VIEWS: 296 PAGES: 10 CATEGORY: Technology POSTED ON: 1/3/2010 Public Domain
Influence of wind characteristics on turbine performance Ioannis Antoniou (1), Rozenn Wagner (1), Søren M. Pedersen (1), Uwe Paulsen (1), Helge A. Madsen (1), Hans E. Jørgensen (1), Kenneth Thomsen (2), Peder Enevoldsen (2), Leo Thesbjerg (3) (1): Wind Energy Department, Risø,DTU, Frederiksborgvej 399, Postboks 49, 4000 Roskilde, Denmark, e-mail: ioannis.antoniou@risoe.dk , (1): Risø,DTU (2): Siemens Wind Power (3): Vestas Wind Systems A/S Summary The uncertainty of power performance measurements is closely related to the uncertainty of the wind speed measurement. An inherent uncertainty in the method of measuring the power curve is by using the wind speed at hub height. The assumption behind this is that the wind speed is representative of the wind over the whole turbine rotor. While this assumption was adequate for smaller wind turbines, this is barely true for the modern multi-MW ones. As a consequence considerable deviations often occur between the expected and the produced power. Wind shear, directional changes and varying turbulence with height as a result of either meteorological and / or terrain conditions call for the adoption of a new measurement method. The size of the rotor combined with the machine hub height means that turbines are often exposed to highly varying wind conditions comprising large wind shear, turbulence variations and direction shear within the rotor. These parameters will affect both the production as well as the structural safety of the turbine. Better predictions of power and loads require a more representative wind measurements over the turbine’s rotor. The wind profiles at the Danish test station for wind turbines in Hovsore, have been analyzed up to a height of 165m for a one year period and for a narrow sector in order to reduce met tower influences. The profiles shapes were found to extend from flat to high shear and on many occasions sharp local maxima were also observed at about 80 to 100m. The measurement heights of the profiles cover the rotor of a Siemens 3.6MW turbine erected at the test site. In order to compare the power in the profiles, all profiles having hub height wind speeds between 6m/s and 8m/s have been normalised at 7m/s and have been grouped to a number of mean shear profiles with varying occurrence frequency. To identify the influence of the wind shear and turbulence on the performance of the wind turbine, the above results are used as input to a simplified version of the aeroelastic model HAWC2, the AE_N_WIND code which contains the same aerodynamic and wind modules as the full HAWC2 code and can be considered as a Blade Element Momentum model operating in the time domain, thus with the possibility to include wind turbulence. The AE_N_WIND code has the option of user defined mean wind speed and turbulence shear. The sensitivity analysis has been carried out as time series simulations with input based on the above wind shear and turbulence profiles and using power as the primary characterization parameter. As a result, the measurements over the whole rotor have allowed the measurement of the electrical power as a function of an “equivalent wind speed” where wind shear and turbulence intensity are taken into account. This reduced the statistical unc uncertainty as compared to the results based on the hub height wind speed. The results are significant and support the introduction of the multi-point measurement concept that should contribute to a more accurate measurement of the power curve and a better basis for validation of the prediction models. Keywords: wind shear, profiles, turbulence, aeroelastic simulations, equivalent wind speed 1. Introduction Power performance measurement have always been an important issue in relation to wind energy. Poul Lacour was one of the pioneers making performance tests back in the 1888. Since then the topic has been investigate through several research projects and papers, see Ref. [2-9]. It is still the same parameters which are relevant for wind power production. The influences from these parameters are even more important today, where the area of the rotor disk is significantly increased since modern wind turbine became commercially available. In order to use information of the wind shear and turbulence over the whole rotor, a new definition of power curve is needed. This paper presents some power performance simulations showing that the measurement uncertainty of the power performance can significantly be reduced if additional information of the wind conditions over the whole turbine rotor is used. This paper describes some results derived from the IMPER research programme [1] with focal point on the problematic use of a single hub-height mounted anemometer for power performance measurements on large wind turbines. The IMPER objectives are: • To provide an evaluation of the influence of wind field characteristics (turbulence, wind shear etc.) which influence the performance of large wind turbines • To develop a new method to characterize the performance of large wind turbines in flat terrain, taking into account the wind field measurement covering the complete rotor area • To create the background to improve the accuracy in wind potential measurement methods • To contribute to further development and use of remote sensing methods in wind energy in order to improve the energy yield and reduce the financial risk • To extend the above method for performance measurements from flat to complex terrain. This paper documents the variation in power in the wind profiles with constant wind speed at a predefined height, which correspond to the hub height of a MW turbine. Subsequently the measured wind profiles are used as input to an aeroelastic model in order to investigate how the power in the wind profiles is captured by the turbine. Different parameters (equivalent wind speed) derived from the input wind field are compared to the power produced by the wind turbine in order to investigate how a better correlation between the wind field and the power production can be archived. The wind profiles were measured at Høvsøre, the National Danish Test Station for Large Wind Turbines 2. The site The wind profiles were measured at the National Danish Test Station for Large Wind Turbines, which is situated in the northwest of Denmark, close to the North Sea. The test site is flat, surrounded by grassland with no major obstacles and is situated a distance of 1.7 km from the west coast of Denmark. The prevailing wind direction is from the west. Figure 1, shows the site layout and the instruments used. The wind profiles have been produced combining the measurements from two met masts at the Høvsøre test site. The two met masts are the aviation light met mast, and the met mast to the right of the picture. Sensor Position Cup anemometer, boom mounted on aviation met 160m mast Cup anemometer top mounted 116.5m Cup anemometer, wind vane, sonic anemometer, 100m temperature, differential temperature, relative humidity, air pressure Cup anemometer, sonic anemometer, differential 80m temperature Cup anemometer, sonic anemometer, differential 60m temperature, wind vane Cup anemometer, sonic anemometer, differential 40m temperature Sonic anemometer 20m Cup anemometer, sonic anemometer, differential 10m temperature, wind vane Figure 1 The test site and the heavily instrumented met tower along with a list of the available instruments 3. The method The electrical power produced by a wind turbine is the result of the wind flow through the turbine rotor. Energy is captured by the rotor and for grid-connected wind turbines; this energy is converted to electrical power. The measured power curve is usually expressed as power vs. wind using the relation below: P = 1 ρ AU 3C P 2 (1) where: P the electrical power, ρ the air density A the rotor area U the wind speed and Cp the power coefficient. In order to measure the turbine’s power curve in flat terrain, the wind speed measurements take place using a cup anemometer mounted at hub height and at a distance of 2.5 rotor diameters away from the turbine. The relation between the electrical power and the wind speed is found by averaging the measured parameters over a 10 minutes period. The wind speed varies with time and its average value is given by the relation: T 1 T∫ U= Udt (2) 0 where T is the averaging period and U the instantaneous value of the wind speed expressed as a mean and a fluctuating part: T 1 T∫ U = U + u , for which u = udt = 0 . 0 Following de Vries [11],the average energy flux within the time period T , over a unit area and assuming unit air density, is given by the following relation: ( ) T 1 3 ∫ U + u dt = U + 3U ⋅ u + 3u ⋅U + u 3 2 = 3 2 3 U T 0 3⎛ u 2 u3 ⎞ 3⎛ u2 ⎞ 3 = U ⎜1 + 3 2 + 3 ⎟ = U ⎜ 1 + 3 2 ⎟ (3) U ⎜ U ⎟ ⎜ U ⎟ ⎝ U ⎠ ⎝ ⎠ Equation (3), takes its final form assuming that the probability distribution for u is symmetrical, in which case the last term in the middle expression becomes zero. Furthermore equation (3) shows that, for a given integration 3 interval, the difference between the true average energy flux U and the energy flux estimated from the average 3 wind velocity U over a rotor area, is a function of the turbulence in the flow. The effect of turbulence intensity on the turbine power performance is complicated and only partly understood. It depends on both the aerodynamics of the rotor and the control of the turbine. It can be positive, negative or neutral. Thus the turbine will not always exploit the additional energy, due to the presences of turbulence. For modern variable pitch / variable speed turbines, this relation depends also on the control algorithm. As a rule the turbine will not respond to turbulent fluctuations which have frequencies above the turbine’s frequency response of the rotor torque. However the turbine always responds to changes in the mean wind speed. That the turbine’s power curve has traditionally been measured as a function of the wind speed measured only at hub height is a simplification based on the assumption that the wind speed at hub height is representative of the wind over the whole rotor. In practice, changes of the wind speed within the rotor will influence the power production. Therefore it is expected that the measurement of the wind at more heights within the turbine rotor, is a better approximation than the measurement of the wind speed at hub height only. In the present case, where the wind speed is measured over a number of heights within the rotor area, a new parameter, the weighted or “equivalent” wind speed, is defined as: Ui ⋅A 1 A∑ U eqM 1 = (4) i i In the case where the turbulence of the flow is accounted for, as in Eq. (3), the above equation take the form: 1 A ∑ Ai ⋅ U i3 U eqT 1 = 3 (5) and i As an alternative, the equivalent wind speed can be expressed as the wind speed which results in the same energy flux over the rotor. In this case equation (5) takes the form: 1⎛ ⎞ A⎜∑ U eqT 3 = 3 U i3 ⋅ Ai ⎟ (6) ⎝ i ⎠ where denotes the specific height, A the rotor area and U i , Ai the wind speed and area corresponding to the i specific height and assuming zero tilt and yaw error. This is schematically shown in figure 2, in the case where the wind speed is measured over a rotor at five heights. U5 U4 U3 U2 U1 Figure 2 The wind profile measured over a number of heights within the rotor area of a wind turbine In the following paragraphs, the power output of a turbine will be simulated for a number of measured profiles which have been measured at the Danish National Test station for large wind turbines of Hovsore. The turbine power will be presented as a function of the wind speed at hub height and as a function of the wind speed as defined through the above equations (4) to (6). The goal is to, through simulations; investigate how a more detailed measurement of the wind speed in front of the rotor may reduce the uncertainty in the power curve measurement. Other issues to be investigated are the influence of turbulence and whether a “true-flux” wind speed is advantageous for the more accurate presentation of the wind speed. 3. Wind profiles in flat terrain The analysis is carried using data from the met-mast at Høvsøre, see Figure 1. For the present analysis only cup wind speeds at 40m, 60m, 80m, 100m, 116m and 165m are used together with the wind direction. To illustrate the importance of the profile as a whole for the power production, Figure 3 and Figure 4 show two examples of wind shear situations at the site. The wind shear is mainly determined by the atmospheric stability. Thus in Figure 3, a situation where the atmosphere is stable during the night (large wind shear) and unstable during the day (well- mixed flat profiles) is shown. In Figure 4, the atmosphere is stable during the whole day and the wind profile has a local maximum at 80m height. Such situations will have an impact on the turbine power curve and it is reasonable to expect that a wind turbine with a hub height of 80m would be seen to under produce if the wind speed is measured at the hub height only. Figure 3. The wind profile during the 29th of March, 2007 Figure 4 The wind profile from the 18th of January 2006 For the needs of the simulation, the wind profiles from 6m/s to 8m/s for the height of 80m, and for one year period were chosen. Within this period, 2340 profiles were found from the easterly directions between 60° and 120°, which were binned and categorized according to their shape into 173 profiles, non-equally weighted, Figure5a. Subsequently all mean profiles were normalized so that the wind speed at 80m became equal to 7m/s, U 80 = 7 m / s , using the ratio Ri = 7 / U i80 , where i is the profile number, see Figure 5(b). The same ratio is used to normalize the standard deviations at all heights; in this way the turbulence intensity at all heights remains the same as in the original profiles. 12 180 13.24% 160 12.13% 10 8.20% 140 8.12% wind speed (m/s) 8 6.92% 120 Height (m) 3.67% 6 100 3.16% 3.08% 80 2.69% 4 60 2.39% 2 wsp116 wsp100 wsp80 40 wsp60 wsp40 wsp165 0 20 1 15 29 43 57 71 85 99 113 127 141 155 169 4 5 6 7 8 9 10 no of profile wind speed (m/s) (a) (b) Figure 5 (a): The classified 173 easterly wind profiles, (b): the 10 most common normalised profiles and their percentage of occurence. Assuming a virtual rotor disk as the one shown in Figure 2, the energy flux distribution per unit of the rotor area following Eq. (3) and (6) is shown in the Figure 6. The variation in energy is significant and in fact following the distribution in Figure 6, the highest energy flux is twice the magnitude of the lowest one. These large flux variations will also result in significant variations in the electrical power produced from the turbine, following the equation below. 1 P = C p ρA U EqT 3 3 (7) 2 However the above results were made possible only because the wind speed is registered at more heights within the rotor disk. In the opposite case, where the wind speed was only measured at hub height, which in the present case is constant, the variations in electrical power would be understood as uncertainty in the measurement. 40 35 30 No of profiles 25 20 15 10 5 0 0.0 0.2 0.4 0.6 0.8 Energy flux per unit area (kW/m2) Figure 6 The energy flux for the 173 normalized wind profiles. 4. Aeroelastic simulations The normalised wind profiles are used as input to the aeroelastic simulations for a wind turbine. The primary objective of this part of the work is to identify the sensitivity of the wind field parameters on the performance of the wind turbine. Relevant wind field parameters are mean shear profiles, turbulence shear profiles and wind vector slope and yaw angles. Wind slope and yaw angles will not be considered in the work presented. A simplified version of the aeroelastic model HAWC2 is used in the analysis, the AE_N_WIND code. It contains the same aerodynamic and wind modules as the full HAWC2 code and can be considered as a Blade Element Momentum model operating in the time domain, thus with the possibility to include wind turbulence. The AE_N_WIND code has the option of user defined mean wind fields and user defined turbulence shear. So the wind speed variations with height are taken into account while for a given height, the mean wind speed is considered constant. The simulation results depend on the aerodynamic model used in the code. The way the induction correction is implemented can have different effects under different situations. Despite this difficulty, the results are considered to be generally applicable. The sensitivity analysis will be carried out as a number of time series simulations with different wind field input. The primary performance characterization parameter will be the power and the ranges of the wind field parameters which are chosen from measurements of the wind profiles as described above. The Mann model of turbulence [11] is used to generate different turbulence fields which are added to the mean wind speed in order to model the random feature of the wind. A number of simulations have been carried out in order to map the statistical variation in normal turbulence fields, see Figure 7. These simulations have illustrated a significant variation across the rotor disk of the wind statistics. On this background 10 realizations were decided adequate in order to reduce the statistical variation to a reasonable level. Figure 7 Variation of mean wind speed and wind speed standard deviation for a 600 s simulation. A power law shear was specified. 5. The wind turbine and the input modifications The sensitivity analysis will be carried out on a model of the Siemens 3.6MW turbine with a hub of 90m height, a rotor diameter of 107.16 meters and a constant rotation speed of 0.99 rad/s. Siemens Wind Power has provided the data for the modeling. The implementation has been compared to other aerodynamic models and simulations carried out by Siemens, Figure 8. In the AE_N_WIND simulations (and WT4 simulations) no loss in drive train or generator has been assumed. 4000 Siemens HAWC2 3500 WT4 3000 2500 2000 1500 1000 500 0 0 5 10 15 20 25 Figure 8 The calculated power curve In order to validate the user defined mean and turbulence shear implementation, a number of simulations with simplified input parameters have been carried out in order to check the implementation of the user defined mean and turbulence shear fields (not presented here). Since the turbine is a variable speed pitch regulated turbine, which operates at variable speed in the low wind speed region (optimal Cp tracking) and the AE_N_WIND code cannot be used together with a controller (no generator or pitch degrees of freedom), it means that for a specific wind speed, one fixed rotational speed must be used. For simulations with turbulence this results in a different behaviour for the model turbine than for the real turbine. In order to investigate the constant/variable speed assumption two power curves have been calculated, one with fixed speed and one with variable speed. For each of these options two different shears have been considered: no shear and an extreme shear with a speed-down of 25% at rotor top and bottom. The results are plotted in Figure 9 below. (a) (b) Figure 9 (a): Calculated power curves with fixed and variable rotational speed. No shear, (b): Calculated power curves with fixed and variable speed. Extreme shear specified. To assess the influence of the two different operational strategies – one fixed speed and one representing an ideal controller – the power curves have been numerically integrated and weighted with a Gaussian wind speed distribution N[6.5;0.65] to calculate the equivalent power. This equivalent power represents a power value, which corresponds to the same accumulated production as if the correct power curve was followed. Due to the curvature of the power curves, the equivalent values are larger than the quasi-steady power value at 6.5m/s. For larger curvatures, larger power values are seen. Var rpm Cnt. rpm No shear 714.888 703.2166 User shear 525.2578 519.7444 ratio: 1.361023 1.353005 Table 1: Equivalent power values The results, Table 1, reveal the same relative reduction in the equivalent power for the two different operational strategies and it is concluded that the constant speed assumption can be applied 6. Results and Discussions In order to use the above profiles for simulations, a parallel translation of 10m (from 80 to 90m) was assumed. The simulations took place for a number of configurations and the results will be presented as a function of a number of equivalent wind speeds as shown below in Table 2. U eqM 1 U eqT 1 U eqT 3 U eqL1 U eqL3 3 profile points √ √ √ √ √ 5-profile points √ √ √ √ √ One Pitot tube on a √ √ √ √ √ blade (at 85% radius) Two Pitot tubes on √ √ √ √ √ a blade (at 45% and 85% radius) Table 2 The simulation cases examined (L=laminar flow) The mean wind speed and standard deviation shear were defined at five points, at the center and at symmetrical positions above and below, along the vertical rotor diameter. The resulted wind speed time series at the corresponding five points of the simulation output have subsequently been used for the calculation of the wind speeds using the equations (4) to (6). In the case of the three-point simulation three points, one at the center and two symmetrically placed between the five-point configurations have been used. In the case of the pitot tube simulations, the five-point profile was used as input and the wind speed output was traced at 45% and 85% radius. 6.a 1, 3 and 5-point profiles In the figure 10 below, the turbine’s power curve and the power coefficient are presented for the cases as described above. As 1-point profiles we consider the wind speed at the center of the turbine which for the case of the normalized profiles is equal to 7m/s. The resulting wind speeds are the weighted results using the equations (4) to (6). The lines in the figures of the power curves and the power coefficients represent the fit of the corresponding power curve or power coefficient points calculated using a flat profile as input in the AE_N_WIND code at the corresponding wind speeds. In the case of turbulent flow, the user defined input standard deviation values, were the weighted mean average of the standard deviation over the number of heights, measured for each wind speed profile. Power curve, laminar flow Cp, laminar flow 0.55 1600 Flat laminar 0.50 1400 1 point El. power (kW) 1200 0.45 3 points Cp () 1000 5 points 0.40 800 1 point 0.35 3 points 600 5 points 400 0.30 flat laminar 200 0.25 5.5 6 6.5 7 7.5 8 8.5 5.5 6 6.5 7 7.5 8 8.5 UeqL1(m/s) UeqL1 (m/s) (a) (b) Power curve, turbulent flow Cp, turbulent flow 1600 0.55 1 point 1400 3 points 0.50 5 points El. power (kW) 1200 0.45 power Poly. (power) Cp () 1000 0.40 1 point 800 0.35 3 points 600 5 points 0.30 flat turbulent 400 Poly. (flat turbulent) 200 0.25 5.5 6 6.5 7 7.5 8 8.5 5.5 6 6.5 7 7.5 8 8.5 UeqM1 (m/s) UeqM1 (m/s) (c) (d) Power_5 points, turbulent flow Cp_5 points, turbulent flow 1600 0.55 UeqT1 UeqM1 1400 0.50 power Poly. (power) 1200 0.45 El. power (kW) Cp () 1000 0.40 800 0.35 600 UeqT1 UeqM1 0.30 400 Cp Poly. (Cp) 0.25 200 5.5 6 6.5 7 7.5 8 8.5 5.5 6 6.5 7 7.5 8 8.5 wind speed (m/s) wind speed (m/s) (e) (f) El. power_5 points_turbulent flow Cp_5 points_turbulent flow 1600 0.55 UeqT3 1400 UeqT1 0.50 power El. power (kW) 1200 Poly. (power) 0.45 Cp() 1000 0.40 UeqT3 800 0.35 UeqT1 600 Cp 0.30 400 Poly. (Cp) 200 0.25 5.5 6 6.5 7 7.5 8 8.5 5.5 6 6.5 7 7.5 8 8.5 wind speed (m/s) wind speed (m/s) (g) (h) Figure 10 (a),(b): laminar flow, (c),(d): turbulent flow, (e), (f): comparison between UeqT1 and UeqM1 expressions, (g), (h): comparison between UeqT1 and UeqT3 expressions. The results of the simulations show clearly that by increasing the number of measurement points over a given profile, it results in a more unambiguous relation between power and wind speed. Among the weighted wind speeds used, the weighted mean wind speed expression represents, for a given number of points, the best approximation to the turbine’s power curve, relative to the rest of the simulations, see Figures 10e to 10h. When accounting for the turbulence in the flow, equation (5), it is expected that the influence of the turbulence will strongly depend on the shape of the power curve and whether it is convex or concave at the specific part. In the case where the wind speed is defined in terms of a wind speed resulting in the same energy flux over the rotor, equation (6), the correlation between wind and power is less good. The reason for this, as shown in Figure 5a, is that the rotor is not able to extract the kinetic energy as efficient when the wind shear over the rotor is large, as the power coefficient at higher wind speeds reduces in value, meaning that the turbine becomes less efficient to extract energy from the wind. In Figure 11, the average standard deviation values between the simulation results for all wind profiles and the polynomial fitted power curves are compared for different number of points. Power Standard Deviation (kW) 120 1 point 3 points 5 points .44 .44 .44 .48 .48 95 95 95 100 87 87 80 .77 .08 60 .56 54 60 .08 46 40 .75 40 .83 .57 26 22 20 0 0 20 9.0 8 8.5 5.5 0 UeqL1 UeqL3 UeqM1 UeqT1 UeqT3 Figure 11 The average standard deviation of the power relative to the flat laminar/turbulent power curve 6.b One-and two-pitot tubes on the blade A mentioned above, in the case of the pitot tube simulations, the five-point profile was used as input and we chose to trace the wind speed output at 45% and 85% radius. The practical value of these simulations which do not account for any rotor or blade induction, is to offer a first estimate of the configuration which could in principle accurate measure the turbine’s power curve without using any ground based equipment. In the case of the one pitot tube, the wind speed is traced at 85% blade radius. Again, this is the wind which the pitot tube ideally “reads” in the absence of any induction or other influences from the presence of the rotor. The results are shown in Figure 12a and b. The line in the figures corresponds to the polynomial fit of the power curve points calculated as above using flat turbulent wind profiles as input to the AE_N_WIND code. The results show that measuring the power curve using one pitot tube is not adequate to correctly reproduce the power curve. The large differences between the UeqM and UeqT expressions of the weighted wind speed are due to wind shear in the wind profiles which is understood as turbulence by the rotating pitot tube. However it is possible that the results are influenced as the integration of the wind speed over the pitot path, does not weight the wind speeds equally over the rotor profile. Some wind speeds are more represented relative to other ones and it is possible that this has resulted to the below shown deviations between the pitot and flat profile power. Pitot tube at 85% Pitot tube at 85% 0.55 1600 UeqT UeqM 0.50 1400 power flat turbulent El. power (kW) 1200 0.45 Cp () 1000 0.40 800 0.35 600 CpT CpM 400 0.30 Cp Poly. (Cp) 200 0.25 5.5 6 6.5 7 7.5 8 8.5 5.5 6 6.5 7 7.5 8 8.5 wind speed (m/s) wind speed (m/s) (a) (b) 2 Pitot tubes (45% & 85%) 2 Pitot tubes at 45% and 85% 0.55 1600 0.50 power 1400 UeqM1 0.45 El power (kW) 1200 Poly. (power) Cp () 1000 0.40 800 0.35 600 CpM1 Cp 0.30 400 Poly. (Cp) 200 0.25 5.5 6 6.5 7 7.5 8 8.5 5.5 6 6.5 7 7.5 8 8.5 wind speed (m/s) wind speed (m/s) (d) (c) Figure 12 The power curve and power coefficient as functions of the weighted wind speeds for two different configurations In Figures 12c to d the power curve and the power coefficient are presented for the case of two pitot tube mounted at 45 and 85% of the blade radius as a function of U eqM 1 .There is a major improvement on the results, yet the wind speed seems still slightly underestimated relative to the flat profile simulations. A parametric study using more pitot tubes, different measurement positions or a combination of both is still needed in order to find whether further improvements in the accuracy of the measurements are possible when using pitot tubes. An alternative approach to the subject of pitot measurements could though be that a new power curve using pitot tubes is defined which includes the induction of the rotor and what remains is to verify this power curve for the specific pitot tube configuration on the blade. 7. Conclusions Large variations have been observed in the wind profiles over a flat terrain site. The results show that the profiles do not follow always the logarithmic law; instead they heavily depend upon the atmospheric conditions. These profiles were used as input to a sensitivity analysis concerning the response of the power production as a function of wind and turbulence variations. Using an increased number of points significantly improved the correlation between wind input and power output. These results support the necessity for the introduction of a new definition for power performance measurements using a distributed measurement of the wind over the rotor area instead of using only the hub-height wind speed. Acknowledgment The authors gratefully acknowledge the financial support of the Danish Energy Agency to the IMPER project (journal no.: 3302-0106), which made possible this paper. 7. References 1. Improved performance measurements: characterization of the wind field over a large wind turbine rotor. IMPER, EFP-2006 Contract # 2. Accuracy of power curve measurements, C. J. Christensen et al. Contract 84/B/7033/II/004/1I7, November 1986 3. Analysis of data from gedser wind turbine 1977-1979, P. Lundsager et al. Risø-M2242, August 1980. 4. Power performance assessment, S. Frandsen et al. Contract JOR3-CT96-0114, JouleIII EU final report, June 1999 5. Identification of variables for site calibration and power curve assessment in complex terrain. I.Antoniou et al., JOR3-CT98-0257, JouleIII EU Final Report 1/8 1998-31.7 2001 6. European Wind Turbine Standards II, Project Results ECN-C-99-073, Joule III EU Contract JOR3-CT95- 0064, June 1998 7. Consequences of Høvsøre meteorology on power production. Workshop on the analysis of wind profiles and turbulence in the boundary layer at the høvsøre test site, Risø (DK), 20 Dec 2005, Schmidt Paulsen, U., Unpublished. 8. Wind flow aberrations at near coastal sites of importance for performance verification measurements, P. B. Enevoldsen, Siemens Wind Power, Denmark, Poster BL3.404, EWEC Conference Athens 2006 9. Sea-land interaction influence on wind turbine power performance. U.S. Paulsen et al. EGU General Assembly 2006, Vienna (AT), 2-7 Apr 2006. Geophys. Res. Abstr. (CD-ROM) (2006) 8 (no.Abstr. EGU06- A-06548) 10. Wind field simulation Mann J, PROBABILISTIC ENGINEERING MECHANICS, 13: (4) 269-282 OCT 1998 11. Fluid Dynamic Aspects of wind energy conversion, O. de Vries AGARD-AG-243, 1978