Posters

Abstract

In this study we present the evolution of the Eolos’ original machine learning (ML) turbulence intensity (TI) correction model. In 2024, Eolos developed a ML model specifically designed to correct TI measurements for the FLS200 buoy, a widely used floating lidar system. This innovative model has undergone validation by DNV, ensuring its reliability and accuracy, is protected by patent and has already reached a commercial stage. However, its application is limited to the FLS200 buoy, restricting its broader use across different buoy types and lidar configurations. To address this limitation, we introduce a new, more versatile ML model capable of functioning across various buoy platforms, regardless of their size or the type of lidar they employ. This advancement was achieved through a two-stage strategy. In the first stage, a transformation layer converts key parameters from other floating lidar systems into equivalent values as if they were measured by the FLS200—such as LiDAR measured 10-minute wind speed standard deviation. The second stage involves a modified version of the original Eolos ML model, adapted to process the transformed data and deliver corrected TI measurements. The results from this new approach are promising, providing confidence in the model’s ability to generalize across different buoy manufacturers. Data from diverse OEMs has been used at a range of different sites and, consistently, the corrected TI values show reduced errors when compared to the uncorrected datasets. By enabling accurate TI correction across diverse FLS types, we contribute to improved wind resource assessment and support the expansion of offshore wind energy projects globally in this challenging times.