Validation of Doppler LIDAR for Wind Resource Assessment

Validation of Doppler LIDAR for Wind Resource Assessment
Applications
Daniel W. Jaynes, B.Sc.
Jon G. McGowan, Ph. D.
Anthony L. Rogers, Ph. D.
James F. Manwell, Ph. D.
Renewable Energy Research Laboratory
Dept. of Mechanical and Industrial Engineering
University of Massachusetts, Amherst, MA 01003, USA
Telephone: +1-413-545-2644, fax. +1-413-545-1027, Email: [email protected]
Abstract
The use of LIDAR (LIght Detection And Ranging) for wind resource assessment is an
attractive technology as hub height wind data are required to predict the power production of
increasingly large wind turbines. Among the appealing features of laser remote sensing is the
ability to measure wind speed and direction over the entire rotor plane of a wind turbine. An
analysis of the LIDAR measurement technique is useful before the instrument is used
autonomously for a wind resource assessment measurement campaign. This paper presents the
results of a long-term data validation experiment where the accuracy of LIDAR wind speed
measurements are compared to data recorded by cup anemometers mounted on a coastal tall
tower at heights of 61 m, 87 m and 118 m. Next, the effects of volume averaging are presented
to explain the behavior of LIDAR wind speed data at ranges over 100 meters. The
performance of the LIDAR during periods of precipitation is also given to demonstrate its
effectiveness in undesirable weather conditions.
1. Introduction
A challenging step in the procedure by which wind energy projects are developed is the
acquisition of accurate wind speed and direction data that can successfully characterize the
wind shear profile at a candidate site. It is well known that the standard method of shear
characterization using data from 40-50 m met towers introduces large levels of uncertainty in
the estimation of long-term averages because the wind shear models (log law and power law)
are often poor predictors of the true wind shear profile [1], [2]. Thus, access to financial
capital for project development can be problematic for sites where the wind shear is not fully
characterized with hub height wind data. In recent years, remote sensing instruments have
become more common due to their improved affordability and ability to directly address the
challenge of wind shear estimation. While SODAR (Sound Detection and Ranging) is a viable
option for remote wind data measurement, it is plagued by noise annoyance issues and echo
interactions with neighboring trees or structures [3], [4]. LIDAR operates silently by the
emission of eye-safe laser radiation rather than high frequency “chirps;” thus it is inherently
immune to echo interactions that may corrupt wind speed measurements.
As part of a suite of wind resource measurement instruments, the University of Massachusetts
Renewable Energy Research Laboratory (RERL) recently acquired a Doppler LIDAR system,
which is capable of remotely measuring wind data at up to five user-programmable heights
above ground level. This system merges established laser technology with widely available
and more affordable internal components to make it commercially accessible. Manufactured
by Qinetiq of England, the ZephIR LIDAR is specifically designed for wind energy resource
assessment applications. This system offers great promise in terms of its ability to provide
accurate wind data at the hub height of a modern wind turbine. Before the instrument is
dispatched for autonomous data collection, the validity of LIDAR measurements are first
explored in a side-by-side data validation campaign. To accomplish this objective, an
experimental test was carried out whereby the data recorded by the LIDAR were compared to
those obtained with cup anemometers that were treated as an experimental control.
2. LIDAR Background
Laser remote sensing technology, as we know it today, was established by the advent of the
laser in the early 1960’s [5]. In recent years, Doppler LIDAR systems have benefited from
robust and inexpensive fiber-optic telecommunication industry components [6], [7]. As a
result, such instruments have become more popular in the field of wind energy research [8],
[9], [10].
The basis of wind speed measurement with Doppler LIDAR involves measuring the shifted
frequency of scattered light that results from the illumination of naturally occurring
atmospheric aerosols. These aerosols are assumed to move at the same approximate speed as
the wind field in which they are found. In accordance with the Doppler phenomenon, the
shifted frequency of the backscattered light is directly proportional to the velocity of target
aerosols. Thus, this technique requires only a simple transformation to obtain the wind
velocity once the Doppler shifted frequency of target aerosols is determined.
The LIDAR system under investigation is a fiber-optic based continuous wave coherent laser
radar (CLR). The laser that is incorporated in the system has an output power of 1-Watt with a
measurement range of 10 m-150 m (according to the manufacturer). The laser transmits at
1.55 µm ( 1.93x1014 Hz), which is an eye-safe wavelength that does not pass through the human
retina. The instrument is a coaxial system in which both the emitted and backscattered light
share common optics. Figure 1 shows the LIDAR instrument. The small protruding mast on
the instrument houses additional sensors as well as a wind direction sensor.
Figure 1: Qinetiq ZephIR Lidar Instrument
2.1. LIDAR Wind Speed Measurement
Power Spectral Density [Arb]
After the LIDAR detector converts the backscattered light to an electric signal, the signal is
digitally sampled at a rate of 100MHz by the data acquisition system that is incorporated in the
system. Next, the signal is sent through a low-pass filter with a cut-off frequency of 50 MHz.
A 512-point fast Fourier transform (FFT) is then applied to the digitized signal to determine its
frequency content. In order to increase the signal to noise ratio, 4,000 of these individual
power spectra are averaged over an integration time of 20 ms to create each wind speed
estimation record. After the averaging step, a clear Doppler frequency peak appears in the
wind spectrum. A generic wind spectrum is shown in Figure 2.
Frequency [Hz]
Figure 2: Doppler-Shifted Wind Spectrum [11]
Turbulent airflow across the LIDAR probe volume can cause more than one peak in the wind
spectrum, but errors can be minimized by calculating a running average [12].
Each wind spectrum is produced and checked for corruption resulting from the presence of
clouds. Then, an algorithm that calculates the centroid of the Doppler spectrum above a predetermined threshold is applied to determine the shifted frequency of the scattered light. The
frequency behavior of the scattered light is then related to the line-of-sight (LOS) wind
velocity by Equation 1
2v
f Shifted = LOS ,
(1)
λ0
where vLOS is the line-of-sight wind speed and λ0 is the wavelength of transmitted light (1.55
µm). Equation 1 can be rearranged to show that the line-of-sight wind speed is determined by
multiplying the shifted Doppler frequency by a simple conversion factor of 0.775 ms-1 per
λ0
. Frehlich and Jorgensen et al. contend that this calibration factor suffers
2
negligible drift (less than 0.2%) over long periods of time because the wavelength of the
emitted light is known to transmit with robust stability [13], [12]. Thus, the LIDAR instrument
does not require calibration.
MHz, or
The LIDAR emits laser radiation in a circular pattern by reflecting the laser beam off of a
spinning optical wedge, via the velocity azimuth display (VAD) scanning technique. The
wedge is positioned such that the beam is transmitted at an angle of 30 degrees from vertical,
thereby creating a cone shaped probe volume. The line-of-sight velocity data measurements
consequently become a function of scan angle as shown in Equation 2 where angle φ is the
azimuth scan angle.
vLOS = a cos(φ − b) + c
(2)
The parameters a, b and c in Equation 2 are obtained by applying a non-linear least squares fit
to the line-of-sight data that are collected by the LIDAR. The wind speed can then be
determined by substitution in Equations 3 - 5 where u is the horizontal wind speed, b is the
direction of approaching wind, w is the vertical wind speed and θ is the beam offset angle of
30 degrees from zenith.
c
cos(θ )
b = Wind Bearing ± 180 o
a
u=
sin(θ )
w=
(3)
(4)
(5)
When the line-of-sight wind speed vs. azimuth angle is plotted on polar axes the result is
shown in Figure 3 for a three second measurement period where 150 data points are plotted.
The relationship of the data to the best-fit approximation (solid line) suggests that the wind
flow across the probe volume is uniform and the slight asymmetry in the lobe sizes indicates
the presence of a vertical wind speed component.
Figure 3: ZephIR Polar Line of Sight Wind Speed Plot in m/s vs. Azimuth Angle
The wind direction shown in Figure 3 is approaching from the NNE direction. The LIDAR
meteorological mast can be used to resolve any ambiguity with respect to a possible 180degree wind direction discrepancy that can occur when the least squares fit to the line-of-sight
data is poor.
2.2. LIDAR Range Resolved Measurements
Wind speed and direction data can be measured at various heights by focusing the laser beam
at a preset range. In order to understand the way in which the LIDAR measures at various
heights, the hourglass-shaped beam geometry must be considered.
πW02
where W0 is the waist
The laser beam can be characterized by its Rayleigh length , z R =
λ
1
radius of the beam in the optical fiber and λ is the wavelength of the laser beam [14]. The
Rayleigh length represents the point where the most amount of laser beam power is
concentrated (in accordance with a Lorentzian distribution of energy in the beam). This
parameter is used to describe the LIDAR’s probe depth along the axis of transmission. An
approximation of the LIDAR probe depth ( ∆z or sometimes, 2ZR) is given by Equation 6
where a0 is the laser beam radius at the output of the lens and p′ is the beam waist position
along the axis of beam transmission [15].
λp′ 2
(6)
∆z ≈
2a02
When the optical fiber near the lens is adjusted along the axis of transmission, the beam can be
focused at a preset height. At a distance p′ from the lens, the beam waist cross-section is given
by Equation 7 [16].
  ′ 2 
p
W ( p′) = W 0 1+   
(7)
  zR  
The geometry of the laser beam is shown in Figure 4 where the position of the optical fiber end
near the lens is defined as p and the beam waist position along the axis of beam transmission is
defined as p′ . In Figure 4, p<< p′ .
LIDAR
Figure 4: Laser Beam Propagation Characteristics [8]
Figure 5 illustrates the behavior of the beam waist at varying measurement height.
1
The Rayleigh length of the laser beam is the distance from the beam waist (in the propagation direction) where
the beam radius is increased by a factor of square root of 2
Figure 5: Beam Waist Detail [8]
As p moves closer to the focal point of the lens, the laser will focus at a larger distance away
from the transmitter with an increasingly large probe volume. As such, the behavior of the
probe volume can be shown to increase as the fourth power of the range [7].
The probe depth and probe volume characteristics at various measurement heights are given
below in Table 1 and in graphical format in Figure 6. Both the probe depth and probe volume
increase as a function of measurement height. The role that this phenomenon plays with
respect to the accuracy of wind speed measurements is explored in greater detail in Section 4.2.
Height
[m]
40
60
80
100
120
140
p' [m]
Probe Depth [m]
46.19
69.28
92.38
115.47
138.56
161.66
2.87
6.46
11.48
17.94
25.83
35.16
Probe Volume
[cm^3]
45.5
230.4
728.2
1777.8
3686.4
6829.5
Table 1: ZephIR Beam Geometry Characteristics at Various Heights
8000
40
7000
35
6000
30
5000
25
4000
20
3000
15
2000
10
1000
5
0
0
50
100
Measurem ent Height [m ]
Probe Volume [cm^3]
Probe Depth [m]
Figure 6: ZephIR Beam Properties
0
150
Probe Depth [m]
Probe Volume [cm^3]
ZephIR Beam Properties
3. Experimental Setup
The LIDAR data validation experiment was performed at a coastal site in the town of Hull,
Massachusetts. Cup anemometers and wind direction sensors were installed on a met mast at
three heights: 61 m, 87 m, and 118 m. The mast is a three-sided lattice tower with an
approximate face width of 1 meter on each side. The tower is surrounded in every direction by
a flat salt marsh. A residential area to the north, south and east borders the vicinity but no tall
objects that would directly obstruct the airflow at any level were present during the experiment.
Furthermore, the terrain surrounding the site is relatively flat and the possibility of abnormal
wind speeding caused by local terrain aberrations is not likely.
While the LIDAR is capable of placement directly next to a tall structure, the closest that the
instrument could be placed to the tower for this experiment was 160 meters to the east. This
location was necessary to acquire a stable power supply from a nearby building.
The placement of the LIDAR during the measurement campaign resulted in the partial
obstruction of its met mast. This situation caused temporary errors in the wind direction data.
Thus, the results of this experiment will primarily focus on the comparison of wind speed data
although wind direction measurements will be presented as well.
3.1. Sensor Equipment
The mast is equipped with wind speed and direction sensors that are manufactured by NRG
Systems. The following equipment is installed on the tower:
• One Y-shaped sensor boom at each level that hosts two anemometers and one wind
direction sensor. The booms face due west and the sensors are located approximately
4.4 meters away from the closest tower leg.
• Six NRG Maximum 40 Anemometers, standard calibration (Slope - 0.765 m/s/Hz,
Offset – 0.350 m/s). Two anemometers are located at 118 m, two at 87 m and two at a
height of 61 m.
• Three NRG 200P Wind direction vanes. One vane is located at each height: 118 m, 87
m and 61 m.
While the NRG Maximum 40 cup anemometer is not IEC approved for power curve
measurements, it has emerged as a popular choice for wind resource assessment in the United
States. Thus, the treatment of the NRG anemometer as the experimental control is designed to
show the extent to which the LIDAR can substitute for standard cup anemometry in a wind
resource assessment campaign.
4. Experimental Results
The 10-minute average wind speed data that are summarized in this paper were collected in the
winter months between December 2nd 2006 and February 13th 2007. Mostly dry, but very cold
conditions were common throughout the measurement campaign. Despite the cold conditions,
the overall data recovery at each height was good. The measurement campaign data recovery
rates are summarized in Table 2.
Height [m]
Data Recovery [%]
61
87
118
98.6
98.1
96.6
Table 2: LIDAR Data Recovery
During the 70-day measurement campaign, a total of approximately 10,500 data records were
available for comparison. The overall measurement period includes a number of days in which
LIDAR operation was interrupted due to problems with the external power source.
4.1. Wind Speed Data Validation
The wind speed comparison at each of the three sensor heights is shown in Figure 7 through
Figure 9. The data presented below have been filtered by:
• Isolating and removing data that were found in direction sectors where consistent
anemometer speed-up or slow-down effects were present
• Eliminating wind velocity data below 1 m/s where the NRG Maximum 40 anemometer
does not accurately record the wind speed [17]
• Removing sporadically spurious cup anemometer measurements that were caused by
extraneous signal noise
As shown in the following figures, the LIDAR achieves excellent correlation with cup
anemometry despite the fact that it was located approximately 160 meters away from the met
mast.
Figure 7: Wind Speed Comparison at 61 m
Figure 8: Wind Speed Comparison at 87 m
Figure 9: Wind Speed Comparison at 118 m
The correlation and linear fit equations for the data at each measurement level are summarized
in Table 3.
Correlation
Coefficient [ ]
Linear Fit
61 m
87 m
118 m
0.984
0.984
0.978
y=0.978x+0.126
y=0.980x+0.160
y=0.957x+0.238
Table 3: Wind Speed Data Correlation Summary
4.2. Volume Averaging Effects
The degree to which the cup and LIDAR wind speed data are correlated is shown to diminish
slightly as measurement range increases. The slight correlation degradation at ranges over 100
meters agrees with the results presented in [18]. This phenomenon is associated with the
increasing size of the circular disc of air that is probed by the LIDAR in its VAD scan and the
increasing size of the LIDAR probe depth [15]. The area of the circular disc of air that is
probed by the LIDAR increases with the square of the measurement height as shown in Table
4.
Measurement
Height [m]
61
87
118
Probe Disc
Diameter [m]
70.4
100.5
136.3
Probe Disc Area
[m2]
3892
7933
14591
Table 4: Probe Disc Diameter at Each Measurement Height
The effects of volume averaging are further illustrated by observing the level of correlation
between wind speed error, ε, and the vertical wind speed gradient. The vertical wind speed
gradient is the ratio of cup measurements at the upper and lower sensor levels while the wind
speed error is defined in Equation 8 where U LIDAR is the horizontal velocity measured by the
LIDAR and UCup is the horizontal velocity measured by the cups.
U
− UCup
ε = LIDAR
(8)
UCup
Figure 10 and Figure 11 show ε as a function of vertical wind speed gradient. In these figures,
LIDAR wind velocity measurements become increasingly correlated with wind speed error as
measurement range increases.
Figure 10: Wind Speed Error vs. Vertical Wind Speed Gradient, 87 m
Figure 11: Wind Speed Error vs. Vertical Wind Speed Gradient, 118 m
The wind speed error associated with LIDAR volume averaging increases when wind shear
across the depth of the beam probe is strong and when the airflow in the circular disc of air is
non-uniform. This relationship illustrates the contrasting measurement characteristics
associated with volume-averaged versus point-averaged wind speed estimations.
4.3. Wind Direction Data Validation
An example of the wind direction time series is shown in Figure 12. This illustrates the
behavior of the LIDAR and tower wind direction data and the isolated instances of
measurement ambiguity that were caused by the partial obstruction of the LIDAR mast during
the experimental measurement campaign. As shown in Figure 12, the LIDAR generally
records the correct wind direction even though its mast was obstructed. Given the unavoidable
circumstance of the LIDAR mast obstruction, the direction data exhibit strong correlation with
tower-mounted wind vane data.
180-Degree Direction Ambiguity
Figure 12: Sample Wind Direction Time Series Showing the Wind Direction Ambiguity
Associated with the LIDAR Mast Obstruction
4.4. LIDAR Performance During Periods of Precipitation
The LIDAR is designed to autonomously report the vertical and horizontal wind speed
components in any environmental condition that occurs in the lower atmospheric boundary
layer. An important presumption in the operation of the LIDAR is that light scatter originates
from atmospheric particles that move at the same approximate speed as the wind field in which
they are found. This assumption fails when the LIDAR beam intersects a solid object such as
falling snow or rain droplets. The downward velocity component of falling objects adds to the
line-of-sight velocity and can therefore skew the vertical and horizontal wind speed
measurement estimations. As the falling object (hail, snow, sleet or rain) increases in size, the
measurement accuracy is affected by a larger degree [10].
During the period of the LIDAR data validation experiment, the rain sensor that is integrated in
the LIDAR system indicated very few instances of rain or snow. However, 406 data points
were identified as events during which rain or snow were present. This corresponds to a total
of approximately 68 hours of precipitation. The concurrent wind velocity data during these 68
hours are compared below in Figure 13, Figure 14 and Figure 15. These figures suggest that
the presence of rain or snow slightly degrades the accuracy of the LIDAR.
Figure 13: LIDAR and Cup Anemometer Wind Speed Comparison During Precipitation
Events at 61 m
Figure 14: LIDAR and Cup Anemometer Wind Speed Comparison During Precipitation
Events at 87 m
Figure 15: LIDAR and Cup Anemometer Wind Speed Comparison During Precipitation
Events at 118 m
Table 5 provides a summary of the data correlation during periods of precipitation. The results
of this inquiry demonstrate that LIDAR measurement accuracy is affected by the presence of
precipitation that disturbs the beam’s probe volume.
Measurement
Height [m]
Wind Speed Correlation
Coefficient During Periods of
Precipitation [ ]
Overall Wind
Speed
Correlation
Coefficient [ ]
Percent Change
During Periods of
Precipitation [%]
61
87
118
0.975
0.951
0.932
0.984
0.984
0.978
-0.90%
-3.35%
-4.64%
Table 5: Summary of Data Correlation During Periods of Precipitation
While the wind speed measurements during such periods are generally accurate, there are
instances in which deviations of up to 7.5 m/s are present. Although rainfall may not
significantly influence the LIDAR’s ability to accurately predict the long-term mean wind
speed, care should be exercised when interpreting individual measurements that are recorded
when precipitation is present.
5. Summary
The ZephIR Doppler LIDAR has proven its ability to accurately measure wind speed and
direction at heights up to 118 meters in a long term comparison to in-situ tower anemometry.
Wind speed data correlations of 0.978 and higher and data recovery rates of 96.6% and better
were achieved throughout the approximately 70-day measurement campaign where the NRG
Maximum 40 cup anemometer was treated as the experimental control.
The performance of the LIDAR during the validation experiment demonstrates its usefulness
for wind power resource assessment applications. In addition to the accuracy of the data
recorded by the LIDAR, it is small enough to be deployed by a team of only two people and
when it is fully assembled, the system does not require permits of any kind. Because the unit
runs silently, it is not affected by echo interactions and its beam has no side lobes, which are
concerns with wind speed measurement where SODAR technology is used. The LIDAR is
also capable of accurate measurements over the entire rotor plane of a modern wind turbine,
thereby drastically reducing the level of uncertainty in long term hub height predications.
6. Acknowledgements
The Town of Hull, Massachusetts, the U.S. Department of Energy (in collaboration with the
Massachusetts Division of Energy Resources) and the Massachusetts Technology
Collaborative provided financial support for this work.
References
1.
2.
3.
4.
5.
6.
7.
Lackner, M.A., A.L. Rogers, and J.F. Manwell, Wind Energy Site Assessment and
Uncertainty 2006, Renewable Energy Research Laboratory.
Ray, M., A. Rogers, and J. Manwell, Accuracy of Wind Shear Models for Estimating
the Wind Resource in Massachusetts. 2005, Renewable Energy Research Laboratory.
Rogers, A.L., et al., Addressing Ground Clutter Corruption of SODAR Measurements,
in ASME Wind Energy Symposium, 2007 AIAA Aerospace Science Meeting. 2007:
Reno, NV.
Hensen, W., Wind Resource Assessment Using SODAR at Cluttered Sites, in European
Wind energy Conference. 2006: Athens, Greece.
Argall, P.S. and R.J. Sica, Lidar, in Encyclopedia of Imaging Science and Technology.
2002, Wiley.
Karlsson, C.J., et al., All-fiber Multifunction Continuous-wave Coherent Laser Radar at
1.5 µm for Range, Speed, Vibration, and Wind Measurements. Applied Optics, 2000.
39(21): p. 3716-3726.
Harris, M., G. Constant, and C. Ward, Continuous-wave Bistatic Laser Doppler Wind
Sensor. Applied Optics, 2001. 40(9): p. 1501-1506.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
Danielian, R., et al., Surface-Layer Wind and Turbulence Profiling from LIDAR:
Theory and Measurements, in European Wind Energy Conference. 2006: Athens,
Greece.
Hansen, R.S., et al., Laser Anemometry for Control and Performance Testing of Wind
Turbines. 2001, Risoe National Laboratory.
Harris, M., M. Hand, and A. Wright, Lidar for Turbine Control. 2005, National
Renewable Energy Laboratory.
Smith, D.A. and e. al. Wind Lidar Evaluation at the Danish Wind Test Site in Hovsore.
in European Wind Energy Conference. 2004. London, UK.
Jorgensen, H., et al. Site Wind Field Determination Using a CW Doppler LidarComparison with Cup Anemometers at Riso. in The Science of Making Torque from
Wind. 2004. Delft.
Frehlich, R. and M.J. Yadlowsky, Performance of Mean-Frequency Estimators for
Doppler Radar and Lidar. Journal of Atmospheric and Oceanic Technology, 1994. 11:
p. 1217-1230.
Goldsmith, P.F., Quasioptical Systems. 1998: IEEE Press.
Banakh, V.A., et al., Representativeness of Wind Measurements with a CW Doppler
Lidar in the Atmospheric Boundary Layer. Applied Optics, 1995. 34(12): p. 2055-2067.
Fowles, G.R., Introduction to Modern Optics, Second Edition. 1989: Dover
Publications.
NRG. NRG Maximum 40 Specifications. [cited; Available from:
http://www.nrgsystems.com/store/product_detail.php?cd=11&s=1899.
Albers, A., Evaluation of ZephIR. 2006, WindGuard.