Posters

Abstract

Background China's ambition for offshore wind power is not ceased by the renewable subsidy sunset, against the backdrop of the country seriously committing to achieve carbon-neutrality by 2060. Amidst this commitment lies a critical challenge: the high construction costs of offshore wind projects juxtaposed with declining feed-in tariffs underscore the necessity of precise wind resource estimation, commencing as early as the project's preliminary planning phase. The swift pace of offshore wind energy development mandates an accurate knowledge of wind resources, often prior to on-site wind measurement campaigns, or only with measurement data from masts or floating lidars geographically removed from the area of interest. While mesoscale models span large areas, they fall short in capturing small-scale phenomena, leading to potentially substantial temporal and spatial errors—thus amplifying risks during the project planning and design stages. Methodology In response, this study spotlights an escalating interest in leveraging satellite data to refine offshore wind speed assessments. We present the SARWind methodology, which utilizes synthetic aperture radar (SAR) data in conjunction with specialized machine learning algorithms and atmospheric models. This approach comprises: Sentinel-1 SAR wind atlas generation * Processing of Sentinel-1 satellite's raw SAR data, inclusive of filtering out bright targets. * Retrieval of the surface wind field from SAR observations, followed by the application of a surface correction model to address biases inherent to SAR sensors. Wind resource assessment derived from Sentinel-1 SAR images * Extrapolation of surface wind speeds from SAR data to target heights, leveraging a machine learning algorithm trained on in-situ observations and high-resolution mesoscale model parameters. * Implementation of a final post-processing step to mitigate the SAR database's low temporal sampling and to compile wind statistics. In our investigation, SARWind methodology is deployed across an offshore region proximal to Zhuanghe in China's Liaoning province. Validation against on-site met mast measurements within the same locale underscores SARWind's robustness. The difference between the SARWind-derived and the mesoscale-based wind atlas is also analyzed, showing the capacity of SARWind solution in predicting wind resources with superior spatial precision. Results and conclusions The affirmative verification that SARWind can accurately assess wind resources—sans integration of any anemometer measurements—is a testament to its performance as a reliable resource assessment method at the early stage of offshore wind power projects. SARWind extends an enhanced level of detail, attributed to the 500m spatial resolution of SAR surface wind measurements. The resultant wind resource maps reproduce the coastal wind speed gradient. Consequently, using SAR data combined with machine learning promises to elevate precision in offshore wind resource assessment at hub heights and yields strategic insights pivotal for optimizing wind farm placement and managing associated risks.