Enhancing offshore wind Resource assessment through neural network-based HF radar data analysis

Abstract

The increasing demand for offshore wind energy underscores the need for accurate wind speed estimation to support the design and operation of offshore wind farms. High-Frequency Radar (HFR), a widely used remote sensing technology in oceanographic research, offers promising potential for wind resource assessment, particularly in areas where conventional measurements are limited. This study explores the application of artificial neural networks (ANNs) for offshore wind speed prediction using HFR-derived data, addressing key challenges in model development and training. A key feature of this approach is the use of a decade-long dataset from the Celtic Sea, off the southwest UK coast, incorporating the full Doppler spectrum and sea surface radial velocity. Model performance was assessed over full-year and seasonally segmented four-month periods, with RMSE values ranging from 1.99 to 2.78 m/s and NRMSE values between 12 % and 20 %, demonstrating the feasibility of HFR-informed ANN models for supporting offshore wind applications.