.

6. Porté-Agel, F., F. Sotiropoulos, R. Conzemius, L. Chamorro, Y.-T. Wu, S. Behara, H. Lu, 2009. ‘Development of a high-resolution Virtual Wind Simulator for optimal design of wind energy projects’. E3 -Energy, Economic and Environmental- Conference. Minneapolis, MN.

5. Porté-Agel, F., Y.-T. Wu, Y.-T.., H. Lu, 2009. ‘Parameterization of turbulent fluxes and wind-turbine forces in large-eddy simulation’. European Geophysical Union. Vienna.

4. Chamorro, L.P., and F. Porté-Agel, 2009. ‘A wind tunnel investigation of wind turbine wakes: Boundary-layer turbulence and surface roughness effects’. European Geophysical Union. Vienna.

3. Wu, Y.-T.., H. Lu, and F. Porté-Agel, 2008. ‘Large-eddy simulation of wind-turbine wakes’. American Geophysical Union. San Francisco, CA.

2. Chamorro, L.P., and F. Porté-Agel, 2008. ‘A wind tunnel investigation of wind turbine wakes: Boundary-layer turbulence and surface roughness effects’. American Geophysical Union. San Francisco, CA.

1. Porté-Agel, F., F. Sotiropoulos, R. Conzemius, L. Chamorro, Y.-T. Wu, S. Behara,

14

H. Lu, 2008. ‘Development of a high-resolution Virtual Wind Simulator for optimal design of wind energy projects’. E3 -Energy, Economic and Environmental- Conference. Minneapolis, MN.