Hai Bui

Low-level jets (LLJs) are wind maxima typically observed within a few hundred meters of the surface. They are often associated with high vertical wind shear, which significantly impacts the performance and loading of modern wind turbines. In this study, we characterize LLJs over the North Sea using 1 year of lidar observations from FINO1 (2015–2016) and compare them with reanalysis (ERA5) and hindcast (NORA3) data. We introduce a log–jet fitting method to represent each observed or modeled wind profile with five parameters, enabling a direct comparison between lidar and model data. Results show that strong LLJs are generally underestimated by reanalysis and hindcast products. A bias-correction procedure based on quantile mapping is then applied to a 50-year ERA5 dataset to improve the long-term representation of LLJs. K-means clustering further reveals distinct directional and stability-dependent LLJ patterns. The findings highlight the need for detailed modeling of near-surface wind structures and motivate future numerical simulations to clarify the underlying mechanisms that govern LLJ development and variability.

Initializing idealized Large-Eddy Simulations (LES) for wind energy applications presents a persistent control problem, typically characterized by slow convergence due to inertial oscillations and the difficulty of matching target height wind targets. To address this, we present Dynamic Geostrophic Nudging (DGN), a method that couples physical fidelity with computational efficiency. Unlike standard velocity nudging, DGN acts on the forcing terms: it dynamically adjusts the geostrophic wind components based on the flow tendency and the error between the mean velocity and the target value. This mechanism allows the controller to efficiently steer the mean wind toward the target while actively damping inertial oscillations in the boundary layer. We employ a one-dimensional model to perform a parameter sweep and investigate the sensitivity of the control parameters before applying the method to a full three-dimensional LES. The results demonstrate that DGN reduces the spin-up time from the standard 12–24 hours to approximately two hours while maintaining the target wind vector with high accuracy. Furthermore, by arresting the unphysical transient growth of the boundary layer, the method allows for the use of vertically optimized domains, representing a significant advancement in the operational efficiency of precursor generation for wind farm simulations.

This paper presents an intuitive application of multivariate kernel density estimation (KDE) for data correction. The method utilizes the expected value of the conditional probability density function (PDF) and a credible interval to quantify correction uncertainty. A selective KDE factor is proposed to adjust both kernel size and shape, determined through least-squares cross-validation (LSCV) or mean conditional squared error (MCSE) criteria. The selective bandwidth method can be used in combination with the adaptive method to potentially improve accuracy. Two examples, involving a hypothetical dataset and a realistic dataset, demonstrate the efficacy of the method. The selective bandwidth methods consistently outperform non-selective methods, while the adaptive bandwidth methods improve results for the hypothetical dataset but not for the realistic dataset. The MCSE criterion minimizes root mean square error but may yield under-smoothed distributions, whereas the LSCV criterion strikes a balance between PDF fitness and low RMSE.

This paper presents an intuitive application of multivariate kernel density estimation (KDE) for data correction. The method utilizes the expected value of the conditional probability density function (PDF) and a credible interval to quantify correction uncertainty. A selective KDE factor is proposed to adjust both kernel size and shape, determined through least-squares cross-validation (LSCV) or mean conditional squared error (MCSE) criteria. The selective bandwidth method can be used in combination with the adaptive method to potentially improve accuracy. Two examples, involving a hypothetical dataset and a realistic dataset, demonstrate the efficacy of the method. The selective bandwidth methods consistently outperform non-selective methods, while the adaptive bandwidth methods improve results for the hypothetical dataset but not for the realistic dataset. The MCSE criterion minimizes root mean square error but may yield under-smoothed distributions, whereas the LSCV criterion strikes a balance between PDF fitness and low RMSE.

Low-level jets (LLJs) are wind maxima typically observed within a few hundred meters of the surface. They are often associated with high vertical wind shear, which significantly impacts the performance and loading of modern wind turbines. In this study, we characterize LLJs over the North Sea using one year of LiDAR observations from FINO1 (2015–2016) and compare them with reanalysis (ERA5) and hindcast (NORA3) data. We introduce a log-jet fitting method to represent each observed or modeled wind profile with five parameters, enabling a direct comparison between LiDAR and model data. Results show that strong LLJs are generally underestimated by reanalysis and hindcast products. A bias-correction procedure based on quantile mapping is then applied to a 50-year ERA5 dataset to improve the long-term representation of LLJs. K-means clustering further reveals distinct directional and stability-dependent LLJ patterns. The findings highlight the need for detailed modeling of near-surface wind structures and motivate future numerical simulations to clarify the underlying mechanisms that govern LLJ development and variability.

In this study, we present the development of a Simple Actuator Disk model for Large-Eddy Simulation (SADLES), implemented within the Weather Research and Forecasting (WRF) model, which is widely used in atmospheric research. The WRF-SADLES model supports both idealized studies and realistic applications through downscaling from realistic data, with a focus on resolutions of tens of meters. Through comparative analysis with the Parallelized Large-eddy Simulation Model (PALM) at resolutions of 10 and 30 m, we validate the effectiveness of WRF-SADLES in simulating the wake characteristics of a 5 MW wind turbine. Results indicate good agreement between WRF-SADLES at 30 m resolution and 10 m resolution and the PALM model. Additionally, we demonstrate a practical case study of WRF-SADLES by downscaling ERA5 reanalysis data using a nesting method to simulate turbine wakes at the Alpha Ventus wind farm in the south of the North Sea. The meso-to-micro downscaling simulation reveals that the wake effect simulated by WRF-SADLES at the FINO1 offshore meteorological mast station aligns well with the cup anemometer and lidar measurements. Furthermore, we investigate an event of farm-to-farm interaction, observing a 16 % reduction in ambient wind speed and a 38 % decrease in average turbine power at Alpha Ventus due to the presence of a wind farm to the southwest. WRF-SADLES offers a promising balance between computational efficiency and accuracy for wind turbine wake simulations, making it valuable for wind energy assessments and wind farm planning.

One of key factors in simulating offshore wind field is con- sidering the effects of waves on momentum and energy ex- changes between the atmosphere and ocean. However, most NWP models traditionally overlook the effects of waves, due to the lack of suitable roughness length parameterization. To address this issue, coupling atmospheric and wave mod- els has emerged which handles the wind-wave interaction through the exchange of a few quantities between two mod- els. On the other hand, integrating atmospheric and wave models requires substantial computational resources and as- sociated expenses. In this study, we developed an offline wave coupling method in the WRF to account for wave ef- fects on wind field simulations. With this method, the WRF is modified to read wave parameters and bathymetry data, and to parameterize roughness length using these data. Our study aims to evaluate the offline method by comparing it to an online coupling system that uses the coupled atmosphere and wave models, as well as observation data. We found that the offline method utilizes almost 17% fewer computational resources and 19% less computational time compared to the online method, while maintaining identical forecasting accu- racy. Evaluation of the results during dominant swell con- dition suggested that both methods are able to reduce the forecast uncertainty compared to LiDAR measurement spe- cially in the lower part of atmosphere.

The development of wind energy in recent years has made the representation of wind turbines in numerical models essential for wind resource and environmental assessment. In this study, we introduce a new wind turbine model, the Simple Actuator Disk for Large Eddy Simulation (SADLES), parameterized in the Weather Research and Forecast (WRF) model. Our goal is to perform realistic downscaling of large eddy simulation, so we integrated SADLES and the cell perturbation method as WRF modules and released the code as open-source for further research. We demonstrate the effectiveness of SADLES by comparing a large eddy simulation of a 5-MW wind turbine using the WRF-SADLES with the Parallelized Large-eddy simulation Model (PALM), showing that SADLES can simulate wind turbine wakes effectively at an intermediate resolution of a few dozen meters. Furthermore, we show the applicability of SADLES in wind farm assessment by downscaling the ERA5 reanalysis data to study turbine-to-turbine and farm-to-farm interactions for the Alpha Ventus wind farm. In our case study, due to the wind farm to the southwest, the ambient wind speed at Alpha Ventus is reduced by 14 % and the average turbine power is reduced by 35 %. SADLES provides a promising balance between computational efficiency and accuracy for wind turbine wake simulations and has potential applications in wind energy assessment and wind farm construction planning.

The development of wind energy in recent years has made the representation of wind turbines in numerical models essential for wind resource and environmental assessment. In this study, we introduce a new wind turbine model, the Simple Actuator Disk for Large Eddy Simulation (SADLES), parameterized in the Weather Research and Forecast (WRF) model. Our goal is to perform realistic downscaling of large eddy simulation, so we integrated SADLES and the cell perturbation method as WRF modules and released the code as open-source for further research. We demonstrate the effectiveness of SADLES by comparing a large eddy simulation of a 5-MW wind turbine using the WRF-SADLES with the Parallelized Large-eddy simulation Model (PALM), showing that SADLES can simulate wind turbine wakes effectively at an intermediate resolution of a few dozen meters. Furthermore, we show the applicability of SADLES in wind farm assessment by downscaling the ERA5 reanalysis data to study turbine-to-turbine and farm-to-farm interactions for the Alpha Ventus wind farm. In our case study, due to the wind farm to the southwest, the ambient wind speed at Alpha Ventus is reduced by 14 % and the average turbine power is reduced by 35 %. SADLES provides a promising balance between computational efficiency and accuracy for wind turbine wake simulations and has potential applications in wind energy assessment and wind farm construction planning.

The development of wind energy in recent years has made the representation of wind turbines in numerical models essential for wind resource and environmental assessment. In this study, we introduce a new wind turbine model, the Simple Actuator Disk for Large Eddy Simulation (SADLES), parameterized in the Weather Research and Forecast (WRF) model. Our goal is to perform realistic downscaling of large eddy simulation, so we integrated SADLES and the cell perturbation method as WRF modules and released the code as open-source for further research. We demonstrate the effectiveness of SADLES by comparing a large eddy simulation of a 5-MW wind turbine using the WRF-SADLES with the Parallelized Large-eddy simulation Model (PALM), showing that SADLES can simulate wind turbine wakes effectively at an intermediate resolution of a few dozen meters. Furthermore, we show the applicability of SADLES in wind farm assessment by downscaling the ERA5 reanalysis data to study turbine-to-turbine and farm-to-farm interactions for the Alpha Ventus wind farm. In our case study, due to the wind farm to the southwest, the ambient wind speed at Alpha Ventus is reduced by 14 % and the average turbine power is reduced by 35 %. SADLES provides a promising balance between computational efficiency and accuracy for wind turbine wake simulations and has potential applications in wind energy assessment and wind farm construction planning.