Search

Data-driven wake models have recently shown a high accuracy in reproducing wake characteristics from numerical data sets. This study used wake measurements from a lidar-equipped commercial wind turbine and inflow measurements from a nearby meteorological mast to validate an interpretable data-driven surrogate wake model. The trained data-driven model was then compared to a state-of-the-art analytical wake model. A multi-plane lidar measurement strategy captured the occurrence of the wake curl during yaw misalignment, which had not yet conclusively been observed in the field. The comparison between the wake models showed that the available power estimations of a virtual turbine situated four rotor diameters downstream were significantly more accurate with the data-driven model than with the analytical model. The mean absolute percentage error was reduced by 19 % to 36 %, depending on the input variables used. Especially under turbine yaw misalignment and high vertical shear, the data-driven model performed better. Further analysis suggested that the accuracy of the data-driven model is hardly affected when using only supervisory control and data acquisition (SCADA) data as input. Although the results are only obtained for a single turbine type, downstream distance and range of yaw misalignments, the outcome of this study is believed to demonstrate the potential of data-driven wake models.

Data-driven wake models have recently shown a high accuracy in reproducing wake characteristics from numerical data sets. This study used wake measurements from a lidar-equipped commercial wind turbine and inflow measurements from a nearby met mast to validate an interpretable data-driven surrogate wake model. The trained data-driven model was then compared to a state-of-the-art analytical wake model. A multi-plane lidar measurement strategy captured the occurrence of the wake curl during yaw misalignment, which had not yet conclusively been observed in the free field. The comparison between the wake models showed that the available power estimations of a virtual turbine situated four rotor diameters downstream were significantly more accurate with the data-driven model than with the analytical model. The Mean Absolute Percentage Error was reduced by 19 % to 36 %, depending on the input variables used. Especially under turbine yaw misalignment and high vertical shear, the data-driven model performed better. Further analysis suggested that the accuracy of the data-driven model is hardly affected when using only SCADA data as input. The outcome of this study demonstrates the enormous potential of data-driven wake models.

This paper presents an analysis of the power loss of the upstream turbine for yaw misalignment depending on the inflow condition. High-quality experimental field data, obtained from a lidar wind profiler, met mast, GPS based yaw sensor, turbine data, and load data is used. To reduce the spreading of the power loss during yaw misalignment, the rotor effective wind speed is estimated using load data. Furthermore, the so-called power loss coefficient currently used in most engineering models, for yaw misalignment, is assessed at various inflow conditions. As main finding, the power loss coefficient increases for stronger vertical wind shear. In addition, there is an asymmetry with respect to the direction of the yaw misalignment that is enhanced by the wind shear exponent and the wind veer. Neglecting these two effects could result in in-conservative wake-steering decisions at stable stratification

Redirecting the wake from an upstream wind turbine by yawing its rotor can reduce the negative impact of the wake on a downstream turbine. The present research investigated wind turbine wake behaviour for three yaw angles [-30,0,30∘] at different inflow turbulence levels and shear profiles under controlled conditions. Experiments were conducted using a model wind turbine with 0.6 m diameter (D) in a large wind tunnel. A short-range lidar WindScanner was used to map the wake with high spatial and temporal resolution in vertical, cross-stream planes at different downstream locations and in a horizontal plane at hub height. The lidar WindScanner enabled fast measurements at multiple locations in comparison to the standard hot-wire measurements. The flow structures and the energy dissipation rate of the wake were measured from 1 up to 10 D, and for one inflow case up to 16 D, downstream of the turbine rotor. A strong dependency of the wake characteristics on both the yaw angle and the inflow conditions was observed. In addition, the curled wake that develops under yaw misalignment due to the counter-rotating vortex pair was more pronounced with a boundary layer (sheared) inflow condition than for uniform inflow with different turbulence levels. Furthermore, the lidar velocity data and calculated quantities such as the energy dissipation rate compared favourably with hot-wire data from previous experiments with a similar inflow condition and wind turbine model in the same facility, lending credibility to the measurement technique and methodology used here. The results of this measurement campaign provided a deeper understanding of the development of the wake for different yaw angles and inflow conditions which can help improve wake models.

This paper aims to develop fast and reliable surrogate models for yaw-based wind farm control. The surrogates, based on polynomial chaos expansion (PCE), are built using high-fidelity flow simulations coupled with aeroelastic simulations of the turbine performance and loads. Developing a model for wind farm control is a challenging control problem due to the time-varying dynamics of the wake. The wind farm control strategy is optimized for both the power output and the loading of the turbines. The optimization performed using two Vestas V27 turbines in a row for a specific atmospheric condition suggests that a power gain of almost 3%±1% can be achieved at close spacing by yawing the upstream turbine more than 15∘. At larger spacing the optimization shows that yawing is not beneficial as the optimization reverts to normal operation. Furthermore, it was also identified that a reduction in the equivalent loads was obtained at the cost of power production. The total power gains are discussed in relation to the associated model errors and the uncertainty of the surrogate models used in the optimization, as well as the implications for wind farm control.

This paper aims to develop fast and reliable surrogate models for yaw-based wind farm control. The surrogates, based on polynomial chaos expansion (PCE), are built using high fidelity flow simulations combined with aeroelastic simulations of the turbine performance and loads. Developing a model for wind farm control is a challenging control problem due to the time-varying dynamics of the wake. Both the power output and the loading of the turbines are included in the optimization of wind farm control strategies. Optimization results performed using two Vestas V27 turbines in a row for a specific atmospheric condition suggest that a power gain of almost 3 % ± 1 % can be achieved at close spacing by yawing the upstream turbine more than 15°. At larger spacing, the power gain the optimization shows that yawing is not beneficial as the optimization reverts to normal operation. Furthermore, it was also identified that a reduction of the equivalent loads was obtained at the cost of power production. The total power gains are discussed in relation to the associated model errors and the uncertainty of the surrogate models used in the optimization, and the implication for wind farm control.

This presentation aims to provide a thorough understanding of the evolution of the wake at different inflow condition and operation condition, by measuring the the flow structures in the wake of a model turbine from 1D up to 10D downstream. The measurement campaign was conducted from May through June 2019 using the MoWiTO 0.6 wind turbine model in the large wind tunnel at ForWind – University of Oldenburg, which can be fitted with a passive or an active grid at the test section entrance to alter the inflow conditions. A short range Lidar WindScanner is used as the primary instrument to map the wind turbine wake at multiple downstream locations.

Due to the interaction between the wake of an upstream turbine on a downstream turbine, power losses and increased fatigue loads occur. By yawing the upstream turbine with regard to the wind direction, one can potentially reduce the power losses of the downstream turbine and reduce the fatigue loads. The evolution of the wake depends on the pressure gradient within the near-wake region and the turbulent kinetic energy and must be incorporated in existing wake steering algorithms to provide an accurate estimation of the wake flow. This paper will show a first approach to implement a near-wake model and a turbulence model in the curled wake model within the controls-oriented framework FLORIS. The near-wake model is based on an analytical expression of the velocity profile to model the pressure gradient. Furthermore, two turbulence models are incorporated within the curled wake model based on a Gaussian-distribution and a mixing length formulation. The adapted curled wake model is then assessed with the measurement data acquired in the wind tunnel at ForWind – University of Oldenburg. The evaluation of the models show good agreement for the velocity deficit and representation of the near-wake region. Furthermore, the turbulent kinetic energy behaved as expected in comparison to other work, showing a ring of high turbulent kinetic energy at non-yawed condition which is deflected to a curled shape at large yaw angles with the turbulence model based on a mixing length formulation.

In this work, a new controls-oriented wake model is modified and compared to an analytical Gaussian wake model, high-fidelity simulation data, and experimental wind tunnel campaign. This model, called the curled wake model, captures a wake phenomenon that occurs behind yawed turbines, modeled as a collection of vortices shed from the rotor plane. Through turbine simulations, these vortices are shown to have a significant impact on the prediction of the wake steering’s performance. Overall, the results support the concept of secondary steering, or a yawed turbine’s ability to deflect the wake of a downstream turbine, and suggest that future turbine wake studies and yaw optimizations should include the curled wake phenomenon.

Wind turbines are typically closely spaced in wind farms, and thus operate in the wake of upstream turbines and experience power losses. Currently, one of the techniques to reduce the wake interaction between turbines within a wind farm is to yaw the upstream turbine with regards to the incident wind direction. The objective is to deflect the wake, which can potentially increase the overall power output and the annual energy production of the wind farm. Experimental data can aid the process to thoroughly analyse the wake deflection under different inflow conditions, which is necessary to apply a yaw-based wind farm control model. The aim of the present research is to investigate the possibility and accuracy of the experimental setup to measure the wake characteristics with a high spatial and temporal resolution through the use of a short-range Lidar WindScanner, within the wind tunnel of ForWind-Oldenburg. This technique provides the opportunity to analyse the flow structures at different operational and inflow conditions in a relative fast manner without disturbing the flow. Experiments were conducted using a model wind turbine in a large cross-section wind tunnel. The short-range Lidar WindScanner is used as the primary instrument to map the wind turbine wake at different downstream locations. The flow structures of the wake were measured from 1 D up to 10 D downstream of the turbine rotor. A stable flow within the wind tunnel segment is measured, which is crucial for the analyses of the evolution of the wake. In addition, a high detailed spatial resolution of the wake profile is observed, showing the symmetric and asymmetric behaviour of the wake, for unyawed and yawed conditions, respectively. Furthermore, the calculation of the thrust coefficient from the velocity data show expected behaviour, giving further credibility to the measurement technique.