Radar Cross Section Analysis of Two Wind Turbines via a Novel Millimeter-Wave Technique and Scale Model Measurements

A novel, low cost, highly accurate, millimeter-wave RCS characterization method is developed and presented in this paper. In order to develop and verify the validity of the proposed method, full scale models and scale models of the horizontal-axis wind turbine (HAWT) and Crossflow turbines have been simulated and compared for a case study. The RCS of a scaled Crossflow turbine model was then experimentally verified using the novel method presented at frequencies of 76-81GHz. The proposed method utilizes the AWR1843BOOST evaluation board and DCA1000EVM real-time high-speed data capture card from Texas Instruments. To the best of the authors’ knowledge, this is the first RCS analysis of a scaled model performed at the mm-wave frequencies of 76-81GHz. This novel method is quick, simple, and fully automated, while maintaining high accuracy. Additionally, this has been achieved at a low cost using commercially available off the shelf parts. Good agreement was observed between the simulated and experimental results. Comparing the RCS data of the two turbines, it appears that the Crossflow turbine geometry offers a lower RCS and Doppler spectrum contamination as compared with a traditional horizontal axis wind turbine structure. These results are necessary and useful in allaying the increasing concerns regarding wind turbine radar interference, which have appeared as a result of the widespread adoption of wind power generation in recent years.


I. INTRODUCTION
Wind turbines are a strategically important renewable energy resource. Recently, there has been an increase in the adoption of renewable energy sources including wind power [1], [2] as part of a push by many governments towards reducing greenhouse gas emissions and the use of fossil fuels. An increase The associate editor coordinating the review of this manuscript and approving it for publication was Cheng Hu . in installed wind energy generation capacity is necessary in order for us to meet current carbon emission reduction targets [1], [2]. As the number of global wind turbine installations has grown, there have been increasing reports of wind turbines interfering with neighboring radar installations in various ways [3]- [8]. Firstly, wind turbines cause shadowing when they are located within the line of sight of a radar installation, reducing the radar's capability to track targets behind a wind farm. Secondly, the multi-path interference generated VOLUME 10, 2022 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ can cause the radar to show wind turbine images in places where they are not located. Additionally, due to their blade length, large wind turbines exhibit large blade tip speeds and significant Doppler spectrum contamination as a result [9]. These can often be above the minimum unambiguous velocity detectable by the radar, making them very difficult to filter, and causing the initiation of false tracks in the vicinity of the wind farm. For all radar wind turbine interference mitigation solutions, it is necessary for the impact on radar installations to be quantified. For this purpose, detailed wind turbine Radar Cross Section (RCS) data is needed. The work in this paper is primarily concerned with monostatic RCS, which can be described as the anisotropic reflectivity of an object in the direction of the incident radiation. As such, the RCS of an object can accurately predict the level of shadowing and clutter that a radar can expect to experience from that object, when it is located in the radar's line of sight. The RCS can be used for the identification and recognition of clutter caused by an object in some cases [12], which makes RCS data useful for the design of filtering algorithms designed to reduce radar interference.
A number of different solutions and mitigation measures have been proposed for the problem of wind turbine radar interference. One of the solutions proposed is the application of radar absorptive coatings and metamaterials to reduce the radar cross section of wind turbines [10], [11]. These solutions however have not been implemented widely at the design and manufacturing stage, and are very costly to apply retroactively to wind turbines that have already been commissioned. Another approach that has been proposed for the reduction of wind turbine-radar interference and that has been used for the RCS reduction of targets generally is that of shaping, i.e. changing the shape of the object in question so that is exhibits a lower RCS. In the case of the wind turbine blades, this is not possible, as aerodynamic performance constraints are the main driver for the blade shape. At the same time, it can be difficult and costly to redesign the tower of the turbine so that it exhibits a lower RCS. Typically, the interference of wind turbines is mitigated via the establishment of exclusion zones around radar installations.
The necessary RCS data can be obtained via simulation, a generally fast, affordable, and accurate solution to the problem [8]. However, it is necessary for simulation results to be verified experimentally. Full scale measurements can be very time consuming and expensive, requiring access to full scale radar installations [5], [13], as well as posing difficulties with controlling the large number of variables. With scale model measurements, it is easier to set up a controlled environment, the experiment takes less time, and is generally cheaper to perform. However, there are still significant costs associated such as that of building or hiring an anechoic chamber and obtaining a scatterometer system that works at a sufficiently high frequency [14]. Detailed RCS data exists for a variety of widely used wind turbine designs, most notably the horizontal-axis wind turbine (HAWT) geometry with three blades. The Crossflow turbine is a novel wind turbine with various advantages over traditional HAWT designs. The entire Crossflow rotor structure is mounted horizontally on a slew bearing and is driven by motors so that the rotor can be rotated to face the wind to maximize power production, and face out of the wind to reduce the frontal area in extreme wind conditions, thus limiting the loads on the tower and foundation.
In this paper, the first simulation results showing the RCS of the Crossflow turbine are presented, and compared against the RCS results of a commercially available HAWT model with a comparable rated power output. In this comparison, clear advantages for the Crossflow turbine are demonstrated. An experimental validation of results is also presented using the proposed novel millimeter-wave scale model RCS measurement technique, which was realized at a much lower cost than traditional methods [13], [15], [16], while maintaining a high level of accuracy and validity. This was achieved using commercially available off-the-shelf hardware, removing the need for a bespoke scatterometer system. The presented methodology can be used to quickly and accurately characterize scale model structures in a cost effective manner, making it easier to design and rapidly prototype low RCS structures for a variety of applications. Crucially, these measurements have been realized at a high frequency of 76-81GHz. This high frequency allows for both the range and scale model structure to have physically small dimensions. On the other hand, if the system functioned at a lower frequency of 10GHz, the size of such a scale model would have to be in the order of several meters. This poses practical difficulties and is an issue that is resolved by the high frequency method presented in this paper.
In section II, the theory behind the proposed novel RCS characterization method is described. In Section III, simulation results of the radar cross section of two different wind turbine geometries are presented, and the methodology for obtaining these results is also described. In the following Section IV, the proposed RCS characterization method is used to produce an experimental validation of the simulation results. In Section V, a detailed analysis of the results is provided, in addition to recommendations for future work, and a conclusion of the paper.

II. RCS CHARACTERISATION METHOD
The method proposed in this paper builds on the existing literature describing the RCS characterization of scale model structures [17]. Crucially, several novel aspects are introduced which make this method distinct from and preferable to methodologies described in the literature to date. In this section, the mathematical background and theory of scale model radar cross section characterization and an explanation of how this was leveraged in this scenario is provided first. Secondly, the novel characteristics of the method in both approach and implementation are covered in detail. Finally, any issues and limitations that may arise from this unique approach are examined in detail, and the measures that have been taken to mitigate their impact on the validity and accuracy of the results obtained are provided.
A scale model RCS testing method via comparison is thoroughly described in [17]. Initially, the RCS of the object used as a reference must be defined via the radar equation form seen in Eq. 1.
where σ Ref is the RCS of the reference object in question, R is the distance between the observing radar system and the object under test, P rRef is the power reflected from the object, P t is the transmitted power from the observing radar system, G is the gain of the transmitting antenna, L is the overall loss in the system, and λ is the wavelength of operation. For the RCS of the object under test, the RCS equation form shown in Eq. 2 can be used.
where P rObj is the power received as a reflection from the object under test. The only difference in these two equations is the presence of the P rObj term in Eq. 2 in place of the P rRef term in Eq. 1. This is to denote that the equations concern different scattering objects. As many of the constants in the equations are the same, and antennas which exhibit identical gain are used for transmitting and receiving, the ratio between the RCS of the object under test and the reference object can be expressed as shown in Eq. 3.
Eq. 4, an expression for the RCS of the model under test, can then be derived.
where a is the side length of the trihedral reflector chosen as the reference object. There are a number of conditions which must be observed when performing a scale model RCS measurement [17], [18]: • Firstly, a minimum distance between the antenna and test subject is specified in Eq. 5 as a function of the largest antenna dimension (D a ) and the largest dimension of the object under test (D ob ).
• Secondly, the relationship between the largest dimension of the object under test (D ob ) and the wavelength displayed in Eq. 6 must be satisfied, thus ensuring that the object under test is in the optical scattering region.
• Thirdly, the measurements must be conducted in a low reflectivity environment.
In Eq. 6, k = 2π/λ and is the wave number in free space. P rObj is backcalculated using the raw ADC voltage input values [19], [20]. This is achieved using Eq. 7: where G LNA is the programmable LNA gain (30dB), G Antenna is the gain of the series fed patch antenna (10.5dBi), and FPL is the free-space path loss (45.9091dB over 0.2m).
The power incident on the scale model, P tObj , can be found using Eq. 8: where P Tx is the maximum transmit power of the board (12dBm). The RCS in m 2 can then be calculated using Eq. 9: where R is the distance between the transceiver and the scale model. Finally, the RCS data for the scale model must have a scale factor applied to it [21] in order for us to obtain the full scale values, as shown in Eq. 10.
where n is the scale factor applied to produce the scale model. The presented novel mm-wave RCS characterization method was enabled by the use of the AWR1843BOOST evaluation module from Texas Instruments [22], which offers a 76-81GHz frequency modulated continuous wave radar implementation. The frequency span of the radar module corresponds to wavelengths of 3.9mm to 3.7mm, which satisfy the conditions required for valid measurements outlined previously.
Additionally, the third condition requiring a low reflectivity environment was satisfied by conducting testing in a wide-open area. At the employed frequencies, this constitutes a low reflectivity environment due to the large free space loss values observed in the mm-wave range. At 75 GHz, the free space path loss is approximately 59 dB/m. The RCS measurements were conducted in 0.9 • increments. The power scattered by the object in the direction of the transceiver, P rObj , and the power incident on the object, P t , are measured using the AWR1843BOOST + DCA1000EVM hardware combination.
The technical characteristics of the AWR1843BOOST millimeter-wave radar transceiver and the DCA1000EVM high-speed data capture are favorable for the application described in this work. Firstly, this hardware combination can produce sufficiently high frequencies, operating in the region of 76-81 GHz [22]. Additionally, the hardware has a low receive channel noise figure of 14-15 dB, built-in calibration, 3 Transmit and 4 Receive channels, and a robust software ecosystem allowing for quick and easy development. While the receive channel noise figure may not be considered low compared with more complex systems, this level is acceptable give the other advantages of the system, namely low cost, high measurement speed, and easy integration. The supplied software also allows the user to configure different transmitted chirp profiles for easy prototyping. The hardware has a configurable transmitted bandwidth in the region of 60 MHz to 5 GHz. The transmitted instantaneous bandwidth used in this application was 1.8 GHz. The range resolution achievable by the hardware is as low as 3.25 cm.
The band of 76-81 GHz was chosen due to the scaling requirement which states that the frequency used must be scaled by a factor n, where n is the scale factor of the model; in addition to the commercial availability of the hardware. This band also ensures that the object under test is in the optical scattering region at the corresponding wavelengths, i.e. l/λ 1, where l is the shortest dimension of the object. The experimental test system setup can be seen in Fig. 1. The performance of an indoor RCS test range is typically measured using the following parameters [23]: • Isolation: The level of isolation which is achieved with regards to external electromagnetic noise. The test area should be as quiet as possible, so that the only frequencies that contribute to the RCS measurement are those generated by the measurement system and reflected by the target within the RCS test range. In addition to the isolation of external electromagnetic noise, the isolation of other undesired signals must be considered. These include transceiver leakage, coupling between the transmit and receive antennas, reflections from features of the chamber, or chamber-target interactions [16].
• Quiet zone: The quiet zone typically dictates the largest target that can be placed within the indoor RCS test range that satisfies the far field conditions, i.e. still allows for a minimum distance between the transceiver and target necessary for the establishment of far field conditions.
• Minimum detectable RCS: The minimum detectable RCS is the smallest RCS that can be measured by the range/system.
• Dynamic range: This performance metric provides the difference between the maximum power transmitted by the transmit portion of the transceiver, and the noise floor of the receive portion of the transceiver.
• Frequency range: The range of frequencies over which the test system can perform measurements.  • RCS accuracy: This performance metric is dependent on the total uncertainty present in the system. The performance of the new system presented in this paper has been evaluated against other works in the literature using these parameters, and the results are presented in Table. 1. In this table ''isolation'' refers to the isolation achieved  towards electromagnetic signals external to the chamber or range.
Overall, it can be seen that the proposed method and system have several key advantages over methods presented to date in the literature. The most important of these are that the proposed method is low-cost, low-complexity, and fast. This system can therefore facilitate easy and fast prototyping for a number of different applications, where measurement accuracy is important but not safety critical. These advantages can serve to enable an RCS driven iterative design process.

III. SIMULATION AND ANALYSIS
In order to validate the proposed methodology, two wind turbine geometries were tested and compared -the Crossflow turbine, and a traditional HAWT geometry. The Crossflow turbine CAD model used for the simulation consisted of 150k faces, while the HAWT model was comparatively simpler at 30k faces.
The Crossflow turbine and HAWT models can be seen in Fig. 2. The HAWT model was obtained from a free CAD model repository [25]. The blades of the HAWT are rotated at an angle of 90 degrees such that the trailing edge is in line with the plane of rotation of the blades. The simulation results were obtained using the electromagnetic simulation software Xgtd from REMCOM [26]. This software uses the Physical Optics and Method of Equivalent Edge Currents (PO + MEC) methodology to simulate electrically large scenarios in a computationally efficient manner. Both models were simulated under illumination by linearly polarized VOLUME 10, 2022 (φ and θ polarized) plane waves at a frequency of 2.82 GHz, which is the main frequency used by primary surveillance radar (PSR) installations in civilian aviation radar applications.
The simulation scenarios for both turbines, including the plane waves used to illuminate the respective structures can be seen in Fig. 3 (a) and (b). The plane waves used to illuminate the two structures are of different areas, in order to ensure that the entire frontal area of each geometry is illuminated while optimizing the simulation process.
This work shows and confirms that the RCS of an object in the optical scattering region is strongly dictated by the geometry of the object. A minima in the RCS of the HAWT at approximately φ = 175 • can be observed in Fig. 4 and 5. This minima can be attributed to the reflectivity characteristics and geometry of the individual components of the HAWT and the way in which their reflectivity changes as the aspect angle changes.
Comparison plots with the RCS results of both the Crossflow turbine and HAWT under each polarization can be seen in Fig. 4 and 5. It can be seen that the obtained HAWT RCS results are similar to those presented in the literature [15], [24], [27], [28]. From the presented simulation results, it can be seen that the Crossflow turbine exhibits lower average monostatic RCS values when compared with a traditional HAWT geometry, while the peak RCS for both turbines occurs at the same aspect angles (φ = 90 • and φ = 270 • ).  These results are also supported by the heat maps of the RCS of the Crossflow turbine and HAWT, which can be seen in Fig. 7. This is evidenced in the results of the RCS simulations that were conducted of the nacelle and the blades of the HAWT individually. The RCS results of the individual components are overlayed on to the RCS results produced by the turbine simulated as a whole, and this can be seen in Fig. 6. These results shown that the local RCS minima of the blades only and nacelle and blades occur in the same place as they do  in the RCS results obtained from the simulation of the entire turbine (φ =~175 • ). With regards to the maxima of the RCS of the two turbines at φ = 90 • , φ = 180 • , φ = 270 • , these can be attributed to the high reflectivity of the nacelles of the turbines, which is owed to the prominence of flat surfaces in these aspect angles. In carrying out this work and explaining the scattering mechanism that generates the observed RCS plots, some advantages for the Crossflow turbine in terms of monostatic RCS and dynamic Doppler returns have been demonstrated in Fig. 4 and 5 when compared with a HAWT design of comparable dimensions. To elaborate on the simulation results presented here, two tables are presented comparing key RCS metrics belonging to the two turbines under different polarizations. It can be seen from Tables 3 and 4 that the Crossflow turbine exhibits lower minimum and mean RCS values than the HAWT under both polarizations tested, however it can be seen that the HAWT exhibits a slightly lower peak RCS than the Crossflow turbine.
As discussed in Section I, Doppler spectrum contamination is one of the major contributors to the interference generated by wind turbines in the vicinity of radar installations. In order to quantify the magnitude of the Doppler returns from the two wind turbine models presented, analytical work was carried out in MATLAB using the Phased Array System Toolbox [29]. The Doppler returns from the two wind turbine models were calculated at the point of maximum rotational velocity, representing a worst case scenario. The reason why maximum rotational velocity is used as the worst-case scenario is that under these conditions, a turbine exhibits its highest blade tip speed, i.e. the maximum linear velocity exhibited by the blade tip in the direction of the radar installation is exhibited under these conditions. The aspect angles at which this is the case are φ = 90 • , θ = 90 • for the case of the HAWT, and φ = 0 • , θ = 90 • for the case of the Crossflow turbine. Table 2 shows a comparison of the physical characteristics of the two wind turbines. It can be seen from the table that the Crossflow turbine has a lower peak rotational speed and shorter blades, resulting in a significantly lower maximum blade tip speed as compared with the HAWT. This lower peak blade tip speed corresponds to a much lower observed peak Doppler shift. The Time-Doppler plots for the case of the HAWT and for the Crossflow turbine are shown in Fig. 8.

IV. EXPERIMENTAL VALIDATION VIA SCALE MODEL MEASUREMENT
The newly proposed Radar Cross Section (RCS) characterization method (shown in Fig. 9) was used to measure the RCS of a scale model of the Crossflow wind turbine kindly supplied by Crossflow Energy. The novel RCS characterization method presented was used to measure the RCS of a 1:25 scale model of the Crossflow turbine. A comparison between the results obtained using the proposed experimental method and those obtained via simulation can be seen in Fig. 10 for the case of φ-polarization and Fig. 11 for the case of θ-polarization. These diagrams show the excellent agreement achieved between the simulated and experimental results. There is good agreement between the average RCS values, as well as between the location and magnitude of the peak RCS values, observed at φ = 90 • and φ = 270 • .
These specific areas (φ = 90 • and φ = 270 • ) have been highlighted because they are considered areas of interest, making them good points at which to perform a direct comparison of the simulated and experimentally obtained results. Generally good agreement is observed in these areas, however there are still some discrepancies that can be observed over certain other aspect angle ranges. The main cause of the discrepancy between the simulated and experimental results is most likely the non-uniform illumination of the target caused by the specific beam pattern of the antennas on the transceiver as opposed to the plane wave illumination used in simulation. To mitigate this limitation of the experimental setup, the scope of the experimental study was limited to the nacelle of the Crossflow turbine only, as the RCS of the tower is consistent across all turbines.

V. CONCLUSION
A novel high-frequency RCS characterization method has been presented. The main advantages of the proposed system are a fast measurement speed, low cost, low complexity, and compact nature. These characteristics make the system ideal for fast and easy prototyping, enabling an RCS driven design process, where measurement accuracy is important but not safety critical. Traditional methods typically require an anechoic chamber, and costly bespoke scatterometer systems. In place of these, a wide open space has been used as a low reflectivity environment, and a cost effective, commercially available radar system have been used in this work. A table comparing this work against others in the literature can be seen in Table 1. Using the proposed method, RCS measurements of a scale model of the Crossflow turbine have been obtained. These results have been compared against the simulation results of the RCS of a traditional HAWT design. Some advantages in terms of lower RCS and lower Doppler spectrum contamination have been demonstrated for the Crossflow turbine, making it more suitable for applications where a low RCS and lower presented radar interference is a concern. Further, good agreement has been observed between the full-scale simulated results and the experimentally obtained scale model results of the Crossflow turbine. This work will also enable the development of novel, low-RCS geometries and designs for a variety of applications such as aviation and ground infrastructure. AMIT MEHTA (Senior Member, IEEE) received the B.Eng. degree in electronics and telecommunication from Savitribai Phule Pune University, India, and the M.Sc. degree in telecommunications and information networks and the Ph.D. degree in smart reconfigurable antennas from the University of Essex, U.K. He is currently the Professor and the Director of the Antenna Research Group, Swansea University, U.K., where his core research interests include wireless communications, and microwave systems and antennas. He is leading large project teams on 5G, adaptive antennas for GNSS, high throughput satellite communications, the IoT, and millimeter waves. He has successfully supervised over 20 post graduate research theses. He has over 100 technical publications. He holds three patents on the invention of the steerable beam smart antennas and concealed weapons detection systems.
WEZI GONDWE received the B.Sc. and M.Sc. degrees in mechanical engineering from Swansea University. Currently, he designs novel micro renewable energy systems for Crossflow Energy Company. Before joining his current employer, he worked for an automotive OEM as a Process Engineer. His experience is varied, it includes controls and instrumentation, data acquisition, wind turbine design, data reduction and analysis, and FEA. He is a member of IMechE.  . His significant contributions are the development of integral equations for line antennas in free space and those for printed antennas, an L-shaped wire/strip antenna feeding method, and the inventions of numerous wideband antennas, including body of revolution (BoR) antennas, natural and metamaterial (NM) curl antennas, NM spiral antennas, and NM helical antennas. His other accomplishments include the design of antennas for GPS, personal handy phones, space radios, electronic toll collection systems, RFID, UWB, artificial satellites, and radars. His parabolic reflector antennas with a center-feed backfire helical radiator and flat antennas with low-profile helical radiators, both for direct broadcast satellite (DBS) programs, have reached a commercial penetration into more than 1,300,000 installations. He served as a member of the IEEE APS Administrative Committee, from 2000 to 2002, and a Region 10 Representative, from 2001 to 2010. He received the H. A. Wheeler Award, in 1994, the Chen-To Tai Distinguished Educator Award, in 2006, and the Distinguished Achievement Award, in 2016, all from the IEEE Antennas and Propagation Society. He was also a recipient of The Prize for Science and Technology from Japan's Minister of Education, Culture, Sports, Science and Technology, in 2010. Most recently, he was selected as a recipient of the Antenna Award of the European Association on Antennas and Propagation (EurAAP), in 2020. He is an Associate Editor of several scientific journals and magazines, including Electromagnetics. VOLUME 10, 2022