Wind turbine noise propagation: Comparison of measurements and predictions by a method based on geometrical ray theory

Abstract

The objective of the present work has been to investigate the accuracy of the Nord2000 propagation model when used to predict noise from high sources like wind turbines. The model is representative to a new generation of �engineering� models developed on basis of analytical and heuristic solutions. The accuracy is investigated by comparison with measurements. Three experiments have been carried out for this purpose: 1) downwind and upwind propagation over flat grass-covered ground from an elevated loudspeaker, 2) downwind and upwind propagation over non-flat ground from a loudspeaker placed on top of a wind turbine nacelle, and 3) downwind propagation over non-flat ground from a single wind turbine. Experiment 1) showed for downwind propagation a very fine agreement between measured and predicted A-weighted noise levels in the considered range of propagation distances (1500 m). The bias error was 0.1 dB with a standard uncertainty of 0.7 dB. The agreement for upwind propagation was not as good. In experiment 2) the accuracy was slightly reduced compared to the downwind results of experiment 1) with a bias error of 0.5 dB and a standard uncertainty of 1.9 dB. In experiment 3) where a wind turbine was used as source the agreement was less than seen in the loudspeaker experiments.