Coordinateurs du projet
Context
As part of their wind farm development/financing/operation activities, developers are regularly faced with the need to reliably and accurately quantify wind turbulence intensity. This assessment can be used to validate the correct positioning and sizing of wind turbines during the development phase, or to better assess the actual atmospheric conditions at hub height in order to optimize the operation of wind farms subject to various turbulence intensity constraints. The current reference measurement method for turbulence intensity is the cup or sonic anemometer, installed on a measurement mast at hub height. However, the costs and constraints of authorizations/installations for this type of solution at sea do not always allow for its deployment. The measurement of resources using Lidars installed on buoys (floating LiDAR) is now recognized and validated for offshore wind projects in terms of the 10-minute average wind resource. However, the measurement of turbulence intensity using floating LiDAR has not yet been validated.
The 24-month MATILDA project aims to quantify the error due to the buoyancy movements of the onboard Lidar system on the measurement of turbulence intensity and to propose data corrections based on sea conditions.
The analysis will be based on a comparison of fixed and floating Lidar measurements taken by EDF Renewables and AKROCEAN during the validation campaign for the WINDSEA floating LiDAR developed by AKROCEAN at the Fécamp site (measurement mast belonging to Eoliennes Offshore des Hautes Falaises, a consortium formed by EDF Renewables, Enbridge, and WPD Offshore). As a preliminary step, a comparison of turbulence measurements between a fixed lidar and conventional instrumentation will be carried out based on measurements taken at a site operated by VALOREM. In parallel with this analysis, the MATILDA project will design a device capable of carrying lidar systems and imposing controlled movements on them in a real environment.
Copyright: WINDSEA 2 @FECAMP copyright J.Vapillon – AKROCEAN
Scientific breakthroughs and innovation
- Processing of statistical wind data obtained in uncontrolled environments (wind and swell)
- Correction laws and processing algorithms enabling reliable estimation of turbulence intensity using floating LiDAR
- Design of a device reproducing the movements of floats (buoys or wind turbine platforms) at full scale
Expected technical and economic impact
Floating lidar is an accepted metrology method for estimating wind resources in offshore wind farm development projects. Floating lidar is also becoming indispensable for floating wind farm development sites where no other means can be deployed. The 10-minute averages of wind speed and direction are the two parameters extracted and are considered reliable. However, while this statistical information is sufficient for estimating annual energy production, it is insufficient for estimating the lifespan of wind turbines. Structural fatigue will be directly dependent on the level of speed fluctuations in the atmosphere, characterized primarily by the intensity of turbulence. The accuracy of turbulence intensity measurements using floating LiDAR and its dependence on sea conditions (wave period, specific wave height, wave direction relative to the wind) have not been clearly established, nor have the limiting meteorological and oceanographic conditions beyond which correction/compensation is essential been clearly identified.
The MATILDA project, led by LHEEA (Centrale Nantes/CNRS), is therefore part of this collective effort by the scientific community working on the characterization of offshore wind resources, contributing to the determination and reduction of the reliability limit. CNRS), is part of this collective effort by the scientific community working on the characterization of offshore wind resources. It contributes to determining and reducing the reliability limit of turbulence intensity measurements by floating lidar systems, based on the analysis of field databases and the design of a device for the controlled reproduction of floating lidar movements.
Results
1/ Technological and bibliographic study
Profiler lidars measure an entire atmospheric boundary layer profile vertically. The number of laser positions used depends on the lidar model. In this study, we focus on the WindCube V2 lidar, which has five positions: one vertical and four at the corners, as shown in Figure 1.
Figure 1: Profiler lidar laser positions [1]
D=2h tanΦ
Thus, the diameter of the measurement circle is 245 m at an altitude of 200 m. The distance between the measurement vectors involves two measurement errors: an averaging error and an error known as cross-contamination. The cross-contamination error is due to the equations of the VAD method, which are based on a first-order Fourier decomposition of the radial velocities. Higher-order terms are associated with divergence and deformation [3].
The estimation error of profiler lidars is low compared to sonic anemometers for measuring average wind speed over 10 minutes [2]. However, it is significant for measuring turbulence intensity, for two main reasons: the filtering of low turbulence scales due to the large measurement volume and contamination by the correlation of phenomena between the two measurement points [4]. Post-processing methods to limit the measurement error of turbulence intensity are proposed in the literature, two of which caught our attention: Squeezing [1] and Six-Beams [4].
The sea motion induced by the buoy causes the lidar to sway and introduces additional uncertainty regarding the position of the measurement points. The influence of the platform’s various movements on lidar measurements can be estimated theoretically [5]. However, an experimental study over a period of time representative of annual conditions would make it possible to accurately assess the extent of the turbulence intensity measurement error as a function of the buoy’s various movements.
Methods for correcting for buoy movement in lidar measurements are known [6] and have been implemented for Zephir-type lidar measurements [7], but no algorithm has been specified in the literature for a floating WindCube V2.
[1] F. Kelberlau and J. Mann, Better turbulence spectra from velocity–azimuth display scanning wind lidar, Atmos. Meas. Tech., 12, 1871–1888, 2019
[2] Sathe, A., Banta, R., Pauscher, L., Vogstad, K., Schlipf, D., & Wylie, S. (2015). Estimating Turbulence Statistics and Parameters from Ground- and Nacelle-Based Lidar Measurements: IEA Wind Expert Report. DTU Wind Energy.
[3] Browning, K. A. and Wexler, R.: The determination of kinematic properties of a wind field using Doppler radar, J. Appl. Meteorol., 7, 105–113, 1968
[4] A. Sathe, J. Mann et al. A six-beam method to measure turbulence statistics using ground-based wind lidars, Atmos. Meas. Tech., 8, 729–740, 2015
[5] J. Gottschall, B. Gribben Floating lidar as an advanced offshore wind speed measurement technique: current technology status and gap analysis in regard to full maturity, WIREs Energy Environ Volume 6, September/October 2017
[6]J. B. EDSON, Direct Covariance Flux Estimates from Mobile Platforms at Sea, JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY, Volume 15, April 1998
[7] F. Kelberlau Taking the Motion out of Floating Lidar: Turbulence Intensity Estimates with a Continuous-Wave Wind Lidar Remote Sens. 2020, 12, 898
2/ Initial analysis of measurements from the Fécamp site
At the Fécamp offshore site (N:5525385.76, E:300078.33), a database containing measurements from a floating lidar (WindCube v2 – 1-second data), a fixed lidar (WindCube – 10-minute data) and cup anemometers on a measurement mast (altitude 58.4 m – 10-minute data) has been compiled thanks to industrial partners AKROCEAN and EDF Renouvelables. An initial study of the measurements has been carried out, and it is considered here that the mast has a negligible effect on the measurements of the two lidars.
Selection of 13 weeks of representative data
Using oceanographic data provided by AKROCEAN and the ERA5 database, 13 weeks of data were selected from the period between April 21 and October 31, 2018. The selection criteria were to maximize data availability and represent all oceanographic conditions over the six-month period. To classify the data, the most representative variables (wave height, period, and direction; wind speed, direction, and shear; precipitation; and temperature) were selected, and a clustering method was implemented. The clustering method identified seven data groups that generally followed the same trends for the eight variables.
Subsequently, 13 weeks were selected to represent all the clusters of weather and ocean conditions and to capture the most unstable periods.
Validation of fixed and floating lidar measurements
As the measurement altitude of cup anemometers is low, they are used to validate the consistency of the measurements of the two lidars (fixed and floating). Here, the lidar measurements are taken from 10-minute files processed by the algorithm implemented by LEOSPHERE. Figure 2 shows that the measurements from the two lidars are consistent with those from the anemometer in terms of average wind speed. The study of turbulence intensity reveals an overall overestimation by the floating lidar compared to the fixed lidar.
Figure 2 – Linear regression of horizontal wind speed from fixed (left) and floating (right) lidars compared to anemometer measurements at an altitude of 58-4m
Evaluation of methods for post-processing measurements per second
Algorithmic corrections for sea motion on floating lidar measurements will be applied to the data at one-second intervals, so it is essential to implement post-processing for calculating 10-minute values. Several methods of peak filtering (Std or Wang) and data sampling (1B and 5B) were therefore applied.
To evaluate these post-processing methods, the relative absolute error compared to the fixed lidar turbulent intensity measurement is compared to that implemented by LEOSPHERE (Process_LEO) in Figure 3.
We can see that the processing applied to the measurements per second and the method used to calculate the turbulence intensity (linear detrend and alignment with the average wind direction) reduce the error of the floating lidar compared to the internal processing of the lidar.
Figure 3: Relative absolute error in the floating lidar’s turbulence intensity estimate compared to the fixed lidar measurement – comparison of several preprocessing methods
Influence of meteorological and oceanographic conditions on floating lidar measurement error
The study of measurements from two lidars (fixed and floating) at different altitudes, combined with meteorological and oceanographic data from ERA5 and the AKROCEAN buoy, provides an initial assessment of the influence of meteorological and oceanographic conditions on floating lidar measurement error. The correlations between meteorological and oceanographic variables and floating lidar measurement error compared to fixed lidar are presented in Table 1.
Atmospheric data are estimated by ERA5 with 1-hourly data, and wave data are estimated by the AKROCEAN buoy with 30-minute data. Thus, the 10-minute turbulent intensity measurement error induced by sea motion is averaged to correspond to the temporality of the meteorological and oceanographic data sources.
This table should be read as follows: the higher the average wave height in the half-hour, the greater the error in the floating lidar’s turbulent intensity measurement, and this trend increases with increasing measurement altitude.
3/ Modeling the behavior of floating lidar
Floating lidar measurements are evaluated in comparison with fixed instrumentation, which is considered to have a lower error rate than floating instrumentation due to the strong influence of waves. However, experimental measurements from lidar and fixed anemometers are contaminated with uncontrolled errors due to the technology used and weather and ocean conditions.
Therefore, the implementation of a model makes it possible to overcome this error and compare measurements from a simulated floating lidar with absolute truth in the numerical simulation of wind conditions.
In this study, TurbSim software is used to simulate turbulent wind slices, which are then stacked to form a turbulence box. In this turbulence box, a lidar is simulated without the optical part (Figure 4).
Figure 4: Schematic representation of the floating lidar model
TurbSim’s turbulent wind modeling is based on a stochastic representation of turbulence, so it does not take into account all of the physics involved. Thus, the use of this model to evaluate fixed lidar measurement correction methods is limited by the turbulence box.
However, the magnitude of the movement of a floating lidar due to the sea is known from experimental measurements, and the model can be used to evaluate sea motion corrections. Thus, the measurements of the motion-corrected floating lidar can be compared to those of a fixed lidar modeled using the same method. The use of the model therefore makes it possible to isolate the effect of wave motion on the theoretical measurement of the floating lidar.
The implemented model presents estimates of measurement error on average wind speed comparable to experimental values (1-2%). The sea motion compensation algorithm has not yet been implemented.