Wake Meandering Model Research Articles

Wake position tracking using dynamic wake meandering model and rotor loads

The wake position is crucial feedback information for closed loop wind farm control. This paper presents a wake center position tracking approach based on turbine rotor loads, which is feasible to track the wake position while the wake is impinging on the downstream wind turbine. The rotor equivalent wind characteristics are estimated based on the blade root bending moments. Furthermore, a wake-wind model is developed to describe the relationship between rotor averaged wind characteristics and wake position, in which the velocity deficit profile of the wake is defined by the Gaussian model. An adaptive extended Kalman filter is proposed to estimate the lateral and vertical wake position states, in which wake dynamics is described by the dynamic wake meandering model. The wake meandering is driven by a colored noise with a covariance matrix that is gain scheduled according to the rotor averaged wind characteristic. Uncertainties of wake position detection are discussed in various combinations of the ambient atmospheric condition. Simulation results show that the proposed tracking method achieves good estimates in low ambient turbulence intensity, while, for a wind condition with high turbulence intensity, relatively longer time averaging needs to be applied to smooth out short-term fluctuations and highlight longer-term wake trends.

FAST.Farm development and validation of structural load prediction against large eddy simulations

AbstractFAST.Farm is a mid‐fidelity engineering tool developed by the National Renewable Energy Laboratory targeted at accurately and efficiently predicting wind turbine power production and structural loading in wind farm settings, including wake interactions between turbines. FAST.Farm is based on several principles of the dynamic wake meandering (DWM) model, but also addresses limitations of previous DWM implementations. Previous FAST.Farm studies have shown the similarities and differences between FAST.Farm and large eddy simulations for rigid turbine cases. The objective of this work is to quantify the ability of FAST.Farm to accurately predict turbine structural response in a small wind farm. This is done by comparing FAST.Farm structural response to SOWFA‐OpenFAST results for three laterally‐aligned turbines. The purpose of this study is to characterize the similarities and differences between FAST.Farm and a higher fidelity model for predicting turbine structural response in a wind farm for differing atmospheric inflows. Strong statistical agreement was found between FAST.Farm and SOWFA‐OpenFAST structural response for the non‐waked upstream turbine, and good agreement was found for the downstream turbines for most structural quantities. Higher differences were seen for downstream turbines with low ambient turbulence intensity or yawed turbines, suggesting areas for FAST.Farm wake dynamics modeling improvements. For all cases and turbines, small statistical differences were seen between blade deflections and bending moments, with larger differences for tower‐top and tower‐base bending moments. Overall, the results establish confidence for applying FAST.Farm to wind farm power and loads analyses and identify areas where further model validations and model improvements should be targeted.

Towards the development of a wake meandering model based on neural networks

In this work, we develop a neural network model for predicting the instantaneous wake position, which is crucial for a wake meandering model. The data used for training are from the large-eddy simulation of a utility-scale wind turbine. A neural network of four hidden layers with 128 units for each layer is found to be effective when training the model. Effects of different input features on the accuracy of the trained model are systematically tested. It is found that the input features including the downwind and crosswind velocities at two locations upwind of the turbine and the thrust and torque acting on the turbine are enough to guarantee the accuracy of the trained model. Without using the thrust and torque as the input features, the accuracy of the model is significantly worse.

Extension of the DWM model towards a static model for site-specific load simulations

The outlined analysis presents a method to extend the Dynamic Wake Meandering (DWM) model to allow a combination with load approximation methods, which are commonly used to estimate the loads during the planning process of a new wind farm layout. Determining a wind farm layout is a highly iterative process, where time-consuming calculations are avoided as much as possible. The DWM model is methodically designed such that it delivers an inhomogeneous wind field, which makes it useless for further application in load approximation tools that generally require a single value for the turbulence intensity. Furthermore, the DWM model is developed in a way that turbulent wind fields need to be generated to define the ambient turbulence as well as the meandering of the wind speed deficit itself. Likewise, this time-consuming process of calculating a turbulent wind field is in strong contrast with iterative wind farm layout optimization processes. Therefore, the developed extension of the model uses a probability density function to describe the meandering and requires no wind field simulations. The static DWM model is finally compared to the commonly used Frandsen turbulence model as well as to the original DWM model and produces less conservative fatigue loads than the Frandsen model.

Effect of varying fidelity turbine models on wake loss prediction

Wind farm simulations are widely used in estimating energy yield and for optimal wind farm layout design. In the early design stages of a wind farm, low fidelity wind turbine models are used to estimate the farm power output, often due to incomplete knowledge of the turbine characteristics and the additional complexities. The discrepancies introduced in a wind farm simulation as a result of using low fidelity models can often be overlooked, leading to a misrepresentation of a wind farm’s yield. In this paper, the discrepancies between five levels of fidelity for two turbine designs are quantified, focusing on the produced wake and the downstream flow effects. Two high fidelity aeroelastic turbine models and three low fidelity models are described, where the wake is produced using the Dynamic Wake Meandering (DWM) model for each turbine and model. The results provide insight into the expected uncertainties in wake simulations as a result of changing the turbine fidelity level.

Validation of FAST.Farm Against Full-Scale Turbine SCADA Data for a Small Wind Farm

FAST.Farm is a new midfidelity engineering tool developed by the National Renewable Energy Laboratory targeted at accurately and efficiently predicting wind turbine power production and structural loading in wind farm settings, including wake interactions between turbines. FAST.Farm is based on some of the principles of the Dynamic Wake Meandering model—including passive tracer modeling of wake meandering—but addresses many of the limitations of previous Dynamic Wake Meandering (DWM) implementations. Previous FAST.Farm verification studies show the similarities and differences between FAST.Farm and large-eddy simulations for rigid and flexible turbines. In this validation study, FAST.Farm turbine responses are compared to multiturbine measurements from a subset of a full-scale wind farm. FAST.Farm predictions of turbine generator power, rotor speed, and blade pitch for five-turbine simulations are compared to supervisory control and data acquisition results. Results reveal that FAST.Farm generator power mean and standard deviation results reasonably match measured data for upstream and downstream turbines, as well as the mean rotor speed and blade pitch above rated wind speeds. However, FAST.Farm generally underpredicts the mean rotor speed and overpredicts the mean blade pitch below rated operation. These errors are likely related to inaccuracies in the generic controller simulated. Despite controller differences, FAST.Farm predicts the same overall relative rotor power trends for all waked turbines at all wind speeds.

Dynamic wake meandering model calibration using nacelle-mounted lidar systems

Abstract. Light detection and ranging (lidar) systems have gained a great importance in today's wake characteristic measurements. The aim of this measurement campaign is to track the wake meandering and in a further step to validate the wind speed deficit in the meandering frame of reference (MFR) and in the fixed frame of reference using nacelle-mounted lidar measurements. Additionally, a comparison of the measured and the modeled wake degradation in the MFR was conducted. The simulations were done with two different versions of the dynamic wake meandering (DWM) model. These versions differ only in the description of the quasi-steady wake deficit. Based on the findings from the lidar measurements, the impact of the ambient turbulence intensity on the eddy viscosity definition in the quasi-steady deficit has been investigated and, subsequently, an improved correlation function has been determined, resulting in very good conformity between the new model and the measurements.

T2FL: An Efficient Model for Wind Turbine Fatigue Damage Prediction for the Two-Turbine Case

Wind farm load assessment is typically conducted using Computational Fluid Dynamics (CFD) or aeroelastic simulations, which need a lot of computer power. A number of applications, for example wind farm layout optimisation, turbine lifetime estimation and wind farm control, requires a simplified but sufficiently detailed model for computing the turbine fatigue load. In addition, the effect of turbine curtailment is particularly important in the calculation of the turbine loads. Therefore, this paper develops a fast and computationally efficient method for wind turbine load assessment in a wind farm, including the wake effects. In particular, the turbine fatigue loads are computed using a surrogate model that is based on the turbine operating condition, for example, power set-point and turbine location, and the ambient wind inflow information. The Turbine to Farm Loads (T2FL) surrogate model is constructed based on a set of high fidelity aeroelastic simulations, including the Dynamic Wake Meandering model and an artificial neural network that uses the Bayesian Regularisation (BR) and Levenberg–Marquardt (LM) algorithms. An ensemble model is used that outperforms model predictions of the BR and LM algorithms independently. Furthermore, a case study of a two turbine wind farm is demonstrated, where the turbine power set-point and fatigue loads can be optimised based on the proposed surrogate model. The results show that the downstream turbine producing more power than the upstream turbine is favourable for minimising the load. In addition, simulation results further demonstrate that the accumulated fatigue damage of turbines can be effectively distributed amongst the turbines in a wind farm using the power curtailment and the proposed surrogate model.

The effect of minimum thrust coefficient control strategy on power output and loads of a wind farm

The use of down-regulation control strategies for wind turbines potentially optimises the wind farm performance. One of the promising strategies to minimise the wind farm loads and maximise the energy production is by down-regulating the upstream turbine with minimum thrust coefficient (Ct ). Nonetheless, a majority of minimum Ct control studies is confined to fatigue load analysis of a single turbine. Therefore, this paper focuses on the investigation of the fatigue loads when operating turbines in a wind farm using a minimum Ct control strategy, as well as extreme loads and energy production. The control strategy is implemented in the DTU Wind Energy controller coupled with the aero-elastic code, HAWC2. The Dynamic Wake Meandering model in HAWC2 is used so that the wake deficits generated from the upstream turbines realistically represent the minimum Ct control strategy, as well as the interactions between wakes. The main contribution of this paper is to investigate the possible wind farm operating strategy using minimum Ct control that leads to decrease the lifetime fatigue loads and increase the overall energy production of a wind farm due to the potential extended lifetime of certain wind turbine components.

Effect of Wake Meandering on Aeroelastic Response of a Wind Turbine Placed in a Park

A wind turbine operating inside a wind farm is subjected to increased turbulence intensity and reduced wind speeds resulting in increased fatigue loadings and reduced power production. Furthermore, meandering of the wakes results in increased dynamic loading of the wind turbine. In the present study, a standalone dynamic wake meandering (DWM) model has been developed and implemented in commercial code SIMA. The standalone DIWA model does not require a direct coupling with aeroelastic code, hence it is computationally fast. Although the standalone tool is a good alternative for power and thrust prediction, it does not have the capability to predict the turbine aeroelastic loads. The new DWM program is referred as “Disturbed Inflow Wind Analyzer” (DIWA). Benchmarking studies of DIWA with the literature data are presented and discussed. Overall the DIWA compares well with the literature data and the discrepancies between DIWA and the literature data are discussed. The present studies show that the wake deficit profiles are very sensitive to the eddy viscosity parameters. Finally, the turbulence boxes generated using DIWA have been used for understanding the aeroelastic behaviour of NREL 5MW turbine and one of the wind turbines from the Lillgrund wind park. The estimated power production using both aeroelastic coupled with DIWA turbulence boxes and standalone DIWA (without aeroelastic) are in good agreement with the literature data. The trends of fatigue loads are predicted well, with a few exceptions.

Surrogate models for parameterized representation of wake‐induced loads in wind farms

AbstractA procedure for mapping wake‐induced load predictions computed with the dynamic wake meandering model to a computationally efficient surrogate model approximation is defined and demonstrated. Using the mapping function, the load variation can be efficiently estimated for a wind farm with arbitrary layout. The resulting load assessment procedure provides continuous, differentiable output with known analytical derivatives and can be used for applications such as wind turbine layout optimization, estimation of turbine lifetime, and uncertainty analysis.

Effects of Inflow Spatiotemporal Discretization on Wake Meandering and Turbine Structural Response using FAST.Farm

FAST.Farm is a newly developed multiphysics, midfidelity engineering tool that can be used to predict turbine power and structural loads of wind turbines in a wind farm with minimal computational expense. Previous studies have shown similarities and differences in wind turbine performance between FAST.Farm and high-fidelity large-eddy simulations (LES) using both LES precursor-generated inflow and TurbSim-generated synthetic inflow. While conservative resolutions have been used to date, no formal spatial or temporal discretization study has been performed for the FAST.Farm model wind domains. This work aims to study the effects of varying the spatial and temporal discretization of the wind domains on wake meandering (low-resolution domain) and turbine structural response (high-resolution domains) and resulting wake and load calculations. The purpose of this study is to establish convergence criteria and recommendations for discretization values in terms of rotor diameter (D, expressed in meters) that will maximize computational efficiency. To ensure a percent error of ≤ 1% in standard deviation of wake center position relative to the finest included resolution, a low-resolution time step of 0.024D s, 0.016D s, and 0.0079D s and a low-resolution spatial discretization of 0.079D m, 0.16D m, and 0.24D m is recommended for 8 m/s, 12 m/s, and 18 m/s mean ambient wind speed at hub height, respectively. These guidelines are likely applicable to other implementations of the dynamic wake meandering model. To ensure a percent error of ≤ 1% in standard deviation of all considered structural response outputs, a high-resolution time step that captures the highest influential excitation and natural frequencies in the system and a high-resolution spatial discretization comparable to the maximum airfoil cord length is recommended. These guidelines are likely applicable to any aeroelastic analysis.

Optimal yaw strategy for optimized power and load in various wake situations

The interaction between nearby wind turbines in a wind farm modifies the power and loads compared to their stand-alone values. The increased turbulence intensity and the modified turbulence structure at the downstream turbines creates higher fatigue loading, which can be mitigated by wind farm and/or wind turbine control. To alleviate loads and maximize power possible strategies such as wake steering, where the turbine is yawed to redirect the wake such that it does not impinge the downstream turbine, have been studied. The work presented here focuses on situations where the wake is nevertheless affecting the downstream turbine, and more specifically how high loads can be avoided by yawing the wake-affected turbine. The analysis is conducted on a 2.3 MW machine, and the flow field is simulated using the Dynamic Wake Meandering model. The study investigates the impact on power and loads for different longitudinal interspacing and turbulence intensities. Optimal yaw strategies are defined for above rated regions where no power loss occurs. The potential load alleviation for different load sensors are studied, but the presentation is focussed on the blade root flapwise fatigue loading. For full wake at 3D interspacing and turbulence intensity of 5 %, around 35 % of load reduction on the 1 Hz Damage Equivalent Loads can be achieved at high wind speeds. Smaller reductions are achieved for higher atmospheric turbulence; the analogue case with 15 % turbulence intensity shows 17 % potential alleviation. The alleviation on the wind turbine lifetime is also calculated and compared for different turbulence intensities and mean wind speeds. Small reductions are achieved for sites with low mean wind speed and high turbulence intensity, but high reductions, of around 19 %, are accomplished in low turbulence intensity with high mean wind speed.

Engineering an optimal wind farm using surrogate models

AbstractA framework for optimal design of wind farm layouts using a surrogate‐based Dynamic Wake Meandering model is presented. The optimization platform is set‐up as a hybrid strategy combining genetic search with the gradient‐based algorithm. The design variables are the number of turbines in the layout and their relative position within the bounded area. The objective function is defined as the net present value of the wind farm's profit, thus including the relevant expenditures throughout the farm's lifespan. Results show that an optimal design is reached by maximizing investment and accepting a minor sacrifice of the wind farm performance.

Validation of wind turbine wake models with focus on the dynamic wake meandering model

This analysis compares current wake models, such as the Sten Frandsen turbulence model, an engineering wind speed deficit and turbulence model developed by G. Chr. Larsen, the N. O. Jensen wind speed deficit description and three variations of the dynamic wake meandering (DWM) model to measurements from onshore wind farms. Special attention is given to the dynamic wake meandering model. Today different versions or implementations of the dynamic wake meandering model exist. These versions differ among others in the calculation of the quasi-steady wake deficit in the meandering frame of reference. The influence of these calculation methods on loads and yield is analysed in this work. The validation of the mentioned wake models is based on turbine load, power and wind measurements from two onshore wind farms. One key result of this work is that the DWM model in most cases coincides superiorly with the measurement results compared to the other engineering wake models commonly used in industry.

Improved modelling of fatigue loads in wind farms under non-neutral ABL stability conditions

The purpose of this study is improve the predictive capability of the Dynamic Wake Meandering (DWM) model generalized to non-neutral atmospheric boundary layer (ABL) conditions in general and under stable ABL stratification in particular. The emphasis is on rotating wind turbine components, and the model improvement in focus is intimately linked to a newly developed refinement of the classic Monin-Obukhov theory, which, for stable ABL stratification, primary results in less pronounced mean wind shear outside the surface layer, where most modern wind turbines are operating. The model improvements are validated against a huge set of full-scale data, which allows for a one-to-one comparison of wind turbine load simulations and measurements conditioned on ABL stability conditions.

Medium fidelity modelling of loads in wind farms under non-neutral ABL stability conditions – a full-scale validation study

The aim of the present paper is to demonstrate the capability of medium fidelity modelling of wind turbine component fatigue loading, when the wind turbines are subjected to wake affected non-stationary flow fields under non-neutral atmospheric stability conditions. To accomplish this we combine the classical Dynamic Wake Meandering model with a fundamental conjecture stating: Atmospheric boundary layer stability affects primary wake meandering dynamics driven by large turbulent scales, whereas wake expansion in the meandering frame of reference is hardly affected. Inclusion of stability (i.e. buoyancy) in description of both large- and small scale atmospheric boundary layer turbulence is facilitated by a generalization of the classical Mann spectral tensor, which consistently includes buoyancy effects. With non-stationary wind turbine inflow fields modelled as described above, fatigue loads are obtained using the state-of-the art aeroelastic model HAWC2.The Lillgrund offshore wind farm (WF) constitute an interesting case study for wind farm model validation, because the WT interspacing is small, which in turn means that wake effects are significant. A huge data set, comprising 5 years of blade and tower load recordings, is available for model validation. For a multitude of wake situations this data set displays a considerable scatter, which to a large degree seems to be caused by atmospheric boundary layer stability effects. Notable is also that rotating wind turbine components predominantly experience high fatigue loading for stable stratification with significant shear, whereas high fatigue loading of non-rotating wind turbine components are associated with unstable atmospheric boundary layer stratification.

Wind farm wake modeling using NWTC design codes

Wind turbines located in a wind farm are subject to a wind field that is substantially modified when compared to the ambient wind field due to wake effects. It is clear that in addition to the power, a tool that could model the loads of the waked turbine is needed. In this work, a wind farm wake model is created using a single wake model based on the dynamic wake meandering model to systematically model the wake effects and thus the power and loads of an entire wind farm with an arbitrary wind turbine layout and wind condition. This wind farm wake model is incorporated into the National Wind Technology Center design codes. Preliminary results demonstrate that this wind farm tool yields satisfactory results when compared with Simulator for Wind Farm Applications, a high fidelity computational fluid dynamics Large eddy simulation model, and with experimentally-obtained field data. The integration of the wind farm model with the National Wind Technology Center design codes is described in detail and validation results of the tool are provided.

Estimation of annual energy production using dynamic wake meandering in combination with ambient CFD solutions

In the present study, combination of the standalone dynamic wake meandering (DWM) model with Reynolds-averaged Navier-Stokes (RANS) CFD solutions for ambient ABL flows is introduced, and its predictive performance for annual energy production (AEP) is evaluated against Vestas’ SCADA data for six operating wind farms over semi-complex terrains under neutral conditions. The performances of conventional linear and quadratic wake superposition techniques are also compared, together with the in-house implemention of successive hierarchical merging approaches. As compared to our standard procedure based on the Jensen model in WindPRO, the overall results are promising, leading to a significant improvement in AEP accuracy for four of the six sites. While the conventional linear superposition shows the best performance for the improved four sites, the hierarchical square superposition shows the least deteriorated result for the other two sites.

Multi-model predictive control for wind turbine operation under meandering wake of upstream turbines

In wind farm operation, the performance and loads of downstream turbines are heavily influenced by the wake of the upstream turbines. Furthermore, the actual wake is more challenging due to the dynamic phenomenon of wake meandering, i.e. the turbine wake often demonstrates dynamic shift over time. To deal with the time-varying characteristics of wake meandering, a multiple model predictive control (MMPC) scheme is applied to the individual pitch control (IPC) based load reduction. The coherence function in the spectral method is used to generate the stochastic wind profile including wake meandering at upstream turbine, and a simplified wake meandering model is developed to emulate the trajectory of the wake center at downstream turbine. The Larsen wake model and Gaussian distribution of wake deficit are applied for composing wind profiles across the rotor of downstream turbines. A set of MMPC controllers are designed based on different linearized state-space models, and are applied in a smooth switching manner. Simulation results show significant reduction in the variation of both rotor speed and blade-root flapwise bending moment using the MMPC based IPC by including the wake meandering, as compared to a benchmark PI controller designed by NREL.