Doppler lidar investigation of wind turbine wake characteristics and atmospheric turbulence under different surface roughness

Abstract

Four field experiments based on Pulsed Coherent Doppler Lidar with different surface roughness have been carried out in 2013-2015 to study the turbulent wind field in the vicinity of operating wind turbine in the onshore and offshore wind parks. The turbulence characteristics in ambient atmosphere and wake area was analyzed using transverse structure function based on Plane Position Indicator scanning mode. An automatic wake processing procedure was developed to determine the wake velocity deficit by considering the effect of ambient velocity disturbance and wake meandering with the mean wind direction. It is found that the turbine wake obviously enhances the atmospheric turbulence mixing, and the difference in the correlation of turbulence parameters under different surface roughness is significant. The dependence of wake parameters including the wake velocity deficit and wake length on wind velocity and turbulence intensity are analyzed and compared with other studies, which validates the empirical model and simulation of a turbine wake for various atmosphere conditions.

Full text

Doppler lidar investigation of wind turbine
wake characteristics and atmospheric
turbulence under different surface roughness
XIAOCHUN ZHAI,1 SONGHUA WU,1,2* AND BINGYI LIU1
1Ocean Remote Sensing Institute, Ocean University of China, Qingdao 266100, China
2 Laboratory for Regional Oceanography and Numerical Modeling, Qingdao National Laboratory for
Marine Science and Technology, Qingdao 266100, China
*wush@ouc.edu.cn
Abstract: Four field experiments based on Pulsed Coherent Doppler Lidar with different
surface roughness have been carried out in 2013-2015 to study the turbulent wind field in the
vicinity of operating wind turbine in the onshore and offshore wind parks. The turbulence
characteristics in ambient atmosphere and wake area was analyzed using transverse structure
function based on Plane Position Indicator scanning mode. An automatic wake processing
procedure was developed to determine the wake velocity deficit by considering the effect of
ambient velocity disturbance and wake meandering with the mean wind direction. It is found
that the turbine wake obviously enhances the atmospheric turbulence mixing, and the
difference in the correlation of turbulence parameters under different surface roughness is
significant. The dependence of wake parameters including the wake velocity deficit and wake
length on wind velocity and turbulence intensity are analyzed and compared with other
studies, which validates the empirical model and simulation of a turbine wake for various
atmosphere conditions.
© 2017 Optical Society of America
OCIS codes: (010.3640) Lidar; (280.3340) Laser Doppler velocimetry; (280.7060) Turbulence.
References and links
1. P. Builtjes and P. Vermeulen, “Turbulence in wind turbine clusters,” in Proc. 4th Int. Symp. on Wind Energy
Systems (1982), pp. 449–464.
2. J. Ross and J. Ainslie, “Wake measurements in clusters of model wind turbines using laser Doppler
anemometry,” in Proceedings of the Third BWEA Wind Energy Conference, Cranfield (1981), pp. 172–184.
3. D. Green and A. Alexander, “Measurement of velocity and turbulence profiles in flow situations relevant to wind
turbine performance. Final Report on ETSU contract E,” (5A/CON/5003/177/026, Loughborough University,
1985).
4. A. Talmon, “The wake of a horizontal axis wind turbine model, measurements in uniform approach flow and in a
simulated boundary layer,” TNO Division of Technology for Society Tech. Rep, 85–01021 (1985).
5. D. Milborrow and J. Ross, “The influence of turbulence and rotor thrust on wind turbine wake characteristics,”
CERL Memorandum No TPRD/L/AP/0098/M83 (1983).
6. P. Alfredsson, F. Bark, and J. Dahlberg, “Some properties of the wake behind horizontal axis wind turbines,” in
3rd international symposium on wind energy systems (1980), pp. 469–484.
7. S. Chowdhury, J. Zhang, A. Messac, and L. Castillo, “Unrestricted wind farm layout optimization (UWFLO):
Investigating key factors influencing the maximum power generation,” Renew. Energy 38(1), 16–30 (2012).
8. D. Elliott and J. Barnard, “Observations of wind turbine wakes and surface roughness effects on wind flow
variability,” Sol. Energy 45(5), 265–283 (1990).
9. U. Högström, D. Asimakopoulos, H. Kambezidis, C. Helmis, and A. Smedman, “A field study of the wake
behind a 2 MW wind turbine,” Atmos. Environ. (1967). 22(4), 803–820 (1988).
10. R. Barthelmie, L. Folkerts, F. Ormel, P. Sanderhoff, P. Eecen, O. Stobbe, and N. Nielsen, “Offshore wind
turbine wakes measured by SODAR,” J. Atmos. Ocean. Technol. 20(4), 466–477 (2003).
11. L. Vermeer, J. N. Sørensen, and A. Crespo, “Wind turbine wake aerodynamics,” Prog. Aerosp. Sci. 39(6-7),
467–510 (2003).
12. F. Bingöl, J. Mann, and G. C. Larsen, “Light detection and ranging measurements of wake dynamics part I: one-
dimensional scanning,” Wind Energy (Chichester Engl.) 13(1), 51–61 (2010).
13. J. Trujillo, F. Bingöl, G. C. Larsen, J. Mann, and M. Kühn, “Light detection and ranging measurements of wake
dynamics. Part II: two-dimensional scanning,” Wind Energy (Chichester Engl.) 14(1), 61–75 (2011).
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A515
#284869
https://doi.org/10.1364/OE.25.00A515
Journal © 2017
Received 17 Jan 2017; revised 21 Apr 2017; accepted 21 Apr 2017; published 1 May 2017

14. Y. Käsler, S. Rahm, R. Simmet, and M. Kühn, “Wake measurements of a multi-MW wind turbine with coherent
long-range pulsed Doppler wind lidar,” J. Atmos. Ocean. Technol. 27(9), 1529–1532 (2010).
15. I. Smalikho, V. Banakh, Y. Pichugina, W. Brewer, R. Banta, J. Lundquist, and N. Kelley, “Lidar investigation of
atmosphere effect on a wind turbine wake,” J. Atmos. Ocean. Technol. 30(11), 2554–2570 (2013).
16. S. Wu, B. Liu, J. Liu, X. Zhai, C. Feng, G. Wang, H. Zhang, J. Yin, X. Wang, R. Li, and D. Gallacher, “Wind
turbine wake visualization and characteristics analysis by Doppler lidar,” Opt. Express 24(10), A762–A780
(2016).
17. R. M. Banta, Y. L. Pichugina, W. A. Brewer, J. K. Lundquist, N. D. Kelley, S. P. Sandberg, R. J. Alvarez II, R.
M. Hardesty, and A. M. Weickmann, “3D volumetric analysis of wind turbine wake properties in the atmosphere
using high-resolution Doppler lidar,” J. Atmos. Ocean. Technol. 32(5), 904–914 (2015).
18. G. V. Iungo, Y.-T. Wu, and F. Porté-Agel, “Field measurements of wind turbine wakes with lidars,” J. Atmos.
Ocean. Technol. 30(2), 274–287 (2013).
19. S. Wu and X. Zhai, “Lidar Investigation of Wind Turbine Wake Characteristics Under Different Surface
Roughness,” in Optics and Photonics for Energy and the Environment (Optical Society of America, 2016), p.
EM2A. 3.
20. D. Medici and P. Alfredsson, “Measurements on a wind turbine wake: 3D effects and bluff body vortex
shedding,” Wind Energy (Chichester Engl.) 9(3), 219–236 (2006).
21. L. P. Chamorro and F. Porté-Agel, “A wind-tunnel investigation of wind-turbine wakes: boundary-layer
turbulence effects,” Boundary-Layer Meteorol. 132(1), 129–149 (2009).
22. R. B. Smith, “Gravity wave effects on wind farm efficiency,” Wind Energy (Chichester Engl.) 13(5), 449–458
(2010).
23. S. Emeis, Wind energy meteorology: atmospheric physics for wind power generation (Springer Science &
Business Media, 2012).
24. B. Campistron, S. Bernard, B. Bénech, F. Ardhuin-Girard, J. Dessens, E. Dupont, and B. Carissimo, “Turbulent
dissipation rate in the boundary layer via UHF wind profiler Doppler spectral width measurements,” Boundary-
Layer Meteorol. 103(3), 361–389 (2002).
25. R. Frehlich, “Effects of wind turbulence on coherent Doppler lidar performance,” J. Atmos. Ocean. Technol.
14(1), 54–75 (1997).
26. A. B. Davis and A. Marshak, “Photon propagation in heterogeneous optical media with spatial correlations:
enhanced mean-free-paths and wider-than-exponential free-path distributions,” J. Quant. Spectrosc. Radiat.
Transf. 84(1), 3–34 (2004).
27. R. Frehlich and L. Cornman, “Estimating spatial velocity statistics with coherent Doppler lidar,” J. Atmos.
Ocean. Technol. 19(3), 355–366 (2002).
28. R. Frehlich and R. Sharman, “Maximum likelihood estimates of vortex parameters from simulated coherent
Doppler lidar data,” J. Atmos. Ocean. Technol. 22(2), 117–130 (2005).
29. R. Krishnamurthy, R. Calhoun, B. Billings, and J. Doyle, “Wind turbulence estimates in a valley by coherent
Doppler lidar,” Meteorol. Appl. 18(3), 361–371 (2011).
30. I. Smalikho, F. Köpp, and S. Rahm, “Measurement of atmospheric turbulence by 2-um Doppler lidar,” J. Atmos.
Ocean. Technol. 22(11), 1733–1747 (2005).
31. B. D. Hirth and J. L. Schroeder, “Documenting wind speed and power deficits behind a utility-scale wind
turbine,” J. Appl. Meteorol. Climatol. 52(1), 39–46 (2013).
32. J. Whale, K. H. Papadopoulos, C. G. Anderson, C. G. Helmis, and D. J. Skyner, “A study of the near wake
structure of a wind turbine comparing measurements from laboratory and full-scale experiments,” Sol. Energy
56(6), 621–633 (1996).
33. M. L. Aitken, R. M. Banta, Y. L. Pichugina, and J. K. Lundquist, “Quantifying wind turbine wake characteristics
from scanning remote sensor data,” J. Atmos. Ocean. Technol. 31(4), 765–787 (2014).
34. P. Sheng, J. Mao, J. Li, A. Zhang, J. Sang, and N. Pan, Atmospheric Physics (Beijing University Press, 2003),
Chap. 9.
1. Introduction
Wind energy industry is growing fast as one of the most promising renewable energy source.
Early investigations were firstly performed on wind turbine static simulators [1]. Then tests
with rotating models were carried out to consider the interaction between different wind
turbine wakes [2]. Several wind tunnel studies have also been performed to investigate the
influence of different parameters, such as the downstream distance between wind turbine
rows [2–4], turbine thrust coefficient [3–5] and ambient turbulence [5–7] on the wind turbine
wake characteristics. With the planning of large offshore wind farms off the coasts of the
continents and larger islands and the need to optimize the turbine arrangement in a wind farm,
various experiments and simulations have been extensively performed to study the influence
of atmospheric conditions such as the wind velocity and turbulence intensity on the wind
turbine wake characteristics, for instance, the velocity deficit and wake length. Elliot and
Barnard [8] collected wind data at nine meteorological towers at the Goodnoe Hills MOD-2
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A516

wind turbine site to characterize the wind flow over the site both in the absence and presence
of wind turbine wakes. Hȍgstrȍm et al. [9] employed four different techniques including
anemometer, balloon sounding, tower-mounted instrument and Doppler Lidar (Light
Detection and Ranging) for investigating the turbine wake. Barthelme et al [10] studied the
offshore wind turbine wakes with the use of a sodar. Vermeer et al [11] comprehensively
reviewed the theoretical and experimental studies on wind turbine wake through the use of
various techniques. Results of a study of the wind turbine wake with the aid of a continuous-
wave Coherent Doppler Lidar (CDL) were presented in [12,13]. Comparing with the turbulent
wake characteristics detection technologies mentioned above, the Pulsed Coherent Doppler
Lidar (PCDL) has higher spatial and temporal resolution and offers a wide range of
possibility to investigate wind turbine wakes by using different scanning modes, providing
multi-dimensional structure of wind turbine wake and turbulence parameters simultaneously
[14–19].
The structure of wind turbine wake is known to be affected by the characteristics of the
incoming flow [20]. The factors that affect the incoming flow characteristics and the wake
structure include the presence of surface inhomogeneity such as surface roughness transitions,
the aerodynamics surface roughness, topography, and the effects of the wakes from other
turbines located upwind in a wind park [21]. Smith [22] considered the wind park as a whole
as an additional surface roughness, an additional momentum sink or as a gravity wave
generator in association with the temperature inversion aloft at the top of the boundary layer
[22,23], and concluded that the surface roughness and the thermal stability of the atmosphere
are erected turns out be decisive parameter governing the efficiency of wind parks. The
dependence of wind and available power reduction as function of surface roughness has
consequences for offshore wind parks which will become the major facilities for wind power
generation in the near future. Therefore, detailed understanding of the turbulent properties of
turbine wakes under different inflow and surface conditions is important to maximize the
energy production and ensure the structural integrity of the wind turbines.
This paper uses a PCDL with relatively high spatial resolution and sampling rate to
investigate the turbulence characteristics of ambient wind area and turbine wake area in
different topographies and surface roughness. Furthermore, the dependence of wake
characteristics on the turbulent state of the atmosphere is studied. This paper is a substantially
extended study of the invited presentation at OSA’s Light, Energy and the Environment
Congress recently held in Leipzig, Germany [19]. The PCDL specifications and field
experiments are described in Section 2. Section 3 presents the retrieval methods of turbulent
characteristics and turbine wake characteristics, respectively. Field experiments results
comparison and diurnal variation case study in Hami Gobi desert wind park are provided in
Section 4. A conclusion is given in Section 5.
2. Experimental setup and field campaigns
PCDL is a scanning, remote sensing instrument that measures aerosol backscatter and the
wind component along the lidar beam. The PCDLs, WindPrint S3000/S4000 that we used for
field campaigns, were from Seaglet Environmental Technology and modified by lidar group
with Ocean University of China (OUC). The laser pulse width can be tuned to 100 ns,
corresponding to spatial resolution of 15 m in the radial direction. A comprehensive
description of PCDL and its specifications are given in [16]. The highlights of the PCDL
include the flexible range resolution from 15 m to 60 m and relatively high sampling rate. The
finest resolution of 15 m and temporal sampling rate of 4 Hz are significant advantages in
studying the turbine wakes and small scale turbulent characteristics.
In order to investigate the wake characteristics in the vicinity of the operating wind
turbine generators (WTGs) under different topographies and surface roughness, four field
experiments were performed both in the onshore and offshore wind parks from 2013 to 2015.
Two lidar scanning modes were applied to visualize the 2D structure of wake. The first mode
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A517

is the Plan Position Indicator (PPI) scanning mode by scanning the laser beam with a fixed
elevation angle and varying azimuth angles, and the second scanning mode is the Range
Height Indicator (RHI) scanning mode by scanning the laser beam with a fixed azimuth angle
and varying elevation angles. In this paper, the wind field in the vicinity of the wind turbine is
studied using PPI scanning mode.
The first onshore field campaign was carried out in Longgang Mountain wind park,
Shandong Peninsula in the winter of 2014. The site experiences complex terrain effects, such
as enhanced turbulence night and day which affects wake characteristics. Figure 1(a) shows
the outfield experiment in Longgang wind park. Figure 2(a) shows the line-of-sight (LOS)
velocity measurement of the wind turbine on 10 Jan 2014 based on PPI scanning
measurement. The ambient wind was blowing from southwest to northeast. The plot colors
indicate the wind speed and direction along the laser beam: positive (red) away from lidar,
and negative (blue) toward lidar. In order to compare the turbulent characteristics in the
ambient wind field and wind turbine wake area, a wind turbine with relatively distinct wake
structures shall be firstly chosen. Then the area with relatively distinct wake structures is
chosen as the wake area, which is circled by green line shown in Fig. 2(a). Conversely, the
ambient wind field which is hardly affected by wind turbines is chosen circled by blue line in
Fig. 2(a).
The second and third field campaigns were performed at Rudong tidal zone wind park,
Jiangsu province in the spring and winter of 2014, respectively. Figure 1(b) shows the
outfield experiment in Rudong wind farm. It is the first intertidal wind energy project in the
world, stretching from a temporary dry coast to the sea horizon. Figure 2(b) shows the PPI
scanning mode result on 10 April 2014. The ambient wind was blowing from southeast to
northwest. Compared with Longgang turbine wakes, the more homogeneous surface in
Rudong intertidal zone produces more distinct turbine wakes. Similarly, the area circled by
blue and green lines in Fig. 2(b) are the ambient wind field and wind turbine wake area,
respectively.
The fourth field campaign was performed at Hami Gobi desert wind park, Xijiang
province in 2014-2015. Figure 1(c) shows the outfield experiment in Hami. It is one of the
largest and fastest growing wind energy facilities in China. Figure 2(c) shows the PPI
scanning mode result on 15 Nov 2015. The ambient wind was blowing from northwest to
southeast. The areas circled by blue and green lines are the ambient wind field and wind
turbine wake area, respectively. Table 1 lists the specific scan parameters during different
field campaigns. Table 2 lists the parameters of wind turbines used in these experiments.
Fig. 1. Wind flow measurement with PCDL in (a) Longgang Mountain, Shandong Peninsula
(b) Rudong tidal zone, Jiangsu Province (c) Hami Gobi desert, Xijiang Province.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A518

Fig. 2. LOS wind field by PCDL PPI scanning measurement at (a) Longgang Mountain from
1148 to 1154 LT 10 Jan 2014 (b) Rudong tidal zone from 1833 to 1835 LT 10 April 2014 (c)
Hami Gobi desert from 1928 to 1931 LT 15 Nov 2015. The area circled in blue and green line
is the chosen ambient wind area and wake area, respectively.
Table 1. PCDL Scan parameters during different field campaigns
Location
Scanning
start time
(LT)
Scanning
end time
(LT)
Azimuth angle
range ( )
Elevation
angle ( )
Scanning
speed
(/
s
)
Range
resolution
(m)
Pulse
width
(ns)
Longgang 20140103
11:48:39
20140103
11:54:38
0~360
(clockwise) 1.96 1 60 400
Rudong 20140410
18:33:56
20140410
18:35:36
120~220
(anticlockwise) 3.00 1 30 200
Hami 20151115
19:28:43
20151115
19:31:12
120~270
(clockwise) 1.30 1 30 200
Table 2. Parameters of wind turbines used in different field campaigns
Location
Wind
turbine
Type
Rotor
diameter
(m)
Hub
height
(m)
Rated
power
(kW)
Cut-in
wind
speed
(m/s)
Rated
wind
speed
(m/s)
Cut-out
wind
speed
(m/s)
Generator
rated
voltage (V)
Longgang 1.5-93 93 80 1500 3 9.5 20 690
Rudong GW
103/2500 103 80 2500 3 10.8 25 690
Hami GW
115/2000 115 85 2000 2.5 9.5 22 690
3. Retrieval method of turbulence and turbine wake characteristics
3.1 Estimation of turbulence parameters from scanning PCDL data
The spatial statistics of the velocity field can be described by the structure function. In many
cases, such as a typical atmospheric boundary layer, the structure function has a simple
description in terms of two parameters: the turbulence energy dissipation rate
ε
and an
integral length scale i
L which is defined as the integral of the normalized correlation of the
radial velocity. The
ε
is difficult to measure directly because of the limited spatial resolution.
The most common remote sensing tools for
ε
measurement include Doppler radar [24] and
Doppler lidar [25,26]. Computer simulations have shown that accurate estimation of
ε
can be
produced from an analysis of the structure function of the radial velocity along each beam
when the range resolution is less than the length scale of the velocity field and the lidar
sensing volume transverse to the lidar beam hΔ is much smaller comparing to the effective
range resolution RΔ [27]. The structure function method has already been developed and
described in many literatures, and it is noted that the algorithm includes the estimation error
of the radial velocity as well as the correction for the spatial averaging by lidar pulse [28].
The retrieval depends on structure functions mainly in two ways. Firstly, the turbulence
parameters can be calculated along the direction of the lidar beam (longitudinal structure
function). Furthermore, it is possible using structure functions of the radial velocity in the
azimuth direction for each range-gate distance R and elevation angle
θ
(transverse structure
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A519

function). Herein the transverse structure function is used to retrieve the
ε
from the 2D wind
field by the PPI scanning measurement since it has fine angular and vertical resolution and
the error introduced by anisotropy is smaller than longitudinal structure function’s [28,29].
The accuracy investigation of turbulence measurement with the structure function method
cannot be complete without inter-comparison with measurement results of independent
turbulence sensors [27,30]. The algorithm for extracting high resolution turbulence
parameters using transverse structure function has been described in detail by Frehlich and
Cornman [27]. From the measured data, the estimates of the raw structure function of the
radial velocity Doppler Lidar estimate ()
raw
Dsand the contribution from the estimation error
of the velocity estimates ()Es can be derived. The measured difference () ()
raw
DsEs− is
fitted to the corrected von Kármán model, in order to find
ε
and i
L in such a way that the
best fitting takes place for both variables, where the genetic algorithm is specially used to
solve the optimum solution [16]. Following this process,we calculate the transverse structure
function at 1.275R=km of the lidar data taken on 10 April 2014 at Rudong tidal zone.
Figure 3 shows the structure function using different models. From the fit to the
()
wgt calculate
Ds
−, we derive the standard deviation of radial velocity 0.54
σ
= m/s and the outer
scale of turbulence 0180.8L= m. The integral length scale 135
i
L= m and the 0.0008
ε
=
m2/s3 can be calculated from Eqs. (1)–(2), respectively. Note that
σ
represents the standard
deviation, usually denoted by v
σ
, of the transverse velocity fluctuations in the hypothesis of
isotropic turbulence.
1/3 3 3
3/2
00
2
[ ] 0.933668 ,
3(1/3)(4/3) LL
πσ σ
ε
==
ΓΓ (1)
00
(5 / 6) 0.7468343 ,
(1 / 3)
i
LLL
π
Γ
==
Γ (2)
where ()zΓ is the Gamma function, and the integral length scale is proportional to the outer
scale 0
L.
Fig. 3. Transverse structure function estimates of turbulence using single PPI scanning on 1833
LT 10 April 2014 at Rudong with an elevation angle 3
θ
= and range-gate distance 1.275R=
km . Curves shows calculations of the corrected structure function wgt calculate
D− (black dots),
the von Kármán model vvonkarman
D−− (black line), the Kolmgorov model v Kolmgorov
D− (blue line)
and the corrected von Kármán model modwgt el
D− (red line) taking the volume average effect of
lidar detection into consideration.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A520

The turbulent intensity is defined as the ratio of the standard deviation of the LOS velocity
to the mean wind speed. In order to estimate the turbulent intensity in the case of the PPI
scans, we can calculate it over a given region at range R and fixed elevation angle
θ
, that is,
()
() ,
()
LOS
LOS
R
IR UR
σ
= (3)
where ()
LOS R
σ
can be estimated from lidar PPI scanning measurement by the use of the
transverse structure method [27], and the spatial averaging effect of the radial velocity over
the sensing volume is taken into account, as it is shown in Fig. 3. The ()URis the mean wind
speed calculated from arc scanning mode [16] and assumes the horizontal homogeneity of the
wind field over probing area with uniform wind speed and direction.
L
OS
I
is simplified as
I
in the following section.
3.2 Estimation of turbine wake characteristics
From a perspective of wind turbine engineering, two characteristics of wind turbine wakes are
essential, one is the turbulence level, which is important in determining fatigue loads on other
wind turbines downwind, the other is the velocity deficit, which is related to the power loss
from the wind turbine [11,23,31]. Definition of velocity deficit in previous studies can be
written as:
() ()
( ) 100%,
()
ambient wake
ambient
uxux
VD x ux
−
=×
(4)
where ()
ambient
ux is the ambient or reference velocity as a function of downstream distance
x
outside of the wake, ()
wake
ux is the wake velocity in downstream distance
x
inside the wake.
Results from previous field experiments vary because of differences in inflow conditions,
atmospheric stability, and surface roughness [8,32]. Furthermore, the different definition of
()
ambient
ux and ()
wake
ux inevitably leads to discrepancies even in the same observation
conditions. Smalikho et al [15] estimated the velocity deficit as:
()
() (1 ) 100%,
(,1.2 )
Ux
VD x Ux D
=− × (5)
where () (,1.2 )
ambient
uxUxD=,
x
is the distance behind the wind turbine along the wind
direction, D is the diameter of the circle described by the outer end of the turbine blade.
1.2D is the distance between the wake location and the candidate ambient wind location at
distance
x
, and different types of WTG and atmospheric conditions exist inevitably
difference for the specific value, where inappropriate choice may result in incorrect
estimation of velocity deficit.
Wu et al [16] defined the upwind velocity of the wind turbine as the mean ambient wind
velocity ambient
u, and the ()
wake
ux is the wind velocity within the wake downstream, so the
velocity deficit can be described as:
()
( ) (1 ) 100%,
wake
ambient
ux
VD x u
=− × (6)
It is noted that the upwind velocity may be affected by the upwind WTG wake, which is
more turbulent and slightly smaller than that in free stream.
Banta et al [17] considered the wake meandering with the mean wind direction and the
slight crosswind angle of the beams, and estimated the minimum value at each distance
x
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A521

and azimuth angle
φ
Δ min (, )ux
φ
Δ as the ()
wake
ux, and the mean upwind velocity of the rotor
plane was regarded as the ambient
u, thus the velocity deficit can be described as:
min (, )
( ) (1 ) 100%,
ambient
ux
VD x u
φ
Δ
=− × (7)
In this case,
φ
Δ should be adjusted based on the different wind direction and distance
x
,
and the determination of the ()
wake
ux is more easily disturbed by turbulent atmospheric
conditions.
In order to achieve batch data processing of wake characteristics from the PPI scanning
measurement, the methods mentioned above have been used and evaluated. Considering the
availability of the automatic processing algorithm, we propose an improved batch data
processing procedure as follows. For the determination of ambient
u, the horizontal wind speed
and direction are firstly obtained using VAD fitting in each range gate, then the wind is
assumed horizontally homogeneous and the mean wind direction can be estimated by
averaging the wind direction in each range gate shown in Fig. 4(a) in blue line. Then, the
ambient
u is calculated as the mean value of the projection of the horizontal wind on the mean
wind direction. In the automatic process of ()
wake
ux, the key point is to obtain min ()ux
quickly and accurately. Since the velocity of wake cross section in each distance
x
has the
distribution of single or double Gaussian shape [33], the wake center 0
p can be firstly
deduced by the least square fit method with the Gaussian curve. The obtained 0
p in each
distance is shown in Fig. 4(a) in black plus signs, which is not necessarily located
symmetrically within the wake. Then the min ()ux is the mean velocity from the position
00
(3,3)pp−+in fitted curve. It is noted that the min ()uxshould be converted into
()
wake
uxusing the geometry relationship and the mean horizontal wind direction. Figure 4(b)
shows the ambient
u, ()
wake
ux and the ()VD x behind WTG A. It can be seen that the ()VD x is
the greatest at around 220 m with the value of 44.95% and gradually diminishes downstream
from the turbine in the far wake. The turbine wake length in this section is defined as the
distance between the WTG A and the point where the ()VD x has minimum value, thus the
wake length in this case is 0.63 km.
Fig. 4. (a) Conical sector scan performed at an elevation angle of 1.5° at 1930 LT 15 Nov 2014
in Hami Gobi desert wind farm, (b) background wind velocity in blue circle, wake area
velocity in red circle (upper), and the corresponding velocity deficit in blue circle (bottom), the
black dash line corresponds to VD = 10%.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A522

4. Results
4.1 Comparison of turbulence estimates under different surface roughness
The wake velocity deficit, the downwind decay rate of the wake, and the added turbulence
intensity within the far wake are largely determined by two factors: the turbine’s thrust
coefficient and the ambient turbulence intensity [23]. Vermeer et al [11] gave three empirical
formulas to describe the added turbulence intensity in the wake. During field experiments, we
used lidar PPI scan to reveal turbulence parameters both from incoming flows and turbine
wakes. This allows us to study the dependence of the wake parameters in a turbulent air.
Furthermore, the three case studies shown in Fig. 2 represent different surface roughness,
giving the highlight to study the effect of surface roughness on the wake characteristics.
The turbulence characteristics in ambient wind (region circled by blue line) and wake
(region circled by green line) over these three wind parks shown in Fig. 2, including the
ε
,
I
and v
σ
, have been retrieved using the transverse structure function method described in
Section 3.1. Figure 5 shows the relationship between the
ε
and the
I
, v
σ
respectively. The
blue circle and red square are the turbulence parameter values in ambient wind and wake,
respectively. It can be seen that the
ε
has a significant positive correlation with
I
and v
σ
,
and it implies that more turbulent atmosphere leads to the faster process that the turbulence
kinetic energy is transferred into internal energy with the action of molecular viscosity. The
averaged
ε
,
I
and v
σ
in wake area are 7.8, 2.6, and 2.3 times of the ambient wind areas’,
respectively, implying that the wind turbine wake has an obviously effect on the atmospheric
turbulence characteristics. Table 3 lists the averaged values of
ε
,
I
and v
σ
in ambient and
wake area.
Table 3. The averaged value of
ε
,
I
and v
σ
in ambient area (A) and wake area (B).
The ratio of the corresponding parameters in A and B is also listed.
Parameters Ambient area (A) Wake area (B) Ratio of A and B
ε
( 23
/ms) 0.0016 0.013 7.8
I
0.087 0.23 2.6
v
σ
( /ms) 0.59 1.37 2.3
Fig. 5. Turbulence energy dissipation rate
ε
versus (a) turbulence intensity
I
and (b)
standard deviation of transverse velocity fluctuation v
σ
from the cases shown in Fig. 2. The
blue and red circles represent the results from ambient wind area and wake area, respectively,
with corresponding averaged estimates in red and blue triangles.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A523

In order to compare the correlation of turbulence parameters under different surface
roughness, the results from Longgang mountain, Rudong tidal zone and Hami Gobi desert are
analyzed and fitted using exponential function exp( )ya bx=, respectively. Figure 6(a) shows
the correlation between
ε
and
I
, and it can be seen that the fitting exponential term
coefficient b is 0.67, 0.77 and 0.90 for Hami Gobi desert, Rudong tidal zone and Longgang
mountain wind farm, respectively, that is, with the increase of
ε
, the
I
in complex mountain
condition has the most obvious rate of increase. On the contrary, the
I
in Gobi desert
corresponds to the lowest sensitivity to the
ε
variation. Figure 6(b) shows the correlation
between
ε
and v
σ
. Similarly to the phenomena in Fig. 6(a), the v
σ
in Longgang mountain
case is more sensitive to
ε
, and the trend in Hami Gobi desert case is less significant
compared to the other two cases. These differences may result from different surface
roughness condition. Specifically, the typical surface roughness for mountain, offshore area
and flat desert are about 1 m, 0.0008 m and 0.0004 m [34].
Fig. 6. The relationship between
ε
and (a)
I
(b) v
σ
under three wind parks cases shown in
Fig. 2. The blue, red and black circles represents the results from Longgang mountain, Rudong
tidal zone and Hami Gobi desert case, respectively, with corresponding fitted curves in blue,
red and black color.
Figure 7 shows the averaged turbulence characteristics in ambient wind region A and
wake region B in the three wind parks, respectively. It can be seen that the experiment was
mostly carried out under weak turbulence condition ( 32
10 10
ε
−−
≤≤ ), and Rudong tidal zone
ambient wind area has relatively higher
I
and v
σ
, since during the measurement, the sea
surface temperature was 14 C
-15 C
 along the coastline and higher than the air
temperature around 13 C
 in the nighttime. This temperature gradient may bring in surface
turbulence created by thermal convection. As for the wake area B, Longgang mountain area
and Rudong tidal zone area have transferred into strong turbulence condition, however,
although the
ε
in Hami Gobi desert wake area is more than 10 times of its ambient wind
areas’, it is still in relative weak turbulence conditions. Different surface roughness may lead
to obviously different turbulence characteristics in these areas. Table 4 lists the averaged
values of
ε
,
I
and v
σ
in ambient and wake area for each case study.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A524

Fig. 7. Correlation analysis of the averaged turbulence characteristics including
ε
versus (a)
I
and (b) v
σ
in background wind area (marked with circles) and wake area (marked with
triangles) under three wind parks, the blue, red and black symbols represent the results from
Longgang mountain, Rudong tidal zone and Hami Gobi desert case, respectively.
Table 4. The averaged value of
ε
,
I
and v
σ
in ambient area (A) and wake area (B) for
each case study.
Parameters Longgang Rudong Hami
Ambient wind area (A)
ε
0.001 0.002 0.0007
v
σ
0.54 0.76 0.35
I
0.09 0.10 0.07
Turbine wake area
(B)
ε
0.015 0.014 0.008
v
σ
1.63 1.49 0.89
I
0.31 0.23 0.16
4.2 Case study in Hami Gobi desert wind park
To analyze the diurnal variation of atmospheric turbulence and wake characteristics and the
effect of wind turbulence on turbine wake, we selected data that were measured in Hami Gobi
desert wind farm from 0000 LT 10 April 2014 to 2350 LT 10 April. The data measured by
PPI scanning mode with an elevation angle of 1.3° every 3 min is used to retrieve the vertical
profiles of the ambient wind velocity A
U and direction V
θ
. The same lidar data is also used
to retrieve vertical profiles of the turbulence parameters such as the
ε
, v
σ
, and i
L obtained
using the transverse structure function method described in Section 3.1. The resulting
temporal profiles of A
U V
θ
at 20H= m are shown in Figs. 8(a) and 8(b). The wind velocity
varies from 1.0 m/s to 8.0 m/s with obvious diurnal variation that wind velocity at nighttime is
higher than the values during the daytime. The wind directions are mostly concentrated near
the dash dotted line that corresponds to the azimuth angle T
θ
during the daytime.
Conversely, it changes into northwest at night with opposite direction with T
θ
. Both of the
cases above can guarantee the accurate wake characteristics retrieval using PPI scanning
mode. The processing procedures described earlier in Section 3.2 are used to determine the
maximum velocity deficit max
VD and the turbine wake length w
L, which are shown in Figs.
8(c) and 8(d). It is noted that the measurement with weak wake visualization characteristics
and lower Signal to Noise Ratio (SNR) is not considered. The result reveals that the max
VD
varies from 33% to 70% and the values during the daytime are integrally greater than the ones
at night. The w
L varies from 0.26 km to 2.1 km. On the contrary, the w
L at night is obviously
longer than the observation during the daytime.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A525

Fig. 8. Diurnal profiles of the (a) ambient wind velocity at 20 m, (b) wind direction at 30 m, (c)
velocity deficit, (d) wake length as obtained from the data measured by CDL using PPI
scanning on 10 April 2014 at Hami wind park.
Before the retrieval of turbulence parameters, the quality control based on the SNR
threshold is used in each PPI scanning measurement to ensure the availability of the data. In
this paper, the SNR is defined as the ratio of the peak value of FFT spectral signal to the
Root-Mean-Square-Error (RMSE) of background noise signal, which indicates the CDL
detection capability, data accuracy and atmospheric tracer particle relative intensity.
Specifically, the SNR at 20H= m is averaged during one PPI scanning and if the averaged
SNR is less than 10, the data is inapplicable. As can be seen in Fig. 9, the data from 0600 to
1000 is screened out because of the lower SNR. The black line with red triangles is the
averaged estimates for each 1-hour interval. It can be seen that the v
σ
and
ε
shown in
Figures. 9(a) and 9(c) has distinct diurnal variation and the values are overall higher at night
compared with the ones during daytime. But the
I
and i
L are relatively less changeable. It is
reasonable since according to the definition of
I
in Eq. (3), the
I
is proportional to the
L
OS
σ
and has an inverse relationship with the A
U. The increase in
L
OS
σ
and A
U at night results in
relative stable variation of
I
.
Fig. 9. Diurnal profiles of the (a) v
σ
(b)
I
(c)
ε
and (d) integral length scale i
L from the
data measured by CDL using PPI scanning on 10 April 2014 at Hami wind park. The black
line with red triangles is the averaged estimates for each 1-hour interval.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A526

To analyze the effect of wind and turbulence on the turbine wake characteristics, the data
measured during the night (from 0000 to 0600 and from 2100 to 2350 LT 10 April) and
during the day (from 1000 to 2100 LT 10 April) are selected to estimate of A
U at 20H= m
versus max
VD and w
L shown in Figs. 10(a) and 10(b). The blue circles and red squares
represent the estimates from daytime and nighttime measurement, with corresponding
averaged estimates in blue and red triangles. It can be seen in Fig. 10(a) that an increase of the
wind velocity from 3 m/s to 6 m/s leads to a decrease of the max
VD , which is consistent with
the previous studies [8] that velocity deficit is notably higher at lower ambient velocity owing
to higher thrust coefficients. The averaged max
VD at daytime and night is 52%, 42%,
respectively. On the other hand, the increase of the wind velocity from 3 m/s to 6 m/s
corresponds to the increase of the w
L. Basically, w
L are less than 0.6 km when the wind
velocity is below 5 m/s, but w
L are larger than 1 km when the wind velocity is above 5 m/s.
The averaged w
L at daytime and night is 0.84 km, 1.21 km, respectively. The results were
compared with other literature [15] and found the similar correlation in the relative lower
wind velocity intervals.
Fig. 10. (a) Velocity deficit and (b) wake length as function of wind velocity at 20 m. Blue
circles and red squares are single estimates from data measured by CDL during the day (from
1000 to 2100 LT 10 April 2014) and at night (from 0000 to 0600 and from 2100 to 2350 LT 10
April 2014), respectively. Averaged estimates are shown as blue (daytime) and red (nighttime)
triangles. The black bars represent the fluctuations of velocity deficit (a) and wake length (b)
for each 1 m/s intervals.
Figure 11(a) and 11(b) shows the correlation of the
I
versus the max
VD , w
L, respectively.
According to the results shown in Fig. 11(a), an increase of
I
leads to an overall increasing
trend of the max
VD . The averaged
I
during daytime and night is comparable and consistent
with the stable diurnal variation of
I
. In Fig. 11(b), an increase of
I
corresponds to a rapid
decrease of w
L, which is consistent with the results in [33] that higher turbulence level should
cause the velocity deficit to recover more quickly as faster-moving air was entrained within
the wake.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A527

Fig. 11. As Fig. 10, but for
I
at 20 m versus (a) velocity deficit and (b) wake length, where
the black bars represent the fluctuations of velocity deficit (a) and wake length (b) for each 0.1
intervals turbulence intensity.
5. Conclusions
The pulsed Doppler lidar has been considered as the most effective remote sensing
technology for wind energy assessment, especially for the offshore wind and wind farm with
complex terrain. Lidar observations have shown that wakes induced by upwind turbines can
significantly influence downwind turbines including reduced wind speeds and enhanced
levels of turbulence which will enhance the fatigue loads and reduce the life time of wind
turbine generator. Therefore, in-depth understanding of the properties of turbine wakes under
different inflow and surface conditions is essential to design the structural integrity of the
wind turbines and to optimize the wind energy production. The PCDL with high spatial and
temporal resolution provides the new perspective to simultaneously detect wind turbine wake
and atmospheric turbulence characteristics, allowing us to study the dependence of wake
parameters in the turbulent air.
Four field experiments with different topographies and surface roughness have been
carried out since 2013 in the onshore and offshore wind parks. The turbulence parameters
including the turbulence energy dissipation rate
ε
, the turbulence integral scale i
L and the
standard deviation of transverse velocity fluctuation v
σ
were retrieved using transverse
structure function method based on lidar PPI scanning observations. The relationship of these
parameters were analyzed and compared in the ambient wind area and wake area during the
analyzed time period. It is found that the
ε
has a significant positive correlation with
I
and
v
σ
, which implies that more turbulent atmosphere leads to the faster process that the
turbulence kinetic energy is transferred into internal energy with the action of molecular
viscosity. Furthermore, the averaged
ε
,
I
and v
σ
in wake area are 7.8, 2.6, and 2.3 times of
the ambient wind areas’, respectively, implying that the wind turbine wake has an obviously
effect on the atmospheric turbulence characteristics.
In order to compare the correlation of turbulence parameters under different surface
roughness, the results from Longgang complex mountain, Rudong tidal zone and Hami Gobi
desert are analyzed and fitted using exponential function, respectively. It is found that with
the increase of
ε
, the
I
and v
σ
in complex mountain condition has the most prominent rate
of increase. On the contrary, the results in Gobi desert corresponds to the lowest sensitivity to
the
ε
variation. These differences may result from different surface roughness condition
where the typical surface roughness of complex terrain, off shore area and flat desert are
about 1 m, 0.0008 m and 0.0004 m. By comparing the averaged turbulence characteristics in
ambient wind region and wake region in these three wind parks case studies, strong
turbulence condition appears at the wake area in mountain and tidal zone wind parks.
Although the
ε
within wake area in Gobi desert is more than 10 times of its ambient wind
areas’, it is still in the relative weak turbulence condition.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A528

For large data analysis, such as wind energy assessment, an automatic processing
approach is very necessary. We find that different definitions of ambient and wake area
velocity inevitably lead to discrepancies even in the same observation conditions. Therefore,
we developed a flexible and robust algorithm to consider the effect of ambient velocity
disturbance and the wake meandering with the mean wind direction. Using this batch
processing procedure, a long time series observation in Hami wind park shows the
dependence of wake parameters on the turbulent state of the atmosphere. According the
diurnal variation of atmospheric turbulence and wake characteristics in this case study, it is
found that the max
VD varies from 33% to 70% and the values during the daytime are
integrally greater than the ones at night. The w
L varies from 0.26 km to 2.1 km, however, the
w
L at night is obviously longer than the observation during the daytime. Furthermore, the v
σ
and
ε
have distinct diurnal variation and the values are overall higher at night compared with
the ones during daytime, but the
I
and i
L are relatively less changeable. In particular, an
increase of the wind velocity at 20H= m from 3 m/s to 6 m/s leads to a decrease of the
max
VD and increase of the w
L. In addition, an increase of
I
leads to an overall increasing
trend of the max
VD and a rapid decrease of w
L in this case, which is consistent with the
previous studies.
The study presented here shows that the optical remote sensing technology opens up a
wide range of possibility to investigate wind energy, practically providing multi-dimensional
dynamics characteristics of ambient flow and turbulent wake. Different atmospheric condition
and surface roughness lead to obvious difference in the wake area turbulence and wake
visualization characteristics, further experiments can yield the results necessary to construct
an empirical model of a turbine wake for various atmospheric and surface roughness
conditions.
Funding
National Natural Science Foundation of China (NSFC) under grant 41471309 and 41375016.
Acknowledgments
We thank our colleagues for their kindly support during the field experiments, including
Kailin Zhang and Jintao Liu from Ocean University of China (OUC) for preparing and
conducting the experiment; Dongxiang Wang for operating the lidar in Rudong; Xitao Wang
and Yilin Qi from Seaglet for preparing and operating the lidar in the Gobi desert; Rongzhong
Li and Xitao Wang from Seaglet, and Changzhong Feng with Ocean University for
performing the wake test in Rudong.
Vol. 25, No. 12 | 12 Jun 2017 | OPTICS EXPRESS A529