Offshore wind farms are a low cost, efficient technology for green energy. They deliver significant economic benefits through manufacturing and operation, and importantly can be deployed at scale. Offshore wind also offers a route to opening up access to renewable energy for a global population, majorly clustered around coastal locations. A few studies have shown that the offshore winds at the hub heights of wind turbines is on average 90% higher than over land (Archer and Jacobson 2005). America’s fledgling offshore wind sector has been growing over the past few years and plans to account for ~22 GW of energy by 2030 and ~86 GW of energy by 2050 (https://www.energy.gov/eere/wind/wind-vision-1). While there are big plans for the future, offshore wind speed data around wind turbine hub heights are only available either through in situ observations viz. wind masts and lidars at selected locations, or forecasting-model based data from NREL’s WIND toolkit. In situ data, which is a good source for validation of wind, are very sparse and costly to install en masse, whereas satellite-derived winds have vast coverage at high resolution. In this study, we show the potential of using machine learning techniques to accurately estimate offshore wind speed profiles from satellite-derived surface wind speeds compared to conventional methods. We use machine learning, in particular the random forest regressor, to estimate wind profiles from the National Oceanic and Atmospheric Administration (NOAA) National Center for Environmental Information’s (NCEI) Blended Seawinds version 2.0 (NBSv2.0) product (Saha and Zhang 2022), which contains satellite-derived 10m global neutral blended wind speed gridded data (0.25 degrees) with a resolution of up to 6 hours dating back to 1987. A single extrapolation model applicable to US coastal regions is developed, instead of an area specific one as attempted in previous studies (Optis 2021).
The study makes use of hundreds of thousands of wind profiles from six publicly available lidar datasets over the Northeast US, California, and Hawaii regions to train and test a random forest model to extrapolate wind speed profiles up to 200m. The final model is implemented on the NBSv2.0 product, to create publicly available wind speed profiles over the US coasts, which are validated against the NREL’s wind resource data for North America. An equally gridded map of wind profile in the wave boundary layer around the USA coastal waters will help develop a suite of wind energy resources (wind speed, wind speed frequency distribution, wind power density, effective wind speed occurrence, and rich level occurrence and their trends) and will help stakeholders in their decision making related to wind based renewable energy development.
References
Archer, C. L., and Jacobson, M. Z. (2005), Evaluation of global wind power, J. Geophys. Res., 110, D12110, doi:10.1029/2004JD005462.
Optis, M., Bodini, N., Debnath, M., and Doubrawa, P. (2021), New methods to improve the vertical extrapolation of near-surface offshore wind speeds, Wind Energ. Sci., 6, 935–948, https://doi.org/10.5194/wes-6-935-2021.
Saha K and Zhang H-M (2022) Hurricane and Typhoon Storm Wind Resolving NOAA NCEI Blended Sea Surface Wind (NBS) Product. Front. Mar. Sci. 9:935549. doi: 10.3389/fmars.2022.935549