Posters

Abstract

In the wind industry, it is common to perform a 12-month wind measurement campaign, as it typically provides sufficient data to determine a long-term wind climate that is representative for normal meteorological conditions. Extreme meteorological conditions, and consequently the design properties of the wind turbine, rely on this one year of data. Representative extreme wind speeds might not occur on an annual frequency, which introduces significant challenges when extrapolating extreme values from one year to 50 years. Another method is to combine this year of data with global atmospheric models (ERA5 or MERRA2 for example). The resolution of these models is too coarse to resolve local effects which is required for an accurate assessment of the extreme events. Eventually, developers might overdesign (or worse, underdesign) their wind farm which would result in additional costs or risks.     We propose a novel methodology where we will select several storm peaks and combine them with reanalysis model data from Whiffle LES. With a 100x100m horizontal model we are able to resolve local effects and improve the representation of extreme events compared to global atmospheric models.Using a 20-year dataset from the Cabauw met mast as reference, we aim at showing that our method can improve the once in a 50 year wind speed (vref) compared with the existing methods. We will also provide a realistic methodology for the uncertainty estimates. The paper will compare results from 4 different methods using: (1) 20 years of measured data at Cabauw, which will be considered as the reference, (2) 1 year of measured at Cabauw, (3) 20 years of ERA5 data calibrated with 1 year of measured data and (4) 45 storms simulated with Whiffle LES, calibrated with 1 year of measurement data. We will discuss advantages and limitations of all four methods. First results show that combining only one year of measurements and Whiffle LES model results improves the vref estimate compared to using only observations. This combination provides a vref within 2% of the full 20 years of observations. Compared to using only a single year of measurements, the ERA5 storm peaks also improved the extreme value analysis. However, compared to the full 20 years from Cabauw and the Whiffle’s LES approach with one year of Cabauw measurements, the ERA5 based extreme value estimates are significantly lower.