Validation and application of advanced sound propagation modeling for optimization of wind farms

Camilla Marie Nyborg*

*Corresponding author for this work

Research output: Book/ReportPh.D. thesis

9 Downloads (Pure)


This thesis introduces the use of an advanced sound propagation model based on the parabolic equation method for noise constrained optimization of wind farms. For this purpose, the WindSTAR (Wind turbine Simulation Tool for AeRodynamic noise) sound propagation model is combined with the wind farm flow modeling tool PyWake and coupled to the wind farm optimization framework TopFarm. Detailed investigations of noise emitted from wind turbines become increasingly important as we experience a continuing demand for onshore wind farms. In this case, more advanced sound propagation models are needed, since they take a broader range of parameters into account in comparison to the simpler models currently used for onshore wind farm planning. Through the parabolic equation method, effects from the wind speed and temperature profiles, the terrain elevation, the acoustic reflection factor of the ground, and the velocity deficits in the wind turbine wake can be studied. In this way, a more detailed sound pressure field can be obtained compared to simpler methods such as the ISO 96132 model. As a result, the parabolic equation method allows for a broader exploration of sound propagation in different atmospheric and terrain conditions.

As a first step, the WindSTAR model is validated against field experiments measuring on shore sound propagation from a loudspeaker and a wind turbine, respectively. As a further validation, the model is compared to other types of the parabolic equation in order to assess the computational bias of using one parabolic equation model over the other. Since WindSTAR can be applied with different complexities (i.e. by including turbulence in the computations), the corresponding computational cost of the model can vary significantly. Hence, the performed validation and comparison studies aim at balancing the computational cost and the modeling accuracy of higher fidelity models like WindSTAR. This is done by, for example, investigating the influence of a multiple noise source representation of the wind turbine rotor in the computations. Throughout the thesis the WindSTAR computations are performed in a steady manner. The unsteady nature of the wind turbine noise originating for example from turbulence effects is therefore neglected in order to obtain a sound propagation model appropriate for wind farm optimization purposes.

As the sound propagation models are carefully validated, the WindSTAR model and the simple ISO 96132 model are applied to a developed optimization algorithm using Top Farm with the aim of performing noise constraint optimization of the operation of onshore wind farms. The power production of a wind farm is optimized by switching between the wind turbine specific noise reducing operational modes until the optimal state is found for the specified flow case while complying with the defined noise regulations. The optimization framework is applied to a few selected wind farm cases considering either flat or complex terrain, and the main (dis)advantages of using the more advanced sound propagation model over the simple ISO 96132 model are evaluated.

In general, this thesis contributes to the understanding of sound propagation from wind farms in various conditions through extensive validation studies. The work further demonstrates the inclusion of an advanced sound propagation model into optimization of wind farms. By this, the presented method contributes with new perspectives to the compromise between wind farm operation and noise, and thus to an increasing development of onshore wind energy.
Original languageEnglish
Place of PublicationRisø, Roskilde, Denmark
PublisherDTU Wind and Energy Systems
Number of pages165
Publication statusPublished - 2022


Dive into the research topics of 'Validation and application of advanced sound propagation modeling for optimization of wind farms'. Together they form a unique fingerprint.

Cite this