Cyclostationary spectral analysis for the measurement and prediction of wind turbine swishing noise

This paper introduces cyclostationary spectral analysis as a new approach to analyzing and predicting the aerodynamic noise generated by wind turbines. This method is able to reveal new insights into the periodic character of the noise signal and is therefore ideally suited to the study of wind turbine noise. A new formulation is presented for the time variation of the noise spectrum due to wind turbines thereby providing insight into the character of the periodic variation in noise referred to as ‘swishing’. The character and mechanism of swishing noise is analyzed in detail.Cyclostationary analysis is applied to noise data obtained on a 1.5 MW wind turbine to predict the time variation of the autocorrelation function and spectral distribution (Wigner–Ville representation). Comparisons of the predictions with measurements are in good qualitative agreement. A swishing index is proposed to quantify the swishing sound of wind turbine noise.