Air Traffic Control at Wind Farms with TERMA SCANTER 4000/5000 A.C.K.Thomsen#1, O.Marqversen#2, M.Ø.Pedersen#3, C.Moeller-Hundborg#4, E.Nielsen#5, L.J.Jensen#6, K.Hansen#7. #1-7

Terma A/S, Hovmarken 4, 8520 Lystrup, Denmark Contact mail: [email protected]

Abstract — The challenges of aircrafts detections in the area of wind farms are addressed. Requirements for a gapfilling radar solution is identified and obtained performance with SCANTER 4000 and SCANTER 5000 are described.

I. INTRODUCTION For many long range ATC-Radars (Air Traffic Control), wind farms reduce the detectability of aircrafts for several reasons [1]. First of all, the blades of wind turbines are moving at high velocities, making it difficult to separate wind turbine radar echoes from aircraft radar echoes by simple velocity filters. Secondly, the radar echo from the tubular wind turbine tower generates range side lobes in a significant range before and after a wind farm. The range side lobes can exceed the echo from small aircrafts in the same area and make the ATC-Radar ‘blind’ in a large area around a wind farm. The ‘blindness’ is a major problem for civil ATC operators and military RAP (Recognized Air Picture) creators and causes wind farm projects to be postponed or cancelled [4,9]. This paper describes the requirements needed for air detection in the area of a wind farms and present coverage performance obtained by Terma SCANTER 4000 and SCANTER 5000 air coverage radars. II. QUANTIZATION OF THE PROBLEM Long range ATC-Radars typically use pulse compression with chirp lengths of about 100µs or more. The pulse compression process generates range side lobes in a range determined by the chirp length. For a typical long range ATC Radar with 100µs chirp lengths the range side lobes will be 15km before and after a wind farm [1]. The peak level of range side lobes is typically -40dBp relative to the echo from the tubular tower (Fig. 5 from [1]). The Radar Cross Section, RCS, of a tubular tower is reported to be about 60dBsqm [10,2]. However, when the actual range and aspect angle at the tubular tower is taken into account as shown in Fig. 4, the RCS will be lower than 60dBsqm, but typically about 30dBsqm at S-band as shown in Fig 1. At X-band, shown in Fig. 2, the peak RCS is higher in the far-field, but the roll-off vs. aspect angle is faster so the actual RCS is expected to be at 20dBsqm. Mind that figures in this paper is not exact figures, but estimation of the order of magnitude. The range side lobes will desensitise the radar due to the signal of the range side lobes. The minimum detectable target in the area of range side lobes will therefore be about 0dBsqm (30dBsqm - 40dB + 10dB) for the typical S-band ATC radar if

978-1-4244-8900-8/11/$26.00 ©2011 IEEE

the radar has the same antenna gain at the target as at the wind turbine tower and 10dB detectability factor is required. For 2D radars (range/azimuth) elevation coverage is often obtained by shaping the antenna elevation pattern.

Fig. 1 RCS calculation of a perfect conical frustum with an upper diameter of 4m, a base diameter of 6m and height of 100m based on [3]. RCS is calculated at S-band in far-field (blue), 60km (purple),20km(magenta), 6km(red) and 2km(green) distance vs. aspect angle to centre line of frustum.

Fig. 2 RCS calculation of a perfect conical frustum with an upper diameter of 4m, a base diameter of 6m and height of 100m based on [3]. RCS is calculated at X-band in far-field (blue), 60km (purple),20km(magenta), 6km(red) and 2km(green) distance vs. aspect angle to centre line of frustum.

247

If the elevation pattern is shaped to a co-secant-squared antenna pattern, optimised for targets at 10,000 feet, the two way antenna gain on the elevated target can be 10-20 dB lower than the gain on the tubular tower, as seen in Fig. 3, where the signal gain for a target at 10,000 feet relative to the main loop signal gain is shown. Two way gain on target [dB]

Antenna gain on target in fixed elevated height of 10.000feet

− 10 − 20 − 30 − 40

20

40

60

80

100

Range [km]

Fig. 3 Two way antenna gain relative to peak antenna gain on an elevated target in 10,000 feet for a co-secant squared antenna as described in [5].

In this case the minimum detectable target due to range side lobes from the tubular tower reflection will be 10 to 20dBsqm.

Aspect angle [deg]

range

2 1.5 1 0.5 0 − 0.5

5

10

15

20

Range [km] Fig. 4 Aspect angle relative to the centre line of a vertical wind turbine tower for an elevated radar antenna positioned at 100mASL.

Fig. 5 Predicted pulse compression range sidelobes from a typical ATC radar performed by [1].

The described desensitisation so far is caused by only one wind turbine. When several wind turbines generate range or azimuth side lobes in the same desensitised cell, the vectors of all contributions shall be added. This means that if 10 wind turbines generate side lobes in the same cell, the desensitisation can in worst case be increased up to 20dB. The minimum detectable signal will in this case be 30 to 40dBsqm,

978-1-4244-8900-8/11/$26.00 ©2011 IEEE

which is far beyond the typical RCS from small aircraft of 0dBsqm. Desensitising can also be caused by dynamic saturation. If the azimuth resolution and the chirp length allow many wind turbines to be present in the same non-compressed pulse, the receiver must be able to handle the voltage addition of all the signals. If saturation occurs, the side lobe level will be increased above the mentioned level and cause additional desensitisation. The moving blades of a wind turbine will not be attenuated by MTI (or MTD) filters. If the range and azimuth resolution causes the MTI signals from the turbines to merge, or almost merge, inter-turbine visibility will be impossible. Even with fine resolution, typical CFAR processing will process the turbines in a large wind farm as background clutter and attenuate the whole wind farm area with some extension at that. III. REQUIREMENTS FOR INTER-TURBINE DETECTION Gap filling radars have earlier been proposed to mitigate desensitisation of prime ATC-Radars [1, 6]. However, in some of the presented configurations, the gap-filling radar must be positioned such that no illumination of the wind farm occurs. This can not always be achieved, especially for offshore wind farms. If the gap-filling radar should be able to detect small aircraft close to or even above the area of the wind farm, a number of requirements must be fulfilled. These requirements for inter-turbine detection can be categorized as follows: 1) Azimuth antenna resolution as fine as possible for the required scanning update rate. 2) Range resolution in the order of the physical extension of the wind turbine. 3) CFAR processing that excludes wind turbines in the background clutter processing. 4) Instantaneous linear dynamic range of the receiver higher than the difference of signals from the vector sum of towers in a non-compressed pulse / azimuth cell and the minimum target to be detected. 5) Adaptive sensitivity control to make the receiver able to place its linear dynamic range right to handle the actual number of turbines in the noncompressed pulse. 6) Pulse compression range side lobes sufficiently low, so that the signal strength in the range side lobes generated by the illuminated number of towers/blades will be sufficiently lower than the signal of target to be detected. 7) Azimuth antenna side lobes sufficiently low, so the vector sum of signals from towers in the antenna side lobes, but only in the pulse compressed range cell, will be sufficiently lower than the signal of target to be detected.

248

8) MTI/MTD processing sufficiently advanced to suppress clutter in the vicinity of the tubular towers. 9) Tracking algorithms which are not seduced by the presence of wind turbines. Quantisation of the listed requirements is of course site and application dependent, but the requirements are tried to be quantified in the following: 1) Antenna beam width shall be as narrow as possible, but is limited by instrumented range, update rate and hits per beamwidth. If instrumented range is 30nmi, update rate is 12 rpm and 12 hit per beamwidth is required for Doppler processing, the beamwidth shall be wider than 0.32°. This size of antenna beamwidth can be obtained by a 7m X-band antenna or a 21m S-band antenna and is therefore more convenient to realize at X-band. 2) Range resolution shall be as fine as possible to get interturbine visibility, but as the turbines can have some physical extension caused by the blades, range resolution do not need to be finer than 10 to 20m. Range resolution about 20m makes inter-turbine visibility possible as the distance between the wind turbines in a wind farm is typically between 200 and 500m. 3) CFAR processing needs to ignore the wind turbines and detect the clutter between the wind turbines. Methods for this have been addressed in several papers, such as [7]. 4) The maximum signal that the receiver shall be able to handle is the vector sum of signals from all towers in an area determined by the non-compressed pulse length times the antenna azimuth beam width. If the antenna beam width is 0.5° and the distance between wind turbines is 200m, only one wind turbine will be present in a beamwidth when the distance from the radar is below 45km. If the non-compressed pulse length is 20µs and the distance between wind turbines is 200m, 15 wind turbines can contribute to the vector sum of the signal. However, this requires that all wind turbines are lined up within an antenna beamwidth (which shall be avoided if possible). If the RCS for a single wind turbine is 20dBsqm, the receiver shall be able to handle the vector sum from 15 wind turbines, equal to 44dBsqm. As the signal from an elevated target can be attenuated due to lower antenna gain, an elevated target of 1sqm will result in a lower signal equals to -20dBsqm in the main beam. The instantaneous linear dynamic range of the receiver shall therefore be more than 64dB. 5) In the dynamic range described in 4), no range dependency is included. To avoid range dependency to be added to the required dynamic range of the receiver, adaptive sensitivity control can be used to place the dynamic range of the receiver so the maximum signal described in 4) just can be handled linearly by the receiver and processing. 6) Range side lobes shall be sufficiently low so the vector sum of range side lobes that can add up in a single cell as described in 4) will make target detection possible. If the target size is assumed to be 1sqm, the number of turbines that can contribute to a signal is 15, the RCS of a wind turbine

978-1-4244-8900-8/11/$26.00 ©2011 IEEE

tower is 20dBsqm, the gain on target is 20dB lower than the gain on the wind turbines and the detectability factor is 10dB, then the required range side lobe level will be 20+24+20+10=74dBp. Mind that this is a worst case factor that seldom occurs. 7) Antenna side lobes shall be sufficiently low so the vector sum of signals from antenna side lobes, that can add up in a single range-azimuth cell, will make target detection possible. If the antenna side lobes roll off as shown in Fig 9, wind turbines in an azimuth band of ±3° can contribute to the vector sum in a single range-azimuth cell. In worst case 24 turbines can contribute to the sum if the distance between turbines is 200m, the range is 45km and the turbines are positioned on an arc with the radar in its centre. This will very rarely happen due to the fine range resolution and the layout of the wind turbines. More than 10 turbines will seldom contribute to the vector sum in the same cell. Given the same assumption as in 6), but the number of turbines being 10, the required two-way antenna side lobe level will be 20+20+20+10=70dB or 35dB one-way. 8) MTI/MTD processing shall be able to suppress rain and sea clutter between the wind turbines without being seduced by the signal spectrum from the wind turbine tower and blades. 9) The tracking algorithm shall be able to distinguish between air targets and wind turbines. As the signals from wind turbines have a broad Doppler spectrum and flashes in amplitude, the signal from a wind farm can look like a target that moves between the wind turbine positions from scan to scan. The tracking algorithm shall therefore be able to classify “flashing signals from a stationary target” as wind turbines and remove the plots from the pool for other track associations. IV. OBTAINED RADAR PERFORMANCE The nine requirements can be used to evaluate how suitable a radar system is for detection of aircrafts in a wind farm or just in the vicinity of a wind farm. As an example the performance of SCANTER 4000 and 5000 is used in this section. 1) Antenna beamwidth close to 0.32°. The Terma produced antenna 15’ LACP-C has an azimuth beamwidth of 0.51°. This antenna is normally used for long range air surveillance, but for a gap-filling application with an instrumented range less than 40km and an update rate that shall match the longrange ACT radar, the 21’ LAHP-C [5] can be used with only 0.36° beamwidth. To minimize the RCS of the wind turbine tower, vertical polarization shall be avoided and therefore also circular polarization. 2) Range resolution less than 20m. Terma SCANTER 4000 has a range quantisation of 6m and the resolution can be configured from 12m and up. In normal air coverage configuration 24m is used to reduce straddling loss on target, but for inter-turbine visibility, 12m is preferred. Terma SCANTER 5000 has a range quantisation of 3m and the resolution can be configured from 6m and up.

249

3) Exclusion of wind turbines in CFAR processing. SCANTER 4000 and 5000 have a fine resolution CFAR map that can exclude cells down to 12x12m from the statistic analysis in the CFAR processing. 4) Instantaneous linear dynamic range higher than 64dB. The linear dynamic range of SCANTER 4000 and 5000 is limited by 14bit AD converters. In principle 14bit is equal to 84dB, but a few dB back-off of the maximum signal is needed to avoid saturation, and the kTB-noise floor needs to be more than 12dB above the minimum detectable signal of the ADC to make its quantisation noise insignificant to the kTB-noise. However, the receiver chain of SCANTER 4000 and 5000 has distributed sensitivity control where the first 25dB of attenuation is performed after the LNA. The kTB noise can thereby be attenuated below the quantisation noise of the ADC, which makes it possible to use the full dynamic range of the ADC. The following digital processing makes no limitation of the dynamic range as all linear processing is performed in floating point format with 138dB linear dynamic range.

Fig. 7 Nuttall Window function with 16 dB/octave decay.

It is important that the side lobes are generated by noise and therefore will be integrated as noise power instead of voltage signals. This reduces the requirement to 20+12+20+10=62dBp.

5) Adaptive sensitivity control to match the receiver dynamic range to the actual signal levels. SCANTER 4000 and 5000 detects the actual signal level and can in the next sweep change the attenuation level of the dynamic STC if saturation is going to be likely. The actual attenuation of the signal is reversed in the digital domain to maintain low range side lobe and MTI attenuation even with fast adjustments of the dynamic STC. To show how the dynamic STC works, Fig. 6 shows an example of attenuation vs. range and azimuth.

Fig. 8 Measured side lobes of a large cylindrical “thermos bottle” at the power plant Studstrupværket. At right the A-scope scale is from 0 to 75dB.

Fig. 6 High resolution dynamic STC map vs. range and azimuth

6) Range side lobes in the order of -74dBp. The pulse compression technique used in SCANTER 4000 and 5000

makes it possible to suppress range sidelobes to the level of the window function applied. To minimize sidelobe addition far from the peak signal, a window function with 16dB/octave roll off in range is used as shown in Fig. 7. However, the measurable side lobe level is limited by noise in calibration signals leaving a sidelobe level of noise just below -60dBp as seen in Fig. 8.

978-1-4244-8900-8/11/$26.00 ©2011 IEEE

7) Two way antenna side lobes below -70dBp. Fig. 9 shows the typical two-way antenna gain vs. azimuth angle of the Large Aperture, Horizontal Polarized Cosecant Squared Antenna LAHP-C produced by Terma A/S [5]. As seen the average peak side lobe level within ±3° is well below the -70dBp.

250

displacement of the wind turbines of only 100x200m, the signals from two turbines merge into one. Fig. 14 shows how the finer range resolution of SCANTER 5000 can be used to separate the wind turbines of Tunø Knob Wind farm and improve the inter-turbine detectability.

Terma 21' LAHP-C 9.17GHz

Two way azimuth side lobe level [dBp]

0

-10

-20

-30

-40

-50

-60

-70

-80 -15

-10

-5

0

5

10

Azimuth [deg], dot: 10x zoom

Fig. 9 Two-way antenna gain of Terma 21’ Large Aperture Antenna vs. azimuth. Dotted line is zoomed 10 times in angle.

15

Gap filling performance can be obtained both with shore based radars as shown, but also with in-farm positioned radars, as described in [6]. The challenges and benefits of in-farm positioned radar are not addressed in this paper. However, the gap-filling solution makes additional services available as – vessel traffic control – collision warning – turning warning flash lights on only on aircrafts approach – monitoring fishing boats in prohibited areas

8) Clutter handling between wind turbines. A clutter map in SCANTER 4000 and 5000 makes it possible for the statistic process to distinguish between clutter in fixed positions, as land and wind turbines, and distributed clutter. The map has a resolution of 12x12m which makes it possible to take out the signals from wind turbines from the statistics and make normal sea and rain CFAR processing between the turbines. 9) Tracker wind turbine handling. By defining a static target type which has zero mean speed and maintain track on this type of target even with low probability of detection (very fluctuating intensity), the Terma Embedded Tracker automatically generates an internal track on each wind turbine and thereby occupies the wind turbine plot to avoid faulty air target plot-track associations. V. RECORDING OF INTER-TURBINE DETECTION Recorded detection performance will be shown in this section. As an example, Fig. 10 shows a small target passing through Nysted Wind Farm [11]. Nysted Wind Farm is one of the largest off-shore wind farms in the world. The video was recorded from SCANTER 4100, a 12kWp TWT based radar, on a test platform on the Danish patrol vessel “Lommen”. The small boat passing between the turbines is fully detected only with a minor signal reduction when the boat is just behind a turbine tower caused by shadowing from the tower. Another example is shown in Fig 11, where a Cessna 172 is passing close to 10 wind turbines at Tunø Knob in a distance of 30km from the radar, a 200Wp solid state SCANTER 5000. As seen the aircraft is just detectable with the same low probability of detection close to the wind farm as before and after the wind farm. This shows that the radar has not reduced sensitivity in the vicinity of the wind farm. A third example is shown in Fig. 12 and 13 where a Bell helicopter type 206L is approaching the wind farm at Tunø Knob. The radar, SCANTER 4000 has been backed off 12 dB to make the target just detectable as seen on the approach in Fig. 12. On its passing of the wind farm, shown in Fig. 13, no indication of desensitivity is seen. Due to the distance from the radar to the wind farm of 30km and the small

978-1-4244-8900-8/11/$26.00 ©2011 IEEE

Fig. 10 Example of radar video at Nysted, Denmark with small target moving through. Yellow is the radar video of the last scan and red is decaying trails. Total size of the wind farm is 6x4km

Fig. 11 Example of radar video at Tunø Knob, Denmark with a small aircraft passing. Yellow is the last video of the last scan and red is decaying trails. Total size of the wind farm is 0.4x0.8km.

251

Fig. 12 Example of radar video at Tunø Knob, Denmark with a Bell 206L helicopter approaching. Distance between turbines is 100m times 200m. Distance from radar to wind farm is 30km. Radar is SCANTER 4000 with 15’LACP antenna, but backed off 12dB by 60m extra waveguide run.

VI. CONCLUSIONS The requirements needed to get aircraft detectability in the area and vicinity of wind farms have been argued and quantified. The requirements have been compared with the actual performance of the Terma developed and produced radar series SCANTER 4000 and 5000 to justify their capabilities in this aspect. Actual recorded inter-turbine detection has been shown and the benefits of this solution compared to an upgrade solution of standard ATC radars are listed. It shall be noted that SCANTER 4000 family is an offthe-shelf products that was installed at a customer for the first time in 2007 and now sold to 6 nations. SCANTER 5000 is a newly developed family of solid state based radars aimed for different applications [8] where SCANTER 6000 is a variant of the SCANTER 5000 complemented with functions for being installed on moving platforms. Article [6] demonstrates that the concept of gap-filling radars have proven to be acceptable by governments, and with the features listed here, this may even be easier to obtain in the future. ACKNOWLEDGEMENT Acknowledgement to all our colleagues who have participated in the development of SCANTER 4000, SCANTER 5000 and LACP antennas. REFERENCES [1] [2] [3] [4]

Fig. 13 Condition as in Fig 9 zoom in on helicopter passing by.

[5]

[6] [7] [8] [9]

[10] [11]

C.A Jackson, Wind farm Characteristics and Their Effect on Radar Systems. IET RADAR 2007, Edinburgh Oct 2007 L. S. Rashid, A.K. Brown., Impact Modelling of Wind Farms on Marine Navigational Radar, IET RADAR 2007, Edinburgh Oct 2007. W.B.Gordon, Far-Field Approximations to the Kirchhoff-Helmholtz Representations of Scattered Fields, IEEE Transactions on Antenna and Propagations, July 1975. M.M. Butler, Humberhead levels wind farms public inquiry (Keadby and Tween Bridge) Proof of Evidence EON/11/1 Radar Mitigation Measures, December 2006 (www.persona.uk.com/humberhead/PROOFS/Eon/eon-11-1.pdf) A. Østergaard, A.Thomsen and M.Løkke, A Low Loss and Low Reflection Duyal Lens for Shaped Pattern Applications, Proceedings of 30th ESA Antenna Workshop on Antennas for Earth Observation, Science, Telecommunication and Navigation Space Missions, 27-30 May 2008, ESA/ESTEC, Noordwijk, The Netherlands, pp. 303-306. E. Aarholt, C.Jackson, Wind Farm Gapfiller Concept Solution, EuRAD, Paris, September 2010. L. Sergey, Advanced Mitigating Techniques To Remove The Effects Of Wind Turbines And Wind Farms On Primary Surveillance, IEEE, 2008 www.terma.com FOX16 News, “Jodziewicz, American Wind Energy Association, said that projects totaling 10,000 megawatts of wind power were built in the U.S. last year, while projects involving another 10,000 megawatts were stalled by the radar issue.” www.Fox16.com March 18th 2010 J. C. G. Matthews, J. Pinto, J. Lord, RCS predictions for Stealthy Wind Turbines, European Conference on Antennas and Propagation, Nice, November 2006 Nysted Windfarm www.dongenergy.com/Nysted/EN/Pages/index.aspx

Fig. 14 Tunø Knob Wind farm on SCANTER 5000 with 6m range resolution in normal radar view without decaying trails. In the right side of the Radar Service Tool, the VRM-scope is shown on the top of the A-scope.

978-1-4244-8900-8/11/$26.00 ©2011 IEEE

252