A noise generation and propagation model for large wind farms

Transcription

1 Wind Farm Noise: Paper ICA A noise generation and propagation model for large wind farms Franck Bertagnolio (a) (a) DTU Wind Energy, Denmark, frba@dtu.dk Abstract A wind turbine noise calculation model is combined with a ray tracing method in order to estimate wind farm noise in its surrounding assuming an arbitrary topography. The wind turbine noise model is used to generate noise spectra for which each turbine is approximated as a point source. However, the detailed three-dimensional directivity features are taken into account for the further calculation of noise propagation over the surrounding terrain. An arbitrary number of turbines constituting a wind farm can be spatially distributed. The noise from each individual turbine is propagated into the far-field using the ray tracing method. These results are added up assuming the noise from each turbine is uncorrelated. The methodology permits to estimate a wind farm noise map over the surrounding terrain in a reasonable amount of computational time on a personal computer. Keywords: Wind Turbine Noise, Wind Farm, Directivity, Propagation, Ray Tracing Method

2 A noise generation and propagation model for large wind farms 1 Introduction Wind farm noise is an important obstacle to the implementation of wind energy on-shore near residential areas. Placing turbines in these locations is surely unfavorable in term of noise annoyance, but significant cost reductions can be obtained, for example when compared to offshore wind farms because of their additional construction and operational costs. Furthermore, proximity to the main electrical grid also contributes to reduce installation costs. In this work we present a new simulation code based on a detailed definition of the noise sources, i.e. each individual turbine in a farm. The noise is propagated to the far-field at a receiver location using a classical ray tracing method linking each turbine to the receiver. The ease of implementation of this method and its low computational cost allow to deal with a large number of turbines. This article is organized as follows. The two components of the proposed wind farm noise model, namely the wind turbine noise generation model and the propagation model are first described. Then, the combined model is used to calculate the noise produced by a fictitious (but realistic) wind farm layout on a fictitious terrain. The results are presented in the form of noise maps highlighting a number of capabilities of this tool. Conclusions and future work are finally discussed. 2 A wind farm noise model 2.1 Single wind turbine noise model The wind turbine noise model used in the present work has been presented and validated in a previous work [1]. The simulation code is based on the in-house aeroelastic code HAWC2 [2, 3]. It is a general purpose simulation code for simulating the aeroelastic behaviour of wind turbines, including exhaustive dynamic and atmospheric flow conditions such as yaw, wind shear/veer, atmospheric turbulence, gusts, wakes from upstream turbines, etc. The aerodynamics of the individual blade sections and of the rotor as a whole is modelled using the classical Blade Element Momentum theory [4] where the flow around each blade is divided into elementary annular stream tubes. For the present purpose, the information extracted from the above aeroelastic model are the airfoil chord and the span length of each blade section used for the rotor discretization, as well as their angular twist around the blade axis, as far as the turbine geometry is concerned. The relative incoming flow velocity and the angle of attack for each airfoil section are recorded as the blades rotate. Their position relatively to any potential observer/listener are also recorded. These data are averaged in time over the rotor disk and subsequently used to perform individual airfoil section noise calculations (see below). The calculated noise spectra are then 2

3 integrated over the rotor disk, resulting in an average spectrum at any given observer location representative of the overall wind turbine rotor noise. Three aeroacoustic noise sources are considered in this work: turbulent inflow noise [5, 6], trailing edge noise [7, 8, 9] and stall noise [10, 11]. Details about each model can be obtained in the above references. Note that in the present work, the wind turbines are assumed to be upwind rotor turbines. Therefore, low-frequency noise originating from the tower wake deficit is not considered here. Tip noise is also not included in this study as modern wind turbines are designed to avoid this phenomenon. 2.2 Definition of wind turbines as individual noise sources The overall wind farm model is based on aero-acoustic calculations of full wind turbine rotors, as described in the previous section. The results of these calculations are used to characterize each individual wind turbine present in the wind farm. This first step of the simulation procedure consists in choosing a wind turbine rotor configuration, its geometry and the inflow wind characteristics. Using the wind turbine rotor noise model described in Section 2.1, the noise is calculated at given mean wind speeds. At the moment, no yaw error of the turbine relative to the inflow is considered but this can easily be included. The emitted noise is calculated at several observer/listener locations evenly distributed on a sphere centered at the rotor center with a radius equal to 5 rotor diameters. This latter radius is chosen such that the main directivity effects associated to the proximity of the wind turbine blades and rotor are removed from the results, however the long-range directivity effects of the noise generated by the turbine are preserved. Note that in practice, some of these observer/listener locations are physically below ground level as a typical wind turbine tower height is of the order of the rotor diameter or less. Nevertheless, these points are fictitious and only used to characterize noise directivity effects at large distances from the turbine (see below). The calculated wind turbine noise pressure Power Spectral Density (PSD) spectra for each observer location are multiplied by the surface of the sphere centered at the rotor center and intersecting the observer location. This is done in order to define an equivalent noise power representative of the emitted rotor noise independent of the observer location. These spectra are stored together with their associated directivity vector defined as the vector from the rotor center to the observer locations. The wind turbine directivity of the wind turbine that will be used later in this article for performing wind farm noise simulations is illustrated in Figs. 1(a) and (b), each with different viewpoint perspectives. The noise spectra are integrated in distinct frequency bands ( Hz, Hz, Hz and 5000 Hz-25 khz) and displayed in colors as Sound Pressure Levels (SPL). A clear directivity pattern is observed in the low-frequency range for which the turbine emits less noise across the rotor plane and more toward the upstream and downstream directions. In the high-frequency range, a noise emission deficiency is observed in the direction toward the left of the rotor (by convention, left is here defined when looking the turbine from upstream). At the present stage, the wind farm model only consider identical wind turbines, but this could 3

4 easily be extended to different turbines. Nevertheless, each turbine may experience different yaw angles and mean inflow velocities. 2.3 Farm layout, surrounding terrain and atmospheric conditions The map used in this article to describe the surrounding terrain is 6 6 km in the latitude and longitude directions. It is defined as an uniform cartesian grid with the altitude (above sea level) specified at each vertex. The step size is 20 m in each direction. The coordinates system origin is located at the grid center. A 3D elevation map is displayed in Fig. 2(a). The surrounding terrain consists of a reference (sea level) westward flat area at an altitude of 0 m, a ridge running from South to North with a 400 m wide plateau at an altitude of 10 m, an eastward flat area at an altitude of 5 m, and a m (in the east-west and north-south directions, respectively) rectangular-shaped object with a top altitude of 100 m. The latter could be representative of a high-rise building. The wind farm is located on the ridge and consists of 7 turbines distributed in a row orientated along the north-south axis and centered at the grid map origin. It is depicted in Fig. 2(b). The distance between each turbine is 100 m. The wind velocity profile is defined using a power law as: V(h) = V H (h/h) 1/7 standard for neutral atmospheric conditions, where V(h) and V H are the wind velocity as a function of the elevation h (i.e. distance from terrain ground level) and the velocity at hub height H, respectively. The temperature at sea level is 20 o C and is decreasing with altitude by 1 o per 100 m. 2.4 Noise propagation using ray tracing method The basic building stone for the propagation model used in the present work is the classical ray tracing method [12]. The main advantages of this method are its flexibility in term of geometry (i.e. ground features) and atmospheric conditions, and its short computational time. However, it suffers a number of deficiencies. For example, it can not model in its simplest basic configuration a number of phenomena such as scattering by objects and turbulence diffraction, although these limitations can be addressed using specific algorithms or parametric models. Note that ground reflection is included in the present model, but so far this feature has not used for the sake of simplicity. In other words, only direct ray paths from wind turbine noise sources and receiver locations in the far-field are considered so far. In the present implementation of the model, the acoustic ray path from each individual turbine to any location of interest in the far-field is computed. This can be used to form a map of noise propagation over a given terrain as it will be illustrated in the next section. Specified as input parameters for the computational model, the following features are also included. Atmospheric conditions are described using specified wind velocity and temperature profiles. In addition, humidity is used to define atmospheric absorption. These conditions are assumed homogeneous across the whole computational domain. The main operational conditions for the present simulations are reported in Table 1. 4

5 Table 1: Main input data for noise propagation simulations. Quantity Value Hub height H 40 m Wind velocity at hub height V H 10 m/s Wind velocity profile Power law as h 1/7 Wind velocity direction 315 o i.e. from North-West Sea level temperature 15 o C Temperature profile Linear upward decrease rate: 0.01 o C/m 3 Model results As mentioned earlier, the results of the model are presented as noise maps. For this purpose, the noise spectra calculated at each vertex of the terrain grid (2 m above the ground) are integrated over frequencies and the resulting noise SPLs are plotted on top of the terrain using colored isovalues. Note that atmospheric absorption has not been included in the present results in order to concentrate the analysis on directivity effects. The noise map for the SPL integrated over the entire frequency range (i.e. from 40 Hz to 25 khz) is displayed in Fig. 3(a), and noise maps for SPL integrated over frequency bands in Fig. 3(b) using the same bands as those defined in Section 2.2. These noise maps clearly show the existence of a shadow zone upwind of the wind farm caused by the bending of the rays resulting from atmospheric inhomogeneities with respect to altitude. A shadow zone behind the rectangular-shaped object is also observed. In the case of the noise maps of the SPL integrated over the entire frequency range or over the low frequency band ( Hz), the directivity pattern of the turbines observed in Fig. 1 at these low frequencies is apparent with a clear noise emission deficit in the rotor planes. Remind that all turbine rotors are assumed facing the wind direction perpendicularly. These two maps also show that most of the propagated noise energy content is concentrated in this low-frequency range. At higher frequencies, the rotor plane noise deficit directivity pattern progressively disappears and the main direction of propagation is along the upwind-downwind axis (also observed at lower frequencies). The directivity pattern observed in Fig. 1 for the 5000 Hz-25 khz frequency band characterized by a noise emission deficit on the left of the rotor is not apparent in Fig. 3(b) for this same frequency band. 4 Conclusions A model for simulating wind farm noise over the surrounding terrain has been proposed. This model is quite general in term of the wind farm layout, the surrounding terrain and the atmospheric conditions. However, the accuracy prediction of noise propagation from wind turbine to receiver is inherently impaired by the use of the ray tracing method. In particular, diffraction and scattering effects (and consequently noise evaluation in shadow zones) are not well han- 5

6 dled by this method. In the future, the latter difficulties will be delt with, as some numerical and modelling techniques exist in order to alleviate these deficiencies. The model has been used to simulate a fictitious wind farm. A shadow zone is observed upwind of the farm as expected, and directivity features are also visible when plotting noise maps. An important step to make such a modelling tool worthwhile is to validate it against experimental results. The main difficulty here will be to get access to such experimental data for which all the model input parameters have been either monitored or can be infered with enough confidence. Such extensive measurement campaigns of wind turbine noise propagation are scarce. Acknowledgements This study was partly conducted as part of the program Cross-Cutting Activities - Wind Turbine Noise internally funded by DTU Wind Energy. References [1] F. Bertagnolio, H. A. Madsen, and A. Fischer. A Combined Aeroelastic-Aeroacoustic Model for Wind Turbine Noise: Validation and Analysis of Field Measurements. Wind Energy. (in revision). [2] T. J. Larsen and A. M. Hansen. Influence of Blade Pitch Loads by Large Blade Deflections and Pitch Actuator Dynamics Using the New Aeroelastic Code HAWC2. In Proc. of the European Wind Energy Conference and Exhibition, Athens, Greece, [3] T. J. Larsen and A. M. Hansen. How 2 HAWC2, The User s Manual. Tech. Rep. RISØ-R- 1597(ver.3-1), Risø-DTU, Roskilde, Denmark, December [4] H. Glauert. Airplane Propellers, volume Aerodynamic Theory Volume IV. W. F. Durand, The Dover Edition (UK), [5] R. K. Amiet. Acoustic Radiation from an Airfoil in a Turbulent Stream. J. Sound Vib., 41(22): , [6] R. W. Paterson and R. K. Amiet. Acoustic Radiation and Surface Pressure Characteristics of an Airfoil Due to Incident Turbulence. In 3 rd AIAA Aero-Acoustics Conference, Conf. Proceedings, Palo Alto, CA, July [7] R. Parchen. Progress report DRAW: A Prediction Scheme for Trailing-Edge Noise Based on Detailed Boundary-Layer Characteristics. TNO Rept. HAG-RPT , TNO Institute of Applied Physics, The Netherlands, [8] P. Moriarty. NAFNoise User s Guide. Tech. Rep., NREL, Golden, CO, July (Available online: 6

7 [9] F. Bertagnolio, A. Fischer, and W. J. Zhu. Tuning of Turbulent Boundary Layer Anisotropy for Improved Surface Pressure and Trailing-Edge Noise Modeling. Journal of Sound and Vibration, 333: , [10] F. Bertagnolio., H. A. Madsen, A. Fischer, and C. Bak. Experimental Investigation of Stall Noise Toward its Modelling. In 6th International Conference on Wind Turbine Noise, Conference Proceedings, Glasgow, UK, April [11] F. Bertagnolio, H. A. Madsen, A. Fischer, and C. Bak. A Semi-Empirical Airfoil Stall Noise Model Based on Surface Pressure Measurements. Journal of Sound and Vibration. (submitted). [12] A. D. Pierce. Acoustics - An Introduction to its Physical Principles and Applications. The Acoustical Society of America,

8 (a) Seen from downstream (b) Seen from upstream Figure 1: Wind turbine noise directivity patterns (Convention: Left of the rotor is defined when looking from upstream) 8

9 (a) Entire terrain map (b) Close-up view of wind farm Figure 2: Wind farm and map of surrouding terrain 9

10 (a) SPL [db 1/3 ] integrated over all frequencies (b) SPL [db 1/3 ] integrated in frequency bands Figure 3: Noise map of Sound Pressure Levels over terrain 10