Presentations

Abstract

The use of Floating Lidars Systems (FLS) is crucial in providing cost-effective and high-quality data for the offshore wind industry. Although the 10-minute average Horizontal Wind Speed (HWS) measured by FLS has been proven to be reliable when compared to traditional cup and sonic anemometers, there are notable differences in the Turbulence Intensity (TI). This discrepancy is due to various factors that affect FLS measurements, including sampling rates, buoy motion, spatial variability, and environmental conditions. The state-of-the-art technique to improve TI estimates is to apply a motion correction of the FLS Line of Sight (LoS) measurements at high frequency. Our study introduces an innovative patent protected Machine Learning framework that enables the correction of Turbulence Intensity (TI) values measured by FLS, allowing for a direct comparison with TI values obtained from fixed Meteorological Masts (MM). The model has undergone extensive training on a large dataset consisting of over 1500 days (~4 years) of data collected from EOLOS FLS200 buoys during approximately 35 validation campaigns, including 4 Stage3 campaigns, conducted at 5 different sites (LEG-NL, IJJM-NL, NOAH-UK, FINO-D, ASIT-US). The data gathered during every validation campaign includes both wind, wave, atmospheric and meteorological features to tackle the full complexity of the problem, a fraction of which is used to train the model. The remaining fraction is then used as a continuous multi-dimensional data stream to correct the FLS TI values. Results are confronted against MM TI, raw FLS TI and high frequency motion corrected FLS TI. The correction algorithm yields TI estimates that show good agreement with MM TI, offering a better correction performance when compared to the LoS data motion correction method. This allows us to conclude that the correction goes beyond the pure motion effect, tackling the whole complexity of the problem, including the Lidar probe volume difference.