Gunner Chr. Larsen

Comparison of Wake Model Simulations with Offshore Wind Turbine Wake Profiles Measured by Sodar

Abstract This paper gives an evaluation of most of the commonly used models for predicting wind speed decrease (wake) downstream of a wind turbine. The evaluation is based on six experiments where free-stream and wake wind speed profiles were measured using a ship-mounted sodar at a small offshore wind farm. The experiments were conducted at varying distances between 1.7 and 7.4 rotor diameters downstream of the wind turbine. Evaluation of the models compares the predicted and observed velocity deficits at hub height. A new method of evaluation based on determining the cumulative momentum deficit over the profiles is described. Despite the apparent simplicity of the experiments, the models give a wide range of predictions. Overall, it is not possible to establish any of the models as having individually superior performance with respect to the measurements.

Comparison of Wake Models with Data for Offshore Windfarms

A major objective of the ENDOW project is to evaluate the performance of wake models in offshore windfarm environments in order to ascertain the improvements required to enhance the prediction of power output within large offshore wind farms [1]. The strategy for achieving this is to compare the performance of the models in a wide range of conditions which are expected to be encountered during turbine operation offshore. Six models of varying complexity have been evaluated initially against the Vindeby single wake data in [2] where it was found that almost all of them overestimate the wake effects and also significant inconsistencies between the model predictions appeared in the near wake and turbulence intensity results. Based on the conclusions of that study, the ENDOW wake modeling groups have already implemented a number of modifications to their original models. In the present paper, new single wake results are presented against experimental data at Vindeby and Bockstigen wind farms. Clearly, some of the model discrepancies previously observed in Vindeby cases have been smoothed and overall the performance is improved.

Simulation of inhomogeneous, non-stationary and non-Gaussian turbulent winds

Turbulence time series are needed for wind turbine load simulation. The multivariate Fourier simulation method often used for this purpose is extended for inhomogeneous and non-stationary processes of general probability distribution. This includes optional conditional simulation matching simulated series to field measurements at selected points. A probability model for the application of turbine wind loads is discussed, and finally the technique for non-stationary processes is illustrated by turbulence simulation during a front passage.

ENDOW: Improvement of Wake Models within Offshore Wind Farms

The partners in the ENDOW (Efficient Development of Offshore Windfarms) project are validating, testing, designing and improving wind farm design tools for the efficient design of offshore wind farms. The different meteorological conditions offshore constitute a challenge for the current design tools and models because they have been developed and tested primarily for the design of land based wind farms. Measured downwind of turbines, wake-affected wind speed profiles at Vindeby offshore wind farm have been compared with the model predictions for single, double and quintuple wake cases. The modelling groups have based on these results adjusted their wake models for offshore wind farm design. This paper presents the data, model comparisons and the improvements to the models.

Implementation of a Mixing Length Turbulence Formulation Into the Dynamic Wake Meandering Model

The work presented in this paper focuses on improving the description of wake evolution due to turbulent mixing in the dynamic wake meandering (DWM) model. From wake investigations performed with high-fidelity actuator line simulations carried out in ELLIPSYS3D , it is seen that the current DWM description, where the eddy viscosity is assumed to be constant in each cross-section of the wake, is insufficient. Instead, a two-dimensional eddy viscosity formulation is proposed to model the shear layer generated turbulence in the wake, based on the classical mixing length model. The performance of the modified DWM model is verified by comparing the mean wake velocity distribution with a set of ELLIPSYS3D actuator line calculations. The standard error (defined as the standard deviation of the difference between the mean velocity field of the DWM and the actuator line model), in the wake region extending from 3 to 12 diameters behind the rotor, is reduced by 27% by using the new eddy viscosity formulation.

Characterising Turbulence Intensity for Fatigue Load Analysis of Wind Turbines

Turbulence in wind velocity presents a major factor for modern wind turbine design as cost reduction as are sort for the dynamic structures. Therefore this paper contains a parametrisation of the turbulence intensity at given sites, relevant for the calculation of fatigue loading of wind turbines. The parameterisation is based on wind speed measurements extracted from the “Database on Wind Characteristics” (www.winddata.com). The parameterisation is based on the LogNormal distribution, which has proven to be suitable distribution to describe the turbulence intensity distribution.

ENDOW: Efficient Development of Offshore Windfarms

The objective of the ENDOW project is to evaluate, enhance and interface wake and boundary-layer models for utilisation in developing offshore windfarms. The model hierarchy will form the basis of design tools for use by wind energy developers and turbine manufacturers to optimise power output from offshore wind farms through minimised wake effects immediately downwind of wind turbines and optimal grid connections. The initial focus of the project was to use databases from existing offshore wind farms (Vindeby and Bockstigen) for the first comprehensive evaluation of offshore wake model performances. The six wake models vary in complexity from empirical solutions to the most advanced models based on solutions of the Navier-Stokes equations using eddy viscosity or k-epsilon turbulence closure. One of the wake models is also being coupled with a full aeroelastic model for the calculation of wind loads on the turbines. Parallel research includes comparison of a local-scale stability/roughness model with a mesoscale model focusing on boundary-layer development within and over a large offshore wind farm, and particularly the influence of large scale thermal flows. A new experiment was conducted using SODAR immediately downwind of offshore wind turbines to examine vertical wind speed profiles to hub-height and beyond in near-wake conditions and wake dispersion to assist in model development and evaluation.

Dependence of offshore wind turbine fatigue loads on atmospheric stratification

The stratification of the atmospheric boundary layer (ABL) is classified in terms of the M-O length and subsequently used to determine the relationship between ABL stability and the fatigue loads of a wind turbine located inside an offshore wind farm. Recorded equivalent fatigue loads, representing blade-bending and tower bottom bending, are combined with the operational statistics from the instrumented wind turbine as well as with meteorological statistics defining the inflow conditions. Only a part of all possible inflow conditions are covered through the approximately 8200 hours of combined measurements. The fatigue polar has been determined for an (almost) complete 360° inflow sector for both load sensors, representing mean wind speeds below and above rated wind speed, respectively, with the inflow conditions classified into three different stratification regimes: unstable, neutral and stable conditions. In general, impact of ABL stratification is clearly seen on wake affected inflow cases for both blade and tower fatigue loads. However, the character of this dependence varies significantly with the type of inflow conditions – e.g. single wake inflow or multiple wake inflow.

Simulation of low frequency noise from a downwind wind turbine rotor

One of the major drawbacks of a wind turbine with a downwind rotor is the generation of considerable low frequency noise (so-called thumping noise) which can cause annoyance of people at a considerable distance. This was experienced on a number of full-scale turbines in e.g. US and Sweden in the period from around 1980 to 1990. One of the common characteristics of this low frequency noise, emerging from analysis of the phenomenon, was that the sound pressure level is strongly varying in time. We have investigated this phenomenon using a model package by which the low frequency noise of a downwind rotor can be simulated. In order to investigate the importance of wake unsteadiness, time true CFD computations of the flow past a 4 m diameter cylinder were performed at 8 m/s, and the wake characteristics were subsequently read into the aeroelastic code HAWC, which finally gives output to the aero acoustic model. The results for a 5 MW two-bladed turbine with a downwind rotor showed an increase in the sound pressure level of 5-20 dB due to the unsteadiness in the wake caused mainly by vortex shedding. However, in some periods the sound pressure level can increase additionally 0-10 dB when the blades directly pass through the discrete shed vortices behind the tower. The present numerical results strongly confirm the experiences with full scale turbines showing big variations of sound pressure level in time due to the wake unsteadiness, as well as a considerable increase in sound pressure level if the blade passing frequency is close to the Strouhal number controlling the vortex shedding from the tower.

Comparison of methods for load simulation for wind turbines operating in wake

For simulation of load response for a wind turbine operating in wake conditions, two different approaches exist. One method – the equivalent turbulence method – is based on the assumption that all load generating mechanisms causing increased loads in wake operation can be merged into an equivalent value of increased turbulence intensity. This method is specified in the IEC61400-1 safety standard. The other method (the new dynamic wake meandering model) is a recently developed more physical approach, taking into account the transversal and vertical dynamics of the wake (i.e. wake meandering). The objective of the work is to compare these two methods for a specific turbine in a specific wind farm environment. Both fatigue loads and ultimate loads are considered. The turbine is a 2.0 MW variable speed/pitch controlled turbine. The aeroelastic model HAWC has been used for the investigation. A number of fatigue load cases are analyzed using the Rainflow counting method, and the combined life time fatigue loads are compared for the two different wake simulation methods at mean wind speeds 10m/s and 20m/s, respectively. The difference in fatigue load is within 20% for most load sensors. Concerning the extreme loads, no differences are expected for the stand still/idling 50-year load case; however, for extreme loads occurring during normal operation the wake method applied influences the results significantly. The main differences between the two wake simulations methods are seen for the extreme yaw loads during operation.