You are on page 1of 11

WIND ENERGY

Wind Energ. (2014)


Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/we.1700

RESEARCH ARTICLE

Turbulence effects on a full-scale 2.5 MW


horizontal-axis wind turbine under neutrally
stratified conditions
Leonardo P. Chamorro1,2 , S-J. Lee1,3 , D. Olsen1 , C. Milliren1 , J. Marr1 , R.E.A Arndt1 and F.
Sotiropoulos1
1 Saint Anthony Falls Laboratory, Department of Civil Engineering, University of Minnesota, Minneapolis, Minnesota 55414, USA
2 Department of Mechanical Science and Engineering, University of Illinois, Urbana, Illinois 61801, USA
3 Research Institute of Marine Systems Engineering, Seoul National University, Seoul, South Korea

ABSTRACT
A field experiment was carried out to study the unsteady behavior of an instrumented full-scale 2.5 MW wind turbine under
neutral conditions. The analysis focused on the structure of the instantaneous turbine power and strain at its foundation.
A meteorological tower located 1.6 rotor diameters upstream of the turbine was used to characterize the turbulent flow.
Mean velocity and temperature were steady during the 1 h period selected. The results suggest that the turbine power and
foundation strain are modulated by atmospheric turbulence in a complex way. The spectral characteristics of both quan-
tities exhibited three distinctive regions. Within the first region, defined by subrotor length scales, the turbine power was
insensitive to the flow turbulence. In the intermediate region, with length scales up to those on the order of the atmospheric
boundary layer thickness, the spectral contents of the power fluctuations ˆP and flow ˆU exhibit a non-linear relationship
of the form ˆP D G.f /ˆU , where G.f / / ./f 2 is a transfer/damping function. In the third region, dominated by the
very large scales of motions, the power fluctuations are found to be directly influenced by the flow. The strain also showed
three regions, similar to the power fluctuations. However, it follows the structure of the inertial subrange of the turbulence
at subrotor scales. Intermittent gusts were able to induce intermittent behavior on the turbine power. Finally, the flow and
power correlation showed that the velocity at the hub height is the best descriptor of the flow turbulence within the rotor
area. Copyright © 2014 John Wiley & Sons, Ltd.
KEYWORDS
atmospheric boundary layer; field experiment; power fluctuations; turbine loading; turbulence; wind energy; wind turbine

Correspondence
F. Sotiropoulos, Saint Anthony Falls Lab, University of Minnesota, Minnesota 55414, USA.
E-mail: fotis@umn.edu

Received 27 September 2012; Revised 11 August 2013; Accepted 19 November 2013

1. INTRODUCTION
Power fluctuations and fatigue loads are among the most significant problems that wind turbines face through their lifetime.
Atmospheric turbulence is the common driving mechanism that modulates the rich dynamics of these quantities. Because
of this, the power and forces on the turbine exhibit fluctuations ranging from synoptic and diurnal to short time scales. The
level of the power fluctuations determines the quality of the electricity produced by the turbine and has direct implications
on the power grid and reliability of transmission systems.1 The fatigue loads, in contrast, affect the various structural and
mechanical components of the turbine. Moreover, with the monotonic increase of turbine rotor size observed in recent
years and projected to continue for the near future blade loading has emerged as a critical element of the cost of energy.2
Parametric load distribution models (e.g., Veers and Winterstein3 ) and probabilistic methods (e.g., Moriarty et al.4 ), among
others, have been proposed to quantify the effect of turbulence-generated loads. Research efforts have also focused on
smoothing the effect of the flow turbulence on the power generated by a turbine (e.g., Ran et al. and Luo et al.5,6 ). Col-
lective and, especially, individual pitch control strategies7 have shown to be very effective in reducing turbine loads.8

Copyright © 2014 John Wiley & Sons, Ltd.


Turbulence effects on a full-scale wind turbine L. P. Chamorro et al.

In particular, several efforts have focused on periodic loads. Among them are the cyclic pitch control (e.g., Geyler9 ) and
higher harmonic control strategies (e.g., Chopra and Gessow10 ).
The complex dynamical behavior experienced by a wind turbine is mostly conditioned by various characteristics of the
atmospheric flow such as short and long-term velocity variations, turbulence structure, intermittency, and thermal strati-
fication, among others. Although turbulence intensity plays a major role in the turbine unsteady behavior,11 it is just one
of the factors determining turbine performance. The spectral characteristics of turbulence are also important since not all
turbulence scales of the flow are able to affect the turbine.12 Therefore, understanding and quantifying the turbine response
(e.g., power, stresses, and moments) to the various scales of atmospheric turbulence are critical prerequisites for improving
the design and enhancing the reliability and life-span of wind turbines. For such studies, however, to make an immediate
impact to the wind energy industry, they need to be carried out at full-scale conditions and under real-life atmospheric
flow conditions.
In this study, we seek to improve our understanding of how real-life atmospheric turbulence under nearly neutral strat-
ified conditions modulates the power fluctuations and loads on a state-of-the-art utility scale wind turbine. The details of
our field site experimental station, including an instrumented 2.5 MW turbine and a meteorological tower, are described in
Section 2. The experimental measurements are analyzed and discussed extensively in Section 3, and the conclusions of this
work are presented in Section 4.

2. EXPERIMENTAL SETUP
The field experimental investigation was performed at the University of Minnesota Eolos Wind Energy Research Field Sta-
tion located approximately 25 miles from the main Twin Cities university campus. The facility consists of an instrumented
Clipper Liberty 2.5 MW horizontal-axis wind turbine and a 130 m meteorological tower. Sampling continuously are five
main sensor systems, amassing nearly 3.9 million data points everyday on secure, backed-up servers. A photograph of the
Eolos Research station is illustrated in Figure 1. The main characteristics of the wind turbine, meteorological tower, the
sensors used in this investigation, and the environmental conditions are described as follows.

2.1. Eolos wind turbine

The Eolos wind turbine is a Clipper Liberty C96 with 2.5 MW rated power. Clipper Liberty turbines are three-bladed,
horizontal-axis, pitch-regulated machines that feature a load-splitting gear box that delivers torque to four permanent mag-
net generators and a power electronics system that allows the wind turbine to operate at various speeds. With a 96 m rotor
diameter, the Clipper Liberty C96 is designed for a IIb wind class. Rated power (2.5 MW) is reached with a wind velocity
near 13 m s1 , whereas the cut-in speed is 4 m s1 . A sustained wind velocity of 25 m s1 causes the wind turbine to
cut-out. Constructed of rolled steel plates, the 80 m tower tapers slightly from bottom to top. The tower is attached to the
5.8 m diameter concrete foundation with 144 high-strength anchor bolts. The Eolos wind turbine is outfitted with a custom
tower and foundation sensor system. The system is designed to measure the overturning moment that is applied to the
concrete foundation of the wind turbine, the foundation’s response to this overturning moment and the settlement of the
entire foundation over time. Measurements of the applied overturning moment are made using a series of 20 strain gauges
mounted to the interior wall of the wind turbine tower near the base. Strain measurements are collected at 20 Hz using

Figure 1. Photograph of the Eolos Wind Energy Research Field Station. As shown, the meteorological tower is in front of the turbine
in the prevailing wind condition.

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
L. P. Chamorro et al. Turbulence effects on a full-scale wind turbine

Table I. Characteristic of the sensors at the turbine tower foundation.


Sensor Model Quantity Sampling rate (Hz) Range

Triaxial DC accelerometer Silicon designs 2460-002 3 20 ˙2G


Single axis strain gauge Vishay MicroMeasurements 20 20 ˙3%
CEA-06-125UN-350/P2
Thermocouple Omega Engineering SA1-T 10 0.0167 (1/min) -60o C to 175o C

two Campbell Scientific CR3000 (Campbell Scientific, Logan, Utah) dataloggers. To ensure that the strain measurements
are not skewed by temperature-induced expansion or contraction in the steel, 10 thermocouples are attached to the steel
next to every other strain gauge. Temperature measurements, made once per minute, can be used to correct the strain mea-
surements. To measure the foundation’s response, three 2G triaxial direct current accelerometers are bolted to the concrete
of the foundation near the base of the tower. Capable of measuring tilt, both static and dynamic, these sensors allow to
estimate the rotational stiffness of the foundation. Measurements of the accelerometers are made at 20 Hz using the same
two Campbell Scientific dataloggers. Table I describes the turbine structure sensors used in this study.

2.2. Meteorological tower

The Eolos meteorological tower (referred to hereafter as the met tower) is specially designed for characterizing the atmo-
spheric boundary layer flow around the turbine and also to measure the turbine wake when possible. Located 160 m south of
the wind turbine, it is upwind of the wind turbine in the predominant wind direction. A system of 10 wind velocity sensors
allows for accurate characterization of the wind resource at elevations from grade to the highest point of the rotor (130 m).
Four of these elevations (129, 80, 30, and 10 m) are instrumented with high-resolution CSAT3 3D Sonic Anemometers,
with a sampling rate of 20 Hz. These four heights were specially selected to match the top and bottom tips of the rotor,
the turbine hub height, and near ground. To ensure accurate measurements, all CSAT3 anemometers are mounted on 18-
foot-long booms that were custom designed to be rigid and limit sway. In addition, all CSAT3’s are paired with a triaxial
accelerometer capable of measuring tower and boom arm movement.
Three meters below each of the CSAT3’s (elevations of 126, 77, 27, and 7 m) and at points representing elevations at
the midpoint between the edge of the rotor and hub height (105 and 55 m) are standard cup and vane anemometers (see the
diagram of Figure 2). Temperature and relative humidity sensors are mounted directly on the tower adjacent to the cup and
vane anemometer booms. Table II summarizes all the the met tower sensors used in this investigation.

2.3. Atmospheric boundary layer characteristics

As a first step in studying the power fluctuations and the dynamic behavior of the turbine induced by atmospheric turbu-
lence, we defined a procedure for selecting appropriate flow conditions. The procedure consists of filtering data obtained
from the Eolos Station to select intervals where the wind speed was stable, the wind direction was predominantly southerly,
allowing the met tower to be upstream of the turbine, and the atmosphere was relatively dry. Wind speed and direction
were observed at 20 Hz from the sonic anemometer mounted at the 80 m elevation of the met tower, i.e., at the turbine hub
height; relative humidity was observed at 1 Hz from the hygrometer mounted just below the 80 m sonic anemometer. We
examined the time series generated from these sensors using a sliding-window approach, where a window of constant width
is used to determine characteristic features of the dataset and then incrementally advanced. For this analysis, a window size
of 60 min at 10 min increments was selected. For each time interval, 20 Hz observations were reduced to 1 Hz averaging
consecutive sets of 20 data points, and the resultant wind vector was generated. Descriptive statistics were then calculated
for all variables to measure central tendency and variation. Finally, stationarity of the time series was determined by calcu-
lating the Augmented Dickey–Fuller (ADF) unit root statistic13 and its associated p-value, over the decimated wind speed
observations. Candidate intervals were ranked using the loss function (L):
"  2 #1
2
ADF o 2
LD a 1 C b  .180  ˛/ C c  u2 C d  RH 2
(1)
ADF max

s.t.: 140o  ˛  220o , and RH  60%. Here, ˛ is the wind direction (0o denotes wind coming from the North), u2 the
velocity variance, and RH the relative humidity. Constants a D 100, b D 50, c D 25, and d D 25 were used to weight
relative importance of dataset features for the analyses set forth herein.
According to this procedure, the data selected for this study spanned 1 h in the period ranging from 05/19/2012 at
11:20:00.000 to 05/19/2012 at 12:19:59.950. Within this period, the mean wind speed at hub height was Uhub  8 m s1 ,
with a wind direction almost coincident with the South-to-North axis. Figure 3 shows the time series of the wind velocity

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
Turbulence effects on a full-scale wind turbine L. P. Chamorro et al.

Figure 2. Schematic of the Eolos Wind Energy Research Field station (top); conceptual plan view of the tower and turbine in the
prevailing wind direction (bottom left); measured wind rose for the period January–July 2012.

Table II. Characteristics of the sensors in the met tower tower.


Sensor Manufacturer Model Sampling rate (Hz) Quantity Location (m)

3D sonic anemometer Campbell Scientific CSAT3 20 4 10, 30, 80, 129


Wind speed (Cup Anemom.) Met One 014-A 1 6 7,27, 52, 77
102, 126
Temperature Met One 083E 1 6 7, 27, 52, 77
and Relative humidity 102, 126
Precipitation gauge Texas Electronics TE525-L 0.0033 1 3

15
12
Uh (ms-1)

9
6
3
0
0 10 20 30 40 50 60
t (min)

Figure 3. Wind velocity at the turbine hub height measured with a CSAT 3D sonic anemometer.

at the turbine hub height for the selected period. It corresponds to an atmospheric boundary layer in nearly neutrally strati-
fied conditions with
 a relative
 humidity of approximately 59% within the turbine height. The bulk Richardson number was
Ri D gT z= T0 Uh2  0.04, measured between near ground (z1 D10 m) and at the turbine top tip (z2 D130 m),

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
L. P. Chamorro et al. Turbulence effects on a full-scale wind turbine

z D z2  z1 , T D T .z2 /  T .z1 / is the temperature difference (in K) between heights z1 and z2 , T0 is a reference
 
2 C U 2 1=2 with U
temperature (in K), g is the acceleration due to gravity, and Uh D UN E N and UE the South-
North and West-East components of the flow velocity, respectively. The Reynolds number based on the rotor diameter was
Re D Uhub dT =  5:1  107 . Figure 4 shows the mean characteristics of the boundary layer upstream of the turbine
at the met tower location. Additional details on the flow velocity distribution approaching to the turbine are described
in Figure 5. It should be noted in this figure that the wind directional
 shear within the rotor area (i.e., between the tur-
bine bottom and top tips heights) is ˇ D d ztop tip  d zbottom tip D 175  170:7 D 4:3o , where d .z/ is the wind
angle at height z. Although not negligible, wind shear should not have a measurable impact on the analysis. This low
value is consistent with the nearly neutral stratification, and much higher wind directional shear is expected in stably
stratified conditions.

3. UNSTEADY BEHAVIOR OF THE WIND TURBINE


In this section, we investigate the features of the power fluctuations and strain at the turbine foundation induced by the flow
turbulence. Cross-correlation between the approach flow at various heights, power, and strain are also discussed.

3.1. Turbine power and torque fluctuations

As pointed in Section 2, the turbine power and its angular velocity were sampled at a frequency of 1 Hz. The time series
of these two quantities are illustrated in Figure 6 for the selected period. During this time frame, the turbine operated
within the so-called Region II. In this region, the turbine should function to capture the maximum power. The mean power

Figure 4. Characteristics of the atmospheric boundary layer at the met tower location. a) Mean temperature, b) relative humidity, c)
mean velocity, d) turbulence intensity, and e) turbulent shear stress.

Figure 5. Polar representation of the wind velocity at the met tower location (left) and variability of the flow direction at various
heights during the time frame selected for the analysis.

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
Turbulence effects on a full-scale wind turbine L. P. Chamorro et al.

generated by the turbine was P D 1207 kW, with a normalized variation of P =P D 0:155, where P refers to the
root mean square (r.m.s.) of the power fluctuations. The mean angular velocity ! T D 1:51 rad s1 (=14.46 rpm) leads to a
tip-speed ratio  D ! T RT =Uhub D 8:7, where RT is the turbine radius. The spectral representation of the turbine power
time series (or its compensated form) reveals significant features and unveils the distinctive role of various turbulent scales
present in the boundary layer flow. Figure 7(a) illustrates the spectra of the incoming flow (ˆu ) at the turbine bottom tip,
hub and top tip heights, whereas Figure 7(b) shows the spectrum of the turbine power (ˆP ). A spectral windowing has
been applied to the spectra to highlight their general trend. As inferred from these plots, the spectral content on the turbine
power is characterized by three regions in the frequency domain. The first region is located on the high frequency range
where the power appears to be insensitive to the turbulent flow structure. The beginning of this region suggests that the
highest frequency (hereon referred to as critical frequency fc ) influencing the turbine is fc  Uhub =.dT =2/, i.e., which
corresponds to a length scale equal to the rotor radius. Here, the flow velocity at the turbine hub height (Uhub ) has been
used as the advection velocity scale of the approach flow since it practically equals the mean velocity across the rotor. In
this sense, the turbine diameter has also been used as a length scale. As small and microscales cannot affect the turbine,
the low energy and nearly flat spectral density in this region are probably signatures induced by mechanical and electrical
components of the turbine structure. A Butterworth low-pass filter with different cut-off frequencies to avoid anti-aliasing
indicated the same spectral trend. Although this finding is consistent with that observed in an axial-flow hydrokinetics
turbine model placed in an open channel flow,12 it needs to be tested and verified at other scales and flow conditions. The
second region is located in the intermediate range of frequencies (fL , fc ). The frequency fL , which marks the start of
this region, coincides with the beginning of the inertial subrange in the velocity spectrum (Figure 7(a)). Assuming that the
velocity at hub height is the appropriate velocity scale, the associated length scale corresponding to frequency fL is of the
order of the boundary layer thickness ı, i.e., Uhub =fL D 640 m  O.ı/. To determine the mechanisms defining fL , which
can be wind-alone or wind-turbine related, tests at different (turbine) scales are required. As inferred from Figure 7(b),
the spectral energy density of the power fluctuations shows a decay of the form ˆP / ./f 5=3 f 2 , which is close
to a f 7=2 scaling. This trend for ˆP , i.e., to be proportional to the power 7=2, has also been reported in Apt14 for
approximately the same range of frequencies. The relation ˆP / ./f 5=3 f 2 suggests complex non-linear interaction
between power and the energy density decay of the flow within the inertial subrange. We postulate that this relation is of
the following general form:

Figure 6. Time history of the turbine power and its angular velocity. Frequency sampling of 1 Hz.

Figure 7. a) Spectrum of the approach velocity at various heights (turbine bottom, hub and top tip) measured in the met tower. b)
Spectrum of the turbine power.

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
L. P. Chamorro et al. Turbulence effects on a full-scale wind turbine

Figure 8. Power spectrum of the turbine power normalized with spectrum of the velocity and G.f / / f 2 .

ˆP D G.f /ˆu ; G.f / / ./f 2 (2)

where G.f / is a damping/transfer function representing the turbine structure effects, including the rotor inertia, blade aero-
dynamics, discrete changes in the blades pitch (not important in this case), and the mechanical components of the turbine.
The non-linear scale-to-scale interaction between flow and turbine leads to a characteristic response of the turbine
 power,
which defines the properties of the transfer/damping function. Figure 8 illustrates the ratio ˆP = f 2 ˆu to highlight
the appropriate form of the damping function defined in Equation 2. Finally, and with reference again to Figure 7, there
is a third region in the power spectra for frequencies f < fL , which is dominated by large-scale oscillations. Within this
region, the dynamics of the power fluctuations appear to be roughly similar to the dynamics of atmospheric turbulence. It is
important to note that atmospheric stability plays an important role in modulating the characteristics of the flow turbulence,
and its influence on each of the spectral regions of the turbine power needs to be investigated.
It is worth noting that the characteristic frequency associated with the rotational motion of the turbine (fT D 0:24 Hz)
does not have a noticeable effect on the spectral content of the turbine power. In principle, the velocity gradient along
the rotor area should contribute to a periodic torque (and power) modulation superimposed to that of generated by the
flow. Although turbulent motions of sizes smaller than the rotor radius appear to have negligible impact on the turbine
power fluctuations (Region 1 in Figure 7(b)), strong and high frequency gusts of non-periodic character can affect the
power fluctuations. Their presence in the flow can be determined through the probability density function (PDF) of the
velocity increments u D u.t C t /  u.t /. For processes, exhibiting intermittency PDFs are characterized by heavy tail
distributions and deviate significantly from the standard Gaussian distribution.15
To determine the potential effects of intermittent wind gusts on the power fluctuations, we computed the PDF of the
power increments P D P .t C t /  P .t / and examined the characteristics of its tail. Figure 9 illustrates the PDF of
both the velocity increments for the approach flow at the turbine hub height (t D 1=20 s) and the the power increments
(t D 1 s). The normal (Gaussian) distribution fit of the two quantities is also included for comparison. The heavy tail of
the PDF distribution of the flow (Figure 9(a)) suggests the existence of intermittent gusts that are able to induce intermit-
tent behavior on the turbine power (Figure 9(b)). It is worth mentioning that the velocity increments with t D 1 s also
shows deviation from the normal distribution. The PDF of the power increments present a clear heavy tail distribution for
P > 40 kW. At negative P , however, the distribution does not deviate noticeably from its Gaussian counterpart. This
non-symmetric behavior at positive and negative increments can be attributed to the effect of the rotor inertia particularly
at sudden decrease in the wind velocity (i.e., u < 0).
The instantaneous turbine torque can be especially useful to infer additional insights of the turbine performance. The
turbine torque can be computed from the relationship:

Q D P ! 1 (3)

where Q is the torque in Nm, P the turbine power in W, and ! the angular velocity in rad s1 . In Figure 10, we plot Q as
a function of ! 2 . The Q D k! 2 relationship is a standard control scheme used in variable speed wind turbines operating
within Region II (for more details, see Johnson et al.16 ). Both quantities are normalized with respect to their time average
counterparts. Two distinctive regions intersecting at Q=Q D .!=!/2  1:12 characterize the Q  ! 2 relationship. As
inferred from the data, the wind turbine operated under the control scheme defined in the region .!=!/2  1:12 during

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
Turbulence effects on a full-scale wind turbine L. P. Chamorro et al.

Figure 9. PDF of the wind velocity (left) and power (right) increments. Normal distribution is included for comparison.

Figure 10. Normalized Q  ! 2 relationship. Variables are normalized by their time-averaged counterparts. Change in the control
characteristics at Q=Q D .!=!/2 1.12.

90% of the time interval selected for this analysis. At .!=!/2 > 1:12, the turbine experienced discrete and collective
changes in the blade pitch. Such effect is illustrated in Figure 10. The significant torque increase (and therefore power)
associated to the blade pitch evidence the critical relevance of the control algorithm for both torque (power) maximization
and turbulence-driven loading reduction.

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
L. P. Chamorro et al. Turbulence effects on a full-scale wind turbine

Figure 11. a) Polar distribution of the mean strain at the turbine foundation; b) mean strain. Error bars indicate the standard deviation.

3.2. Fluctuating strain at the turbine foundation

As indicated in Section 2.2, 20 single axis strain gauges are used to determine the characteristics of the fluctuating forces
experienced by the turbine at its foundation. Such forces are largely the result of the thrust experienced by the turbine rotor.
Unlike the turbine power and its rotational speed, the strain () was measured at a frequency of 20 Hz. The average distri-
bution of the normalized strain along the turbine foundation is depicted in Figure 11. As expected, half of the foundation
structure is under relative tension (with respect to the no wind case). The strain distribution shows a slight departure of
roughly 8o between the nacelle direction and the strain symmetry axis, which is the 0o  180o line in Figure 11(a). The
offset of the symmetry axes between the nacelle and strain is probably due to the Coriolis effect within the turbine rotor,
which is depicted in the polar representation of the wind velocity profile of Figure 5. As expected, the maximum strain
variations (Figure 11(a)) and standard deviations (Figure 11(b)) are located in the vicinity of the 0o and 180o angles.
To determine the governing mechanisms that modulate the strain fluctuations, we show in Figure 12 the spectra of the
strain at 0o and 180o (i.e., in the maximum relative tension at 180o and compression at 0o ). This figure suggests that the
forces experienced by the turbine are modulated by several features of the flow turbulence and the turbine. Like the turbine
power, the spectral content of the strain is characterized by three regions. One is modulated entirely by the large-scale fea-
tures of the flow with length scales on the order of the boundary layer thickness and larger. The intermediate region follows

Figure 12. Spectrum of the strain at the turbine foundation (0o and 180o ).

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
Turbulence effects on a full-scale wind turbine L. P. Chamorro et al.

Figure 13. Correlation function between the turbine power and flow at bottom tip, hub, and top tip heights (left) and between power
and strain at 0o and 180o angles (right).

the same scaling law of the turbine power in practically the same intermediate range (fL ; fc =2), i.e., ˆ / ./f 5=3 f 2 .
The third region roughly follows the structure of the inertial subrange of the flow turbulence. This region is modulated by
length scales smaller than the rotor radius. The fluctuating forces induced by the blade rotation, which falls within this
region, are clearly felt by the turbine structure. Strong peaks at the turbine rotation (fT ) and blade passing (fbpf D 3fT )
frequencies and their harmonics are evident from the spectrum in Figure 12. The major difference between the spectral
characteristics of the strain and the turbine power are at length scales smaller that the rotor size. Here, the strain shows to
be particularly sensitive to the small-scale features of the flow and to the blade rotation. Small and fast changes on the flow
velocity can induce measurable torque at the turbine foundation because of the large lever-arm distances from this location.
Although not shown herein because of space considerations, the PDF of the strain increments indicates that intermittent
gusts also affect the strain in a similar manner as that shown in the power PDF shown in Figure 9.

3.3. Correlation between flow and turbine quantities

Additional understanding of the complex interaction between the flow, power, and strain can be obtained by examining
how these quantities are cross-correlated. The correlation function between two signals A and B is defined as follows:

A.t /B.t  /
RA;B . / D  1=2 (4)
A02 B 02

where  is the time delay between the A and B measurements, primes indicate fluctuations with respect to the time averaged
values, and overbars indicate time averaging.
Figure 4(a) illustrates the cross-correlation between the turbine power and flow velocity at various heights (bottom
tip, hub, and top tip), and Figure 4(b) depicts the correlation between the power and strain at the front (180o ) and back
(0o ) of the turbine foundation. The power and velocity correlations suggest that the velocity at the turbine hub height is
the
 best descriptor of the flow turbulence within the rotor area. The corresponding maximum correlation at hub height
RP ;Uhub  0:55 is significantly higher than that at the other heights examined (RP ;U  0:3 and 0.28 at the top and
bottom tips, respectively). The maximum correlation is achieved at a time lag consistent with the separation L (=160 m;
Figure 2) between the met tower and the turbine, i.e.,   L=Uhub . The power and strain at both 0o and 180o are highly
correlated (RP ;  0:95). As observed in Figures 7 and 12 from the spectral features of these quantities, the differences
between the behavior of these quantities are confined mostly on the contribution of the small scales, where the turbine
power is not affected to flow motions with length scales lower that the rotor size. The high correlation between power and
strain necessarily implies that the correlation function between flow and strain is almost identical to that obtained for the
power and flow.

4. SUMMARY AND CONCLUSIONS


The fluctuating power generated by a full-scale 2.5 MW horizontal-axis wind turbine and the strain at its foundation are
shown to be modulated by the scales of the turbulence in a complex way. Three distinct regions in the spectral domain
define the flow and structure interaction. In the first region, the turbine power was insensitive to turbulent scales smaller

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we
L. P. Chamorro et al. Turbulence effects on a full-scale wind turbine

than the rotor radius. In the intermediate region, the spectral content of the power fluctuations (ˆP ) and flow turbulence
(ˆU ) evidenced a non-linear relation of the general form ˆP D G.f /ˆP , where G.f / / ./f 2 is a transfer/damping
function suspected to depend on the turbine characteristics (e.g., inertia). Direct influence of the flow on the turbine power
fluctuations occurred at the very large scales of motion (with sizes on the boundary layer thickness or larger). This behavior,
however, needs to be verified at other scales and flow conditions as well as with numerical models. In particular, future
efforts consider the evaluation and comparison with FAST (fatigue, aerodynamics, structures, and turbulence) model to
complement the findings obtained in this research.
The particular behavior of the power fluctuations in the spectral domain and its relation with the spectral structure of the
turbulence enables the development of improved control strategies for reducing power variability. It also provides insights
for engineering design. For example, numerical simulations dealing with turbine unsteady behavior can eventually focus
on turbulence scales larger than the rotor size, or equivalently, a survey of flow turbulence at a specific location can limit
the temporal resolution to those on the order O.dT =Uhub / if the turbine power output is of primary interest. Like the tur-
bine power, the strain fluctuations at the turbine foundation also displayed three characteristic regions. However, the region
defined by subrotor scales appeared to be responsive to the flow and followed the inertial subrange of the turbulence. Veloc-
ity and power increment analysis demonstrated that intermittent gusts are able to produce intermittent power output and
forces at the turbine foundation. Finally, computed cross-correlations between the turbine power and flow at various heights
indicate that the turbulence around the rotor area is best characterized by the wind measured at the turbine hub height.

ACKNOWLEDGEMENT
Funding was provided by the US Department of Energy (DE-EE0002980).

REFERENCES
1. Ernst B, Wan YH, Kirby B. Short-term power fluctuation of wind turbines: analyzing data from the german 250-mw
measurement program from the ancillary services viewpoint. NREL Report, 1999. Presented at the Windpower ’99
Conference Burlington, Vermont June 20-23.
2. Barlas TK, van Kuik GAM. State of the art and prospectives of smart rotor control for wind turbines. The science
of making torque from wind. Journal of Physics. Conference Series: The Science of Making Torque from Wind 2007;
75(012080): 233–239.
3. Veers P, Winterstein S. Application of measured loads to wind turbine fatigue and reliability analysis. Journal of Solar
Energy Engineering 1998; 120(4): 233–239.
4. Moriarty P, Holley W, Butterfield S. Extrapolation of extreme and fatigue loads using probabilistic methods.
NREL/TP-500-34421 32, 2004.
5. Ran L, Bumby J, Tavner P. Use of turbine inertia for power smoothing of wind turbines with DFIG, Proceedings 11th
International Conference Harmonics Quality Power, Cairo, Egypt, 2004; 106–111.
6. Luo C, Shen B, Ooi B. Strategies to smooth wind power fluctuations of wind turbine generator. IEEE Transactions on
Energy Conversion 2007; 22(2): 341–349.
7. Bossanyi EA. Individual blade pitch control for load reduction. Wind Energy 2003; 6(2): 119–128.
8. Bossanyi EA. The design of closed loop controllers for wind turbines. Wind Energy 2000; 3(3): 149–163.
9. Geyler M. Advanced pitch control for wind turbines. Technical Report 2001.001, Delft University of Technology -
DUWIND, 2001.
10. Chopra I, Gessow A. Recent progress on the development of a smart rotor system, In 26th European Rotorcraft Forum,
The Hague, Netherlands, 2000; 26–29.
11. Rosen A, Sheinman Y. The power fluctuations of a wind turbine. Journal of Wind Engineering and Industrial
Aerodynamics 1996; 59: 51–68.
12. Chamorro LP, Hill C, Morton S, Ellis C, Arndt REA, Sotiropoulos F. On the interaction between a turbulent open
channel flow and an axial-flow turbine. Journal of Fluid Mechanics 2013; 716: 658–670.
13. Elliott G, Rothenberg T, Stock J. Efficient test for an autorregressive unit root. Econometrica 1996; 64(4): 813–836.
14. Apt J. The spectrum of power from wind turbines. Journal of Power Sources 2007; 169: 369–374.
15. Peinke J, Barth S, Bottcher F, Heinemann D, Lange B. Turbulence, a challenging problem for wind energy. Physica A
2004; 338: 187–193.
16. Johnson K, Fingersh L, Balas M, Pao L. Methods for increasing region 2 power capture on a variable speed hawt, 23rd
ASME Wind Energy Symposium, Reno, Nevada, 2004; 103–113.

Wind Energ. (2014) © 2014 John Wiley & Sons, Ltd.


DOI: 10.1002/we

You might also like