Wind resource at deep-sea sites applying remote sensing and numerical models

In order to achieve a green energy transition, countries in Europe and Worldwide are pushing to increase offshore wind energy. The planned expansion of offshore wind glob ally is massive. Wind farms are being planned with larger and larger wind turbines and in places with ever deeper seas and further from land. In those areas there is a great lack of wind measurements. The traditional measurements with cup anemometers installed on meteorological masts are no longer economically feasible and have technical problems associated with installation.

This PhD investigates the use of remote sensing tools and numerical models applied to the deepsea wind resource. As wind varies in time and place, the PhD study is divided into three specific objectives which reflect the wind’s variations on different temporal and spatial scales.

The first objective examines the longterm wind resource and the consequences due to the changing climate according to historical reanalysis data. In the North Sea there is a netincrease in the wind energy potential due to the differentially increasing air and sea temperatures in the winter months. The US East Coast on the other hand sees a consistent decline in the wind energy potential over the last 60 years.

The second objective explores the midrange temporal scales, applying machine learning (ML) techniques to extrapolate 12 years of satellite ocean wind observations to wind turbine hub heights. The results from the ML are found to have less uncertainty in predicting the wind at hub heights than mesoscale and reanalysis models. All results are compared with in situ meteorological observations from tall measuring masts. When extended to the area surrounding the model training site, the model fails to fully capture the spatial wind variations and may require additional input information. Furthermore, due to the nature of the polarorbiting satellite only measuring at discrete times, the model fails to fully rep resent the diurnal wind variability and thus overpredicts the wind resource statistics. For the investigated location, it appears that the ML model thereby overestimates the wind resource.

The third and final objective looks towards shorter time spans, investigating short buoy based and ferrymounted lidar campaigns. The buoybased system with additional sensors can properly reproduce atmospheric stability conditions in the surface layer and accurately measure wind speeds at multiple heights without needing to account for flow distortions. Wind lidars mounted on ferries are an interesting new option but the data is difficult to validate against other data sources such as satellite data and numerical models due to the ferry’s route and that data is collected while in motion.

The overall conclusions from the PhD are that the research results show strengths as well as limitations for each of the remote sensing techniques studied. The methods have thus different applicability in relation to wind resource statistics at different times such and spatial scales. It is recommended that the techniques be used in conjunction with each other for evaluation, training and/or validation in order to achieve as detailed as possible that map the wind resource over the sea.