A Novel Approach for the Extrapolation of Wind Speed Time Series to Higher Altitude Using Machine Learning Model

Mapping the wind speed profile at higher altitudes over long periods is an essential requirement in the wind energy sector and the erection and maintenance of tall wind measuring masts for this purpose pose many practical difficulties from the engineering perspective. A rather simplified method in use for this purpose is to estimate the wind profile at relatively lower heights and estimate the same to the required heights using mathematical equations formulated using a combination of theoretical and empirical research, which are often subject to uncertainties caused by the complex nature of turbulent flows and the terrain. A new approach of employing a machine learning model built on Support Vector Machine algorithm is attempted in this work to extrapolate the wind speed profile at a higher altitude from the measured values at lower heights. The measured values at 50 m and 20 m above the ground level have been separately used to train the model to predict the values at 80 m height and the comparison of the model prediction with the measured values at higher altitude shows remarkable agreement.