Posters

Abstract

Introduction Yaw misalignment, defined as the average difference between the orientation of a wind turbine rotor and the wind direction, can be a potential cause for wind turbine underperformance and an opportunity for optimization [1]. We introduce a top-down approach that leverages satellite imagery, numerical weather models, and operational data to determine and optimize yaw misalignment on all turbines of an operating wind farm. In this presentation, we first describe the different steps of the method, including physics and data-driven modeling elements, and model their uncertainty. Then, we apply and validate the method on test sites where external sensors, such as nacelle lidars, have been installed to measure the actual yaw misalignment. This enables us to estimate the end-to-end uncertainty of the process. Method Traditionally, reanalysis time series such as ERA5 are used to verify and calibrate nacelle positions, which are often subject to errors. This is a valid first step to correct large offsets and detect jumps but but does not guarantee the resulting accuracy compared to ground truth. To overcome this, we introduce the use of high-resolution (1 m or finer) satellite imagery as additional observations. In these images, we identify markers such as blade tips and nacelle edges for each turbine and its corresponding shadow. Then we determine the expected yaw position by minimizing the difference in positions compared to markers obtained from a 3D projection model, accounting for the satellite and sun positions and the terrain slope. Once reliable 10-minute time series have been created for nacelle yaw positions, we use a numerical weather prediction (NWP) tool to generate reanalysis wind direction time series at each turbine location. The wind direction timeseries are calibrated using on-site wind measurements, if available, and additional adjustments are determined using the observed wake patterns, which are compared to the predicted wake patterns according to the simulation. Finally, the yaw misalignment time series is calculated for each turbine as the difference between the nacelle direction and the wind direction. The optimal yaw misalignment, where maximum power is produced, may not be 0° and is potentially dependent on the aerodynamic design of the turbine, its operation strategy, and the site conditions [2]. Thus, to optimize the park, it is important to determine it. To achieve this, we evaluate the power performance of each turbine relative to an ensemble of its neighbors as a function of the misalignment. Validation and discussion We validated the method through blind experiments conducted on three full-scale operational sites, each involving several turbines equipped with independent sensors. The results demonstrate that it achieves a standard uncertainty of 3° on the average yaw misalignment. We discuss key challenges, including access to regular high-resolution satellite imagery, and evaluate the potential performance gains from correcting yaw misalignment errors. [1] Fleming et al. (2014). Field-test results using a nacelle-mounted lidar for improving wind turbine power capture by reducing yaw misalignment.  [2]  Hulsman et al. (2022). Turbine power loss during yaw-misaligned free field tests at different atmospheric conditions.