WIND ENERGY Wind Energ. 2006; 9:267280 Published online 13 October 2005 in Wiley Interscience (www.interscience.wiley.com) DOI: 10.1002/we.

170

Research Article

Wind Turbine Response to Parameter Variation of Analytic Inow Vortices

M. Maureen Hand* and Michael C. Robinson, National Wind Technology Center, National Renewable Energy Laboratory, 1617 Cole Blvd, MS 3811, Golden, CO 80401, USA Mark J. Balas, Electrical and Computer Engineering, University of Wyoming, PO Box 3295, Laramie, WY 82071-3295, USA

Key words: vortex; modelling; coherent turbulence; stable boundary layer

As larger wind turbines are placed on taller towers,rotors frequently operate in atmospheric conditions that support organized, coherent turbulent structures. It is hypothesized that these structures have a detrimental impact on the blade fatigue life experienced by the wind turbine. These structures are extremely difcult to identify with sophisticated anemometry such as ultrasonic anemometers. This study was performed to identify the vortex characteristics that contribute to high-amplitude cyclic blade loads, assuming that these vortices exist under certain atmospheric conditions. This study does not attempt to demonstrate the existence of these coherent turbulent structures. In order to ascertain the idealized worst-case scenario for vortical inow structures impinging on a wind turbine rotor, we created a simple, analytic vortex model. The Rankine vortex model assumes that the vortex core undergoes solid body rotation to avoid a singularity at the vortex centre and is surrounded by a two-dimensional potential ow eld. Using the wind turbine as a sensor and the FAST wind turbine dynamics code with limited degrees of freedom, we determined the aerodynamic loads imparted to the wind turbine by the vortex structure. We varied the size, strength, rotational direction, plane of rotation, and location of the vortex over a wide range of operating parameters.We identied the vortex conformation with the most signicant effect on the blade root bending moment cyclic amplitude. Vortices with radii on the scale of the rotor diameter or smaller caused blade root bending moment cyclic amplitudes that contribute to high damage density. The rotational orientation, clockwise or counter-clockwise, produces little difference in the bending moment response. Vortices in the XZ plane produce bending moment amplitudes signicantly greater than vortices in the YZ plane. Published in 2005 by John Wiley & Sons, Ltd. Received 9 April 2004; Revised 14 March 2005; Accepted 19 August 2005

Introduction
The current design trend for wind turbines is towards large-diameter rotors on tall towers, each in excess of 100 m. At this scale, fatigue loads become critical design loads for wind turbine blades.1 Blade fatigue loads are driven by the stochastic nature of the turbulent wind inow. Damage density for composite materials is dominated by low-cycle, high-amplitude stresses.2 Thus turbulent inow that causes large, cyclic blade root bending moment uctuations contributes to fatigue damage. Understanding the rotor/inow interaction driving the stochastic wind turbine loads has proven elusive to researchers. Several researchers have studied the effect of inow turbulence on the structural response of wind
* Correspondence to: M. M. Hand, National Wind Technology Center, National Renewable Energy Laboratory, 1617 Cole Blvd, MS 3811, Golden, CO 80401, USA E-mail: maureen_hand@nrel.gov This article is a US Government work and is in the public domain in the USA.

Published in 2005 by John Wiley & Sons, Ltd.

268

M. M. Hand, M. C. Robinson and M. J. Balas

turbines38 in various inow environments, e.g. multirow wind parks, near complex terrain, smooth terrain. Coherent turbulence (as revealed by the Reynolds stress eld) and atmospheric stability were identied as parameters contributing to large load excursions. Researchers have measured instantaneous Reynolds stresses that exceed those generated with normal turbulence models.9 All these studies were performed on wind turbines that operated primarily in the surface layer of the planetary boundary layer (050 m above ground level). The planetary boundary layer is divided into several atmospheric layers that radically change throughout the diurnal cycle. The layer near the ground is the surface layer, bounded above by the mixed layer during the day and by the nocturnal, or stable, boundary layer at night. The depth of each layer grows and shrinks as the dynamic stability promotes or restricts the development of turbulence.10 Now that wind turbines are operating at heights in excess of 100 m, they frequently operate within the stable boundary layer. Atmospheric phenomena such as low-level jets, gravity waves and KelvinHelmholtz instabilities occur within this boundary layer.1113 Coherent turbulence may result from these atmospheric phenomena, and these structures may cause blade load amplitudes that contribute to fatigue damage. Although low-level jet formation has been observed from Texas to Minnesota,10 the most frequent low-level jet activity occurs near the Oklahoma panhandle and extends into Kansas and Texas.14 Researchers detected the existence of the low-level jet near Lamar, Colorado with a highly instrumented, 120 m tower as well as a SODAR (sonic detection and ranging) system.13 The jet formed nearly every night during the summer months along with associated coherent turbulence. Because this Great Plains region of the USA has a signicant documented resource15 and is currently being considered for wind energy development, it is important to explore the implications of operating wind turbines in the presence of coherent turbulence. Very few experimental data that include detailed turbulence measurements and corresponding wind turbine response measurements exist for large wind turbines. Initially we used experimental data obtained from the National Renewable Energy Laboratorys (NRELs) long-term inow and structure test (LIST) to establish a correlation between inow parameters, such as Reynolds stresses, and wind turbine response, such as blade root ap bending moment. This correlation was not clearly identiable with standard 10 min record statistics.16 Additional complicating variables included rotor teeter dynamics and blade pitch motion for power regulation. Although the unique inow array, consisting of ve sonic anemometers, provided more spatial resolution than previous experiments, the delity was insufcient to produce the detail necessary to establish this causal relationship. Atmospheric conditions suggested that the turbine layer frequently exhibited characteristics of the stable, nocturnal boundary layer. The turbine operated in the stable boundary layer for 30% of the entire wind season for which data were collected.16 However, this turbine operated in a layer from 15 to 58 m, which is much lower than the turbines proposed for the Great Plains region. Finally, the normal turbulence model specied by the International Electrotechnical Commission standard IEC 61400-117 assumes surface layer scaling and neutral atmospheric stability, which may not always apply at heights exceeding 50 m. Measurement campaigns to date39 have relied upon turbines that operate primarily in the surface layer. These experiments found that Reynolds stresses, which suggest coherent turbulent ow, and atmospheric stability are related to large load excursions. However, the spatial delity of these experiments was not sufcient to establish a causal relationship between turbulent uctuation parameters and wind turbine response. As turbine hub heights approach 100 m, operation in the stable, nocturnal boundary layer is more likely to occur. Within the nocturnal boundary layer, coherent turbulence can be generated from atmospheric phenomena that are not currently modelled with turbulence models that rely upon surface layer scaling parameters. Experiments with spatial delity to establish a direct relationship are very difcult and expensive to conduct. Until such data are available, this analytic study was designed to begin to understand the conformation of coherent, vortex structures that cause high-amplitude cyclic blade loads, assuming that these structures exist under certain atmospheric conditions. We developed a simple vortex model that convects at a specied mean wind speed. This analytic model was interfaced with the commonly used wind turbine aerodynamics model AeroDyn.18 The wind turbine structural dynamics code FAST19,20 simulated the wind turbine response to the vortex. By restricting the degrees of freedom in the model to rotor rotation only, the turbine model isolated the aerodynamic load introduced by the vortex to the wind turbine blade. We used experimental data to establish an upper bound of circulation strength
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

Wind Turbine Response to Inow Vortices

269

and vertical wind speed. We created response surfaces illustrating the blade root bending moment induced by vortices of varying radius, circulation, location with respect to the wind turbine hub, rotational direction and plane of rotation. The similarity between aerodynamic response to a vortex from two- and three-blade turbines is clear. This study identies the size, circulation strength, orientation and position of vortices that cause highamplitude cyclic loads.

Rankine Vortex
The Rankine vortex is modelled as a potential vortex with a vortex core of variable radius undergoing solid body rotation to avoid a singularity at the vortex centre. Parameters that are varied include the vortex radius, the circulation strength, the location of the vortex centre with respect to the turbine hub, the orientation of the vortex ow, i.e. clockwise or counter-clockwise, and the plane in which the vortex rotates, i.e. XY, XZ or YZ. Figure 1 is a diagram of the wind turbine co-ordinate system with a counter-clockwise rotating vortex in the XZ plane. Note that the origin is located at the intersection of the wind turbine tower centreline and the hub height. If the vortex centre is (x0, y0, z0) in the aerodynamics code co-ordinate system, and the co-ordinates of the blade elements provided by the aerodynamics subroutine are (x, y, z), then the following equations describe velocity components at the (x, y, z) location that result from a vortex rotating clockwise in the XZ plane for r R: u = -(z - z0 )w + V v=0 w = (x - x 0 )w The following equations describe the velocity components when r < R: (1) (2) (3)

z x y

R (xo, yo, zo)

w
u

Wind

Figure 1. Wind turbine co-ordinate system with counter-clockwise rotating vortex of radius R in the XZ plane. The vortex convects to the left by adjusting x0 by VDt at each simulation time step.
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

270

M. M. Hand, M. C. Robinson and M. J. Balas

R z y x (xo, yo, zo)

w v

u Wind

View of rotor, looking downwind

Figure 2. Wind turbine co-ordinate system with counter-clockwise rotating vortex of radius R in the YZ plane. The vortex convects to the left by adjusting y0 by VDt at each simulation time step.

u=

-G z - z0 + V 2p (x - x 0 )2 + (z - z0 )2 v=0

(4) (5) (6)

w=

G 2p

x - x0 (x - x 0 )2 + (z - z0 )2

The vortex convects at a specied mean wind speed from -x to x by shifting x0 by VDt at every simulation time step. In the case of the vortex in the YZ plane, shown in Figure 2, the vortex convects laterally across the rotor from -y to y by shifting y0 by VDt at every simulation time step. We designed this numerical study to quantify the wind turbine response to the passage of an isolated vortex through the rotor. Thus we selected 10 m s-1 for the convection speed for all simulations in this study. At this wind speed the wind turbine operates below the rated wind speed where the aerodynamic performance is optimized. To map the response of the wind turbine to vortices in the inow, we varied the radius of the vortex and its circulation strength. The radius was varied from 1 to 100 m to encompass a range that included scales of the turbine such as the blade chord (0.75 m) and the rotor radius (21.5 m). Using available experimental data, we developed an upper bound for the circulation strength.

Experimental Data
We used data collected from the NRELs LIST to bound the simulated vortex circulation strength and vertical wind speed component. We conducted this eld experiment at the National Wind Technology Center (NWTC) near Boulder, Colorado from October 2000 to May 2001. A planar array consisting of ve sonic anemometers located 1.5 diameters upwind of the advanced research turbine (ART) collected three-component wind velocities in 10 min records at 40 Hz.9 Figure 3 shows the inow array. The ART is a two-blade, upwind, 43 m diameter wind turbine with a teetered hub at 37 m.21,22 Strain gauges measured the blade root bending moments.
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

Wind Turbine Response to Inow Vortices

271

FT

FT

T , DP, BP
Legend BPBP
cup anemometer hi-resolution sonic anemometer/thermometer wind vane

T FT BP

temperature

DP dew point temperature

temperature difference fast-response temperature barometric pressure

Figure 3. Inow instrumentation array viewed from upwind

The spatial resolution of the four anemometers corresponding to the circumference of the rotor permits computation of the vorticity in the YZ plane according to the equation r w v r wx = i y z (7)

The entire database of 10 min records in which the turbine operated throughout the record, consisting of 1941 records, was processed to compute vorticity at each time step. The instantaneous maximum and minimum values (positive suggests clockwise rotation and negative suggests counter-clockwise rotation from a reference point upwind of the turbine) were 0.38 and -0.43 s-1 respectively. These vorticity values correspond to circulation using the following relation, where RS represents the radius of the circle formed by the sonic anemometers, 21 m: r r 2 G = w x n dA = w x pRS
A

(8)

The highest circulation strength for a vortex in the YZ plane observed in the LIST experimental data was then 526 or -582 m2 s-1. Limiting this study to a maximum circulation of 1000 m2 s-1 provides a conservative overestimation of the circulation observed in the data. Because experimental data were not available to compute vorticity in other planes, our conservative assumption was that the circulation strength of vortex structures in the XY and XZ planes would not be substantially different from the circulation in the YZ plane. Another restriction to the simulation parameters was based upon the vertical velocity component measured in the data. The maximum, instantaneous, vertical velocity, or w component, in the database of records where the turbine operated throughout each record was 12.2 m s-1. By restricting the simulation to vortices where the vertical velocity component did not exceed 16 m s-1, another conservative overestimation of observed ow parameters created a limit for simulation conditions. This limited the circulation for vortices of radius 1 m to
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

272
800 Root Flap Bending Moment Range (kN m) Root Flap Bending Moment Range (kN m) 700 600 500 400 300 200 100 0 800 700 600 500 400 300 200 100 0

M. M. Hand, M. C. Robinson and M. J. Balas

All Wind Classes

10 m s-1 Wind Bin

Figure 4. Root ap bending moment ranges from LIST experiment for all wind speed bins and for 10 m s-1 wind speed bin

600

600

YZ Peak Circulation Strength (m2 s-1 )

400

YZ Peak Circulation Strength (m2 s-1 )

CW CC All Wind Classes

500

500

400

300

300

200

200

100

100

CW

CC

10 m s-1 Wind Bin

Figure 5. Circulation strength measured in YZ plane from LIST experiment for all wind speed bins and for 10 m s-1 wind speed bin. CW, clockwise rotation; CC, counter-clockwise rotation.

100 m2 s-1, vortices of radius 3 m to 300 m2 s-1 and vortices of radius 5 and 7 m to 500 m2 s-1 in addition to the restriction for all other radii introduced by the measured circulation strength, which was restricted to 1000 m2 s-1. Figure 4 graphically depicts the population of root ap bending moment range (difference between 10 min record maximum and minimum) measurements for the entire database containing those records where the turbine operated throughout each record. The subset of those records corresponding to an average wind speed within the range 10 1 m s-1 is also included. The line in the centre of the box represents the median value of the population. The box represents the lower and upper quartiles of the population. The whiskers extend to 150% of the inner quartile range. The outliers are represented by the + symbol. A similar graphical representation of the population of circulation strength maxima and the absolute value of circulation strength minima is shown in Figure 5. Again, the entire database is contrasted with those records that fall within the 10 m s-1 mean wind speed bin for comparison.
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

Wind Turbine Response to Inow Vortices

273

Wind Turbine Simulation

The wind turbine geometry simulated using the FAST code was that of the ART, a two-blade, teetered hub, constant speed wind turbine with blade pitch control to regulate power above rated wind speeds.21,22 This turbine has a rotor radius of 21.5 m and produces 600 kW at rated wind speeds. This simulation study restricted the modelled degrees of freedom to rotor rotation at a constant speed only. This limited degreeoffreedom turbine model allowed isolation of the aerodynamic forces imparted to the turbine by the vortex structure. Because wind turbines are designed for optimal aerodynamic performance at rated wind speeds or lower, we selected an operating point of 10 m s-1. The rated wind speed for the ART is 12.9 m s-1. If the ART operated at variable speed, the corresponding rotational speed at a wind speed of 10 m s-1 would be 37.1 rpm and the corresponding pitch angle to achieve maximum power capture would be 1. We specied the convection speed for the vortex to be 10 m s-1 for all simulations in this study to correspond to this turbine operating point. Under these conditions the ow remains attached across the majority of the blade span except for the root section. Little lift is generated in this section owing to the relatively small chord length and generally cylindrical shape. The study estimated aerodynamic loads using the AeroDyn subroutines incorporated in the FAST code. For simplicity, we did not simulate dynamic stall and we used the equilibrium wake model. These simplifying assumptions did not affect the relative response from one simulation condition to another, but the magnitude of the response may be underpredicted. Minor modications to the code permitted the inclusion of the vortex model in a fashion similar to the hub height wind le option. In other words, the velocity components at each blade element were computed analytically at each time step using equations (1)(6). We ran simulations at the operating point mentioned above with several inow vortex permutations. We varied the vortex radius from 1 to 100 m. We varied the circulation strength from 10 to 1000 m2 s-1 with the additional restriction that resulted from the vertical velocity limitation for radii 7 m. The centre of the vortex was varied from hub height to the top of the rotor. We performed this series of 455 permutations using the models representing a vortex rotating in the YZ and the XZ plane for both clockwise and counter-clockwise orientation. We assumed that the XY vortex would be symmetric to the XZ vortex and so it was excluded. The predicted blade root ap bending moment statistics, including mean, range (difference between maximum and minimum values) and standard deviation, were obtained for each simulation run.

Wind Turbine Response to Vortex

To visualize the turbine response to relative vortex scale variations, we generated surfaces illustrating the blade root ap bending moment statistics. Figure 6 shows the mean, range and standard deviation for the clockwise rotating vortex in the YZ plane. The vertical gradation of the surfaces represents the effect of shifting the vortex centre with respect to the turbine hub height. The bottom surface represents a vortex centred at the hub (z0 = 0 m); the top surface represents a vortex centred at the top of the rotor plane (z0 = 21.5 m); intermediate surfaces correspond to a vortex centred at equally spaced vertical positions along the blade between the hub and the tip. Similar surfaces representing a clockwise rotating vortex in the YZ plane and a counter-clockwise rotating vortex in the XZ plane are shown in Figures 7 and 8 respectively. All surfaces show similarities in turbine response. The mean bending moment shows almost no variation with the scale and position of the vortex. The bending moment cyclic range and standard deviation, which contribute to fatigue damage, both increase as the radius of the vortex decreases and the circulation strength increases. Vortices with radii greater than 2030 m appear to contribute very little to load variation. This suggests that vortices on the scale of the rotor (turbine radius 21.5 m) or smaller produce the load variation that leads to fatigue damage. The clockwise rotating vortex in the YZ plane has the same rotational orientation as the wind turbine, so the rotational velocity of the vortex adds to the rotational velocity of the wind turbine. Intuitively, one would expect higher loads when the velocities add as compared with the case where the velocities subtract, the counterclockwise rotating vortex in the YZ plane. However, comparison of the bending moment range and standard
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

274
500 0 500 0 500 0 500 0 500 0 100 50 Radius (m) 1000 500 0 Gamma (m2 s-1) 200

M. M. Hand, M. C. Robinson and M. J. Balas

20 Blade Root Flap Bending Moment Std. Dev. (kN m) 100 50 Radius (m) 1000 500 0 Gamma (m2 s-1) 10 0 10 0 10 0 10 0 10 0 100 50 Radius (m) 0 1000 500 Gamma (m2 s-1)

Blade Root Flap Bending Moment Range (kN m)

100 0 100 0 100 0 100 0 100 0

Blade Root Flap Bending Moment Mean (kN m)

Figure 6. Clockwise rotating vortex in YZ plane with 10 m s-1 convection speed

500 0

200

20 10 0 10 0 10 0 10 0 10 0 0 1000 500 Gamma ( m 2 s -1) 100 50 Radius ( m) 0 1000 500 Gamma ( m2 s-1)

500 0 500 0 500 0 500 0 50 Radius ( m) 100 0

0 100 0 100 0 100 0 100 0

500 Gamma ( m2 s -1)

1000

50 Radius ( m)

100

Figure 7. Counter-clockwise rotating vortex in YZ plane with 10 m s-1 convection speed

deviation for both the clockwise and counter-clockwise rotating vortex in the YZ plane (Figures 6 and 7 respectively) shows similar magnitudes for the vortex centred at the hub. The counter-clockwise rotating vortex produces bending moment ranges exceeding those produced by the clockwise rotating vortex by approximately 20 kN m when the vortex is centred at the top of the rotor plane. In general, the loads produced when the vortex is centred at the hub are slightly higher than at any other location.
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

Blade Root Flap Bending Moment Std. Dev. (kN m)

100

Blade Root Flap Bending Moment Range (kN m)

Blade Root Flap Bending Moment Mean (kN m)

Wind Turbine Response to Inow Vortices

500 0 Blade Root Flap Bending Moment Mean (kN m) Blade Root Flap Bending Moment Range (kN m) 500 0 500 0 500 0 500 0 100 50 Radius (m) 0 500 Gamma (m2 s-1) 1000 500 0 500 0 500 0 500 0 50 25 0 25 0 25 0 25 0 25

275

500 0 100

0 Radius (m)

50

0 1000 100 500 50 Gamma (m2 s-1) Radius (m)

Blade Root Flap Bending Moment Std. Dev. (kN m)

1000 500 Gamma (m2 s-1)

Figure 8. Clockwise rotating vortex in XZ plane with 10 m s-1 convection speed

Figure 9 illustrates time-series traces of the blade bending moment for the most extreme hub-height vortex in each rotation orientation. The centre of the vortex passes through the wind turbine rotor at 120 seconds. The symmetric nature of the response explains the relatively small differences between the clockwise and counterclockwise rotating vortices. For comparison, Table I lists the magnitudes for both orientations at the edge of the surface with the highest magnitude. For vortices centred at hub height, the counter-clockwise rotating vortex in the XZ plane has a steeper slope of blade bending moment range at the edge corresponding to small radius and high circulation strength. On average, both clockwise and counter-clockwise rotating vortices produce similar magnitude bending moment ranges for all vortex centre locations. The vortex centred at the hub produces higher magnitude range loads than at any other vertical location. The bending moment range for a vortex centred at the top of the rotor plane produced about half the cyclic load as the vortex centred at the hub. However, the vortex centred at the top of the rotor could contribute to extreme peak loads. The vortex in the XZ plane produces cyclic range variations that exceed those produced by a vortex in the YZ plane by 210 times. The highest magnitude bending moment range of 629 kN m that results from a vortex of 10 m radius and 1000 m2 s-1 circulation strength rotating counter-clockwise in the XZ plane compares favourably with the extreme range measured on the operating turbine in the 10 m s-1 wind speed bin shown in Figure 4. There are several complicating factors in this comparison. For example, the operating turbine has a teetered hub and the measured blade bending moment could include the result of teeter stop impact. Also, the blade bending moment responds with higher cyclic uctuation when the blade pitch is adjusted to regulate power production. Lastly, the mean wind speed over 10 min can uctuate signicantly and the range statistic would not necessarily represent a single inow event. The simulation does not include any pitch control or turbine dynamics in order to isolate the aerodynamic load induced by a vortex in the inow. Regardless, the relative magnitude of bending moment ranges produced by the simulated vortices compares favourably with those observed in the experimental data. We performed a preliminary investigation of approximating the vortex ow eld by using the hub height wind input le capability of the AeroDyn code. The horizontal, or u, velocity component produced by a 10 m radius, 500 m2 s-1 circulation vortex at the top and bottom of the rotor was used to compute a time-varying,
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

276
Blade Root Flap Bending Moment (kN m)

M. M. Hand, M. C. Robinson and M. J. Balas

800
YZ Vortex, CW YZ Vortex, CC XZ Vortex, CW XZ Vortex, CC

600

400

200
R = 10 m, G = 1000 m2 s-1; centred at hub height

0 110

112

114

116

118

120 Time (s)

122

124

126

128

130

Blade Root Flap Bending Moment (kN m)

800
YZ Vortex, CW YZ Vortex, CC XZ Vortex, CW XZ Vortex, CC

600

400

200
R = 10 m, G = 1000 m2 s-1; centred at top of rotor

0 110

112

114

116

118

120 Time (s)

122

124

126

128

130

Figure 9. Time series traces of simulated root ap bending moment that result from vortex passage

linear, vertical shear across the rotor. Similarly, we extracted the hub height vertical, or w, velocity component from the vortex at each simulation time step. We used these values to create hub height wind les that varied the vertical shear only, the vertical wind component only and the combination of the two. Note that AeroDyn assumes specied vertical wind velocity components to be uniform across the entire rotor. Figure 10 illustrates the time-varying bending moment response to the vortex as well as these three approximations. The amplitude of the bending moment produced by the vortex exceeds all three approximations. Of the three approximations the vertical shear alone most closely reproduces the cyclic response and amplitude that result from vortex passage. However, the vortex model induces some cyclic bending moment variation at the peak amplitudes that is not captured by the approximations. The aerodynamic response of a wind turbine to a vortex in the ow eld should not be substantially different for two- or three-blade machines. This is demonstrated in Figure 11, where a dimensionless representation of the response surfaces for the bending moment coefcient range for the two-blade turbine is compared with that of a three-blade turbine where the vortex was centred both at the hub and at the top of the rotor. The threeblade turbine model is a 46 m, 750 kW rotor based on an initial iteration from the WindPACT study.1 It was modelled with only the rotor degree of freedom in the same manner as the two-blade turbine model. The vortex radius was normalized with the turbine rotor radius (R/RT). The vortex circulation was normalized with the parameter Go = 2pRT V The blade root bending moment was normalized with the parameter
3 Bo = QpRT

(9)

(10)

where
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

Wind Turbine Response to Inow Vortices

277

Table I. Simulated root ap bending moment values that result from vortex passage: (a) z0 = hub height; (b) z0 = top of rotor planea Vortex in XZ plane, CW Range (kN m) (a) R = 1 m, G = 100 m2 s-1 R = 3 m, G = 300 m2 s-1 R = 5 m, G = 500 m2 s-1 R = 10 m, G = 1000 m2 s-1 270 507 514 551 64 164 226 285 (SD) (kN m) 12.4 26.9 36.7 48.3 8.2 17.8 26.0 39.8 Vortex in XZ plane, CC Range (kN m) 223 367 459 629 84 194 226 282 (SD) (kN m) 10.2 23.2 32.6 47.4 8.5 20.1 29.6 47.2 Vortex in YZ plane, CW Range (kN m) 27 68 105 199 38 57 76 123 (SD) (kN m) 5.6 7.5 10.3 17.4 5.2 5.9 7.3 11.5 Vortex in YZ plane, CC Range (kN m) 28 71 101 185 53 81 93 140 (SD) (kN m) 5.7 7.5 10.2 16.8 5.9 7.6 9.6 14.5

(b) R = 1 m, G = 100 m2 s-1 R = 3 m, G = 300 m2 s-1 R = 5 m, G = 500 m2 s-1 R = 10 m, G = 1000 m2 s-1

a

CW, clockwise rotation; CC, counter-clockwise rotation; SD, standard deviation.

600
XZ vortex, R = 10 m, G = 500 m2 s-1

Blade Root Flap Bending Moment (kN m)

550 500 450 400 350 300 250 200 114

Vertical shear w component Vertical shear and w component

115

116

117

118

119 Time (s)

120

121

122

123

124

Figure 10. Comparison of wind eld approximations of vortex with actual vortex-induced bending moment

2 Q = 1 rV 2

(11)

The bending moment ranges for the two- and three-blade turbines are similar whether the vortex is centred at the hub or at the top of the rotor. The three-blade turbine has a slightly steeper slope as the normalized radius is decreased and the normalized circulation is increased. However, the similarity in the aerodynamic response of the two turbines is conrmed.
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

278

M. M. Hand, M. C. Robinson and M. J. Balas

2-blade turbine 0.2 Bending Moment Coefficient Range Bending Moment Coefficient Range 0.1 0 0.4 0.3 0.2 0.1 0 5 R/RT 0 0 0.5 G/G o R/R T 0.2 0.1 0 0.4 0.3 0.2 0.1 0 5

3-blade turbine

0.5 G/Go

Figure 11. Comparison of dimensionless bending moment response to vortex passage from two- and three-blade wind turbine simulations

Summary
This analytic study identies the aerodynamic response of a wind turbine to analytic vortex structures in the inow. The vortex variation included radius, circulation strength, location with respect to the hub height, orientation and plane of rotation. We obtained blade response under each permutation through simulation using the FAST wind turbine dynamics code. We presented the blade responses in surface plots that illustrate the variations in blade load as a result of vortex passage through the rotor. Owing to symmetry, the turbine response to clockwise and counter-clockwise rotating vortices was similar for vortices in the YZ plane as well as for vor-tices in the XZ plane. We assumed that this would also be true for vortices in the XY plane. A counterclockwise rotating vortex in the XZ plane and aligned with the hub produced the greatest load variation. We observed decreased load variation as the vortex was moved vertically to the top of the rotor plane, but these loads still exceeded those observed as a result of passage of a vortex in the YZ plane. We assumed that vortices in the XY plane would be symmetric to those in the XZ plane. Vortices of radii less than 2030 m, on the scale of the rotor or smaller, produce substantial, cyclic, bending moment uctuations. Similar bending moment response is produced by a three-blade turbine, as would be expected. Vertical shear across the rotor approximates the vortex-induced bending moment cyclic variation more closely than the hub height vertical wind component or a combination of vertical shear and vertical velocity. However, the amplitude of the vortexinduced bending moment is underestimated and some higher-frequency dynamics are not captured.

Conclusions
Because high-amplitude cycles contribute to high damage density in composite materials, coherent vortex structures cause cyclic blade amplitudes that contribute to fatigue damage. This analytic study identies the
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

Wind Turbine Response to Inow Vortices

279

most likely conformation of vortices to produce large, cyclic blade amplitudesa vortex on the scale of the rotor or smaller rotating in the XZ plane. This type of vortex is representative of that which could occur in association with atmospheric phenomena such as KelvinHelmholtz instabilities. Recent work by atmospheric scientists suggests that coherent turbulence may frequently form at heights wind turbines now occupy. This parameter variation study of the wind turbine/vortex interaction provides a starting point for further research into the implications of wind turbine operation in the presence of coherent turbulence. This study demonstrates that purely analytic vortices can cause blade load cycles commensurate with those observed in wind turbine measurements. Obviously coherent turbulent structures would be superimposed on normal turbulence in the atmosphere. Experimental data consisting of detailed inow measurements and the corresponding wind turbine response at a height of 80 m are currently being collected. Until additional data are available, this analytic model will be used for the design of wind turbine control systems that mitigate the effect of coherent vortices impinging upon the wind turbine blades.

Appendix: Nomenclature
Dt A Bo CC CW G or G Go r r n w Q r r R RS RT u, v, w V x, y, z x0, y0, z0 simulation time step, 0.004 s area of integration constant for normalization of blade root bending moment (kN m) counter-clockwise rotation clockwise rotation vortex circulation strength (m2 s-1) constant for normalization of vortex circulation (m2 s-1) unit vector along x-axis unit normal vector vorticity (s-1) dynamic pressure over rotor based on mean convection speed (Pa) air density radial distance from vortex centre to (x, y, z) position (m) radius of vortex (m) radius of circle formed by sonic anemometers, 21 m radius of wind turbine (m) wind velocity components corresponding to x, y, z co-ordinates respectively (m s-1) convection speed of vortex (m s-1) blade element co-ordinates in aerodynamics code co-ordinate system (m) position of vortex centre in aerodynamics code co-ordinate system (m)

References
1. Malcolm DJ, Hansen AC. WindPACT turbine rotor design study: June 2000June 2002. NREL/SR-500-32495, National Renewable Energy Laboratory: Golden, CO, 2002. 2. Sutherland HJ. On the fatigue analysis of wind turbines. SAND99-0089, Sandia National Laboratories: Albuquerque, New Mexico, 1999. 3. Kelley ND. The identication of inow uid dynamics parameters that can be used to scale fatigue loading spectra of wind turbine structural components. NREL/TR-442-6008, National Renewable Energy Laboratory, Golden, CO, 1994. 4. Glinou G, Fragoulis A, (eds). Mounturb nal report. JOU2-CT93-0378, Center for Renewable Energy Sources Wind Energy Department Pikermi, 1996. 5. Kelley ND, McKenna HE. The evaluation of a turbulent loads characterization system. NREL/TR-442-20164, National Renewable Energy Laboratory, Golden, CO, 1996. 6. Sutherland H. Inow and the fatigue of the LIST wind turbine. In Collection of the 2002 ASME Wind Energy Symposium Technical Papers Presented at the 40th AIAA Aerospace Sciences Meeting and Exhibit, 1417 January, Reno, Nevada. ASME: New York, NY, 2002; 427437.
Published in 2005 by John Wiley & Sons, Ltd. Wind Energ 2006; 9:267280 DOI: 10.1002/we

280

M. M. Hand, M. C. Robinson and M. J. Balas

7. Sutherland HJ, Kelley ND, Hand MM. Inow and fatigue response of the NWTC advanced research turbine. In Collection of the 2003 ASME Wind Energy Symposium Technical Papers Presented at the 41st AIAA Aerospace Sciences Meeting and Exhibit, 69 January, Reno, Nevada. ASME: New York, NY, 2003; 214224. 8. Kelley ND, Bialasiewicz JT, Osgood RM, Jakubowski A. Using wavelet analysis to assess turbulence/rotor interactions. Wind Energy 2000; 3: 121134. 9. Kelley N, Hand M, Larwood S, McKenna E. The NREL large-scale turbine inow and response experimentpreliminary results. NREL/CP-500-3-917, National Renewable Energy Laboratory, Golden, CO, 2002. 10. Stull RB. An Introduction to Boundary Layer Meteorology. Kluwer Academic Dordrecht, 1988. 11. Blumen W, Banta R, Burns SP, Fritts DC, Newsom R, Poulos GS, Sun J. Turbulence statistics of a KelvinHelmholtz billow event observed in the nighttime boundary layer during the CASES-99 eld program. Dynamics of Atmospheres and Oceans 2001; 34: 189204. 12. Banta R, Newsom RK, Lundquist JK, Pichugina YL, Coulter RL, Mahrt LD. Nocturnal low-level jet characteristics over Kansas during CASES-99. Boundary-Layer Meteorology 2002; 105: 221252. 13. Kelley N, Shirazi M, Jager D, Wilde S, Adams J, Buhl M, Sullivan P, Patton E. Lamar low-level jet project interim report. NREL/TR-500-34593, National Renewable Energy Laboratory: Golden, CO, 2004. 14. Bonner WD. Climatology of the low-level jet. Monthly Weather Review 1968; 96: 833850. 15. Elliot DL, Holladay CG, Barchet WR, Foote HP, Sandusky WF. Wind energy resource atlas of the United States. DOE/CH10093-4, Pacic Northwest Laboratory, Richland, WA, 1987. 16. Hand MM. Mitigation of wind turbine/vortex interaction using disturbance accommodating control. NREL/TR-50035172, National Renewable Energy Laboratory, Golden, CO, 2003. 17. International Electrotechnical Commission. Safety of Wind Turbine Conversion Systems. IEC 61400-1, International Electrotechnical Commission, Chicago, IL, 1998. 18. Laino DJ, Hansen AC. Users guide to the wind turbine Aerodynamics Computer Software AeroDyn v.12.50. [Online]. http://wind.nrel.gov/designcodes/aerodyn/aerodyn.pdf [24 September 2003]. 19. Buhl Jr ML, Jonkman JM, Wright AD, Wilson RE, Walker SN, Heh P. FAST users guide. NREL/EL-500-29798, National Renewable Energy Laboratory, Golden, CO, 2003. 20. Wilson RE, Freeman LN, Walker SN, Harman CR. Final report for the FAST advanced dynamics code: two bladed teetered hub version 2.4 users manual appendix for three and four bladed versions. NREL/SR-500-23563, National Renewable Energy Laboratory, Golden, CO, 1996. 21. Snow AL, Heberling II CF, Van Bibber LE. The dynamic response of a Westinghouse 600-kW wind turbine. SERI/STR217-3405, Solar Energy Research Institute, Golden, CO, 1989. 22. Hock SM, Hausfeld TE, Thresher RW. Preliminary results from the dynamic response testing of the Westinghouse 600-kw wind turbine. SERI/TR-217-3276, Solar Energy Research Institute, Golden, CO, 1987.

Published in 2005 by John Wiley & Sons, Ltd.

Wind Energ 2006; 9:267280 DOI: 10.1002/we