DFFV Forest workshop, Vattenfall, Fredericia, March 2008. Black Law wind farm in central Scotland (Fig [1]) is large, consisting of over 40 turbines in complex.
Wind Modelling Evaluation Using an Operational Wind Farm Site
PO.141
Roy Spence1, Christiane Montavon2, Ian Jones2, Chris Staples2, Callum Strachan3, David Malins3 1 SgurrEnergy Ltd., 2 ANSYS UK Ltd., 3 Scottish Power Renewables.
Abstract The occurrence of open areas suitable to the siting of wind farms is limited. As a consequence, more recent wind farms are being built in regions close to complex topography and/or forestry. Such features influence the wind regime characteristics around the wind farm and affect not only the wind farm energy yield, and uncertainties in that value, but also the load on the wind turbines. This preliminary study explores the use of three commercial software modelling packages (WAsP, Meteodyn and ANSYS-CFX), which range in numerical complexity and encompass the extent of wind farm modelling tools currently available to model an operational wind farm. Employing different software packages highlights the effectiveness of the different modelling approaches taken in each case. Particular attention is paid to the methods employed in the software packages for modelling the wind turbine generator wakes and the surrounding forestry. The output from these three models is compared to the operational performance of the wind farm.
Background Black Law wind farm in central Scotland (Fig [1]) is large, consisting of over 40 turbines in complex surroundings due to the presence of significant forestry. The study utilises data from two on-site meteorological masts to generate consistent data for the models; i.e. topographical data, forestry tree heights and datum wind regime at mast height and locations. Statistics from the on-site meteorological masts were fitted to sector-wise Weibull distributions for the first year of operation. The statistics demonstrated a typical UK wind resource with prevailing winds from the west and southwest, with average wind speeds at both masts in excess of 7m/s.
Methods Used The WAsP software is generally considered as the "industry standard" software package for the simulation of wind flow for onshore wind farms. It is based on a linearised approach, solving equations for the perturbations due to the effects of terrain. This software was not designed for the highly non-linear simulation of complex wind regimes arising from complex forestry and wakes but is included in this study to provide a comparison of Computational Fluid Dynamics (CFD) with the standard methods in WAsP, [1,2,3] e.g. the zero-plane-displacement-height method for roughness. SgurrEnergy has had success in using modified WAsP roughness values to model forestry-influenced wind flows [4], but these are not used here. Meteodyn is a 3D CFD software package designed to solve the full Reynolds-Averaged Navier-Stokes equations [5,6,7] using a rectangular grid topology and a single-equation turbulence model. Forest canopies are modelled through the introduction of a sink term at computational cells at ground level. The tree height is taken from roughness values (Z0*30), with a number of cells appropriate to the tree height having the sink term applied. Meteodyn analyses are conducted at 10° sector increments with grid sizes between 1.12 and 1.52 million cells depending on the sector direction. A synthesis process, using the reference wind rose and model simulation results, provides the final output. ANSYS CFX is a state-of-the art CFD package, tailored for the wind power community through the WindModeller front end. As standard it uses 2-equation turbulence models, with the Shear Stress Transport (SST) model, Menter [8], an extension of the wellknown k – ε model to better account for separated flows, as default. It has an advanced wake model based on actuator disk theory, [9], and a non-linear forestry volumetric resistance model. The model of Lopes da Costa [10] was used for the forest resistance and turbulence. An adaptive gridding approach was used to automatically refine the region around each wind turbine. This approach has been previously used on Black Law wind farm [9] to successfully predict wind speeds at each turbine hub, for a fixed reference wind speed and direction. The current study builds on this work, to predict the capacity factor of the wind farm. Actuator disk theory is also used to obtain the „undisturbed‟ flow upstream of each turbine, which is used with the manufacturer‟s power curve, to get the power from the wind turbine for a given speed.
Figure 1: Map of Black Law. Map Image: Ordnance Survey © Crown Copyright 2008, License number 100048580
Results The different approaches have been applied to the same data as consistently as possible. Normalised capacity factors are shown in Fig 5. These show similar trends between the models and the data, with significant effects from the wakes and forestry. The ANSYS CFX model shows much greater fall-off in power from forestry and wakes, and is in closer agreement with the operational data for those turbines most affected. However, it does show greater power generated from the better performing turbines. This may be consistent with the omission of turbine generated turbulence in its wake model, which would tend to increase the spread of the wakes, reducing the wake length. There are also other details which may have a bearing: some turbines were down for a part of the period of the study, corrections applied to the met data for wake effects, forestry density and assumed free stream turbulence levels.
Figure 5: Normalised Capacity Factors: Measured and Predicted .
Figure 2 shows the best and worst performing turbines. The results clearly indicate that the worst performing turbines are located close to the forestry areas. One wind turbine shows a significant efficiency reduction for almost all wind directions due to a combination of low hub height and position close to forested areas. This is well captured in the ANSYS CFX results.
Conclusions A preliminary study has been carried out to explore the use of three different modelling packages to model an operational wind farm with significant forestry and wake effects. These models were applied in a consistent fashion as far as possible, to predict the capacity factor of the wind farm over approximately one year. All models show very similar overall trends but with some differences between them. The more advanced predictive model in ANSYS CFX represented well the drop off in performance of the more poorly performing turbines, while over-predicting the power from the better turbines, consistent with the omission of turbulence generated by the turbines. Given that there are still differences in the underlying assumptions in the approaches, which may influence the comparison, then a more detailed study would be worthwhile, focussing on the sensitivity of the results to the key assumptions.
References
3a) 3 b) Figure 3: ANSYS CFX a) Background grid b) adapted mesh in the region of the wind turbines.
Figure 2: Forestry representation and location of the wind turbines and met masts. The five best performing turbines are circled in green, the five worst are circled in red, and the next five worst are circled in orange.
4a) 4b) Figure 4: Example ANSYS CFX results a) wind speed and b) turbulence intensity, reference speed 6 m/s from sector 210.
1. http://www.wasp.dk/ 2. http://www.risoe.dk/vea/storpark/presentations/WAsP8%20Wakeeffect%20model.pdf 3. Dellwik, E, Landberg, L. And Jensen, N O. “WAsP in the forest”, European Wind Energy Conference and Exhibition 2004, London, England, 22-25 November. 4. Boddington, R, “The Effect of Forestry on Wind Regime and Energy Yield”. BWEA Tree Workshop 2006 5. http://www.meteodyn.com/en/software/meteodyn-wt.html 6. http://www.dffv.dk/vindkraftworkshop%202008/Aur%C3%A9lien%20Chante lot%20%20Meteodyn%20WT's%20approach%20to%20forest%20modelling.pdf 7. http://www.ewec2007proceedings.info/allfiles2/409_Ewec2007fullpaper.pdf 8. Menter, F. R, "Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications", AIAA Journal, vol. 32, pp. 269-289, 1994. 9. Montavon C., Jones, I. Staples. C., Strachan, C., Gutierrez, I., Practical issues in the use of CFD for modelling wind farms, EWEC Proceedings, Marseille, 2009 10. Stuart, P., Hunter, I., Lopes da Costa, J.C.P., da Palma, J. M.L.M. Wind Flow Over Forested Hills: Mean Flow and Turbulence Characteristics, DFFV Forest workshop, Vattenfall, Fredericia, March 2008.
European Wind Energy Conference & Exhibition 2010, Tuesday 20 - Friday 23 April 2010, Warsaw, Poland