Psychoacoustic Evaluation of Modelled Wind Turbine Noise

Current validation of wind turbine noisemodels primarily focuses on sound levels averaged over time, typicallyexpressed in metrics such as the equivalent A-weighted sound pressure level(LA,eq). Whereas valid for regulatory purposes, these methods are notsufficient for psychoacoustic research, as time-averaged levels alone do notfully explain the measured noise annoyance. Therefore, this research aims toestablish whether psychoacoustic sound quality metrics (SQMs) provideadditional value when analysing wind turbine noise models. This work employsthe Horizontal Axis Wind turbine simulation Code 2 (HAWC2) to generate noisesource spectrograms, which are propagated through the atmosphere with aGaussian beam-tracing approach. The final sound signals are retrieved throughan inverse Short-Time Fourier Transform (iSTFT). This methodology is applied toa case study featuring a stall-controlled, horizontal-axis, Nordtank NTK 500/41wind turbine. The results are evaluated against measurements by consideringLA,eq, SQMs, and a comparative listening experiment. The LA,eq metric shows aconsistent underprediction of the simulations with respect to measurements,which is partly explained by a ground reflection modelling error. In thehigh-frequency range, stall noise is known to be significantly underpredictedby the aero-acoustic simulation model. This usually translates in increasingdiscrepancies between measurements and models as the wind speed increases. Thecomparative listening experiment confirms that participants experience thesimulations and the measurements as significantly different. The differenceratings show a good agreement with the differences in the psychoacousticannoyance and loudness metrics. It is more difficult to relate the results fromthe listening experiment to LA,eq. These findings confirm that an evaluationwith psychoacoustic metrics next to conventional methods provides additionalvalue in validating wind turbine noise models for human perception research.