Presentations

Abstract

The mesoscale Weather Research and Forecasting (WRF) model is a crucial tool for assessing wind energy resources, particularly in offshore environments where direct measurements are limited. Wind farm parameterizations (WFPs) are utilized to simulate turbine power and flow effects. Since the initial implementation of the Fitch parameterization (Fitch et al., 2012), various modifications and alternative models have emerged, though few studies have rigorously compared these WFPs. Existing comparisons against measurements and between the models show that Fitch tends to produce higher wake losses than alternatives like the Explicit Wake Parametrisation (EWP, Volker et al., 2015), and that the spread of results between models is significantly large (Pryor et al., 2022; Ali et al., 2023). Here, we present a validation study of seven different WFPs using production data from three wind farms in the German Bight over several years. The model setup is based on previous work and is optimized for offshore wind conditions, utilizing one-way nested domains. Two different spatial resolution of the innermost domain are compared: 2 km and 0.7 km. The simulation spans the years 2017 to 2021. Production data was collected from all turbines at the selected wind farms, with a minimum of two years of data for each. The SCADA data underwent quality control and periods of production losses due to various reasons (technical issues, maintenance, curtailment etc.) were identified and gap-filled with theoretical power calculations to ensure a fair comparison with simulations. The results indicate that the Fitch WFP consistently underestimates the energy yield across all farms by 3-5% in the low-resolution case and by even more at higher resolution (5-8%).  EWP meanwhile shows the opposite trend with an overestimation of energy yield at low resolution (0-5%) and quite a good match (0-3%) in higher resolution. Adjusting the Fitch WFP for rotor equivalent wind speed or density has minimal impact on the results. The correction for inner-grid layout effects increases the bias by even more, while the correction for local turbine induction (Vollmer et al., 2024) provides the closest match to the observations (0-4% overestimation at 2km and 0-3% underestimation at 0.7km).  The study highlights the challenges in generalizing results from case studies, as the accuracy of WFPs varies across different wind farms. More detailed analyses will be presented, such as gross yield estimates of undisturbed turbines, wind direction dependency of internal wake losses (aligned or staggered layout) and overall patterns of production as well as the representation of external wakes. The split into the contributing factors reveal that some models replicate the observations at individual wind farms purely due to error compensations. The research acknowledges the contributions of Vattenfall and RWE for data provision and recognizes funding support from the German Federal Ministry for Economic Affairs and Climate Action for the C²-Wakes project (grant no. 03EE3087A).