Power Performance Measured Using a Nacelle-mounted LiDAR

Transcription

1 Power Performance Measured Using a Nacelle-mounted LiDAR R. Wagner, M. Courtney, T. F. Pedersen; DTU Wind Energy, Risø Campus, Roskilde, Denmark R. Wagner External Article English Introduction Wind turbine power performance requires the measurement of the free wind speed at hub height upstream of the turbine. For modern, multi-megawatt wind turbines, this implies heights in the range from 80m to 150m. Using the standard method with a cup anemometer mounted at the top of a met mast, this becomes ever more challenging and expensive, especially offshore. A nacelle based, forward looking lidar appears as a very attractive alternative since it requires no structure other than the wind turbine itself. Nacelle mounted pulsed lidars have already been demonstrated to be suitable for power curve measurements [1]. Although the scatter in the power curve was reduced in comparison to a simultaneous power curve based on a traditional mast-mounted cup anemometer, the experiment identified a discrepancy between the cup anemometer and lidar wind speeds that was not immediately easy to resolve. This highlighted the need for a traceable calibration procedure for the nacelle lidar that could form the basis of an uncertainty budget. This paper presents the results from a measurement campaign where a Wind Iris nacelle lidar was used to measure the power curve of an offshore multi-megawatt turbine. The lidar measurements were directly compared to simultaneous measurement from a cup anemometer mounted on the top of a met mast at hub height in front of the turbine. Prior to being installed on the turbine the lidar was calibrated at DTU s test site for large wind turbines. Here the lidar was installed on a fixed platform and the speed measured along each line of sight was compared separately to a sonic anemometer. This required several preliminary steps, such as the calibration of the lidar tilt, roll and lidar beam s opening angle. The paper structure is as follow: first the lidar used in this measurement campaign is briefly presented, then the methodology and the results of the calibration are described, and finally the power curve measurement is discussed. Description of the LiDAR The lidar used in this measurement campaign was a Wind Iris from Avent Lidar Technology [2]. This lidar system has been specifically designed to measure horizontally, so that it can measure remotely the wind speed and the wind direction at hub height once mounted on the nacelle of a turbine. This pulsed system is able to measure at several ranges from 50 m to 400 m in front of the turbine. The lidar uses two lines of sight (LOS) which are separated by a horizontal angle of 30. The lidar emits a stream of pulses per measurement in each LOS, switching between the two alternatively. This provides consecutive measurements of two radial speeds (projection DEWI MAGAZIN NO. 43, AUGUST

2 Fig. 1: A tilted and rolled lidar beam (red) shown in relation to the zero position (black). Point A is the origin of the beams (at the lidar telescope), point B is the detected position of beam 0 at distance L 0 and point C is the detected position of beam 1 at distance L 1. Due to the tilting (β) and the rolling (φ), the beam 0 and beam 1 positions are lifted by heights H 0 and H 1 respectively. L 2 is the distance between the detected beam positions. The full opening angle of the lidar beams is α. Fig. 2: Picture of two Wind Iris installed for LOS calibration on the platform at 9 m a.g.l. Fig. 3: Line of sight calibration seen from above - first beam. of the wind speed along the LOS) along each direction, at several distances. The horizontal wind speed and direction are retrieved from two consecutive radial wind speeds, one along each LOS, based on the assumption that the wind speed is horizontally homogeneous. Lidar Calibration The nacelle lidar calibration aims at ensuring the traceability of the measurements and quantifying their accuracy of the system during its operation. The set up should therefore ideally be identical to that of a power curve measurement, i.e. the lidar should be mounted at a height corresponding to wind turbine hub-height and shoot the beams towards an equally high mast situated at a distance of m (a typical value for 2.5D). However, such a facility was not available to us, so we opted for an alternative based on placing the lidar on a low platform (at 9 m a.g.l.) and shooting the beams towards a distant mast. Since horizontal homogeneity is impossible to achieve at such a low height, the lidar radial speeds were compared individually to reference speeds measured by a sonic anemometer. The measurement accuracy of the nacelle lidar is dependent on accurate measurement of the tilting and rolling of the lidar beams since these deformations can alter the effective sensing height of the instrument. The accuracy of the horizontal wind speed resulting from the radial speed calibration also depends on the accuracy of the beam opening angle. All three angles were calibrated before installing the lidar on the platform. Calibration of the tilt, roll and beam opening angle In order to calibrate the lidar beam geometry it is necessary to precisely determine the position of the two lidar beams in relation to the origin of the beams at the lidar telescope. The beam position was identified by an iterative process of blocking and un-blocking of the beam as identified from the reported signal strength (CNR). The end result of this process was a wooden target with a small hole through which the beam is known to pass. With the help of a theodolite the height of the beam positions is determined relative to a horizontal plane passing through the telescope origin and the distances from the telescope were measured with a laser distance meter. By repeating this process several times with several (small) tilt and roll displacements, the gain and the offset of the tilt and roll sensors could be determined and the opening angle calculated. Fig. 1 shows the geometry of the lidar beams exactly levelled in the horizontal plane (black) and after a tilt displacement β and a roll displacement φ. The inclinometers calibration revealed an offset of 0.4 in the tilt angle. The offset could have been accounted for in the lidar software, but it was chosen to leave unchanged for more transparency. The offset was accounted for (manually) during the lidar installation on the turbine (see section 4) and when the tilt measurements were used in the analysis of the power curve. 50 DEWI MAGAZIN NO. 43, AUGUST 2013

3 Werbung Kunde 1/1 s/w oder 4c Knowing is better As an offshore wind farm developer and operator you need to know about the marine conditions and the performance of your wind turbine. Based on our long term experience and the engagement in German research initiatives like FINO and RAVE since 2003 DEWI offers a wide variety of offshore services according to BSH standards and internationally accepted guidelines: Monitoring of offshore foundations Underwater noise measurements and prognosis Meteorological and oceanographical measurements Offshore LiDAR wind measurements Scour monitoring Consulting services and studies Energy yield assessment and micrositing As one of the leading international consultants in the field of wind energy, DEWI offers all kinds of wind energy related measurement services, energy analyses and studies, on-/ offshore wind turbine and component certification, further education, technological and economical consultancy for industry, wind farm developers and banks. DEWI GmbH is a member of MEASNET and is recognized as an independent institution in various measurement and expertise fields. /

4 Fig. 4: Lidar radial wind speeds divided by the projected vector wind speed (blue). A cosine fit for the nominal LOS direction is also inserted (red). Fig. 5: Plot of binned lidar radial wind speed versus binned projected reference wind speed with a forced and free linear regression inserted. LOS calibration procedure The lidar was mounted horizontally on a platform in a mast (as shown in Fig. 2), and oriented so that the first beam was aligned to pass closely a sonic anemometer in a second mast at 276m from the lidar, as represented in Fig. 3. The sonic was mounted at the same height as the lidar beam (8.5m a.g.l.). Having performed a calibration of the first beam, the lidar was turned and the second beam aligned to the same reference instrument. Although the LOS direction was given geometrically by the position of the two masts, our approach has been to determine this direction from the data since exact alignment of the sonic (to within a few tenths of a degree) is not practical. This was achieved by plotting the ten minute lidar radial wind speeds normalized by the vector mean speed of the sonic anemometer; an example is given in Fig. 4. Here we can see a maximum at an angle of around 290 where the wind direction coincides with the Wind Iris LOS. The accuracy of the lidar radial wind speed measurements were assessed by performing comparison to the wind speed measured by the sonic anemometer and projected along the line-of sight direction. This comparison was performed on the data after being filtered as follow: Lidar measurements availability1 within 10 minutes > 0.95; Sonic data with no error indication; Wind direction within 290 ±90 ; Sonic horizontal wind speed between 4 and 16 m/s. Fig. 5 shows the results from the linear regression obtained for the first beam. The regressions were performed on the data, which were binned according to the reference wind speed, so that the results are somewhat less sensitive to the exact data distribution. The wind speed measurement showed very good agreement to the reference sonic anemometer. The results for each beam are summarized in Tab. 1. Therefore, no correction was applied to the lidar wind speed given by the lidar in this measurement campaign. The calibration method is described in detail in [4]. Power Curve Measurement Experimental setup After calibration the Wind Iris lidar was installed on the nacelle of a multi-megawatt turbine, located in Avedøre, at the coast, approximately 10 km south-west from central 1 The lidar radial wind speed availability is the ratio between the valid radial speed measurement and the number if expected measurements (correspond to the number of streams of pulses emitted) within 10 minutes. The availability does not have a significant influence on the lidar deviation (i.e. difference between the lidar horizontal wind speed and the cup anemometer horizontal wind speed), except for very low availability (below 0.4). 52 DEWI MAGAZIN NO. 43, AUGUST 2013

5 Regression with offset Regression forced through the origin Gain Offset R 2 Gain R 2 Beam Beam Tab. 1: Regression results between the lidar radial speed and the projection of the sonic anemometer speed along the LOS for each beam. Fig. 6: Avedore Google Earth picture. The three turbines are denoted Ti ( with i between 1 and 3) and the met mast is denoted M. The nacelle lidar was mounted on turbine T3. (T3 : E N WGS84/ETRS89 Zone 33; M: E N WGS84/ETRS89 Zone 33) Fig. 7: Lidar measurement height at 2.5D relative to hub height. 10 minute data within hub height +/- 2.5% (black),10 minute data outside the range hub height +/- 2.5% (red). Fig. 8: Lidar 10 min radial wind speed availability for beam 0 (RWS0 availability) vs horizontal wind speed measured by the mast cup anemometer at hub height. Black: data with more than 0.75 availability; gray: data with less than 0.75 availability Fig. 10: Bin averaged power curve: with the cup anemometer measurements (black); with the lidar measurements (red). Fig. 9: Linear regression between the bin-averaged wind speed measured by the lidar at 2D and the mast cup anemometer at 2D. DEWI MAGAZIN NO. 43, AUGUST

6 Copenhagen. The location of the wind farm is shown in Fig. 6. The nacelle lidar was installed on the wind turbine labeled T3 in Fig. 6. Two other turbines ( T1 and T2 ) were located on the west side of T3 at approximately 450m and 800m. On the north east of T3 was the Avedøre power station with tall and large buildings. There was a met mast ( M ) at 2 rotor diameters (2D) to the south west (210 ) of T3. It was a lattice mast instrumented with a top-mounted Thies cup anemometer at hub height and a boom-mounted wind vane 4 m below hub height according to the IEC [3]. The south of the site is facing open sea, it was therefore considered as offshore wind conditions. The lidar optical head was installed on the roof of the nacelle 5 m behind the rotor plane and 1.8 m above the turbine hub. It was tilted downward by 0.7 in order to account for the height of the optical head above hub height and for the backward tilting of the nacelle when the turbine is operating. This pre-inclination angle was calculated so that the laser beams reached hub height at 2.5D for the wind speed giving the maximum power coefficient. The lidar was configured to measure at 10 ranges from 220m to 400m, including 2D and 2.5D (where D is the turbine rotor diameter) in front of the turbine. The sampling rate was set so that for each ten minute period 427 scans were achieved along each LOS. The effective probe length was 60m. The lidar was synchronized to all other measurements through an NTP server. Meteorological and wind turbine data were acquired by the turbine manufacturer s data acquisition system. This system was also regularly synchronised to an NTP server. Data from both systems were combined in a DTU database in such a way that the 10 minute lidar and wind turbine data were synchronised. Measurement height The nacelle of the turbine tilts backwards when the thrust increases, making the lidar beam tilt upwards, and thus measuring higher. Fig. 7 shows the variation of the lidar beam sensing height relative to hub height at 2.5D, which was derived from the 10 min mean tilt angle measured by the lidar inclinometer. As expected, the lidar was measuring at hub height for a wind speed corresponding to the maximum Cp (about 70% of rated speed). It was measuring below hub height for lower wind speeds, down to 97% of hub height for 20% of the rated speed. It was measuring above hub height for higher wind speeds, up to 101.6% of hub height for 95% of the rated power. For low wind speeds, the measurement height falls outside the range required by the IEC , within 2.5% of hub height. Data selection Before deriving the power curve, the following data were excluded: turbine status indicating a failure ; Wind direction outside the sector (in order to be able to make a comparison with the measurements from the met mast); Lidar radial wind speed availability below 0.75; a generator rotation speed below x max. RPM. The 10 min lidar data availability increases with the turbine rotor speed, therefore with the wind speed (see Fig. 8). As the blades passed faster in front of the lidar beam, a larger number of pulses could get through in between the blades. From a certain rotor speed (0.7 x rated speed), the blades passed fast enough to ensure a valid measurement for every stream of pulses and therefore get an availability of 1. To exclude data with an availability higher than 0.75 would have removed low wind speed data and therefore would have truncated the resulting power curve. However an availability of 0.75 only ensures that every 10 min average radial speed was calculated with a minimum of 320 valid measurements (along one LOS). There is no clear and direct way to know how these measurements were distributed in time within the 10 min period; and therefore there might be some large gaps in the 10 minute time series. For this reason the generator speed filter was also applied in order to avoid large gaps due to one blade continuously blocking the laser beam. Results The 10 minute average horizontal wind speed and direction were retrieved from the averaged radial wind speeds, based on the assumption that the wind speed was horizontally homogeneous. Average longitudinal and transversal wind speed components V x and V y are first computed as: and The average horizontal wind speed <V> is then retrieved by computing the vector modulus: The lidar mean horizontal wind speeds were compared to the mast cup anemometer 10 minute wind speed. Fig. 9 shows the linear regression performed on the bin-averaged data. The data were bin-averaged according to the cup anemometer wind speed. The lidar measurements compare well to the cup anemometer. Fig. 10 displays the bin-averaged power curves obtained with the lidar at 2D and with the cup anemometer for the same dataset. The same air density correction, using the mast temperature, was applied to both measurements, according to the IEC [3]. The two power curves are very similar. The lidar power curve, measured at 2D, results in an underestimation of the annual energy production (AEP) by 0.6% compared to that obtained with the cup anemometer power curve (for an average wind speed of 8 m/s). Discussion This paper presented the results of a power curve measurement using a 2-beam nacelle mounted lidar. Prior to being installed on the turbine, the lidar was calibrated in order to ensure the lidar wind speed measurement accuracy and provide the basis for an uncertainty budget. The technique adopted to calibrate the lidar was to compare the radial wind speed measured by along each LOS with the projection of the wind speed measured by the sonic anemometer along the LOS direction. This method has the disadvantages of being time consuming since two individual lines of sight must be calibrated separately. Alternatively, both lines of sight could be calibrated simultaneously with a setup using one platform and two masts but it would be necessary to use individual reference sensors. A second reason for long test 54 DEWI MAGAZIN NO. 43, AUGUST 2013

7 durations is that the comparison is made between projected wind speeds and lidar radial wind speeds. High values of these parameters can only be measured when high wind speeds occur in a wind direction close to the LOS direction. How to use the individual LOS calibrations is not immediately apparent, especially if they are significantly different. However a consensus is emerging that it is more accurate to use vector mean wind speeds from nacelle lidars since the cross-contamination of the turbulence components can give rise to over-estimation of especially the transverse turbu lence component, leading to significant errors in scalar aver ages. In the case that vector averaging is chosen, LOS calibra tions can be readily applied since the vector averages can be derived directly from the ten minute averaged and calibra tion corrected, individual radial wind speeds. After the calibration provided satisfactory results, the lidar was installed on a multi-megawatt wind turbine to measure the power curve. The main deviation from the IEC requirements was that the lidar cannot always measure at hub height +/- 2.5%. For this measurement campaign, only a few data around cut-in wind speed were outside this range and could have been excluded without causing major detri ment to the resulting power curve. More generally, the amount of data outside the permissible sensing height range probably depends on the flexibility of the turbine tower and in the case of a very flexible tower; data around rated wind speed (where the highest tilt is expected) could be outside the range upper limit. Removing these data from the power curve would not be acceptable and an alternative would be to increase the uncertainty of these data, as suggested in [5]. In order to obtain a complete power performance measure ment, the uncertainty should be assessed, requiring to first identify all the uncertainty sources and secondly to quantify them. The uncertainty sources should at least include the calibration uncertainty and the uncertainty related to beams tilting during the power curve measurement. This will be reported on in the near future. Conclusions We have performed a power curve measurement with a 2-beam nacelle mounted lidar. In order to assess the accu racy of the lidar measurement, the lidar was calibrated before being installed on the turbine. A method to calibrate directly the radial wind speed along the two lines of sights was tested, giving satisfactory results. During the measure ment campaign on the turbine, the 10 minute mean wind speed measured by the lidar compared very well to the mea surements of the cup anemometer mounted on a mast in front of the turbine, within a horizontally homogeneous flow. Consequently, the bin-averaged power curve obtained with the lidar was very similar to that obtained with the IEC stan dard set up, resulting in a difference in AEP of only 0.6%. The main challenge in using this technology for power curve mea surement is that the lidar tilts back and forth due to the motion of the turbine nacelle. In this campaign, most of the data were within the range recommended by the IEC stan dard: hub height plus or minus 2.5%. Only few points were outside this range, for low wind speeds. This analysis demon strated that the nacelle mounted lidar is a promising technol ogy for power curve verification if the deviations from the standard approach are handled with care. In order to com plete the power curve measurement, the complete uncer tainty budget must be assessed. Acknowledgements The work described in this report has been carried out under an EUDP funded project (journal no ). The au thors acknowledge Avent Lidar, DONG Energy, and Siemens Wind Power A/S for their fruitful collaboration within this project. The authors are grateful to the DTU s technicians for the installing and operating the instruments during the cali bration campaign. References 1. Wagner R. et al, Power curve measurement with a nacelle mounted lidar, Wind Energ. (2013), Published online, DOI: /we IEC standard :2005, Power performance measurement of electricity producing wind turbines. 4. Courtney M., Calibrating nacelle lidars, DTU Wind Energy E-0020, Janu ary Wagner R. et al, Procedure for wind turbine power performance measurement with a two-beam nacelle lidar, DTU Wind Energy E-0019, January 2013 ANEMOMETERS & WINDVANES A100 Series The First Choice for First Class ACCURATE Proven "First Class" instrument for WIND ASSESSMENT with SUPERIOR PERFORMANCE IN THE FIELD CALIBRATED Available calibrations: MEASNET, IEC "/PRO bundles" with MEASNET calibration certificates usually available from stock ROBUST & RELIABLE Engineered using Anodized Aluminium, Stainless Steel & 40 years of experience in the field NEW: A100AC Low Level AC Output Anemometer joins First Class A100 Series line up...now everyone can upgrade to proven FIRST CLASS performance REPAIR/RE CAL WINDVANES W200P Series Windvanes OFFSHORE We can Service/Repair and Recalibrate your existing Vector Sensor inventory making them "as good as new" to install at your new sites Offshore/Marine versions of A100 series "First Class" Anemometers now available DEWI MAGAZIN NO. 43, AUGUST