Study of Backscattering from Small Wind Turbines Powering BTSs: RCS Simulations and Measurements

Transcription

1 Study of Backscattering from Small Wind Turbines Powering BTSs: RCS Simulations and Measurements 1 Tran Vu La, 1 François Le Pennec, Fabrice Comblet, 3 Serge Elenga 1 Lab-STICC (UMR CNRS 685), MW Dpt., Telecom Bretagne, Brest, France, Francois.LePennec@Telecom-Bretagne.eu Lab-STICC (UMR CNRS 685), ENSTA Bretagne, Brest, France 3 Bâtiment Gérard Megie, IDSUD ENERGIES, Aix-en-Provence, France Abstract Hybrid-powered base stations have become an effective solution to reduce fossil fuel consumption, when power demand for future mobile networks (LTE/4G, 5G) dramatically increases. Such base stations are powered by small wind turbines (SWTs) having nominal power in the range of kw. Due to the close proximity between SWTs and BTSs, questions have been raised about possible radio interference. In the context of the OPERA-Net European project, the main objectives of this original work concern cellular radio / SWT compatibility as well as recommendations to mitigate possible related disturbances. Compared to megawatt WTs, SWTs have a variety of shapes, materials and orientations (horizontal / vertical). Thus, a simple representative model (RSWT-u) was first designed. It is fully metallic to correspond to the worst radio case and its dimensions are scalable to SWT nominal power. The RSWT-u dimensions have since been optimized to improve the RCS match between the RSWT and realistic SWT models having the same nominal power. To verify, comparisons have been made between the RSWT and 3.5 kw Nheowind-100 model (IDSUD ENERGIES). A scaled prototype of the 3.5 kw RSWT with a 1:10 ratio was manufactured. Its RCS was measured in an anechoic chamber at 9 GHz and 17 GHz. A comparison between simulation and measurement at these frequencies shows good agreement. Taking into account the frequency and dimension scale factors, the results obtained at radio cellular frequencies are validated. Keywords hybrid/wind-powered base station, small wind turbine (SWT), base transceiver station (BTS), radar cross section (RCS), Doppler shift, scattering, Luneberg Lens calibration. I. INTRODUCTION The importance of hybrid energy has been identified in recent years due to the dramatic increase in power demand. Coupled with solar panels, a number of wind farms have been deployed in the world. However, this has caused various unwanted effects, such as acoustic disturbance, increased mortality of birds, etc. More particularly, the electromagnetic (EM) interference produced by such wind farms can disturb radars, televisions (TV), GPS, etc. The initial studies on this subject were carried out in the 1970s [1]. They used analytical EM calculations to evaluate the impact of a simple wind turbine (WT) model on analogue TV. Since the 000s, studies concerning the radio impact of WTs on radars have exploded. The CAD models used in these studies were nearly identical to real WTs []. Scattering from such models was simulated and analyzed by EM computation software. Additionally, various Fig. 1. Different installations of hybrid-powered base stations. WT on a BTS mast in Vietnam. WT separate from a BTS mast in Turkey. solutions have been proposed to mitigate the EM impact of WTs, including improvement of radar signal processing, modification of nacelle and tower shapes [3], and use of radar absorbent material [4]. Coupled with megawatt WTs, recently smaller WTs (SWT) have been used more and more to power factories, homes, etc. In radio mobile communications, due to demands in ultra broadband and high data rate, power for future networks (e.g. 4G/LTE, 5G) is increasing very rapidly. In this context, base transceiver stations (BTSs) powered by hybrid energy (solar/wind/fuel) have become an effective solution to help reduce fossil fuel consumption. Additionally, it is an example of energy independence for BTSs that are too far from the power supply network. Such BTSs have been deployed in many regions of the world in two main installation scenarios: SWT on a BTS mast (Fig. 1a) and separate from a BTS mast (Fig. 1b). It should be noted that differently from MW wind turbines, many different manufacturers and models exist for SWTs (horizontal, vertical, shapes and materials, etc.). Typical SWT power for most mobile BTSs is in the range of kw. Hence, compared to megawatt WTs, the dimensions of SWTs are much smaller. However, because of close proximity between SWTs and BTS antennas (as in Fig. 1a), questions have still been raised about the EM impact of SWT on BTS, and thus on radio mobile communications. Additionally, the moving blade of an SWT can produce Doppler shift (frequency deviation) and possibly degrade the

2 demodulation of the received signal. In the context of the OPERA-Net European project, a study is being carried out on the interactions between cellular radio transmissions and SWTs that contribute to power BTSs. Depending on the intensity of these interactions, related installation guidelines for SWTs will be proposed. This paper is organized as follows. Section II reminds the reader of the indicators (Radar Cross Section, Doppler Shift) used to analyze EM scattering from SWTs. It also presents the design and dimension optimization of a representative SWT model (RSWT). This model is then compared to the realistic 3.5 kw Nheowind-100 one (IDSUD ENERGIES). In Section III, near-field coupling between the RSWT and a panel antenna in the worst case scenario at 900 MHz is studied. Section IV describes the scaled prototype of the RSWT which was manufactured, as well as the instrumentation and calibration for RCS measurement. Comparisons between these simulated and original results of measured RCSs are also presented and discussed. II. SWT MODEL STUDIES A. Indicators Radar Cross Section (RCS) is used to define the intensity of energy backscattered to the wave source (monostatic RCS) or to the receiver (bistatic RCS). For an SWT, complex backscattering is caused by all components (blade, nacelle, tail, etc.). From the well-known radar equation, RCS is described as σ Es ( θ, ϕ) = lim 4πr r Ei ( θϕ, ) where E s (θ, φ) is the scattered field and E i (θ, φ) is the incident field illuminating the target located at distance r from the receiving antenna. For an incident wave having linear polarization (vertical or horizontal), there are two polarizations appearing in the scattered signal. However, because the cross polarization is generally negligible, only the co-polarization is considered in the determination of RCS [5]. Thus, in this case depending on the linear polarization of antennas, monostatic RCS is taken into account as σ VV (vertical-vertical) or σ HH (horizontal-horizontal). Coupled with RCS in a WT scattering analysis, Doppler shift describes a frequency deviation observed in the backscattered signal due to the moving of rotor blades. It is defined in relation to the tangential velocity of blades v R (m/s) and the frequency of transmitting signal f 0 (Hz). Obviously v R depends on the considered radial position and rotational speed of the blades. f D v R (1) = f 0 () c During rotation, if blades move away from receiving antenna (v R > 0), the Doppler shift is negative. On the contrary (v R < 0), if blades move toward it, the shift is positive. Fig.. Geometries of uncalibrated RSWT (RSWT-u), RSWT and Nheowind-100 models having the same 3.5 kw nominal power. B. Design, Comparison and Calibration SWTs powering most BTSs have nominal power in the range of kw. Compared to megawatt WTs, their shapes, materials and orientations are much more diverse. Thus, a representative model (uncalibrated RSWT in Fig. ) has been studied in the other work of the same authors [6]. It is fully metallic to correspond to the worst radio case. The model consists of three blades, one nacelle and one tail. Its dimensions have been computed from the calculation formulas for big WTs. They are proportional to SWT nominal power. In addition to the RSWT-u, the 3.5 kw Nheowind-100 realistic model made entirely of metal was studied in [7] (Fig. ). With noticeably curved blades, this model is a significant example of the diversity of SWTs. Compared with the original CAD design, the Nheowind-100 model in this study has been simplified to mitigate the unnecessary meshes due to small holes, screws, etc. This simplification did not have any effect on RCS pattern, while it helped to save a lot of simulation time and computer memory during EM simulations (FEKO version 6.3 CAD Software with MoM/MLFMM numerical method). As shown in Fig., the RSWT-u having the same 3.5 kw is much smaller than the Nheowind-100. One reason is that the dimensions of the RSWT-u were calculated based on the formulas for megawatt WTs. This makes them less accurate compared to realistic models. On the other hand, to increase efficiency and reduce noise, the size of the Nheowind-100 is much larger than the other models having the same nominal power. Consequently, in Table I, significant differences in RCS mean value of two models are noted. They are 4.8 db in the horizontal plane, 7 db and 4 db in the vertical front and side planes, respectively. These results suggest that a calibration of the RSWT-u dimensions would be useful in reducing the difference in the RCS mean values of the two models. TABLE I. RCS MEAN VALUES OF UNCALIBRATED RSWT, RSWT AND NHEOWIND-100 MODELS AT 900 MHZ RCS Mean Value (dbm²) Plane Nheowind- RSWT-u RSWT 100 Horizontal Vertical front Vertical side Such a calibration was obtained based on the comparisons with the RCS mean value of the Nheowind-100 model in three

3 model). It is presented in Fig. 4. In this spectrogram, maximum Doppler shift has been scaled to correspond to the maximum value in (). It corresponds to the worst case of Doppler spread, if we consider the negligible effect of the tail on RCS. In a nominal rotational period of blades (T = 0.48 s), six flashes having maximum power spectral density (PSD in dark red) and Doppler shift are noted. This signifies that the backscattering level at these positions is the most significant. At 900 MHz, maximum Doppler shift is 146 Hz, there are three negative and three positive shifts in a rotational period. The flash period T bl counts 80 ms corresponding to the time between straight up and down blade positions. Fig. 4. Doppler spectrogram describing the RCS observed on the side (y axis in Fig. 5) of the RSWT at 900 MHz, ω R = 15 rpm. Fig. 3. RCS pattern of the RSWT-u, RSWT and Nheowind-100 models at 900 MHz in horizontal plane and vertical front plane. planes at 900 MHz. Indeed, in Table I, the RCS mean values of the calibrated model (RSWT in Fig. ) are very close to those of the Nheowind-100. Compared to the RSWT-u, they have been improved 5 db in the horizontal plane and 6.4 db in the vertical front plane. The geometries of these three models are laid out in Fig.. In comparison to the dimensions of the RSWT-u, those of the RSWT are much improved. However, they are not comparable to the dimensions of the Nheowind- 100, due to the specific shape of the latter. Meanwhile, they are very close to the other realistic SWT models having the same nominal power and similar shape (e.g. Kingspan models in Fig. 1 scaled from 3 kw to 3.5 kw nominal power). The RCS comparison of the RSWT-u, RSWT and Nheowind-100 is shown in Fig. 3. In both planes, the RCS peaks of the RSWT-u and RSWT are more noticeable than those of the Nheowind-100, because the backscattered signal of the latter one spreads in all directions due to the particularly curved blades. This also causes the RCS peaks of the Nheowind-100 to increase while the variation of RCS pattern is less than that of the other models, notably in the vertical front plane (Fig. 3b). C. Doppler Spectrogram Following the references [8], The Doppler spectrogram is calculated using a Fourier function to transform the monostatic RCS variations observed in the plane containing moving blades (yz in Fig. 5),. From EM simulations at 900 MHz, the Doppler spectrogram of the RSWT is determined at nominal rotational period ω R = 15 rpm (nominal speed of the Nheowind-100 III. NEAR FIELD COUPLING WITH A PANEL ANTENNA Based on the wind-powered base stations in Fig. 1, Fig. 5 models the disposition between the RSWT and a panel antenna studied in [9]. The panel antenna is a generic four dipole model with maximum gain 16 dbi and aperture angles of 65 in horizontal plane and 1 in vertical plane. The horizontal distance between them is 0.5 m while the vertical one, indicated as d r, is variable. θ rot represents the rotation angle of the blades. If d r is small enough and one blade is in a straight down position, all antenna dipoles are obstructed. Obviously, the radio impact of the RSWT in this case is the most significant. This corresponds to the worst case scenario. Fig. 5. Disposition of the RSWT (Fig. ) above a panel antenna of BTS to study near field coupling.

4 A. Scaled Prototype Manufacturing Scaled prototypes are manufactured with the respect of fundamental requirements, which have been described in [8]. The most important requirement is the proportion of scale ratios of dimension and frequency. k d k = 1 (3) where k d is the ratio of dimension between the scaled prototype and full model, while k f is the frequency ratio between them. The RCS of scaled prototype σ s at frequency f s in relation to that of full model σ 0 at frequency f 0 is described as f σ s f s k = σ (4) k d 0 f f0 Fig. 6. Radiation pattern of the panel antenna affected by the RSWT at 900 MHz in the horizontal plane (xy) and in the vertical front plane (xz). The radiation pattern of the panel antenna in the worst case scenario of the RSWT impact is shown in Fig.6. In straight up position (θ rot = 0 ), the impact of the RSWT is not present. However, at θ rot = 30, the radiation pattern of the antenna is degraded. This is confirmed at θ rot = 60, where all dipoles are obstructed by a blade (in straight down position). The antenna gain is reduced by 7.5 db. The radiation pattern in the horizontal plane is significantly deformed, while a back lobe occurs in the vertical front plane. Its level is 3.6 db compared to 8.3 db in the main lobe. Due to gain reduction, the efficiency of BTS cellular coverage can be significantly reduced. Likewise, the back lobe with higher level can degrade the performance of cells located on the back of the panel antenna. IV. EXPERIMENTAL VALIDATION Various methods are used to measure the RCS of large objects: either to measure a full model in real conditions, or to take measurements with a scaled model in an anechoic chamber. The first method is complicated to set up because it has to be able to deal with weather conditions and interfering signals. Additionally, many factors (e.g. ground, objects, etc.) can affect measurements. Meanwhile, a scale model in an anechoic chamber can be used easily multiple times and is highly cost-effective compared to the full-model method. Fig. 7. Instrumentation in the anechoic chamber. (Left) Two horn antennas as transmitter and receiver. Prototype rotated by positioners and drivers. Fig. 7 shows the scaled prototype of the 3.5 kw RSWT model in aluminium. It is placed on a plastic support. This model is ten times smaller than the full size one (k d = 1/10), so with k d = 1/10 and k f = 10, the RCS of the prototype is expected to be 0 db less than that of the full RSWT model. The blades of this prototype are easily replaceable, which makes it possible to study with different blade materials in future work. The tail is not considered here for RCS measurement because its impact is predicted to be much less significant than that of the blades and nacelle. B. Instrumentation and Calibration The RCS of the prototype is measured in an anechoic chamber (ENSTA Brest Facilities) measuring 8 m length 5 m width 4.5 m height. The basic instrumentation for RCS measurement consists of three subsystems controlled by a central computer. These are transmitter / receiver (Fig. 8a), positioners / drivers (Fig. 8b) and a data acquisition system (Anritsu Network Analyzer). Transmitter and receiver are dual- Fig. 8. Scaled prototype of the RSWT model in full metal (k d = 1/10).

5 Fig. 9 describes the RCS pattern of this lens measured in the anechoic chamber in Fig. 8 at 9 GHz. As expected, it is quasi stable around 7 10 dbm² between the aspect angles of 60 and +60. Using (5) and the measured RCS of the lens, we can calculate the RCS of the prototype. Fig. 9. RCS pattern of the Luneberg Lens Reflector XBR10 at 9 GHz, used as a reference target for direct calibration. polarization wideband circular horn antennas. They operate at wideband frequency ranges ( 17 GHz). To measure bistatic RCS, the two antennas can be separated to a maximum angle of 4. Monostatic RCS is measured with a 10 separation angle. It ensures low coupling between Tx/Rx antennas and it is quasi monostatic RCS. To measure RCS as a function of azimuth angle, the model is rotated around its vertical axis by positioners and drivers. The maximum rotation range is from - 95 to +95. This range was fully used for measurements in the horizontal plane with a 0.1 step. RCS measurement was done with four blade positions corresponding to a rotational period: 0 (or straight up) as in Fig. 8b, 15, 30 and 45 (repeating position of 15 ) to evaluate the symmetry of the model. All blade positions were measured with monostatic RCS at 9 and 17 GHz (maximum usable frequency during measurements). Based on (3), these two frequencies correspond to 900 MHz and 1700 MHz of the RSWT model with realistic dimensions. 900 MHz corresponds to the widely used GSM frequency band, while 1700 MHz was selected as the closest frequency to another GSM band (1800 MHz).. As discussed in Section II.A, RCS is determined from polarization VV (co-polarization) while VH is the cross polarization. Calibration is the next important step to have a reliable measurement. As shown in [10], there are two main methods for calibrations: direct and indirect calibrations. The direct calibration is used in this study. It involves the use of reference targets with well-known RCS values such as sphere, flat plate, dihedral and dielectric lens. In the case of monostatic RCS, where transmitting and receiving antennas are similar in terms of polarization, gain, loss, the RCS of target under measure σ mes is defined in relation to that of reference target σ ref as: σ mes = ησ (5) where η is received power ratio between reference target and target under measure. Luneberg Lens XBR10 (Lun tech) is selected as a reference target for direct calibration because its RCS value has been certified by the manufacturer (Lun tech). Additionally, the lens has a region with the almost constant RCS level. This makes the measured results become more accurate when using (5). ref Fig. 10. RCS simulated by MLFMM method vs. RCS measured in the anechoic chamber calibrated by Luneberg Lens. At 9 GHz. At 17 GHz C. Experimental vs. Numeric Results The measured RCS of the prototype after calibration is shown in Fig. 10. It is also compared with the simulated RCS. In comparison to the RCS level in Fig. 3a, that in Fig. 10a is 0 db smaller. This is due to the correspondence shown in (4). At 9 GHz, the main lobe and side lobes of two RCS patterns are very close. In the axis (0 ), the deviation of RCS level is only 0.6 db. Likewise, a good agreement in side lobe level is noted. Additionally, the main lobe of the measured RCS is a little bit wider than that of the simulated result (about 1.4 at 3dB HPBW). This difference can be explained by two reasons. First, although the support in Fig. 7 is made from plastic, it still has an impact on RCS. To verify, we have simulated the RSWT with and without the support made from used material (ε r =3). Consequently, the RCS comparison of two models (not presented here) shows that the RCS main lobe of the RSWT with the support is wider than that of the model without the support. Second, some unwanted EM coupling between the prototype and objects inside the chamber (e.g. plastic support, positioners, absorbent) which has not been taken into account by the direct calibration may be present. At 17 GHz, a good agreement is also noted in the two RCS patterns. However, the difference in the 3dB beamwidth of two RCS main lobes is still

6 present. In addition to co-polarization, we also consider the cross component of RCS (VH). Its level is 0 db lower than the co-polarization RCS at 9 GHz (approximately 30 db at 17 GHz). This confirms the definition of RCS as in Section II.A. V. CONCLUSION Because radio systems (radar, etc.) and wind turbines interfere, questions have been raised about possible EM disturbances from small wind turbines on the cellular BTSs they supply. This original subject is under study in this work. To better address the diversity of SWTs, a representative SWT model (RSWT) scalable with nominal power has been designed. In a first version (RSWT-u), its dimensions were too small compared with realistic models of SWTs. This was corrected by scaling its dimensions to match RCS results between the RSWT and 3.5 kw Nheowind-100 realistic models. The remaining differences in size between the two models are due to the very particular shapes of the blades in Nheowind-100 model. Indeed these differences are weak with more conventional shaped SWT models, like the Kingspan model for example. RCS results in horizontal and vertical planes have been presented and discussed for RSWT-u, optimized RSWT and Nheowind-100 models. Additionally, the Doppler shift and blade flash of RSWT have been described by a spectrogram at 900 MHz. Due to close proximity between the RSWT and a panel antenna, the near field coupling has been studied. In the worst case scenario, all antenna dipoles are obstructed by a blade in a straight down position. The radiation pattern of the antenna is significantly degraded. The maximum gain decreases 7.5 db compared with that not affected by the SWT. Notably, a significant regrowth of 15 db is noted in the back side lobe. It could create unwanted intra-cell interference in such cases. To validate the simulations, a full metallic scaled prototype of the RSWT has been manufactured, with a ratio of 1:10 compared to the original model. The RCS measurements for this prototype were done in an anechoic chamber at 9 GHz and 17 GHz (maximum usable frequency). Because of the ratios in dimensions and frequency, this corresponds respectively to radio cellular frequencies 900 MHz and 1700 MHz, close to 1800 MHz. The Luneburg Lens is used as a reference target to calibrate the measurements. Comparisons between measured and simulated RCS in the horizontal plane show quite good agreements and validate the simulations carried out. Future work will include the effect of SWT materials on back scattering. ACKNOWLEDGMENT This work is supported by the European project OPERA- Net (Celtic-plus). REFERENCES [1] D. L. Sengupta, T. B. A. Senior, Electromagnetic interference to television reception caused by horizontal axis windmills, IEEE Proceedings, vol. 67, no. 8, p , July [] Gavin J Poupart, Wind farms impact on radar aviation interests, Final Report FES W/14/00614/00/REP DTI PUB URN 03/194, Sept [3] J. Pinto, J. C. G. Matthews, C. Sarno, Radar signature reduction of wind turbines through the application of stealth technology, 3rd European Conference on Antennas and Propagation (EuCAP 009), p , Berlin (Germany), March 009. [4] L. Rashid, A. Brown, Partial treatment of wind turbine blades with radar absorbing materials (RAM) for RCS reduction, Proceedings of 4th European Conference on Antennas and Propagation (EuCAP 010), p. 1 5, Barcelone (Spain), Apr. 010 [5] P. Blacksmith, R. E. Hiatt, R. B. Mack, Introduction to radar crosssection measurements, Proceedings of the IEEE, vol. 53, no. 8, p , [6] T. Vu La, F. Le Pennec, C. Vaucher,"Small Wind Turbine Generic Model Design for BTS Radio Interaction Studies", 4th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications 013, p , London (UK), Sep [7] T. Vu La, F. Le Pennec, S. Elenga, Modelling the Scattering and Coupling of Small Wind Turbines Located Above BTS Antennas, 8th European Conference on Antennas and Propagation 014,The Hague (Nertheland), Apr. 014, in press. [8] Y. Zhang, A. Huston, R. D. Palmer, R. Albertson, F. Kong, S. Wang, Using Scaled Models for Wind Turbine EM Scattering Characterization: Techniques and Experiments, IEEE Transactions on Instrumentation and Measurement, vol. 60, no. 4, p , Apr [9] F. Le Pennec, M. Roques, S. Germaine, Dosimétrie des antennes relais des systèmes de troisième génération, RNRT ADONIS sous-projet ISIS, Rapport d étude Lot 3 / Tâche 3, Nov [10] Marcelo Miacci, Mirabel Rezende, Basics on Radar Cross Section Reduction Measurements of Simple and Complex Targets Using Microwave Absorbers, Applied Measurement Systems, Chapter 16, InTech., 01.