Presentations

Abstract

In this study, we first tackle the challenge of processing and filtering SCADA data to enable an accurate turbine interaction (wakes and blockage) modelling validation. Secondly, we compare the performance of various engineering wake models and different high-fidelity CFD simulations to production data of the Arkona wind farm. The Arkona wind farm is located in the Baltic Sea and comprises 60 6MW Siemens turbines. With its irregular layout, the presence of the neighbouring Wikinger wind farm, the multiple turbine add-ons, and the absence of free wind measurement, Arkona is a challenging case for a wake model validation. Three years of raw SCADA data, from 2020 to 2023, is averaged and filtered using FLASC, an open-source tool developed by NREL. The filtering steps include the removal bad or recurrent data points, correction of yaw calibration errors, curtailment periods, low power and wind speed, and non-operational turbines. Data is kept for 10-minute timestamps where, at least, 57 out of 60 turbines are under normal operation. The free wind conditions are estimated using a combination of an average of the wind farm and a selected set of front-turbines. One of the objectives of the study is to assess the performance of selected engineering wake models and different high-fidelity tools for modelling the annual energy production. The engineering wake models compared include Park2, TurboPark, Fuga, and RWE's in-house VV model. The high-fidelity models are Equinor's RANS simulations from DTU PyWakeEllipSys and RWE's simulations from Whiffle's GRASP LES model. The results show variable performance of the wake models compared to the data, specifically in some specific wind direction sectors, suggesting that parameters that are not properly accounted for in the wake modelling tools may play an important role. The uncertainty of the SCADA data and processing methodology will be discussed, as well as the difference in model performance.