Lidar uncertainty and beam averaging correction

Remote sensing of the atmospheric variables with the use of Lidar is a relatively new technology field for wind resource assessment in wind energy. A review of the draft version of an international guideline (CD IEC 61400-12-1 Ed.2) used for wind energy purposes is performed and some extra atmospheric variables are taken into account for proper representation of the site. A measurement campaign with two Leosphere vertical scanning WindCube Lidars and metmast measurements is used for comparison of the uncertainty in wind speed measurements using the CD IEC 61400-12-1 Ed.2. The comparison revealed higher but realistic uncertainties. A simple model for Lidar beam averaging correction is demonstrated for understanding deviation in the measurements. It can be further applied for beam averaging uncertainty calculations in flat and complex terrain.


Introduction
Lidar is an acronym for light detection and ranging.Lidars are laser based systems working on principles similar to that of Radar or Sodar, see Boquet et al. (2011).There have been studies performed to validate the Lidar with metmast sensors by Smith et al. (2006), Lang and Mckeogh (2011), Westerhellweg et al. (2010) and more.In this study, the aim was to extend the filtering criteria to include the Lidar CNR, availability and Lidar's sensitivity in foggy and rainy conditions.Albers et al. (2009), Lindelöw-marsden (2009) and Albers et al. (2012) have worked extensively on evaluation of uncertainty in wind speed measurements with Lidar.This study aimed at identifying the sources of deviations in the measurements and quantifying the uncertainty in wind speed.Thus, significant atmospheric variables can be considered for inclusion in the CD IEC 61400-12-1 Ed.2, see Albers et al. (2012).
Lidars can detect wind speed without getting affected by the metmast and wake shadowing, one of the reasons is due to averaging of the 4 lidar beams separated spatially into one measurement.This study was aimed at evaluating the effects of Lidar beam averaging on wind speed measurement which can be further used for uncertainty calculations.A simple model for Lidar beam averaging is discussed here.These understandings would help the inclusion of Lidars and their appropriate uncertainties for the reference wind speed measurement into the power curve standard CD IEC 61400-12-1 Ed.2, IEC (2005).

Site description
The site considered for the comparison is the ECN windturbine test site at Wieringermeer, EWTW in North Holland as shown in Fig. 1.The test site is characterised by flat terrain, consisting mainly of agricultural area with single farmhouses and rows of trees.The lake IJsselmeer is located at a distance of 2 km East of metmast 3, MM3 (Lat: 52 • 50 N, Lon: 5 • 5 E).The data from MM3 and the two ground based vertical scanning Lidars are considered for the comparisons which is calibrated and installed using the IEC and Measnet guidelines, see IEC (2005) andMeasnet (2009).The relevant obstacles affecting the MM3 are the Nordex N80 wind turbines in the North direction at a distance of 283 and 201 m at the direction angles of 30 and 315 • respectively with the North.The cup, sonic and Lidar provide measurements at 3 common heights i.e. 52, 80 and 108 m heights.The Lidar provides measurements at additional heights which are not considered in this study.
The MM3 is a lattice tower mast with guy wires for support.The mast is constructed with tubular elements making

Data description
The data from the MM3 at EWTW used for this study is collected for the period 1 July 2013 to 26 January 2014, approximately 30 weeks (Bergman et al., 2014).The data includes 10 min averages amounting to roughly 30240 data points measured with the existing measurement standards (Measnet, 2009).The wind direction measurements from two wind vanes is combined into one wind direction measurement considering wake free sectors.Similar method is applied to the wind speed measurements from the cup anemometers for reducing the metmast shadowing effects (Bergman et al., 2014).The sonic anemometer at 52 m is available until mid October.2010) discuss about filtering of the Lidar which is however restricted to the continuous wave Lidar and parameters like low wind speeds, CNR values, wind directions and wake sectors.The extended data filtering is performed according to the procedure as shown in Table 1.The metmast shadow effects are valid for sonic anemometers as the cup anemometers are already averaged to minimize those effects.Further considerations to limit the Carrier to Noise ratio, CNR between −17 and 0 dB are being made since, only wind speeds lower than 10 m s −1 were found between 0 and 20 dB.While, the scatter of wind speed between −22 and −17 dB was high (R 2 = 0.65).D s is the standard deviation lower limit while the D δ is the minimum allowable difference between two consecutive measurements as shown in Table 1.The method according to CD IEC 61400-12-1 Ed.2 determines the accuracy classes of the Lidar measurement based on the predetermined atmospheric variables affecting the Lidar measurement.The accuracy classes are derived using the verification tests, see Fig. 3, sensitivity tests, Noise tests and Control test using a small mast for correlations, see Albers et al. (2012).The results of the tests for the EWTW test site for MM3 are shown in Table 2.The verification test results as in Table 2 match the ranges also suggested by Albers et al. (2012).Here, the procedure was extended to include wind shear and wind direction also.The Sensitivity test results are derived by the deviation of wind speed as a function of the variables listed.The accuracy classes are then derived by summing the sensitivity test uncertainties and dividing by √ 2, see Albers et al. (2012) for details.The accuracy classes are found to be in higher ranges than found by Albers.Uncer-

Effects of wakes on Lidar beam averaging
The Windcube probes the atmosphere with 4 beams in the directions North, South, East and West direction.The North beam of the Lidar is calibrated with the magnetic North and the Lidar self calibrates itself regularly to check for errors.
When one or more of the beams are in the wake sector, this results into an additional uncertainty into the Lidar measurements.To quantify this uncertainty, different scenarios of beam under wake are analytically calculated and applied as a correction on the measurement data.The uncertainty due to the atmospheric lapse rate, Lidar availability and CNR, and precipitation account for an independent uncertainty which cannot be neglected and shall be introduced into the CD IEC 61400-12-1 Ed.2 for power curve standard.The uncertainty in wind speed measurement of Lidar was around 3-4 % and that of sonic anemometer was between 4-5 %.The uncertainty calculated as a result is on the higher side when compared to the ones in the literature.This was expected as more uncertainties have been included into the uncertainty analysis.
The simple Lidar beam averaging correction method employed in this study resulted into 15 % improvement in the mean wind speed ratio within the wake sectors.A combination of bin-wise correction along with a wake model will provide better quantification of flow modelling uncertainty in Lidars.

Table 1 .
Filters used for data filtering.

Table 2 .
Albers et al. (2012)nd Sensitivity tests for MM3 for Lidar and sonic anemometer when compared against the reference cup anemometer.The accuracy classes are approximated according to the CD-IEC 61400-12-1 Ed.2 description inAlbers et al. (2012).The variables counted in the uncertainty are crossed for reference.