Sensitivity analysis of wind turbine broadband noise estimation to semi-empirical models parameters

The continuous increase of energy demand and the rising concerns on climate change, are pushing the European Union decarbonization strategies and transition toward renewable based energy systems, with wind energy playing a leading role. It is therefore necessary to have a better understanding of how wind turbines (WTs) impact on their surroundings, including their noise emissions. Among the different methods to compute noise emissions of WTs, semi-empirical models are a valid choice to have a-priori estimations of noise spectra and sound pressure levels. These models are based on correlation laws for different physical mechanisms that contribute to noise generation. Popular models for dominant noise sources include the Amiet approach for inflow turbulence noise and the Lowson model for turbulent boundary layer-trailing edge noise. Determining the parameters involved in these models can be challenging, potentially leading to significant errors in noise prediction. In this study, we conducted a novel sensitivity analysis of the models by varying different parameters such as turbulent intensity and dissipation, boundary layer thickness, and temperature. The selected test case is the reference multi-MW horizontal axis wind turbine Neg-Micon 80. The results of the multilevel-multivariate analysis, involving 63,360 combinations of the input parameters, clearly demonstrate a significant dependence of these models on atmospheric turbulence parameters. Furthermore, these models exhibit an higher sensitivity to input parameters at lower frequencies of the noise spectrum, which are generally associated with higher values of sound pressure level.