The long-range and short-range WindScanner systems (LRWS and SRWS), multi-Doppler lidar instruments, when combined together can map the turbulent flow around a wind turbine and at the same time measure mean flow conditions over an entire region such as a wind farm. As the WindScanner technology is novel, performing field campaigns with the WindScanner systems requires a methodology that will maximize the benefits of conducting WindScanner-based experiments. Such a methodology, made up of 10 steps, is presented and discussed through its application in a pilot experiment that took place in a complex and forested site in Portugal, where for the first time the two WindScanner systems operated simultaneously. Overall, this resulted in a detailed site selection criteria, a well-thought-out experiment layout, novel flow mapping methods and high-quality flow observations, all of which are presented in this paper.
In wind energy research, field experiments are important for wind resource evaluation but also to establish, validate and improve theories and wind flow models. If experiments are well planned, designed, executed and reported, the field datasets have a long lifetime and are a firm basis for the advancement of our knowledge on atmospheric flows.
A large number of field experiments addressing flows over hills
Due to the costs of tall meteorological masts, especially in complex terrain,
it is unrealistic to sample the wind within an entire region occupied by
today's largest wind turbines or farms with traditional anemometry. This is
exactly what can be achieved with multi-Doppler lidar systems. The reason for
using multiple lidars is that a single lidar can directly sense only
line of sight (LOS) or radial wind speed, which is a projection of the wind
vector on a laser light propagation path
Employing at least two lidars and intersecting their beams at a point of
interest is required to directly measure two components of the wind at that
point, while to fully characterize the wind vector at least three laser beams
should intersect at a given point. By moving the beam intersection over an
area or volume of interest, the flow can be mapped and resolved in two or all
three dimensions. Dual-Doppler setups consisting of two scanning lidars have
been used in several large atmospheric studies
As discussed in
Efforts put into the WindScanner.dk project have resulted in two specific and
highly configurable multi-Doppler lidar instruments, known as the long-range and
short-range WindScanner systems
In this paper, we propose such a methodology. The methodology will be discussed through its application to a pilot experiment, Perdigão 2015, that took place in a complex and forested site in Portugal, where the two WindScanner systems were operated simultaneously for the first time. In addition, the methodology and operation of the WindScanner systems, this paper also presents a review of the two WindScanner systems, novel scanning methods, observational highlights of multi-lidar measurements of a single turbine wake and inflow conditions in complex terrain, multi-lidar measurements of wind resources along a ridge and observations of valley flows. It should be pointed out that the data analysis or discussions of particular flow situations are not the purpose of the present paper and are included as a result and as an illustration of the presented methodology.
The paper is organized as follows: Sect.
A WindScanner system consists of two or more spatially separated scanning
lidars (long- or short-range WindScanners, Fig.
DTU Wind Energy scanning lidars. Long-range WindScanner
The WindScanners have been specifically tailored to perform user-defined and
time-controlled scanning trajectories, known as complex trajectories, either
independently or in a synchronized mode. The lidars in the long-range
WindScanner system are usually connected to the master computer using a 3G
network
SRWSs provide high-frequency measurements (up to 400 Hz) of the flow while
probing the atmosphere with a relatively small probe length. This allows
resolving small length scales (down to a few cm) and short time scales (down
to 2.5 ms) of the flow. As SRWSs are based on the continuous wave (CW) lidar technology there
are several limitations. Since the LOS speed at a given point is resolved by
focusing the laser beam, the probe length increases with range. This limits
the maximum range of a SRWS to about 150 m. Furthermore, since the beam can
only be focused at a single point in time, radial velocity from one
single range can be resolved. However, a high measurement rate compensates
for this limitation in ranging. Based on these characteristics the short-range
WindScanner system is ideal for mapping of turbulent flow features around a
single wind turbine rotor
LRWSs have a larger probe length (minimum 25 m) and lower measurement
frequency (10 Hz at best, typically 1 Hz) than SRWSs. Since LRWSs are based
on the pulsed technology their probe length is constant with range.
Furthermore, LRWSs can simultaneously retrieve radial velocity from a number
of ranges along the laser light propagation path. This number is limited by
the computational power of the lidar. This particular characteristic of
ranging compensates for a lower measurement frequency since at any given
measurement rate LRWS can provide a “snapshot” of the atmosphere up to
several kilometers along a single LOS. The maximum range of LRWSs is about
8 km, which has been typically observed in offshore conditions
A summary of the characteristics of the two types of WindScanners is given in
Table
Based on the characteristics of the two multi-lidar instruments, the long- and short-range WindScanner systems are complementary to each other. Combining these two systems into a hybrid system unlocks a possibility of simultaneously observing mean flow features over a large region and turbulent characteristics and fine flow structures within a preselected area (not larger than a rotor-swept area) of this same region. At the current state of the lidar technology, we cannot have all these capabilities within one single system.
The campaign Perdigão 2015, described in the present paper, was the first attempt to simultaneously operate both long- and short-range WindScanner systems with an aim of combining them into a hybrid WindScanner system and acquiring both mean and turbulent flow features of the site.
Characteristics overview of the WindScanners.
The methodology for WindScanner-based experiments consists of 10 steps: (1) definition of scientific objectives, (2) site selection, (3) site characterization, (4) experiment layout design, (5) scanning modes design, (6) infrastructure planning, (7) deployment and calibration, (8) execution and data collection, (9) decommissioning and post-calibration, and (10) data archiving and dissemination.
Defining scientific objectives is related to outlining scientific questions of interest that can be addressed with WindScanner observations. According to the scientific objectives, the site selection is made and followed by a detailed site characterization (e.g., wind conditions, terrain). Sometimes the order is reversed – a known site can stimulate a scientific question.
In the following steps, the experiment layout is made and physical infrastructure is planned (e.g., power, network, access roads). Afterwards, given the now-established logistical constraints, the scanning modes to be implemented during the campaigns are designed. Once the physical installation commences, the deployment and calibration procedural steps are applied (e.g., leveling and orientation of WindScanner, assessment of pointing accuracy). Following the start of the campaign, the execution and data collection procedural steps are put in action (e.g., experiment monitoring and information logging). The decommissioning and post-calibration procedural steps are applied at the end of the campaign. In the last procedural step, all the data regarding the campaign are collected together with the acquired datasets and uploaded to an online information system, making them available for end users.
The aforementioned steps will be demonstrated in the content of the Perdigão-2015 experiment in the following section.
There were multiple reasons in favor of the Perdigão-2015 experiment: there was the need to test both the equipment and our human resources in a demanding field experiment. The question was whether the new scientific equipment, which is expensive, fragile and sensitive, and has been developed and tested previously in a laboratory environment or in short-duration field campaigns was robust enough to withstand realistic conditions; for instance, high temperatures and remote locations with no power or network grid. The equipment, transported by road between Roskilde (Denmark) and Serra do Perdigão (Portugal) and remained in the mountains without surveillance for long periods.
WindScanner.eu (2012–2015) is a European Union project of the European
Strategy Forum on Research Infrastructures program, under which a 3-year
preparatory phase for establishing a pan-European research
infrastructure was financed. Among the project deliverables, the methodology for
WindScanner-based field experiments (which we report in the present
paper) was developed based on the previous extensive work done under the
WindScanner.dk project
Perdigão 2015 was a last demonstration campaign within the WindScanner.eu
project that served as the preparation for the larger experiment conducted
within the NEWA project
In addition, the WindScanner.eu project, Perdigão 2015 was funded by the NEWA, Unified Turbine Testing (UniTTe), and Wind Farm Layout Optimization in Complex Terrain (FarmOpt) projects. The latter three projects have specific research goals that assisted in defining scientific objectives for Perdigão 2015.
The NEWA project aims to improve wind resource modeling for different site
conditions. Areas with steep ridges and forested terrain are of particular
interests since the current engineering (linear) flow models are unable to
correctly predict the behavior of the flow over the sites with these features
For Perdigão 2015, we selected several flow aspects to investigate and addressed them with WindScanner measurements. To assist the NEWA project, we chose to measure wind resources along a ridge, occurrences of flow separations on lee sides of hills (i.e., recirculation zone) and valley flows. For the FarmOpt project, we intended to characterize a single wind turbine wake in horizontal and vertical planes. Specifically, we aimed to provide measurements for studying the wind speed deficit up to five diameters downstream of the wind turbine, the wake position in a vertical plane close to the wind turbine, and the wake geometry in a horizontal plane with center at the wind turbine hub. Similarly, for the UniTTe project, the inflow conditions were intended to be characterized in the same planes. The objective was to derive datasets for a detailed investigation of the wind turbine induction zone in complex terrain. The ultimate objective was to create a dataset that could be used in the appraisal and development of computational models for wind resource, wind turbine design and wind farm layout optimization in complex terrain.
According to the scientific objectives, the site of interest should be in a terrain consisting of a hill with a steep ridge where an isolated wind turbine is operating. The site complexity should be within manageable levels, i.e., yielding a flow complexity that we can still understand. The hill size should be such that the flow clearly separates. Ideally, the hill should be in an environment surrounded by flat terrain, on which a well-defined flow would impinge, providing clear boundary conditions. Providing that this requirement is almost never met in nature, realistically, the surrounding terrain's complexity should be significantly lower than the selected site's complexity. A quasi two-dimensional, or a long-ridge, hill is a logical choice, and, to assure a two-dimensional flow, dominant winds should be perpendicular to the ridge. Land cover, particularly forests, would add to the flow complexity and is considered desirable since many wind farms are installed near or within forested regions. According to these criteria, the Perdigão site was selected. The presence of a wind turbine at the site provided the opportunity for wake and inflow measurements.
It should be noted that the site selection is typically made considering specific criteria or the site itself triggers ideas for experiments. In the case of the Perdigão site, it was both. In 2009, the site visit initiated the idea for a double-hill experiment. In 2014, due to the presence of the wind turbine, the site became an eligible location for a measurement campaign addressing the wake and inflow conditions of wind turbines in complex terrain.
Perdigão site in September 2014
Perdigão (Fig.
Perdigão is an ideal site in terms of the orography and main road access,
but with a difficult access to the ridges. The access to ridges is mostly
through poorly maintained, steep and narrow unpaved roads, which creates
difficulties for the equipment installation. During late springs and summers,
the daytime temperatures are frequently above 30
The site orography and canopy were mapped in March 2015 during a helicopter
laser mapping mission. The area of 20 km
Wind characteristics (January 2002–December 2004).
Wind regime at met mast station for 3-year period at 40 m a.g.l.:
Computer simulations
Simulated wind flow on a surface 80 m a.g.l.:
Simulated wind flow in a vertical plane indicated by the dashed
line in Fig.
The two WindScanner systems comprised a total of six scanning lidars: three
LRWSs, named Koshava (LR1), Sterenn (LR2) and Whittle (LR3); and three SRWSs,
named R2D1 (SR1), R2D2 (SR2) and R2D3 (SR3). The lidar locations were
selected with the aim of sampling the flow field along the south ridge
and within the transect perpendicular to the ridges that goes through the
wind turbine (Fig.
We define the intersecting angle as the smallest angle between the
projections of two intersecting laser beams in a horizontal plane. The
intersecting angle can take any value between 0 and 90
The previously acquired point cloud and orthophotos assisted in choosing the
most accessible locations for the WindScanners' installation with respect to the
previously established criteria. LR3 was located on the south ridge next to
the wind turbine, and LR1 and LR2 were located on the north ridge
(Fig.
Perdigão site:
Position of instruments and landmarks given in datum ETR89/PTM06 (m).
The field campaign's control center, consisting of the LRWS and SRWS master
computers, data processing computer, network appliances, WiFi antennas,
security camera and a small work space, was installed in an office container,
20 m northeast from the wind turbine. Power and internet connections were
provided from the wind turbine substation by pulling 200 m long power and
fiber-optic cables from the substation to the control center, from which the
wired power and internet connections were further distributed to the nearby
WindScanners. For network redundancy, a secondary Internet connection for all
devices was provided via mobile network connections. For the uninterrupted
communication and power of LR1 and LR2 on the opposite (north) ridge, two
unidirectional WiFi links from the control center were configured and a power
container was located on the northeast slope of the north ridge approximately
1 km from each lidar unit (Fig.
After all scanning lidars were positioned, oriented and leveled at the
designated locations
For assessing the pointing accuracy of LRWSs, selected landmarks were mapped
with the LRWS laser beams using the CNR (carrier-to-noise) mapper
Since the DC component in the SRWS Doppler spectra is notched out, non-moving
hard targets cannot be used to assess their pointing and sensing range
accuracy. Instead of non-moving targets, for the SRWSs we employed a 12 m
calibration pole with two motor-driven balls on the top made of low-density
polyether (Fig.
Calibration of the SRWS:
Integrated spectrum versus focus distances for SR1 (R2D1, black line), SR2 (R2D2, green line) and SR3 (R2D3, red line).
Five scanning modes, Table
The ridge scan was designed to address the flow above the south ridge
(Table
In addition to the large-scale flow observations, two scanning modes for the
short-range WindScanner system (Table
To address the wake conditions with the short-range WindScanner system, the
vertical scan mode consisting of vertical lines was employed
(Fig.
While executing the second version of the
Scanning patterns of short-range WindScanners:
The transect scan was employed to investigate the flow field within a
vertical plane perpendicular to the ridges entailing the wind turbine
(Table
Transect scan: red lines indicate LOS measurements acquired by LR3, the vertical dashed line shows the position of the virtual mast, and the inclined horizontal dashed line indicates the position of the diamond scan plane.
The long-range WindScanner system's scans were run in a batch mode, where each strategy was executed over a 10 min period. The sequence of the ridge, diamond and transect scan was executed over a 30 min period and then repeated. Because only LR1 and LR2 were needed to execute the ridge and diamond scans, LR3 continued to perform the transect strategy throughout the entire campaign.
The Portuguese Institute for Sea and Atmosphere (IPMA) provided a daily forecast for the Perdigão site using a non-hydrostatic numerical weather prediction model of a limited area (AROME, the numerical prediction model of Meteo-France) for variables such as temperature, relative humidity, rain, wind direction and velocity at 10 m and 80 m a.g.l. The wind direction was used to decide the scanning mode of the short-range WindScanner system: the vertical mode if the wind was coming from the southwest direction (wake scanning) or the T-scan mode in the case of northeast wind (inflow scanning).
Scanning modes.
For the experiment, visits to the WindScanners and
hardware checks were performed several times per day in the first 3 weeks
of the experiment, as fine adjustments of the scanning modes were needed, and
then every other day in the last stage of the experiment. Also, the number of staff at
the site during the first 3 weeks decreased from 10 to 1 single
person, who managed the whole setup until the end. The status of the
experiment was reported daily on the dedicated web blog
At the end of the measurement period the pointing and sensing range accuracy
were reassessed with the LRWSs only. We found that the new positions of the
landmarks matched the previously mapped position within 0.05
The short-range WindScanner system was operational from 8 May until 3 June. During this period, 110 h (665 runs with a duration of 10 min each) of data were acquired. A total of 407 runs were made with the T-scan mode, addressing inflow conditions, whereas the remaining 258 were made by employing the vertical planes mode, addressing wake conditions. The long-range WindScanner system was operational from 19 May until 26 June. Overall, the long-range WindScanner system acquired 528 h of data, from which about 180 h of data were recorded with each scanning strategy. A special case is the transect scan, where in addition to the 180 h of concurring data acquired with all three LRWS there is also the additional 360 h of data collected only by LR3. There are 6 h of simultaneous measurements of the inflow conditions with both WindScanner systems.
Also, for the entire measurement period (May–June) the owner of the wind turbine (Generg) provided 10 min means of the wind turbine SCADA (supervisory control and data acquisition) data.
The acquired datasets of radial velocities were entirely processed and data
artifacts removed. The LRWS dataset was filtered on the CNR values of each
individual measurement point (CNR
The SRWS data were first filtered to remove signals that originated from hard
target motions (i.e., wind turbine blades). These signals are relatively easy
to detect due to the corresponding high-intensity power spectral density that
appears in the Doppler spectra, as a result of the increased backscattered
light from hard targets, relative to the one from aerosols. Afterwards, the
data are filtered by using an adaptive threshold of the acceptable maximum and
minimum of the total energy of the laser Doppler spectra. The threshold value
is calculated by the lower and upper outer fences of the distribution of the
total energy per spectrum. Consequently, the data were spatially averaged
after being grouped in cubic grid cells both of
4
Where possible, the processed radial velocities were combined to calculate
two (diamond, ridge and transect scanning methods) or all
three components of
the wind vector (
Data availability expressed in hours.
Both raw and highly processed datasets have been uploaded to a MySQL
database and are currently available for the participants of the projects
that funded the experiment. We intend to release the entire dataset for
public use during the second half of 2017 through the e-WindLidar web
platform (
Figures
Wind vectors, wind speed and direction (10 min averaged)
80 m a.g.l. along the south ridge for wind dominant directions acquired
using the ridge scan performed by the long-range WindScanner system:
The ridge scan, 2 km along the south ridge, shows no turning of the wind for
the northeast winds (Fig.
The initial data analysis of the wake measurements indicates a clear diurnal
dependence of the wake characteristics
The wake in a horizontal plane observed with the diamond scan
(10 min averaged) performed by the long-range WindScanner system for
different atmospheric conditions:
The wind speed (10 min averaged) wake in a vertical plane one rotor
diameter away from the turbine observed with the vertical plane scans
performed by the short-range WindScanner system:
The objective of the transect scans (Fig.
Radial flow fields (10 min averaged) in a vertical plane (left
figures) and corresponding virtual mast measurements of the horizontal wind
speed and wind direction (right figures) acquired using the transect scan
performed by the long-range WindScanner system:
Figure
Inflow conditions (10 min averaged) measured by a hybrid WindScanner
system on 3 June 2016 from 07:30 to 07:40 UTC:
In this paper, we presented the methodology for conducting field studies with multi-Doppler lidars. This was a preliminary attempt at outlining and defining systematic steps that can lead to the acquisition of high-quality datasets from field studies. Despite being developed for multi-Doppler lidar experiments, the methodology can be used for any field campaign. The majority of the presented steps are relevant for all field experiments (e.g., defining scientific question, planning infrastructure), while some are WindScanner specific (e.g., scanning mode design). In this paper, we presented a sequential execution of the steps. However, some steps can and should run in parallel (e.g., data archiving and dissemination with execution and data collection) and some steps can be merged together (e.g., experiment layout design and infrastructure planning).
The application of the methodology in the Perdigão-2015 experiment helped in
devising future improvements. Simulation results could be used to assess
whether WindScanners can capture the desired flow features by basically
simulating the flow study itself. This calls for the integration of an
end-to-end WindScanner measurement process simulator with a flow model
As the WindScanner wind vector retrieval accuracy is influenced by the LOS
uncertainty, pointing accuracy and retrieval technique, extending the
simulator with the lidar measurements uncertainty model would provide grounds
for a preliminary accuracy assessment and an optimization of the WindScanner
installation locations
To better assess pointing accuracy, more than one hard target should be used
and mapped prior, during and after the field study. The number of hard
targets could correspond to the number of error sources (e.g., leveling,
mirror alignment), since this would allow calculating the coordinate
system for steering the laser beams that can compensate for pointing errors
The methodology for atmospheric multi-Doppler lidar experiments was applied
in the pilot study Perdigão 2015, which serves as an introductory
campaign to more extensive and longer planned field studies within the NEWA
project. In this field campaign, the long-range and short-range WindScanner
systems were used simultaneously for the first time. Also, it was the first
time that these instruments were deployed in a challenging site. Over a short
period of time, both systems were installed at the designated locations. The
established WindScanner calibration and configuration procedures, which were
previously used in flat terrain, were successfully applied in the heavily
complex terrain of Serra do Perdigão. A pointing accuracy of
0.05
An important step has been made towards the coupling of the long-range and short-range WindScanner systems into a hybrid WindScanner system. Namely, the scanning strategies were designed and implemented in such a way as to achieve a symbiosis of the two WindScanner systems. However, in this first attempt to realize a hybrid WindScanner system, we can report that we managed to simultaneously operate a hybrid WindScanner system for several hours. A longer operation was hindered by the issues earlier mentioned in Sect. 4.8. The coinciding measurement periods will be thoroughly addressed in forthcoming publications. An obvious focus of future campaigns with a hybrid WindScanner system will be to achieve a full synergy between the long- and short-range WindScanner systems, thus to acquire longer simultaneous measurements of flow fields.
Despite the environmental conditions, challenges imposed by the site and
technical issues with the instruments, high-quality observations of the
various flow aspects of the Perdigão site have been acquired, with a few
highlights outlined in this paper. Also, the datasets and presented
methodology have been used for the design of a longer and,
instrumentation-wise, more extensive Perdigão-2017 field campaign, which
took place in the first half of 2017 over a period of 6 months
The methodology for atmospheric multi-Doppler lidar experiments was developed and applied to the Perdigão-2015 field campaign. We described the 10 steps which constitute the methodology and explained how each step was implemented in the field campaign. The application of the steps resulted in a high pointing accuracy and temporal synchronization of the WindScanners. Five novel scanning modes were designed, three in the case of the long-range and two in the case of the short-range WindScanner system, with the purpose of characterizing the overall flow pattern over the double-ridge site. Each scanning mode followed a specific purpose, disclosing a particular aspect of the flow. Because the larger area was covered by the WindScanner measurements in comparison to the standard tower based anemometry, a more detailed view of the atmospheric flow was possible, which increased our understanding of the interplay between the large synoptic scales associated with the weather conditions and the site. An important step was made towards the realization of a hybrid WindScanner system, based on both pulsed and CW lidar technology. Overall, the methodology and its application lay out the foundations for much larger future endeavors that will take place within the NEWA project.
The datasets presented in this paper are currently under conversion to the format which will make them simpler for sharing and reuse. They will be accessible throughout the web page (URL):
Based on the meteorological convention, the wind vector is defined by the three velocity components:
The LOS or radial speed,
By measuring three independent radial velocities (
If the elevation angle is low (e.g.,
Therefore, the radial velocity can be treated as the projection of the
horizontal components of the wind vector on the laser light propagation path.
Accordingly, by measuring two independent radial velocities (dual-Doppler
retrieval) with laser beams directed at low elevation angles, the horizontal
components of the wind vector can be retrieved:
The authors declare that they have no conflict of interest.
The authors are grateful to the following institutions, without which the
Perdigão-2015 field campaign would have not been possible. The
municipality of Vila Velha de Ródão was extremely useful and always
available to answer all our requests, providing solutions to many simple and
practical aspects of the experiment that otherwise would have been
insurmountable difficulties. We thank Generg for collaboration and support in
the field experiment, namely providing power and data connection and for
making the wind turbine SCADA data available. We thank the Portuguese
Institute for Sea and Atmosphere (IPMA) for the daily forecasts. We also
thank Regacentro Comércio de Representações Lda for the ingenious
solution that enabled off-grid power to two scanning lidars in remote
locations. Particularly, we would like to express special gratitude to Per
Hansen and Claus Pedersen from DTU Wind Energy for their dedicated work and
positive spirit during the experiment deployment and execution. We are
grateful to the FarmOpt, UniTTe, NEWA and WindScanner.eu projects for the
financial support of the Perdigão-2015 field campaign. FarmOpt
(