Beamforming applied to Small Wind Turbines

Abstract

The increase of energy consumption as well as the need to reduce green-house emissions trigger the shift from a fossil-fuel based society towards a sustainable society. With a constant increase in the installed power capacity, wind plays a key role in the global energy framework. The improvement of the technology opens up new targets for the wind development [28]. In particular, the installation of small wind turbines in the urban environment represents a promising solution to supply residential power. Whilst, some of the problems faced by large-scale wind farms, like losses in the electrical distribution and transportation system, can be reduced. However, the immaturity of the technology leads to elevated capital costs and technical challenges [41]. Among these disadvantages, the siting near human activities is limited by the sound emitted by small wind turbines [65].
The current work focuses on the aerodynamic noise produced by horizontal-axis small wind turbines. For this purpose, the rotational beamforming algorithm, ROSI (i.e. ROtating Source Identifier), together with microphone array measurements, have been used. The ability of the method to follow the movement of the sound source is essential to correctly localize and quantify the rotating sound source. ROSI is applied in the time domain: first, for every source position a time history reconstruction of the signal recorded at each microphone is performed. Then, the microphone signals are sampled for every emission time, leading to the de-Dopplerisation of the signals. Last step is to sum all de-Dopplerised signals and to repeat this procedure for each scan point [33].
In order to validate the technique, the ROSI performance has been investigated with an increasing degree of uncertainty: initially, under ideal and know conditions with simulated datasets and, then, applying ROSI to experimental data. The results of the simulations show that ROSI is not influenced neither by the rotational speed of the source or by the selected time snapshot. Furthermore, the differences in the beamformed results always agree with the differences in the source level, proving that ROSI can be used to compare different blade or turbine geometries. Additionally, the ROSI performance is in agreement with the Rayleigh limit. In fact, if the spacing between the sound source is below the Rayleigh limit, the sources are not distinguished as separated and their SPLs are not correctly estimated. As regards the experimental approach, a small-scale prototype simulating the acoustic of a small wind turbine was tested in the Anechoic Tunnel facility while an upwind turbine was measured in the Open Jet Facility, varying both the wind speeds of the tunnel and the rotational speed of the blades. The two facilities are located at the TU Delft University. During the first experiment, the measurement geometry is verified and the resulting outcomes are compared with the simulations. It is confirmed both that the rotational speed does not limit the ROSI performance and that the source localization and quantification are affected by the Rayleigh limit. As regards the last experiment, the ROSI uncertainty is investigated in a non-anechoic environment. In the range of frequencies where the wind turbine noise is expected (i.e. 500 Hz - 2000 Hz [60]), the high levels of background noise do not allow to localize and to quantify the main noise sources in the small wind turbine. Thus, the wind turbine noise is not assessed and the experimental data are not compared with the analytical noise models, provided in literature.