INTERFERENCE effects of wind turbines on communication

Transcription

1 TCOM-TPS R 1 A Measurement-based Multipath Channel Model or Signal Propagation in Presence o Wind Farms in the UHF Band Itziar Angulo, Member, IEEE, Jon Montalbán, Graduate Student Member, IEEE, Josune Cañizo, Graduate Student Member, IEEE, Yiyan Wu, Fellow Member, IEEE, David de la Vega, Member, IEEE, David Guerra, Member, IEEE, Pablo Angueira, Senior Member, IEEE, and Amaia Arrinda, Senior Member, IEEE Abstract Scattering signals on wind turbines may lead to degradation problems on the communication systems provided in the UHF band, such as terrestrial television broadcasting, broadband wireless systems or public saety services. To date, despite the continuous requests rom the International Telecommunication Union or studies on this ield, no channel model has been developed to characterize signal propagation under these particular conditions. In response to this necessity, this paper presents a complete Tapped Delay Line (TDL channel model to characterize multipath propagation in presence o a wind arm, including novel scattering modeling and Doppler spectra characterization. As proved later, this channel model, which is based on both theoretical development and empirical data obtained in the surroundings o a real wind arm, is adaptable to the particular eatures o any case under study: wind turbine dimensions, working requency, and relative location o the wind arm, transmitter and receivers. Index Terms Channel models, Doppler measurements, Multipath channel, TSR Channel Model, Wind arms. I. INTRODUCTION INTERFERENCE eects o wind turbines on communication systems, such as analogue or digital television broadcasting, have been largely reported in the literature. These degradation eects are due to the signal rom the transmitter impinging on the wind turbine structure and being scattered in all directions. Moreover, these scattering signals are dynamic in nature because o the rotation o the blades, leading to a time varying propagation channel even or ixed reception [1]-[4]. Up to now, there are no available channel models to characterize the particular propagation conditions encountered in presence o a wind arm in the UHF band, where terrestrial television broadcasting and other mobile communication services are provided. Regarding international regulations, the Manuscript received February, 13; revised April 18, June 18 and September, 13. This work was supported in part by the European Union FP7 (grant agreement n 96164, by the Spanish Ministry o Economy and Competitiveness (project TEC1-337, and by the Basque Government (SAIOTEK program. I. Angulo, J. Montalbán, J. Cañizo, D. de la Vega, D. Guerra, P. Angueira and A. Arrinda are with the Signal Processing and Radiocommunications research group (TSR o the University o the Basque Country (UPV/EHU, 4813 Bilbao, Spain (phone: ; ax: ; itziar.angulo@ehu.es. Y. Wu is with the Communications Research Centre, Ottawa, Canada ( yiyan.wu@crc.ca International Telecommunications Union proposes in Recommendation ITU-R BT.85 [5] and in Recommendation ITU-R BT.1893 [6] simpliied methods to estimate the potential degradation due to a wind turbine on analogue and digital television, respectively. However, these suggested methods have two main disadvantages: they account or only one wind turbine, without considering the contributions o all the turbines that compose a wind arm; and they do not model the time varying nature o the scattering signals. Thus, they cannot be considered as complete channel models. In act, the ITU maintains an open question where urther studies about these aspects are requested [7]. In response to this necessity, in 9 and 1 the University o the Basque Country UPV/EHU carried out a measurement campaign in the surroundings o a wind arm in order to a characterize signal scattering rom wind turbines and b evaluate the potential degradation o the DVB-T service due to the propagation conditions in presence o a wind arm. A description o the ield trials and the methodology to analyze the scattering signals rom wind turbines based on measured DVB-T data can be ound in [8], whereas the reception thresholds o the DVB-T service in presence o a wind arm are analyzed in [4]. However, the main contribution o the research eorts that have been made since then is the adaptable channel model presented in this paper. The two main issues that should be dealt with or the channel model under consideration are the scattering model to account or the relative power o the scattered signals, and the Doppler spectrum model to characterize signal variability due to blade rotation or the dierent working regimes o the wind turbine, i.e., dierent rotor orientations, rotation speed o the blades, etc. This work addresses both issues, providing signiicant results that are validated against empirical data rom the extensive measurement campaign. With respect to the irst, the classical scattering models used in the UHF band were both theoretically and empirically evaluated in [9], concluding that they do not provide realistic estimations o scattered signals rom modern wind turbines. Accordingly, the scattering pattern o a wind turbine was analyzed by means o physical optics-based simulations in [1], obtaining that the scattered signals o higher amplitude are due to the mast. The obtained conclusions justiy the necessity o proposing the novel scattering model o Section V.

2 TCOM-TPS R Regarding the time variability o the channel model, although several research studies about the spectral characteristics o the signals scattered by wind turbines are ound in the literature [11]-[1], these studies are ocused on higher requency bands and solely consider monostatic Doppler conditions. Hence, an analysis o the spectral characteristics o the signals scattered in bistatic conditions in the UHF band is needed. In this respect, a generic Doppler spectrum model that characterizes the speciic variability o the scattering signals rom wind turbines with rotating blades was developed by the authors in [13]. Nevertheless, this model needs to be adapted or the estimation o new working conditions, as explained in Section VI. This document is organized as ollows. Section II gives a brie description o the acquisition and processing o the empirical source data. Next, Section III introduces the basic concepts o the analyzed propagation channel and the Tapped Delay Line (TDL scheme. Section IV presents the proposal o an adaptable channel model to characterize signal propagation in presence o wind arms in the UHF band. This channel model requires the development o a new scattering model, described in Section V, and the characterization o the time variability by means o representative Doppler spectra, included in Section VI. Section VII describes the step-by-step implementation o the channel model, whereas Section VIII is ocused on the practical application o the novel channel model. Finally, the conclusions o the paper are gathered in Section IX. II. EMPIRICAL SOURCE DATA The ield trials in which the ollowing analysis is based were carried out in two stages during spring 9 and spring 1. Regular DVB-T (Digital Video Broadcasting Terrestrial transmissions in the UHF band were recorded in the surroundings o a wind arm installed close to two television transmitters [8]. Thanks to the pilot carriers included in the DVB-T signal, the Channel Impulse Response (CIR that characterizes the propagation channel can be estimated [14]. Due to the constant delays o the multipath components at each ixed reception point, the signals scattered by the dierent wind turbines can be identiied and obtained rom the CIR. Accordingly, the time variability o the scattered signals as blades rotate can be analyzed by considering successive CIRs [4],[8]. Measurements were taken in 6 dierent reception locations around the wind arm. In order to account or dierent wind conditions, signal recordings were carried out in these reception locations through various days. From each recorded DVB-T signal, scattering signals corresponding to the wind turbines installed closest to the transmitters were detected and stored. In total, the empirical data base is ormed by 38 signals scattered by the wind turbines as their blades rotate, each o these scattering signals lasting 1 seconds. More detailed descriptions o the ield trials and the methodology to obtain the components scattered by each wind turbine rom the CIR o the DVB-T signal can be ound in previous reerences rom the authors [4],[8]. III. BASIS OF THE PROPOSED CHANNEL MODEL A. Propagation Channel Characteristics The propagation channel to be modeled corresponds to static reception in rural or semi urban environments in the surroundings o a wind arm. In presence o wind turbines, the propagation channel shows dierent characteristics whether the receiver is located in the orward scattering zone or in the backscattering zone [4],[15]. It has been proved that the propagation channel in the backscattering zone might be more demanding in terms o carrier-to-noise ratio (CNR than the channel models proposed to test the DVB-T system. By contrast, the eects on quality degradation do not seem to be signiicant in the orward scattering area [4]. In the backscattering zone, the signals scattered on the wind turbines reach the receiver as attenuated, delayed and phase shited replicas o the direct signal rom the transmitter, i.e., a discrete multipath channel is encountered. For static reception in a certain location, the delays o the multipath components are constant. Nevertheless, the amplitude o each o these components depends on a set o ixed actors such as the relative position o the transmitter, the wind turbine and the receiver or the dimensions and materials o the wind turbine components; but also on a set o varying actors such as the orientation o the rotor with respect to the wind or its rotational speed. In act, the rotation o the blades causes periodic variations in the amplitude and phase o the multipath components [8]. Thereore, it is necessary to propose a novel channel model to characterize signal propagation in the backscattering zone o wind turbines. Indeed, the typical channel models used to represent time-varying channels normally account or the mobility o the receiver [16]-[17], and thus do not apply to the particular eatures o signal propagation in presence o a wind arm. It should be noted that although the term backscattering is sometimes used to reer to monostatic situations, the term backscattering zone as used in this paper reers to bistatic conditions, and covers about 8 percent o the region around the wind turbine [15],[18]. B. Tapped Delay-Line Model We are dealing with a multipath channel composed o a constant number o discrete resolvable components with constant delay and variable tap gains. The lowpass-equivalent impulse response o a discrete multipath channel composed o N components is given by [19] c(t, τ = N ã k (tδ(τ τ k (1 k=1 where, in our case, ã k (t represents the time-varying scattering signal rom k-th wind turbine measured in the receiver position, and τ k is the relative delay o this scattering signal, except or k=, which corresponds to the direct signal rom the transmitter. The time-varying multipath channels can be described by TDL channel models. In a TDL model, a set o discrete paths is deined, each o which has a certain delay and

3 TCOM-TPS R 3 attenuation level. Furthermore, each path has an associated Doppler spectrum to account or the channel variability due to the movement o the transmitter, the receiver, or the environment [19]-[1]. Usually, TDL channel models provide just a set o ixed paths with representative delays, amplitudes and Doppler spectra or a general characterization o a certain reception environment [16]-[17]. Based on this generic TDL scheme, the main advantage o the channel model proposed in this paper is that it can be adapted to the particular eatures o any case under study. IV. A CHANNEL MODEL FOR SIGNAL PROPAGATION IN PRESENCE OF WIND FARMS As previously mentioned, the proposed channel model is based on the TDL scheme, and it can be suited to a certain transmitter potentially aected by a wind arm and or each reception location within its coverage area according to the ollowing parameters. Number o paths: The number o multipath components is ound rom the number o wind turbines whose backscattering zones include the reception location, plus the direct path corresponding to the signal rom the transmitter. Relative delays o the paths: The relative delay o each multipath component is calculated as a unction o the distance dierence between the direct path (transmitterreceiver and the path o the scattered signal (transmitterwind turbine-receiver. Mean amplitude o the paths: The amplitude o each o the multipath components is the mean power o the scattering signal rom the corresponding wind turbine measured in the receiver. The signal scattering on the wind turbine can be modeled by means o a scattering pattern which relates the incident signal on the turbine to the signal scattered in all directions. Doppler Spectrum: The Doppler spectrum represents the power spectral distribution o the scattered signals, and determines the temporal variability o the multipath components due to the rotation o the blades. According to the previous approach, two main issues should be addressed: obtaining a scattering model that will truly characterize signal scattering rom modern wind turbines, and inding a bistatic Doppler spectrum model to account or the movement o the blades and the nacelle or the dierent working regimes o the wind turbine. V. SCATTERING MODEL In the literature, several scattering models to account or the eect o a wind turbine on a signal transmitted in the UHF band are ound: the scattering models included in the Rec. ITU-R BT.85 [5] and in the Rec. ITU-R BT.1893 [6], and the scattering models proposed by Van Kats [], Sengupta [18], and the British Broadcasting Corporation (BBC []. However, these scattering models suer rom several theoretical limitations. Furthermore, the empirical evaluation o these scattering models proves that they ail to provide accurate estimations o the signal scattered by modern wind turbines [9]. Moreover, in contrast to the assumptions o the above mentioned theoretical scattering models, accurate simulations o the scattering pattern o an actual wind turbine, based on physical optics theory, show that the metallic mast contribution is signiicantly higher than the contributions o its other components [1]. For this reason, the scattering model proposed in this paper is based on the signal scattered by the mast. This does not only relect better the actual design and composition o modern wind turbines, but also allows considering the eect o the rotating blades separately, by means o the characterization o the Doppler spectrum o the scattering signals (as shown in Section VI. A. Theoretical Basis o the Proposed Scattering Model The signal scattered by the mast is represented by the Radar Cross Section (RCS o a circular cylinder. The RCS is the projected area required to intercept and isotropically radiate the same power as a scatterer (target scatters toward the receiver, and thus it is normally expressed in db with respect to a square meter (dbsm [3]. The ormal deinition o radar cross section states that the distance between the radar and the object must become ininite in order to eliminate any distance dependence in the RCS characteristics, i.e., ar ield condition in the context o signal scattering must be ulilled [3]. The ar ield condition requires that the object is illuminated by a plane wave. The ar ield distance R is normally expressed as a unction o the lateral dimension o the object D and the wavelength λ [9], according to ( R = D λ For the case under study, where transmission requency is 8 MHz and the mast length is 55 m, the ar ield distance is approximately 16 km. However, the cases o impact in the UHF band occur or quite shorter distances between transmitters and wind turbines [4]. Thereore, near ield eects must be taken into account and included in the scattering model. 1 Bistatic RCS o a Circular Cylinder: The proposed model is based on the physical optics approximation. Considering the assumptions o this high requency method, in [4] the RCS pattern o an elliptic cylinder is obtained as a unction o its dimensions and the angular positions o the transmitter and receiver in both the vertical and the horizontal planes (θ t, φ t, and θ r, φ r respectively. However, this expression is o indeterminate orm or some combinations o transmitter and receiver angular positions. The spherical coordinate system used to speciy incident and observation (scattering directions is shown in Fig. 1. The expression proposed in [4] was adapted to a circular cylinder and simpliied to avoid indeterminate orms, as included in Appendix A. The resulting expression or the calculation o the bistatic RCS o a circular cylinder is given by (3, (

4 TCOM-TPS R 4 r z TABLE I TSR SCATTERING MODEL T Near ield condition R < L, being R the transmitter to wind turbine λ distance and L the cylinder length θ t 1 + cos σ(φ r, φ t n = krl φr n sin θ t x φ r φ t θ r R y being k = π/λ, λ the transmission wavelength, r the cylinder radius φ r the receiver angular position in the horizontal plane, θ t the transmitter λr angular position in the vertical plane and L n = Application limits Fig. 1. Spherical coordinate system or the cylinder 7 < θ t < 1 1 < φ r < 1 16 θ t < θ r < θ t 1 + cos σ(φ r, θ r, θ t =krl φr sin θ t ( sin θ t + sin θ r kl(cos sinc θt + cos θ r where k = π/λ, λ is the wavelength, r is the cylinder radius and L is the cylinder length. According to the physical optics theory, the accuracy o RCS estimations err by wider margins as the direction o observation moves arther away rom the specular direction [3]. In order to deine the application limits o (3, the characteristics o the shapes corresponding to the horizontal and vertical planes o the cylinder have been considered. For the horizontal plane, the limit established or a sphere is used, in such a way that 1 < φ r < 1 [5]. For the vertical plane, a lat plate can be taken as a reerence, or which the physical optics theory perorms quite well in predicting the returns in the region at within or 3 degrees to either side o normal incidence, i.e., 7 < θ t < 11. Considering a similar margin or the bistatic case, the receiver angular position should be to either side o the specular direction (θ r = 18 θ t, in such a way that 16 θ t < θ r < θ t. The graphic representation o the model proposed in (3 is given in Fig. or the dimensions o the wind turbine under study [4],[8], where application limits are depicted in red. It can be observed that the RCS pattern in the horizontal plane is almost constant, whereas the scattering pattern in the vertical plane has a directive narrow lobe or the specular direction [1]. Near Field Eects: Near ield eects in the context o signal scattering and or monostatic reception result in a RCS reduction [6] which can be equivalent to consider a near ield size instead o the real size o the object [7]. This near ield size L n is calculated according to (4, where R is the transmitter to the scattering object distance and λ is the wavelength, L n = λr (3 (4 Nonetheless, under near ield condition, some additional eects on the scattering pattern are observed: not only the above mentioned reduction in the amplitude o the main lobe o the object s pattern, but also the periodic nulls o the sinc squared pattern are smoothed away with decreasing distances, and the amplitudes o the sidelobes o the pattern increase [15],[8],[9]. Taking into account these eects, the scattering pattern o the cylinder in the vertical plane represented in Fig. will lose the directivity typical o the sinc unction. That is to say, in near ield conditions, the RCS value corresponding to the specular direction can be used or wider angular positions in the vertical plane [15],[8],[9]. Thereore, the proposal or applying near ield eects to our novel scattering model is to estimate the RCS value corresponding to the near ield size L n in the specular direction in the vertical plane (θ r = 18 θ t, and use it within the application limits established by the physical optics theory in this plane. B. Proposed Scattering Model The novel scattering model or a wind turbine that is proposed or its implementation in the proposed channel model is based on the scattering rom the mast under near ield condition. An accurate and easy-to-implement expression or the assessment o the mean values o the multipath components due to the scattering signals rom wind turbines has been obtained. Table I shows the expressions o the proposed model called TSR, which stands or Signal Processing and Radiocommunications Research Group in Spanish to characterize signal scattering by a wind turbine in the UHF band. C. Validation o the Proposed TSR Scattering Model In order to check the validity o the proposed model, it is compared to the empirical data obtained rom the measurement campaign. A similar study was carried out and reported in [9] or the pre-existing scattering models prior to the proposal o this new TSR scattering model.

5 which the physical optics theory perorms quite well in predicting the returns in the region at within or 3 degrees to either side o normal incidence, i.e., 7 < θ t < 11. TCOM-TPS R n 5 Considering a similar margin or the bistatic case, the receiver object distance and λ is the wavelength, R L (4 θ r ϕ r Fig.. Scattering pattern (RCS, dbsm o the proposed expression or r=.8 m, L=55 m and λ=.38 m. Incident direction is represented Fig.. Scattering by a red pattern arrow. (let (RCS, vertical dbsm plane: o thercs proposed as a unction expression o θ r or or r=.8 ϕ r = m, (right L=55horizontal m and λ=.38 plane: m. RCS Incident as a unction direction o is ϕ r represented or θ t = θ r = by 9º a red arrow. (let vertical plane: RCS as a unction o θ r or φ r = (right horizontal plane: RCS as a unction o φ r or θ r = θ t = 9 TABLE II STATISTICAL CHARACTERIZATION OF THE PREDICTION ERRORS OF THE SCATTERING MODELS Mean Std. dev. 5 th / 95 th %±6 db error percentiles Rec. ITU-R BT.85 [5] 6.4 db 6.1 db -3.6 / 17. db 45% BBC [5] 14.5 db 6.1 db 5.9 / 4.7 db 6% Van Kats [] 3. db 6.1 db -5.5 / 13.3 db 63% Sengupta [15] -5. db 6.1 db / 5. db 49% TSR 1.4 db 5.8 db -6.5 / 11.4 db 7% The power o the signal scattered by a wind turbine is expressed as a carrier to intererence ratio (C/I: the ratio in db between the direct signal rom the transmitter (the desired signal, reerred to as carrier C and the signal received ater being scattered by the wind turbine (the interering signal I [9]. The results provided by all the scattering models are compared to the empirical values registered in the measurement campaign [8],[9]. Finally, a statistical analysis o the dierences between the C/I pred predicted by the theoretical models and the measured C/I meas obtained by the estimation o the CIRs rom the DVB-T signal is carried out. The proposed TSR model has been developed in order to be independent o the variable conditions o blades rotation and dierent rotor orientations; all these variable conditions have been characterized by the Doppler Power Spectral Density o the scattered signal, analyzed in Section VI. Thereore, the validity o the scattering model will be assessed by evaluating the prediction error or each reception location, and then analyzing the global prediction error. For the statistical characterization o the prediction errors, the mean value, the standard deviation, the 5 th and 95 th percentiles and the percentage o errors between ±6 db are calculated. This statistical characterization o the prediction errors is shown in Table II. From the results, it can be concluded that the TSR scattering model provides more accurate results than the pre-existing scattering models, with a mean error o approximately 1 db. This is also noticeable when the ±6 db criterion is considered: 7% o the predicted values show absolute errors lower than 6 db with respect to the empirical data. This result is almost 1% better than the next more accurate model proposed by Van Kats. It should also be noted that the standard deviation values are similar or all the analyzed models. This is probably due to the act that the variability o the prediction errors is related to the inherent variability o the measurements and not to the scattering models [9]. Apart rom providing more accurate results when compared to empirical data, it should also be highlighted that the TSR scattering model makes coherent assumptions with respect to the geometry and composition o the masts. The classical scattering models, by contrast, assume simpliied geometries or the blades, representative materials and blades positions that do not relect real conditions. Furthermore, in contrast to the computationally complex RCS estimation algorithms, the proposed TSR scattering model consists o a simple expression, which is suitable or its implementation in planning tools. VI. DOPPLER SPECTRUM MODEL The time variability o the channel is normally represented by Doppler spectrum models. In a previous reerence rom the authors, a generic Doppler spectrum model that characterizes the speciic variability o the scattering signals rom wind turbines with rotating blades was proposed and developed [13]. In this section, this novel spectrum model is briely described, and the adaptation or its use in the proposed channel model is explained. A. Characterization o Doppler Spectra o Scattering Signals rom Wind Turbines with Rotating Blades The novel spectrum model proposed in [13] is based on the empirical data obtained rom the measurement campaign in the

6 TCOM-TPS R 6 surroundings o a wind arm, and applies to dierent working regimes and rotor orientations [8]. The estimated Doppler Power Spectral Densities (PSDs correspond to near ield condition with respect to transmitter-wind turbine distances, which dier rom the spectral characteristics o the scattering signals in the ar ield [3]. However, as previously commented, it is precisely when near ield conditions apply that more severe degradations on the telecommunication services may occur, and these conditions are o particular interest to the proposed channel model [4]. The common eature o all the estimated PSDs is the main component at Hz, and lower power spectral densities or higher requencies. The presence o this main component at Hz, which is used or the normalization o the spectra, is due to the static part o the wind turbine, which validates our proposal o basing the scattering model on the mast [1]. The decreasing power levels or the higher requencies are related to the movement o the blades, and are not necessarily symmetric with respect to Hz due to their complex aerodynamic design [13]. In order to account or these special variability eatures o the scattering signals due to the rotating blades, a new exponential model was proposed in [13]. This model is composed o a Dirac delta or the zero Doppler requency, and side components o decreasing power spectral density or the lowest and highest requencies, according to (5 a exp(b c, min < S( = δ(, = (5 d exp( e g, < max where S( is the PSD expressed in db/hz, a, b, c, d, e and g are positive constants, stands or requency (Hz, δ( = db/hz and min and max are the minimum and maximum observable Doppler shits respectively. b and e represent the exponential decay (wider spectral characteristics or lower values and vice versa, c and g are related to the asymptotic values or ininite negative and positive requencies, respectively, and a and d account or the relative value o the curve or the y-axis with respect to the asymptotic values or ininite requencies given by c and g. The goodness o it o this Doppler spectrum model has been validated by means o the estimation and characterization o an empirical data set o more than 3 complex scattered signals [13]. B. Minimum and Maximum Observable Doppler Shit due to Blade Rotation For the model given by (5, the minimum and maximum observable Doppler shits min and max were empirically obtained or each estimated PSD according to their spectral shape [13]. It should be taken into account that the theoretical bistatic Doppler shit B when the transmitter and receiver are stationary depends on the bistatic angular separation (transmitter-turbine-receiver, the rotor orientation with respect to this bistatic angular separation, the rotational velocity o the blades, the blades length, and the transmission wavelength. That is to say, the maximum Doppler shit varies rom place to place or the same wind conditions, and also varies in a certain reception location or changing conditions o rotor orientation and/or rotational speed. Accordingly, a detailed calculation o the bistatic Doppler shit and consequently o the shape o the Doppler spectrum under the particular conditions o new situations and locations is not easible. However, or a given bistatic angular separation transmitterwind turbine-receiver φ r, the maximum bistatic Doppler shit B max can be calculated as a unction o the maximum rotation rate o the wind turbine ω max and the blade length l, as given by (6 B max = ω maxl λ cos(φ r / (6 Normalizing the estimated Doppler PSDs as a unction o their corresponding maximum bistatic Doppler requency shit B max allows the estimation o new situations o time variability due to dierent working requencies, wind turbine dimensions and relative locations transmitter-wind turbinereceiver. That is to say, in order to extend the applicability o the results, the maximum bistatic Doppler requency shit B max corresponding to the speciic characteristics o a new case under study should be calculated according to (6, and the empirically estimated Doppler spectra adapted accordingly. C. Doppler Spectra in the Proposed Channel Model As previously commented, the Doppler spectrum o the signal scattered by a wind turbine depends on the relative location transmitter-wind turbine-receiver, the rotor orientation with respect to transmitter and receiver, and the blades rotational speed. Considering that both the rotor orientation and the blades rotational speed are dependent on the wind conditions, dierent levels o channel variability will be aced or a certain reception location in case o changing weather conditions. For this reason, the aim o this study is to provide the user o the channel model with a set o representative Doppler spectrum examples corresponding to the dierent levels o variability that will be aced in a ixed reception location. The criterion used or the selection o representative cases is related to the impact o the time variability characterized by the dierent Doppler spectra on the potential degradation o the services provided in this requency band. The potential degree o aection o the scattering signals on communication systems is directly connected to their level o variability. This degree o variability o the scattering signal is related to the spectral width o its Doppler spectrum, given by the dierence between the minimum and the maximum observable requency shits min and max, and to the power spectral density values or the end requencies [13]. Bearing this in mind, three representative PSDs were selected according to their spectral width and the relative power level or the end requencies, and thus, to their potential degree o inluence on the transmitted signals: low, medium or high. Taking into account that, or a given reception location, the most critical case would be encountered when the maximum observable Doppler requency shit is B max, the selected PSDs represent dierent levels o variability corresponding to approximately 9%, 65% and 3% o the maximum Doppler

7 TCOM-TPS TCOM-TPS TCOM-TPS R 7 location. TABLE III location. DOPPLER PSDS SELECTED FOR THE CHANNEL MODEL The criterion used or the selection o representative cases is related The criterion to the impact used or o the the selection time variability o representative characterized cases by is -1-1 High the related variability dierent to the Doppler impact o spectra the time on the variability potential characterized degradation by o - the dierent services provided Doppler spectra in this on requency the potential band. degradation The potential o - degree the services o aection provided o the in 19.7 exp(4.5 scattering this requency 38., signals.9 band. on B communication max The potential < -3 systems degree o is aection directly connected o the B max scattering to their signals level o on variability. communication -3 S high ( = δ(, = This degree systems o is variability directly connected o the scattering to their level signal o is variability. related to This -4 the -4 spectral degree o 1.4 width variability exp( 4.8 o its Doppler o the scattering 38.1, spectrum, signal < given by is.9 the related B max B max dierence to the between spectral width the minimum o its Doppler and the spectrum, maximum given observable by the dierence requency Normalized -.5 requency (/.5 shits min and max and to the power spectral density values or Bmax 1 Medium between variability the minimum and the maximum observable requency Normalized requency (/ the end requencies shits min and max, and [13]. to the power spectral density values or Fig.3. Estimated PSD and exponential itting corresponding Bmax to high variability Fig. conditions. R =.89. to variabil- the Bearing end requencies this [13]. Fig.3. Estimated PSD and exponential itting corresponding to high variability. exp(6.1mind, three 3.4, representative.7 B max PSDs < were ity R =.89. selected Bearing conditions. R according this in to mind, B max =.89. their three spectral representative width and PSDs the relative were S med ( = δ(, = power selected level according or the end to requencies, their spectral and width thus, and to their the potential relative degree power level 5.1 o inluence or exp( 8.7 the end on the requencies, 9.5, transmitted and signals: thus, < to.6 low, their B medium potential max B max or -1 high. degree Taking o inluence into account on the that, transmitted or given signals: reception low, medium location, or -1 Low the high. variability most Taking critical into account case would that, or be a given encountered reception when location, the - maximum the most observable critical case Doppler would requency be encountered shit is when B_max the selected maximum PSDs observable represent Doppler -.9 exp(17.9 dierent 4.9, requency levels.3 shit B max o is variability B_max <, the -3 corresponding selected PSDs to represent B max approximately dierent 9%, levels 65% and o 3% variability S low ( = δ(, = o the -3 maximum corresponding Doppler to approximately requency shit. 9%, Fig. 65% 3, Fig. and 3% and o Fig. the show maximum 3. the Doppler exp( 17.6 estimated requency PSDs shit. 5., and Fig. the 3, < exponential Fig..3 4 B and max ittings -4 B max Fig corresponding show the estimated to high, PSDs medium and the and exponential low variability, ittings -4-1 Normalized -.5 requency (/.5 Bmax 1 respectively. corresponding The to normalized high, medium exponential and low ittings variability, o the Normalized requency (/ Fig.4. Estimated PSD and exponential itting Bmax corresponding to medium selected respectively. PSDs The (expressed normalized in db/hz exponential are shown ittings in Table o III. the Fig.4. variability Estimated conditions. PSD R =.93. and exponential itting corresponding to medium selected PSDs (expressed in db/hz are shown in Table III. Fig. variability 4. Estimated conditions. PSD R and =.93. exponential itting corresponding to medium parameters which are obtained rom the input data o Table IV variability conditions. R =.93. VII. IMPLEMENTATION OF THE TSR CHANNEL MODEL are included VII. IMPLEMENTATION in Table V. OF THE TSR CHANNEL MODEL According In this section, the practical implementation o the TSR -5 to these data, the parameters o the channel channel In this model section, or the certain practical case implementation under study is o described. the TSR -5 model are adapted as ollowing described: -1 More channel precisely, model or the a methodology certain case to under obtain study the parameters is described. o -1 the More channel precisely, model the or methodology certain reception to obtain location the parameters is detailed. o -15 This the channel adaptation model o or the a certain channel reception model location to the is particular detailed. -15 A. Number o Paths - characteristics This adaptation o o case the under channel study model requires to some the input particular data, - Regarding -5 which characteristics the gathered o number a case o in Table under paths, IV. study on Accordingly, requires a irst approach, some the input the necessary data, total parameters which number is gathered o wind which are in turbines obtained Table IV. o the rom Accordingly, wind arm the input data the should o necessary be -5 Table IV -3 considered are parameters However, included which in Table are it V. obtained should berom checked the input thatdata the reception o Table IV -3 Normalized requency (/ Bmax location are According included is the in to Table backscattering these V. zone o each wind turbine, data, the parameters o the channel Normalized requency (/ Fig.5. Estimated PSD and exponential itting corresponding Bmax dismissing to low variability model According the wind are adapted to these turbines as ollowing data, out the o described: parameters the angular o application the channel conditions. R =.96. limits model o are Table adapted I. Fig.5. Estimated PSD and exponential itting corresponding to low variability as ollowing described: Fig. conditions. 5. Estimated R =.96. PSD and exponential itting corresponding to low variability A. Number o paths conditions. R =.96. TABLE III A. Number o paths DOPPLER PSDS SELECTED TABLE FOR III THE CHANNEL MODEL Regarding the number o paths, on irst approach, the total B. Relative number Regarding Delays o wind the number o the Paths DOPPLER PSDS SELECTED FOR THE CHANNEL MODEL turbines o paths, o the on a wind irst approach, arm should the total be High considered. number o However, wind turbines should o be the checked wind that arm the should reception be requency High shit. Figs. 3-5 show the estimated PSDs and the The signal rom the transmitter is taken as the reerence exponential 19.7 ittings 4.5 corresponding / B_max to 38.high, medium.9 B_max andlow orlocation considered. calculating is in However, the the relative backscattering it should delays be ozone checked eacho multipath each that wind the component. Thus, or each wind turbine, the relative delay o the reception turbine, Shigh ( 19.7 exp 4.5 / B_max B_max variability, dismissing location is the in the wind backscattering turbines out zone o the o angular each wind application turbine, respectively. The normalized exponential ittings o S high 1.4 exp ( 4.8 / B_max the selected B_max limits dismissing o Table the I. wind turbines out o the angular application PSDs (expressed in db/hz are shown in Table III. scattered signal is calculated as a unction o the dierence in 1.4 exp4.8 / limits o Table B_max B_max distance between I. the direct path (transmitter-receiver and the Medium B. Relative Delays o the Paths Medium VII. IMPLEMENTATION OF THE TSR path B. o Relative the scattered Delays signal o the (transmitter-wind Paths turbine-receiver CHANNEL MODEL. exp(6.1 / B_max 3.4 B_max according The signal to (7 rom the transmitter is taken as the reerence or SIn this medium section,. exp(6.1 ( / the practical B_max implementation 3.4.7o B_max the TSR calculating The signal the rom relative the transmitter delays o is each taken multipath as the reerence component. or channel S 5.1exp( 8.7 / B_max medium model or a certain ( case under study is described. B_max Thus, calculating or each the wind relative turbine, delays the o relative each multipath delay o the component. scattered More precisely, 5.1exp( 8.7 / B_max B _max the methodology to obtain the parameters signal Thus, or τ is calculated i each = R T x W T i + R W Ti Rx R T x Rx wind as turbine, unction the relative o the dierence delay o the in scattered (7 c distance o Lowthe channel model or a certain reception location is between signal is the calculated direct path as a (transmitter-receiver unction o the dierence and the in distance Low path o detailed. This adaptation o the channel model to the particular where.9 exp(17.9 / B_max B_max between τ the i is scattered the thedirect relative signal path delay (transmitter-receiver o the signal scattered (transmitter-wind turbine-receiver and the bypath i-th o characteristics Slow.9o exp(17.9 a case ( under / study B_max 4.9 requiressome.3 input B_max data, wind the turbine, according scattered R stands to (7 signal or (transmitter-wind distances (see Tableturbine-receiver V and c is which S low isgathered 3. exp( 17.6 / B_max 5..3 in ( Table IV. Accordingly, the necessary theaccording speed o to light. (7 B_max 3. exp( 17.6 / B_max 5..3 B _max Normalized Power/requency (db/hz Normalized Power/requency (db/hz Normalized Power/requency (db/hz

8 TCOM-TPS R 8 For each wind turbine Transmitter Receiver TABLE IV INPUT DATA TO ADAPT THE CHANNEL MODEL TO THE SPECIFIC FEATURES OF A CASE UNDER STUDY Type Position Mast dimensions Blades length Maximum rotation rate, ω max Position Radiating pattern Antenna height Frequency, Power, P t Position Radiating pattern Antenna height Symbol Description UTM (m coordinates, including terrain height Vertical dimension o the mast (m Lower and upper diameters o the mast (m Longitudinal dimension o the blades (m Maximum rotation rate o the blades (rpm UTM (m coordinates, including terrain height Radiating pattern o the radiating system o the transmitter Radiating system height within the telecommunication tower where it is allocated (m Working requency within the UHF band (Hz Maximum transmitter power (W UTM (m coordinates, including terrain height Radiating pattern o the radiating system o the receiver Radiating system height (m TABLE V DATA CALCULATED FROM THE INPUT DATA OF TABLE IV R T x W Ti R W Ti Rx R T x Rx G T x W Ti G Rx W Ti G T x Rx G Rx T x r L φ r θ t θ r Description For each wind turbine, transmitter to wind turbine distance (m For each wind turbine, wind turbine to receiver distance (m Transmitter to receiver distance (m Radiation pattern gain o the transmitter toward i-th wind turbine Radiation pattern gain o the receiver toward i-th wind turbine Transmitter to receiver gain o the transmission radiation pattern Maximum gain o the receiver radiation pattern Mean radius o the mast Length o the slanted surace o the mast, which is a truncated right circular cone Bistatic angle in the horizontal plane (transmitter-wind turbine-receiver, or each wind turbine Angular position o the transmitter in the vertical plane, with respect to each wind turbine Angular position o the receiver in the vertical plane, with respect to each wind turbine C. Mean Amplitude o the Paths The mean amplitude o the path corresponding to the direct signal rom the transmitter is also taken as the power reerence ( db. Hence, the mean amplitude o the remaining multipath components is given by the ratio, measured in the reception location and expressed in db, between the power o the signal scattered rom the corresponding wind turbine P T x W Ti Rx and the power o the direct signal rom the transmitter P T x Rx. The direct power rom the transmitter in the reception location, P T x Rx, is calculated as a unction o the transmitter to receiver distance R T x Rx, the transmitter to receiver gain o the transmission radiation pattern G T x Rx, the gain o the receiver radiation pattern G Rx T x and the wavelength λ, including the corresponding additional propagation losses L prop (such as diraction losses due to terrain eatures, as shown in (8 P T x Rx = P tg T x Rx G Rx T x λ L prop (4π R T x Rx Then, or each wind turbine, the power o the scattered signal in the receiver location is calculated using the bistatic radar equation [15],[3],[8], according to (9 P T x W Ti Rx = P tg T x W Ti G Rx W Ti λ σ i (4π 3 R T x W T i R W T i Rx where P t is the transmitted power, G stands or the dierent radiation pattern gains, R stands or distances (see Table V, and σ i is the RCS o the mast in the receiver direction [1]. To calculate the RCS o the mast in the receiver direction, the incidence angle in the vertical plane (θ t as well as the reception angles in the horizontal and vertical planes (φ r, θ r are calculated as a unction o the position o the transmitter, the wind turbine and the receiver. The application limits o these angular positions and the near ield condition should be checked (see Table I. For the transmission requency and the corresponding distances and angular positions o the dierent elements, the bistatic radar cross section o that wind turbine or that reception location is obtained, as indicated in Table I. Finally, the mean amplitude o each path P i is given by the ratio o both powers: P i = P T x W Ti Rx/P T x Rx. Ratios lower than -45 db can be neglected, as well as scattered power levels below the noise threshold. D. Doppler Spectrum For the characterization o the Doppler spectrum, representative PSDs corresponding to dierent levels o variability have been selected, as described in Section VI. These examples characterize increasing levels o variability due to dierent rotation rates and orientations o the wind turbine with respect to the transmitter and the receiver. These Doppler spectra need to be suited to each reception location o new cases under study by calculating the corresponding maximum bistatic Doppler requency B max, which will depend on the relative position o the transmitter, the wind turbine and the receiver, the transmission wavelength, the maximum rotation rate, and the blade length (see (6 and Table V. This way, the Doppler spectra o Table III are to be adapted to the particular conditions o the new case under study by means o the speciic value o B max. To account or the dierent wind conditions that will probably be aced or a certain reception location, it is recommended that the three PSDs provided in Table III are considered in the (8 (9

9 TCOM-TPS R 9 channel model. In this way, the user o the channel model can obtain an overview o the dierent situations that may be encountered without the need or accurate estimations o wind directions or wind speeds. VIII. PRACTICAL APPLICATION OF THE TSR CHANNEL MODEL As previously mentioned, the parameters o the TSR channel model have to be adapted to each reception location o the coverage area. To do so, a digital terrain database can be used to divide the coverage area o a potentially aected transmitter into small grids o a given accuracy. For each o the center locations o these grids, the parameters o the channel model or those speciic conditions would be obtained, as explained in the previous section. This process is easily implementable in planning tools, and provides a ast overview o the potential degradation due to the wind arm. Once the channel model has been adapted, the most complete way to estimate the impact on a certain service is to develop some simulations o the eect o the resulting time-varying channel model on the corresponding reception threshold. This implies obtaining a realization o the channel model, i.e., obtaining the successive channel impulse responses that characterize the signal propagation in presence o a wind arm. Getting back to Eq. (1 in Section III.B., the tap-gain processes ã k (t are obtained generating a set o white Gaussian processes, whose power spectral densities are shaped by a shaping ilter whose amplitude transer unction is H( = S(, where S( is the Doppler power spectrum [19],[31],[3]. The resulting ilter must have a normalized power o 1, so that the individual path gains have to be properly scaled to account or the dierent powers o the taps. Frequency-domain simulators use the Fast Fourier Transorm (FFT to perorm convolution or the iltering operation o the generated complex Gaussian process [33]. A. DVB-T Case One o the most important services provided in the UHF band is television broadcasting. Digital TV service coverage is characterized by a very rapid transition rom near perect reception to no reception at all, and it thus becomes much more critical to be able to deine which areas are going to be covered and which are not [14]. For the speciic case o DVB-T, the potential increment in the CNR threshold or Quasi Error Free (QEF reception with respect to the typical Ricean channel used or the planning o ixed services [14] can be estimated as proposed in [4]. To do so, based on the estimated complex tap-gain processes obtained as mentioned above, the multipath energy and the mean standard deviation can be calculated as [4]: P mult = N P i (1 i=1 where i=1 and i = N are the indices o the irst and last paths and P i is the normalized mean amplitude rom path i (in linear units, and N std i i=1 std mean = (11 N where i=1 and i = N are the indices o the irst and last paths and std i is the standard deviation o the time varying complex tap gain rom path i (in linear units Based on the obtained parameters, the maximum increment o the CNR thresholds over the theoretical Ricean CNR threshold can be predicted according to the results presented in [4]. Depending on the calculated multipath energy and mean standard deviation values, these increments can range rom being negligible to being as high as 9 db [4]. Thereore, this channel model provides a valuable tool to estimate the potential degradation on the DVB-T service, a potential degradation that is diicult to evaluate a priori due to the multiple actors that should be taken into account. However, it should be remarked that although this study has been supported by measured DVB-T signals and this application case is also based on DVB-T, the proposed channel model is independent o the standard, and as such it is applicable to any service provided in the UHF band: terrestrial television broadcasting, broadband wireless systems or public saety services. B. Common Application Scenarios The application o this channel model is considered or two possible scenarios: planning o a new service on an area where a wind arm exists, or evaluation o the potential inluence o a wind arm on a pre-existing service during the wind arm design process. For the planning o a new service, according to the obtained results and depending on the quality target o that speciic service, the required transmission power should be estimated. It should be also noted that increasing the transmitted power will not always mean to avoid the eect o the wind arm. Instead, it is necessary to diminish the power that reaches the wind turbines and increase the power transmitted towards the potential users o the service. In this respect, appropriate modiications in the radiation pattern o the transmitter may be more eective than an increase o the transmitted power. In case o evaluation o the possible impact o a wind arm beore its installation, i the quality is ound to be potentially degraded, some preventive measurements can be taken, such as the relocation o some wind turbines, changes in the radiation pattern o the transmitter, etc. IX. CONCLUSION This paper proposes a channel model to characterize multipath propagation due to the presence o a wind arm in the UHF band, a study requested by the ITU since the irst cases o impact [7]. To date, no channel model to describe signal propagation under these particular circumstances had been deined. The most important eature o the TSR channel model is that it is adaptable to the particular eatures o any case under study, as the number o paths and their delays are calculated or