Posters

Abstract

Yaw misalignmentin wind turbines, categorized as dynamic and static, occurs when the turbines don't face directly into the wind, impacting power generation. Any measured difference between wind direction and nacelle position is calledyaw error. When the turbine is pointing directly into the wind, yaw error should measure 0°. Controllers do not immediately respond to changing wind directions, waiting for yaw error to exceed a set threshold over time before signaling nacelle movement—known asdynamic yaw misalignment. On the other hand,static yaw misalignmentoccurs when the measured yaw error is 0°, yet the turbine isn't facing directly into the wind. This misalignment remains invisible to the turbine controller, compromising turbine performance by reducing Annual Energy Production (AEP) and increasing loads. Consequently, the wind industry has put much effort into resolving static yaw misalignment with LiDARs, spinner anemometers, met towers, and data-driven approaches. A major question using data-driven approaches is determining the data resolution needed to discern static yaw misalignment accurately. Some solutions in the market rely on 10-minute data, as seen in NREL's open-source Open Operational Assessment (OpenOA) software, recently released. The OpenOA algorithm assumes static misalignment as the difference between the mean 10-minute yaw misalignment and the average yaw error value at peak power, determined by a nonlinear curve fit.In contrast, WindESCo's static yaw misalignment algorithm utilizes high-frequency upwards of 1-second data from 10-15 SCADA tags, identifying the optimal yaw offset for maximum power production. A comparative analysis was performed among the open-source 10-minute algorithm, WindESCo's 1-second algorithm, and a LiDAR campaign for three turbines. Partial results of the analysis are presented below: * Turbine 1, Period 1: LiDAR -6.7, WindESCo1sec -6.6 +/- 2.5, Open-Source10min +22.2 +/- 10.9 * Turbine 1, Period 2: LiDAR 0, WindESCo1sec +0.8 +/- 2.8, Open-Source10min +15.6 +/- 13.5 * Turbine 2, Period 1: LiDAR +2.7, WindESCo1sec +1.7 +/- 2.1, Open-Source10min +14.3 +/- 12.1 * Turbine 2, Period 2: LiDAR 0, WindESCo1sec +0.8 +/- 3.7, Open-Source10min +19.2 +/- 15 * Turbine 3, Period 1: LiDAR -2.2, WindESCo1sec -0.8+-2.1, Open-Source10min +14.8 +-9.3 * Turbine 3, Period 2: LiDAR 0, WindESCo1sec +1.7 +/- 2.9, Open-Source10min +17.4 +/- 10.8 Further analysis on a 4-turbine site, following a similar approach to the 3-turbine test, compared the results from LiDAR measurement, OpenOA (Single Method and Monte Carlo), and the WindESCo algorithm. The open-source 10-min-based algorithm showed differences in static yaw misalignment values between 20 and 36 degrees compared with the LiDAR results. Despite having 15 months of data, the 10-min-based algorithm exhibited an uncertainty of up to ±10 degrees. In contrast, the WindESCo 1-second algorithm demonstrated a very high correlation with LiDAR results, achieving an R2 of 0.9. In summary,this study centers on comparing data-driven methodologies to determine the sign and magnitude of static yaw misalignment, crucial for enhancing the AEP of wind turbines. The analysis contrasts static yaw misalignment values derived from the open-source 10-minute algorithm, the WindESCo 1-second algorithm, and LiDAR results. Notably, a robust correlation is observed between the 1-second-based algorithm and LiDAR, while substantial disparities emerge between these approaches and the open-source algorithm founded on 10-minute data.