Energy losses due to wind farm clustering and wind farm interaction are rarely well represented in the wind farm design process because of the lack of fast models that can accurately account for neighboring wind farm wakes. A recently developed solution is the actuator wind farm (AWF) model, which is a Reynolds-averaged Navier-stokes (RANS) based wind farm parametrization that models a wind farm as a distributed thrust force and applies a global wind farm thrust coefficient controller. We propose an improved version of the AWF model, where each turbine employs a local thrust force controller and uses turbine thrust and power coefficients as input to better handle inhomogeneous inflow conditions. The proposed AWF model shows improved performance compared to the original AWF model in terms of predicted wind turbine power of a downstream wind farm that operates in a partial wake of an upstream wind farm, without significantly increasing the computational effort. However, the annual energy production (AEP) wake losses of a large wind farm cluster are nearly unaffected by using local or global control and input because the largest impact is found near the cut-in wind speed, which does not contribute much to the AEP wake losses.

Artificial Neural Networks (ANNs) are being applied as a faster alternative to Computational Fluid Dynamics (CFD) for wind turbine engineering wake models. Unfortunately, ANNs can fail to generalize if the data is insufficient. Physics-Informed Neural Networks (PINNs) can improve convergence while lowering the required data amounts. This paper investigates the PINN methodology systematically by considering varying amounts of data and physics collocation points. This work considers the rotationally symmetric Reynolds Averaged Navier-Stokes (RANS) formulation. Initially, a baseline fully data-driven ANN is studied to determine a suitable network size. Then, multiple PINN-based wake surrogates are trained with continuity and momentum conservation knowledge, varying amounts of data, and physics collocation points. It was found that including physics information under the best circumstances could improve accuracy by 18% at the cost of increasing the training time by a factor of 116. The findings imply that physics information can improve neural network based wake surrogates.

Inside wind farms, wake effects are the primary source of turbine interactions, and as such, they constitute one of the most important aspects of wind farm operations. The two most widespread methods for calculating the wind farm wake flow are computational fluid dynamics (CFD) methods and engineering models. Both methods have drawbacks; CFD methods can be very accurate but are computationally expensive. Vice versa, engineering models sacrifice accuracy by simplifying the physics, thereby improving computational efficiency. In cases where many evaluations of the flow are needed, this trade-off is a hindrance. One such case is wind farm layout optimization problems. It has been shown that the estimation of wake flows can be improved by surrogate modeling. Recently, Artificial Neural Networks (ANN) have been demonstrated to predict accurate wakes over a mesh. In this work, a new mesh-free ANN-based wake model is proposed. This new model can predict the flow everywhere in the domain, and as it employs smooth activation functions, it is suitable for gradient-based optimization. Two ANNs were trained with data generated by Reynolds-Averaged Navier-Stokes with Actuator Disc simulations for several yaw angles. The first ANN predicts streamwise wake velocity induction/deficit, the second ANN predicts added wake turbulence intensity. Both ANNs predict with a low error.

A new wake surrogate model based on Reynolds-averaged Navier-Stokes (RANS) single rotor simulations is presented. The model relies on a series of three-dimensional pre-calculated deficit and added turbulence intensity flow fields, stored in a look-up table (LUT) as a function of the thrust coefficient and the ambient turbulence intensity. For any combination of these parameters, the flow around a wind turbine can be predicted by linearly interpolating within the look-up table. Furthermore, the resulting three-dimensional flow fields from different turbine sources can be superposed linearly to calculate the total wind farm flow. The model is implemented in PyWake and benchmarked against other, commonly employed engineering wake models, namely the Gaussian-Bastankhah, the N. O. Jensen and the Zong models, where RANS wind farm simulations are used as reference. In both full and partial wake cases, the surrogate model achieves a higher accuracy than any other model. Besides providing an accuracy comparable to a full RANS solution, the model can compute a flow case in the order of 1 s on a single processor. The main disadvantage is that the generation of the look-up table is time consuming, computationally expensive and can be memory demanding (especially if more inputs, such as the yaw misalignment angle, stability, etc. are added). Nevertheless, generating the LUT only has to be done once per wind turbine type.

Wind farm Annual Energy Production (AEP) calculations are required to design energy efficient wind farm layouts. We investigate methods that can reduce the computational effort of AEP calculations using Reynolds-averaged Navier-Stokes simulations of an idealized atmospheric wind farm setup. In addition, we introduce a generalized wind turbine model that compares well with wind turbine aerodynamic data covering a large range of wind turbine sizes. We apply the general wind turbine model to reduce the computational effort of the AEP calculations by decreasing the number of independent wind speed flow cases. Furthermore, we apply Reynolds-number similarity to compute the wind speed flow cases faster and we show how wind farm layout mirror- and rotational-symmetry can reduce the number of independent wind direction flow cases.

The article presents an approach to combine wake models of multiple levels of fidelity, which is capable of giving accurate predictions with only a small number of high fidelity samples. The G. C. Larsen and k-ε-fP based RANS models are adopted as ensemble members of low fidelity and high fidelity models, respectively. Both the univariate and multivariate based surrogate models are established by taking the local wind speed and wind direction as variables of the wind farm power efficiency function. Various multi-fidelity surrogate models are compared and different sampling schemes are discussed. The analysis shows that the multi-fidelity wake models could tremendously reduce the high fidelity model evaluations needed in building an accurate surrogate.

A research agenda is described to further encourage the application of Multidisciplinary Design Analysis and Optimisation (MDAO) methodologies to wind energy systems. As a group of researchers closely collaborating within the International Energy Agency (IEA) Wind Task 37 for Wind Energy Systems Engineering: Integrated Research, Design and Development, we have identified challenges that will be encountered by users building an MDAO framework. This roadmap comprises 17 research questions and activities recognised to belong to three research directions: model fidelity, system scope and workflow architecture. It is foreseen that sensible answers to all these questions will enable to more easily apply MDAO in the wind energy domain. Beyond the agenda, this work also promotes the use of systems engineering to design, analyse and optimise wind turbines and wind farms, to complement existing compartmentalised research and design paradigms.

This paper describes a consistent algorithm for eliminating the numerical wiggles appearing when solving the finite volume discretized Navier-Stokes equations with discrete body forces in a collocated grid arrangement. The proposed method is a modification of the Rhie-Chow algorithm where the force in a cell is spread on neighboring cells by applying equivalent pressure jumps at the cell faces. The method shows excellent results when applied for simulating the flow through an actuator disk, which is relevant for wind turbine wake simulations. (c) 2015 Elsevier Ltd. All rights reserved.

SCADA data, recorded on the downstream wind farm, has been used to identify flow cases with visible clustering effects. The inflow condition is derived from a partly undisturbed wind turbine, due to lack of mast measurements. The SCADA data analysis concludes that centre of the deficit for the downstream wind farm with disturbed inflow has a distinct visible maximum deficit zone located only 5-10D downstream from the entrance. This zone, representing 20-30% speed reduction, increases and moves downstream for increasing cluster effect and is not visible outside a flow sector of 20-30°. The eight flow models represented in this benchmark include both RANS models, mesoscale models and engineering models. The flow cases, identified according to the wind speed level and inflow sector, have been simulated and validated with the SCADA results. The model validation concludes that all models more or less are able to predict the location and size of the deficit zone inside the downwind wind farm.

Wind farm flow models have advanced considerably with the use of large eddy simulations (LES) and Reynolds averaged Navier-Stokes (RANS) computations. The main limitation of these techniques is their high computational time requirements; which makes their use for wind farm annual energy production (AEP) predictions expensive. The objective of the present paper is to minimize the number of model evaluations required to capture the wind power plant's AEP using stationary wind farm flow models. Polynomial chaos techniques are proposed based on arbitrary Weibull distributed wind speed and Von Misses distributed wind direction. The correlation between wind direction and wind speed are captured by defining Weibull-parameters as functions of wind direction. In order to evaluate the accuracy of these methods the expectation and variance of the wind farm power distributions are compared against the traditional binning method with trapezoidal and Simpson's integration rules. The wind farm flow model used in this study is the semi-empirical wake model developed by Larsen [1]. Three test cases are studied: a single turbine, a simple and a real offshore wind power plant. A reduced number of model evaluations for a general wind power plant is proposed based on the convergence of the present method for each case.

A Reynolds-averaged Navier-Stokes code is used to simulate the interaction of two neighboring wind farms. The influence of the Coriolis force is investigated by modeling the atmospheric surface/boundary layer with three different methodologies. The results show that the Coriolis force is negligible for a single wind turbine, small for a single wind farm, but important for simulations of wind farm wake interaction.

ABSTRACTFor wind resource assessment, the wind industry is increasingly relying on computational fluid dynamics models of the neutrally stratified surface‐layer. So far, physical processes that are important to the whole atmospheric boundary‐layer, such as the Coriolis effect, buoyancy forces and heat transport, are mostly ignored. In order to decrease the uncertainty of wind resource assessment, the present work focuses on atmospheric flows that include stability and Coriolis effects. The influence of these effects on the whole atmospheric boundary‐layer are examined using a Reynolds‐averaged Navier–Stokes k‐ ε model. To validate the model implementations, results are compared against measurements from several large‐scale field campaigns, wind tunnel experiments, and previous simulations and are shown to significantly improve the predictions. Copyright © 2013 John Wiley & Sons, Ltd.

A detailed study of the device angle dependency of a single vortex generator (VG) is presented in this paper. A single Vortex Generator on a test section wall case, with four different positions of the device angle to the incoming flow, has been designed and solved by computational methods. The computational fluid dynamic (CFD) simulations have been com- pared with a wind tunnel experiment, where the corresponding parametric study was per- formed over a single vane mounted on the test section wall in low-speed wind tunnel. In this experiment the flow was recorded using Stereoscopic Particle Image Velocimetry (S-PIV) in cross-planes at various positions downstream of the vane. The main goal of this article is to study the angle dependency of a single VG mounted on a test section wall; for this purpose CFD simulations have been carried out and compared with a wind tunnel experiment and an analytical model.

A new integrated design tool for optimization of offshore wind farm clusters is under development in the European Energy Research Alliance - Design Tools for Offshore wind farm Cluster project (EERA DTOC). The project builds on already established design tools from the project partners and possibly third-party models. Wake models have been benchmarked on the Horns Rev and, currently, on the Lilgrund wind farm test cases. Dedicated experiments from 'BARD Offshore 1' wind farm will using scanning lidars will produce new data for the validation of wake models. Furthermore, the project includes power plant interconnection and energy yield models all interrelated with a simplified cost model for the evaluation of layout scenarios. The overall aim is to produce an efficient, easy to use and flexible tool - to facilitate the optimised design of individual and clusters of offshore wind farms. A demonstration phase at the end of the project will assess the value of the integrated design tool with the help of potential end-users from industry.This abstracts summarizes the objectives and preliminary results of work package 3. In order to provide an accurate value of the expected net energy yield, the offshore wind resource assessment process has been reviewed as well as the sources of uncertainty associated to each step.Methodologies for the assessment of offshore gross annual energy production are analyzed based on the Fino 1 test case. Measured data and virtual data from Numerical Weather Prediction models have been used to calculate long term wind speed, wind profile and gross energy. (C) 2014 Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

ABSTRACTIn the present paper, Reynolds‐averaged Navier–Stokes predictions of the flow field around the MEXICO rotor in yawed conditions are compared with measurements. The paper illustrates the high degree of qualitative and quantitative agreement that can be obtained for this highly unsteady flow situation, by comparing measured and computed velocity profiles for all three Cartesian velocity components along four axial transects and several radial transects. Copyright © 2012 John Wiley & Sons, Ltd.

Accurately quantifying wind turbine wakes is a key aspect of wind farm economics in large wind farms. This paper introduces a new simulation post-processing method to address the wind direction uncertainty present in the measurements of the Horns Rev offshore wind farm. This new technique replaces the traditional simulations performed with the 10 min average wind direction by a weighted average of several simulations covering a wide span of directions. The weights are based on a normal distribution to account for the uncertainty from the yaw misalignment of the reference turbine, the spatial variability of the wind direction inside the wind farm and the variability of the wind direction within the averaging period. The results show that the technique corrects the predictions of the models when the simulations and data are averaged over narrow wind direction sectors. In addition, the agreement of the shape of the power deficit in a single wake situation is improved. The robustness of the method is verified using the Jensen model, the Larsen model and Fuga, which are three different engineering wake models. The results indicate that the discrepancies between the traditional numerical simulations and power production data for narrow wind direction sectors are not caused by an inherent inaccuracy of the current wake models, but rather by the large wind direction uncertainty included in the dataset. The technique can potentially improve wind farm control algorithms and layout optimization because both applications require accurate wake predictions for narrow wind direction sectors. (C) 2013 The Authors. Wind Energy published by John Wiley & Sons, Ltd.

For wind resource assessment, the wind industry is increasingly relying on Computational Fluid Dynamics models that focus on modeling the airflow in a neutrally stratified surface layer. So far, physical processes that are specific to the atmospheric boundary layer, for example the Coriolis force, buoyancy forces and heat transport, are mostly ignored in state-of-the-art flow solvers. In order to decrease the uncertainty of wind resource assessment, the effect of thermal stratification on the atmospheric boundary layer should be included in such models. The present work focuses on non-neutral atmospheric flow over complex terrain including physical processes like stability and Coriolis force. We examine the influence of these effects on the whole atmospheric boundary layer using the DTU Wind Energy flow solver EllipSys3D. To validate the flow solver, measurements from Benakanahalli hill, a field experiment that took place in India in early 2010, are used. The experiment was specifically designed to address the combined effects of stability and Coriolis force over complex terrain, and provides a dataset to validate flow solvers. Including those effects into EllipSys3D significantly improves the predicted flow field when compared against the measurements.

The MEXICO Experiment is reproduced in CFD, including the geometry of the wind tunnel and the wind turbine rotor. The wind turbine is modelled both as a full rotor and as an actuator disc. Various questions regarding the wind tunnel effects on the measurements are investigated. As in a previous work carried out without modelling the wind tunnel, the CFD methods are found to give satisfying agreement with the axial velocity deficit in the wake. However, confirming the previous work, the blade loadings estimated from CFD are found to be consistently larger than the one estimated from measurements. In order to investigate further this issue, the loadings estimated from measurement are used with an actuator disc model. This approach gives a too small wake deficit in comparison with the measurements, which tends to agree with the full rotor computation results.

Bolund measurements were used for a blind comparison of microscale flow models. Fifty-seven models ranging from numerical to physical were used, including large-eddy simulation (LES) models, Reynolds-averaged Navier–Stokes (RANS) models, and linearized models, in addition to wind-tunnel and water-channel experiments. Many assumptions of linearized models were violated when simulating the flow around Bolund. As expected, these models showed large errors. Expectations were higher for LES models. However, of the submitted LES results, all had difficulties in applying the specified boundary conditions and all had large speed-up errors. In contrast, the physical models both managed to apply undisturbed ‘free wind’ boundary conditions and achieve good speed-up results. The most successful models were RANS with two-equation closures. These models gave the lowest errors with respect to speed-up and turbulent kinetic energy (TKE) prediction.