Poster presentations

Poster presentations

Abstracts

Topic 1 – Advanced Flow Modelling

High resolution mesoscale modelling of the Egmond aan Zee wind farm

Presenting author: Mohamed Safwan Elmahdy, Vrije Universiteit Brussels (VUB)

Co-author(s): Mohamed Safwan
Vrije Universiteit Brussel, Department of Mechanical Engineering, FLOW Research Group
Mark C. Runacres
Vrije Universiteit Brussel, Department of Industrial Engineering, Faculty of engineering, Pleinlaan 2, 1050 Brussels, Belgium

Abstract: In this study, the Weather Research and Forecasting model (WRF) is used to simulate the atmospheric flow within the Egmond aan Zee wind farm at very fine resolution (150 m). Short and medium range simulations are conducted for both summer and winter weather conditions. The simulation results are validated using a set of SCADA data provided by the wind farm owner and operator NoordzeeWind. The main goal of this work is to push WRF to the limits in terms of grid resolution and investigate the performance of the Fitch wind turbine parameterization scheme – already implemented in WRF – to account for the wind turbines’ effect on atmospheric flow.
Different atmospheric variables and PBL parameters (wind speed, wind direction, potential temperature, relative humidity and PBL height) are validated and compared to the available data. The initial results show a good matching with the SCADA data, with error values almost the same as the results of previous mesoscale simulations with large grid sizes. The simulations show no clear improvement using fine resolutions. This is mainly due to the physical parameters neglected in the model assumptions based on the large grid sizes typically used in mesoscale simulations. A more comprehensive study is in progress in order to investigate the effect of the neglected physics and whether a change of the model is required for a high-resolution grid.

Wind resource error estimation from mesoscale modeling for the Wind Atlas for South Africa

Presenting author: Andrea Hahmann, DTU Wind Energy

Co-author(s): Niels Gylling Mortensen, Patrick Volker, Chris Lennard and Jens Carsten Hansen

Abstract: The Wind Atlas for South Africa (WASA) project created a wind atlas database of winds for the Western Cape and parts of the Northern and Eastern Cape. The wind atlas was created using mesoscale modeling with the Weather, Research and Forecasting (WRF) model, and afterwards the Wind Atlas Method was used to process the data from the simulations.
The atlas was verified against measurements from 10 wind masts, which were also part of the project, and gave a mean absolute error in wind speed of about 5%. The verification errors are useful to assess the possible errors around the observation sites, but only sample the large variety of wind climates and terrains across South Africa providing an illustration of the range of errors and uncertainties that generally would be expected at similar sites. Away from these 10 sites, it is not straightforward to estimate the possible errors in wind resource assessments made from the wind atlas at sites that are either far from verification masts or at sites of a different climate or topography.
We explore new methods that try to estimate errors based on ensemble WRF simulations. These ensemble simulations are created by perturbing parameters of the simulations themselves and by introducing variations in the initial atmosphere and surface conditions. The results of the ensemble simulations can be processed to give a ‘map’ of the spread of the wind resource estimation. By comparing these ‘maps’ with the observed wind resources at the sites and by relating these to the terrain complexity and wind climate complexity, it might be possible to diagnose the geographic distribution of possible errors in the wind resources away from the observation sites. If successful, the new method will be able to assess the error in wind resources at a geographical scale.

Can a Mesoscale Model Reasonably Estimate Turbulence Intensity?

Presenting author: Santi Vila, AWS Truepower

Co-author(s): Philippe Beaucage, AWS Truepower
Michael C. Brower, AWS Truepower

Abstract: In wind resource assessment, the mean meteorological fields such as the mean wind speed are often of primary interest, but their associated turbulent fields such as turbulence intensity (TI) are a critical component of site suitability analyses (i.e. IEC turbine class) and energy assessment. TI can have major impacts on the fatigue loading, the wear and tear of wind turbine components as well as on the wind turbine power generation. Mesoscale numerical weather prediction (NWP) models have been in use for wind resource assessment since the 1990s. They are not often regarded as reliable for TI estimates given their coarse resolution (> 1 km). However, they have proven advantageous for wind resource modeling, especially for sites with strong influences of vertical and horizontal temperature gradients. Like RANS/URANS models, NWP models take into account the mechanical production of turbulence (shear). Unlike some RANS/URANS models, they also consider the buoyant production or consumption of turbulence. Thus, while sacrificing spatial resolution, NWP models may capture more of the physics of turbulence production and dissipation. The present study aims to evaluate the ability of mesoscale NWP models to predict the variation in TI across 13 sites in the US with a wide range of terrain complexities, surface characteristics, and wind climates. Overall, the WRF mesoscale model shows decent skills at predicting the hourly turbulence intensity (TI), more so in fairly flat terrain conditions where at least half of the variance in the observed TI can be explained by the WRF-derived TI. For site suitability analyses, the WRF-derived TI seems to be an adequate tool as the modeled TI versus wind speed curves align very well at almost all the sites. In order to put these results in perspective, a comparison against other numerical models such as WAsP Engineering and RANS/URANS would be beneficial.

AEP Estimation Improvement Through Full Atmospheric Boundary Layer Stability CFD Modelling

Presenting author: Gregory Oxley, Envision Energy

Co-author(s): Kyle Hutchings, Envision Energy USA
Jun Li, Envision Energy USA
Guolei Wang, Envision Energy China
Xing Sun, Envision Energy China

Abstract: Although ever improving, wind resource assessment (WRA) continues to be fraught with uncertainties and approximations, with developers and financial institutions continuing to demand more accurate WRA assessments to better qualify the business case certainty of wind projects. To meet these needs, Envision Energy has developed the Greenwich cloud platform to ensure the economic indicators of wind power assets and investments, provide developers with comprehensive technological solutions to wind farm planning, wind resource assessment, micro-siting, layout optimization, assessment of economic viability and post asset evaluation analysis. With its digital model structure, the Greenwich cloud platform controls and lowers risk arising from investment in wind farms, thereby significantly reducing the uncertainties of investing in wind energy.
The primary engine of Greenwich is the GWCFD CFD code, which is a fully automated CFD model leveraging the top super computing resources in the world. As it’s workhorse, GWCFD employs the standard steady state RANS neutral approach to CFD wind modelling. Despite widespread industry acceptance, the neutral approximation can be quite limited, in particular in areas where local site climatology is dominated by strong influences of atmospheric stratification. This work will demonstrate Envision Energy’s implementation of a stratified CFD modelling approach representing the entire atmospheric boundary layer, rather than the more common, but limited, atmospheric surface layer approaches. In addition, we will examine a proposed approach to aggregate neutral and stratified CFD simulations into an effective wind field for AEP estimations, and compare results to operational wind farm data as part of our ongoing internal AEP benchmarking studies.

The uncertainty of vertical wind speed extrapolation as a function of terrain type

Presenting author: Wiebke Langreder, EMD A/S

Co-author(s):

Abstract: Hub height measurements of wind speeds performed with traditional masts are rare. The reason is mainly that masts high enough for the hub height of today’s turbines are very expensive.
The estimate of wind resources and consequently AEP in these cases then requires vertical extrapolation from measurement height to hub height. Similarly, vertical extrapolation also becomes necessary, when production data from existing WTGs is used for estimating the AEP for future WTGs typically with higher hub heights.
The vertical extrapolation of wind speed or production data results in an increased uncertainty. As a rule of thumb 1% uncertainty of the wind speed per 10m vertical extrapolation is often assumed. However, this number lacks validation. Additionally, it can be shown that the process of vertical extrapolation can be biased, which is often not considered.
This paper describes a more sophisticated approach taking the topography and roughness of the site into account. The vertical wind speed changes were predicted using WAsP and compared with actual measured wind speeds. 300 measurement cases from masts equipped with cup anemometry in various climatic and topographic conditions were used. If found reasonable, the stability settings within WAsP were adjusted relative to the local conditions. Also, displacement heights were introduced if required. Empirical values for the bias and the uncertainty of the vertical wind speed extrapolation were found for complex and flat terrain, with and without forests in the vicinity.
Using these results allows for a more comprehensive estimate of the contribution of the vertical prediction uncertainty on the P50/P90 and enables you to quantify the bias.

Mesoscale vs. Linearized and CFD Modelling: A Case Study for Wind Flow Modelling in Complex Terrain in East Africa

Presenting author: Meltem Duran, WSP | Parsons Brinckerhoff

Co-author(s):

Abstract: Traditional wind flow models, both linear and CFD-based, fail to capture wind regime variations on some sites where factors other than the terrain and overall stability drive the flow, e.g. thermal gradients such as downslope katabatic winds. Unless meteorological measurements are available across the site, this issue is difficult to identify and could lead to potential bias and increased uncertainty in the final yield results.
We present a case from a proposed wind farm site in East Africa. The site is located on a high mountain plateau with proposed turbine elevations varying by up to 300 m. The mast coverage on site was relatively good with three meteorological masts sampling wind speeds throughout the project lands. Elevations drop rapidly to the west of the plateau where the observed wind speeds were considerably higher than measurements further atop the plateau.
The wind flow at the site was modelled using,(1) WAsP, (2) WAsP CFD in WindPro, (3) Meteodyn CFD, and (4) WRF mesoscale modelling downscaled with Meteodyn. Cross prediction errors between masts are presented using all four methodologies.
Predicted long term wind speeds at turbine locations using the four methods and the standard deviation of modelled wind speeds are compared. Strengths and weaknesses of the different modelling approaches are presented. Implications for yield predictions and uncertainty analysis are finally discussed.

CFD simulations of the Egmond aan Zee wind farm using high resolution mesoscale predictions

Presenting author: Nikolaos Stergiannis, Vrije Universiteit Brussels (VUB)

Co-author(s): Mark C. Runacres (Vrije Universiteit Brussel, Belgium) ; Jeroen van Beeck (von Karman Institute for Fluid Dynamics)

Abstract: Due to the complexity and uncertainty involved in the process of making power from wind, more and more advanced tools are being developed to maintain the sustainability and the growing trend of the wind industry. Prior to the development of a wind farm project, measured data are provided by limited installed wind masts at the site under investigation and by nearby weather stations. Therefore, the wind resource assessment depends on the uncertainties and limitations of those measurements. To improve the reliability and limit the risks, weather prediction forecasting models can be employed in parallel with measurements, to investigate the local wind map and the potential wind power. Nevertheless, the physics involved at the inter-turbine or smaller scales cannot be captured by mesoscale modelling. To obtain predictions of such scales, high-resolution meso-scale models are coupled with micro-scale computational fluid dynamics (CFD) simulations.
Results of neutral atmospheric stability over a defined wind sector have been averaged in time and extracted from meso-scale simulations using the Weather Research and Forecasting model (WRF) with a very fine resolution (150 m). Those results were used to provide the inlet plane of the micro-scale CFD simulations that were performed using the open-source CFD software OpenFOAM. The predicted time-averaged atmospheric flow within the Egmond aan Zee wind farm is compared for both numerical approaches. The wind farm’s total power estimations are compared to operational SCADA data.

Why Wind Resource Engineers should not Compromise with the Number of Directional CFD Calculations?

Presenting author: Tristan Clarenc, MeteoPole Zephy-Science ZephyCloud

Co-author(s): Theo Reffet MeteoPole Zephy-Science ZephyCloud ; Karim Fahssis MeteoPole Zephy-Science ZephyCloud

Abstract: Generally speaking, decision making is influenced by the complexity of the problem, the skill of the decision maker, and the time pressure. The global wind industry has become more mature and sophisticated in the past half-decade, with a better understanding of wind farm modeling at multiple scales. However, the increasing complexity of the sites along with the lack of appropriate calculation power still seriously affects quality decision-making as engineers often have to compromise on simulations accuracy levels in order to be able to deliver results with tight project timelines. Thanks to on-demand elastic cloud computing and freely available opensource solvers, all the wind simulation jobs can be run at the same time on powerful servers so that engineers are left with more time to tune models and to improve accuracy. Case studies on projects located on complex sites in Europe and Asia clearly show that accuracy in decision-making is a growing function of calculation power. Deploying simultaneously opensource solvers on an unlimited number of elastic cloud servers allow getting almost instant results with no limitation in number of cells (i.e. resolution levels) while discretizing the atmospheric boundary layer, with no limitation in number of directional runs while discretizing the wind directional scenarios and with no limitation in number of solver iterations while solving the flow equations. The presented white paper shows the benefits of increasing the number of calculated directions in terms of both accuracy and reliability. An extensive parametric analysis is performed on an extremely complex Indian complex site (maximal slope higher than 23 degrees and maximum elevation gradient higher than 500 meters) with high-quality measurements available at two IEC-compliant mast locations, and for which quality-control and long-term adjustment processes were performed by a globally reputed third party consulting firm. The number of considered CFD directional computations is varied to assess its influence on modeling accuracy, reliability and wind phenomena spotting. Results provide guidelines to define a minimum number of directional computations to be considered to perform a bank-grade CFD site modeling in highly complex sites. The impact of this minimal setting on the total computation time needed along with available solutions to tackle this issue are also discussed.
The interpolation errors related to lower numbers of computed directions can be crucial in terms of both energy yield assessment and IEC site suitability assessment.
Open Source solvers and Cloud Computing capabilities can be leveraged to address wind flow modeling requirements in the most complex sites and to deliver accurate results without affecting overall calculation durations. This study is part of a larger project to generate a model allowing quantifying wind modeling uncertainties depending on site complexity and modeling configurations.

Topic 2 – Siting and layout design optimisation

We minimise CoE by AUTO selecting the ideal geography for wind-farms

Presenting author: Marcello Deplano, Qurus Europe Ltd.

Co-author(s): Lucy Kennedy, Paolo Senes

Abstract: The Space Enabled Wind Installation Site Screening (SEWISS) service has been conceived to put wind-farm developers in the (so far) unprecedented position of being able to automatically scout for the best onshore wind farm locations, anywhere in the world. So far, most automation R&D in this sector has been dedicated to the prediction and modeling of wind resource availability. Far less effort has thus far been invested in helping developers to automate the assessment of the many other resources (geographical, environmental, logistical, and so on) which are in fact as critical as wind when it comes to minimising CoE (minimising it by optimising the choice of wind farm location). As a matter of fact, the workshops we have conducted over the last 3 years have demonstrated that even large, multi-national developers still manage this whole process in an overwhelmingly manual fashion.
SEWISS is therefore a breakthrough proposition for developers, who now have the possibility to automatically analyse very high resolution multi-spectral satellite imagery, combined with other relevant data sets to screen these global ‘non-wind resources’ with a view to reducing both the cost and duration of the site screening and selection phase by >10%.
Technical and commercial viability of the concept was validated by means of a successful feasibility study with the active involvement of developers and co-funding from the European Space Agency (ESA). The SEWISS service has now moved into a development and trial phase, again with the active involvement of developers and co-funding from the European Space Agency.
The SEWISS project is now being evaluated by 4 Users who will be testing its final configuration. All of our Users have identified the innovations SEWISS is intending to introduce as potentially crucial means to minimising CoE and increasing the chances of auction success.

Wind speed measurements above a noise barrier next to a highway for urban wind turbine installation

Presenting author: Nikolaos Chrysochoidis-Antsos, TU Delft

Co-author(s): Ad van Wijk

Abstract: In this study the preliminary experimental results of a near-highway wind resource assessment will be presented. The measurement anemometers have been placed near a highway site and in particular close to a noise barrier in order to assess the effect of the noise barrier on the wind flow. Noise barriers and similar windbreak alike structures result in flow acceleration when the flow is perpendicular to them and this might pose a prosperous installation scenario for urban wind turbines. Already wind tunnel experimentalists have indicated flow acceleration in the range of 20-30% for locations above noise barriers, but an assessment with real on-site data would deliver valuable knowledge for practical applications like urban wind turbines. This set-up consists of 8 sonic anemometers installed on 3 poles on different heights so that wind can be measured. The data output of the anemometers is being recorded with 4 Hz sampling frequency. The data recorded can reproduce mean wind speeds, turbulent intensity and vertical flow angles for different wind orientation and speeds above the noise barrier. The results could indicate the wind potential and address other challenges for urban wind turbine installations above noise barriers. Finally, this dataset could be used for improving tools and modelling of local micro wind resource assessments.

Wind resource assessment over a complex terrain covered by forest using CFD simulations of neutral atmospheric boundary layer with OpenFOAM and MeteoDyn

Presenting author: Nikolaos Stergiannis, Vrije Universiteit Brussels (VUB)

Co-author(s): Baris Adiloglu (3E S.A., Brussels, Belgium) ;  Mark C. Runacres (Vrije Universiteit Brussel, Belgium) ; Jeroen van Beeck (von Karman Institute for Fluid Dynamics)

Abstract: Following the growth of the wind energy sector and the corresponding decrease of land availability, the market share of onshore wind energy installation on complex terrain is expected to increase. Reliable wind resource assessment is crucial for the successful development of a wind farm project in complex terrain. To estimate the future energy production on a specific site, developers investigate the potential wind power which is related with the local winds. For cases of complex terrain with significant changes in roughness due to vegetation or buildings, local winds can vary considerably across a wind farm site, resulting in a potentially inaccurate energy estimation.
The site under investigation in the proposed contribution is on an island, in complex terrain covered by 70% of thick forest with trees of roughly 15 to 20 m height. The atmospheric boundary layer stability is mainly neutral with a very unidirectional wind direction from ESE. Two meteorological masts have been installed providing measurements of more than a year. Average wind speeds at 78 m height and turbulence intensities at 72 m height were measured to be 7.3 m/s and 9% for the first met mast and 6.2 m/s with 19% for the second, respectively. Wind resource assessment has been performed using 12 wind sectors and the commercial software meteodyn WT. Results are compared to computational fluid dynamics (CFD) simulations of the steady state 3-D Reynolds-Averaged Navier Stokes (RANS) equations, using the open-source CFD software OpenFOAM and the met-mast measurements.

Turbine performance and wind veer

Presenting author: Przemek Marek, Prevailing

Co-author(s):  Neil Atkinson, Prevailing
Matthew Colls, Prevailing
Joel Manning, Prevailing

Abstract: Real world wind conditions frequently fall outside those considered for turbine power curves, which are calculated for a specific range of conditions (turbulence, shear, inflow and veer). As a result, actual turbine performance often deviates considerably from that indicated by the sales power curve. This effect is particularly pronounced for relatively low wind speed locations, such as inland France and Germany, and the forests of Sweden.
Turbine performance prediction in pre-construction energy yield assessments has progressed significantly over the past few years, moving away from the use of a single adjustment factor towards multiple parameter models taking into account site specific wind conditions.
Prevailing continues to advance this field by developing methods based on the physical drivers of performance variation. These drivers have been identified through the application of well established aerodynamic theories as well as extensive consideration of measured data, leading to the previously presented three dimensional turbine performance matrix. This matrix provides accurate turbine performance predictions for known wind conditions (normalised values of wind speed, shear and turbulence intensity) and is applicable globally across different turbine types, rotor diameters and hub heights. The matrix contains data from some 60 power performance tests, covering a wide range of atmospheric conditions.
Building on this turbine performance matrix, Prevailing has identified wind veer (variation in wind direction with height) as a further key parameter affecting turbine performance variation. Here, real world wind and production data are used to illustrate the effect of wind veer on turbine performance. This helps to explain the significant drop in turbine performance observed for low turbulence wind conditions at some sites and it is clear that the consideration of wind veer has the potential to improve the accuracy of current turbine performance models.

Topic 3 – Outputs in real-world conditions

Modeling the dynamic behavior of wind farm power generation : building upon SCADA system analysis

Presenting author: Michael C. Brower, AWS Truepower

Co-author(s): Philippe Beaucage, AWS Truepower, LLC
Nick Robinson, AWS Truepower, LLC
Brian Kramak, AWS Truepower, LLC

Abstract: Time-series energy modeling represents the next frontier of wind plant design and energy estimation. Its potential advantages over existing methods based on static frequency distributions include capturing a wider range of dynamic wind characteristics such as shear and turbulence, more accurate wake simulations, and better estimation of losses that depend on time and prior conditions, such as icing and directional and environmental curtailment. Modifications were made to the Openwind software to implement a full time-series energy estimation model. Advances were made possible through the analysis of SCADA data from 18 operational wind farms in the Quebec province of Canada to support grid operations, integration and reliability. Effort was devoted to understanding the time-varying plant losses related to wakes, availability, environment, and electrical systems and developing ways to model them. Particular attention was paid to icing losses, which are severe in Quebec. Historical time series of wind power production and associated plant losses were generated for the 1979-2015 period. The long-term, hourly meteorological time series were created with the Weather Research and Forecasting (WRF) model initialized by the ERA-Interim reanalysis data set. The meteorological time series were then converted into wind power generation in Openwind, taking into account plant losses on an hourly basis. All plant losses were tracked separately and at the turbine level, providing the ability to make detailed comparisons with actual operation. Strong agreement is observed between the actual and modeled net power generation even though icing-related losses add to the complexity. The average hourly coefficient of determination (R2) was 0.80, while the mean daily R2 was 0.88. Our analyses also indicated that the monthly/seasonal trends in net power are well captured by the simulation system.

Continual Improvement of Assessment Methods

Presenting author: Scott Eichelberger, Vaisala

Co-author(s): Celeste Johanson, Vaisala, Dr. Mark Stoelinga, Senior Scientist, Vaisala

Abstract:
Project stakeholders, and investors in particular, require a high level of comfort with due diligence wind energy assessment reports. Stakeholders develop an experiential calibration as a result of long accumulated familiarity with an established assessment method and its performance over a large number of cases and under a wide array of conditions. While familiarity can provide confidence in methods, there is a risk that process stability will be favored over innovation. Maximum accuracy can only be achieved by innovating with the best science and latest technology available. An intelligent use of validation is therefore key to more accurate assessments that simultaneously achieve stability and innovation.

Vaisala’s continuous validation feedback process is illustrated here through a case study involving an evaluation of several algorithm improvements and upgrades to numerical weather prediction model configurations. These methodology improvements were developed in order to improve project accuracy rather than reducing mean calibration bias. Although each innovation is expected to reduce project error, the effect across a suite of projects may be to shift the mean error bias away from a zero-centered distribution. A set of acceptance criteria and evaluation metrics to aid a continual validation process is presented. These criteria may be applied to evaluate innovations in energy assessments as well as associated uncertainty models.

Project Transparency: Wind turbine performance evaluation in all terrains

Presenting author: Henrik Pedersen, ROMO Wind

Co-author(s): Sowjanya Subramaniam Iyer
Data Analyst
ROMO Wind

Abstract: One hundred power curve (PC) evaluations following IEC 61400-12-1 and -2, for three different wind turbines types, comparing 30 of each type to each other, from three wind farms with minimum 10 turbines from different terrain classes, flat terrain, complex, offshore. All results to be shared with the public under the Danish EUPD funded project between ROMO Wind and DTU RISØ.
Problem: It is not possible using a WTG’s own nacelle anemometer to accurately measure and monitor the turbines PC, because the wind measurement equipment located behind the rotor is heavily disturbed by rotor turbulence and other unpredictable flow conditions created on the site and/or by other WTGs in the surroundings.
Objective: Based on previously obtained measurement data from iSpin EUDP projects, the hypothesis of this project is that iSpin can function as a, universal tool for wind turbine PC verification and monitoring.
A tool, which after one WTG type calibration, in flat terrain only :and with no need for further site calibration: can be used for verifying and monitoring PCs of all WTGs in a wind farm and in all wind farm across a fleet of turbines independent of site location or climate conditions. Thus enabling turbine owners to monitor individual turbines performance and compare them to each other.
All results will be evaluated by independent global 3’rd party consultant houses.
Results: We will show the initial results which started project transparency, Showing the results from power curve comparison on 40 wind turbines in a complex, cold climate and forested site, using a NTF for iSpin developed from a flat site in Southern Europe. And also show that the added uncertainty from making 360 degree PC’s using wake sectors with iSpin yields very limited additional uncertainty. And give the latest status on the transparency project itself.

Validating turbine performance predictions

Presenting author: Taylor Geer, DNV GL

Co-author(s): Josiah Mault, Carl Ostridge, Jose Francisco Herbert, Xiomara Herrera

Abstract: DNV GL will update the industry on the current method for understanding turbine performance in preconstruction energy assessments, including the addition of over 70 new power performance tests from Europe. We will then present a comprehensive validation of DNV GL’s current approach to better understand the accuracy of current methods and to highlight areas for the improvement.

Topic 4 – Measurement and analysis

Long term prediction : impact of the choice of the reference sources

Presenting author: Marion Jude, Eoltech

Co-author(s): Habib Leseney, Eoltech

Abstract: Context
The long term adjustment of wind data is one of the key issues in wind potential assessments and can lead to significant deviations in terms of estimated production depending on the methodology, the length of the long-term period, but also the choice of the sources considered as references, main focus of this study.
Two datasets were compared: data from MERRA-2 reanalysis and an average wind index based on consistent data from at least 4 meteorological stations in the area. The latter, called multisource index, is considered as the reference (resulting from the combination of fully independent and convergent sources).
Objectives
-To compare the long term adjustments performed using either MERRA-2 data or the multisource index.
-To get orders of magnitude of the deviations obtained depending on the considered source or the length of the long term period (from 10 to 17 years).
Results
Analyses carried out over 4 areas in France, have shown that even if the wind trends since 2000 seem quite similar between both sources (MERRA-2 and multisource index), some specific patterns can be observed, and could lead to significant differences in terms of long term prediction for 3 of those areas.
Example: for a measurement data set recorded during year 2015, the gap between the long-term wind speeds adjusted using either MERRA-2 data or the multisource index can vary from 1% to 4% depending on the region.
Generally, the influence of the length of the long term period seems lower than the choice of the source considered itself. Thus, using MERRA-2 could lead to significant differences in terms of production estimation, depending on the year of measurement and the area considered.
Similar analyses are currently carried out in other areas in France and in Europe (complementary results available within a few weeks).

Turbine performance analysis using CFD and machine learning algorithms

Presenting author: Iain Dinwoodie, The Natural Power Consultants Ltd.

Co-author(s): Selena Farris, Natural Power

Abstract: Advanced turbine performance analysis techniques offer the opportunity to accurately predict future performance based upon past performance and also to identify and rectify any underperforming turbines ensuring that future performance is maximised and reduce cost of energy. Traditional power curve analysis approaches are qualitative and don’t consider the complex flow conditions in the real world and have significant associated uncertainty. A new methodology based on CFD modelling coupled with machine learning algorithms is utilised to normalise the power for complex flow conditions such as turbulence, shear, veer and inflow angle through in order to generate a site specific multi-parameter power curve. Flow parameters are modelled through both the use of VENTOS CFD modelling coupled with on-site observations concurrent to wind farm operation or coupled mesoscale-CFD model VENTOS/M when no on-site observations are available. This innovative approach identifies the impact of adverse flow conditions on turbine performance, allowing mechanical underperformance to be accurately identified.

After this initial analysis, additional machine learning tools are used in a deeper investigation into the SCADA data to identify the key indicators leading to individual turbine underperformance. Strategic maintenance campaigns are derived through benchmarking of turbine performance indicators to determine an optimised maintenance strategy that will deliver the highest potential gains in performance, reducing OPEX and improving cost of energy. A case study is presented to show the results and describe the on-going maintenance campaign. As a greater volume of data is generated, a wider range of machine learning algorithms can be considered to maximise the benefit from this approach.

Wide performance test campaign of a model based LiDAR turbulence intensity estimation

Presenting author: Fabrice Guillemin, IFP Energies nouvelles

Co-author(s): Paul Mazoyer, LEOSPHERE

Abstract: In recent years, LiDAR sensors emerged as a reliable and accurate remote sensing technology for wind speed measurements. In particular, ground based pulsed LiDAR sensors get more and more used for site assessment applications, as a measurement solution for 10 minutes mean wind measurement.
The LiDAR provides raw data for each line of sight, which corresponds to the wind velocity projected along the direction of the laser beam, so called “radial wind speed”. These measurements are provided within a scale of limited spatial and time resolutions, thus small eddies and spatial events may be filtered.
With the use of standard reconstructions on radial wind speed measurements, 10 minute mean horizontal wind speed estimates has been proven to be robust and are used widely.
However, the evaluation of turbulence intensity, calculated by dividing the standard deviation of 10 minute wind speed series by its mean wind speed, appears more challenging, as the LiDAR sensor filters the wind information and add some measurement noise. Additional post-processing is recommended to obtain more accurate turbulence intensities (cf. IEAWind task 32 report).
A model based optimal filtering approach has been developed by IFPEN to fulfill these requirements (cf. EWEA2015 publication). The algorithm has been implemented as a cloud application service, to be accessed easily. A whole test campaign has thus been carried on by LEOSPHERE, involving different end users of windcubeV2 LiDARs. To quantify the added value, turbulence intensity from standard WINDCUBE processing and from model based processing are compared to turbulence intensity from meteorological mast cups.
The presentation exposes the performances of the model based algorithm through the results of this test campaign and the feedback of the end-users.

Reducing offshore modeling uncertainty by using scanning Doppler Lidar

Presenting author: Guillaume Terris, La Compagnie du Vent

Co-author(s): La Compagnie du Vent : Benoit Buffard, G. Terris, P. Alexandre
Leosphere : P. Royer, M. Boquet

Abstract: The accuracy of wind models in offshore conditions are generally observed to be higher, due to lack of appropriate parameterization schemes for resolving the land-sea interaction effects. To assess the accuracy of the offshore models, a Windcube 400S scanning Doppler Lidar was deployed on the coast of southern France to measure accurate wind speed and direction at multiple distances and heights above mean sea level. A windcube V1 vertical profiler was deployed on the coast to correlate the measurements from offshore to onshore conditions, to assess the land-sea interaction effects. Profiles of wind speed measurements at 5 km & 9 km from the scanning Doppler Lidar was assimilated to the offshore wind model to reduce the bias and improve the overall assessment of the annual energy production (AEP) of the future wind farm. In this paper, preliminary results from the offshore measurement campaign will be provided, along with the setup & calibration procedures for measuring accurate wind speed and direction from a long-range scanning Doppler Lidar. The use of scanning Doppler Lidars for reducing the uncertainty of offshore wind resource assessment is studied.

Could wind turbines fuel up our future hydrogen refueling stations? – A GIS-based methodology

Presenting author: Nikolaos Chrysochoidis-Antsos, TU Delft

Co-author(s): prof. Ad van Wijk

Abstract: 400 hydrogen refueling stations have been announced to be by 2023 in Germany while transport energy for passenger vehicles is near 20% of an average European energy mix. Wind energy resources have a great potential to be converted into hydrogen fuel for our future refueling stations. Beside the centralized solutions this could be done in a distributed and local manner right next to the refueling stations located near highways and regional roads. Wind energy could provide all the electricity needed for hydrogen production, compression, storage and refueling with small wind farms located next to hydrogen stations or along highways. However, the suitability of wind turbine installation is restricted. In this study a methodology is presented for the identification of the suitability of these locations based on the local wind resource and other factors. For this, handling big data could be of great use. Many GIS databases could provide wind, land, population, airport, highway and refueling station data to be used for obtaining an overall picture and insight into the real potential of wind turbine installation next to these refueling stations. A methodology is presented for handling the reanalysis climate data, road linestring data, built-up area data, airport locations, land use/cover data and fuel demand data while some first results for the case of Germany are shown.

The use of Computational Fluid Dynamics to post-process ZephIR 300 wind speed data in complex terrain

Presenting author: Matt Smith, ZephIR Lidar

Co-author(s): Alex Woodward, Stephane Sanquer and Cyrille Vezza

Abstract: Data from remote sensing devices (RSD) are now widely considered as being bankable for use in wind resource assessment campaigns. Once installed, both the position and height of conventional anemometry such as cups are fixed, limiting their measurements to certain turbine locations and dimensions. These benefits, including the ability of RSD to measure at greater heights than current masts are generally capable of, can help to reduce project development risks and secure more favourable investment.
The current draft of the IEC 61400-12-1 ed.2 includes guidance on how to deploy a RSD for the purpose of turbine power performance testing. This version of the standards limits the use of RSD to flat terrain. Ground-based vertically-scanning RSD, whether sodar or lidar, calculate the mean wind speed based on measurements taken from the circumference of a scanned area. This process relies on the assumption that the line-of-sight velocities measured around the scan are representative of the wind speed at the centre of the scan. It is possible for this assumption to break down in strongly non-uniform flow, which can lead to possible differences in measured wind speeds between RSD and conventional anemometry.

By using a flow model, such as Computational Fluid Dynamics (CFD), it is possible to compute a set of factors that enable the conversion of RSD measurements to ones comparable with those from the point measurements sampled by conventional anemometry. This process is key to ensuring continued project financing based on data from RSD alone by reducing the uncertainty between a RSD and conventional anemometry in complex terrain.
In this work, CFD conversion of measurements taken by a Continuous Wave lidar, ZephIR 300, in varying terrain types and complexity is demonstrated, highlighting a transparent methodology that is capable of producing bankable measurements in terrain not considered to be simple.

Turning the tides on wind measurements: The use of lidar to verify the performance of a meteorological mast.

Presenting author: Scott Wylie, ZephIR Lidar

Co-author(s): Michael Harris and Alex Woodward

Abstract: The wind industry routinely compares the measured energy yield of constructed wind farms with the energy yield prediction from pre-construction analysis based on a set of measured wind speeds. Typical results published recently by consultants have shown a general over-estimation in pre-construction energy yield predictions in the region of 10%, reducing to 7% after availability and windiness normalisation.

The correct outcome of a wind resource assessment campaign is crucially dependent on the quality of the anemometry data. Cup anemometers, mounted on meteorological masts (met masts), are the industry standard for measuring wind speed at wind farms sites. Measurements from cups are, therefore, considered the “norm” against which any alternative measurement device is judged. Remote sensors, such as lidar, are continually judged against data from cup anemometers and as a result are now becoming increasingly used and accepted by the wind industry.
Even though cups are still considered the industry standard, there are several ways in which data from met masts can become compromised, e.g. flow distortion from the mast itself, out of date instrument calibrations, mounting irregularities, etc. Such problems, if undetected and uncorrected, will result in a degradation of the overall wind resource assessment at a site, with potentially damaging financial consequences for the project.
This work outlines, through a number of case studies, a cost-effective approach for verifying the performance of a met mast using a continuous wave lidar, ZephIR 300, ensuring that calibration of the data is both accurate and problem-free. Once deployed for mast verification, ZephIR 300 also provides the ability to validate the wind shear model used for a particular deployment, demonstrating a further reducing in the uncertainty associated with the extrapolation of mast data within the wind resource campaign.

Deep learning for short term wind power production forecast

Presenting author: Alla Sapronova, Center for Big Data Analysis, Uni Research Computing

Co-author(s):

Abstract: An accurate prediction of wind power output is crucial for efficient coordination of cooperative energy production from different sources. When a forecast for short time horizon (less than 1 hour) is anticipated, an accuracy of predictive model that utilizes the data from hourly mesoscale numerical weather prediction (NWP) is decreasing as the higher frequency fluctuations of the wind speed are lost when data is averaged over an hour. The analysis of wind speed fluctuations over periods from minutes to hours shows that higher frequency variations of wind speed and direction have to be taken into account for accurate short-term ahead energy production forecast.
In this work big data mining technique, known as Deep Learning, has been used to develop a new model to forecast wind park energy yield 5 to 30 minutes ahead. The model uses historical park production time series and hourly NWP data to predict the total power production of the wind park.
The model shows higher accuracy comparing to models based on feed forward neural network or linear regression methods.
The new model shows that use of NWP data does not significantly improve the accuracy for both very short time ahead forecast (5-10 minutes ahead) and 30 (and longer) minutes ahead forecasts.
It is also observed that when categorization of one of the input variables is used (wind speed, for instance), the accuracy of the forecast can be improved even with shorter time window for historical time series variables.

Stability parameters in Onshore wind farm sites

Presenting author: Elena Cantero, CENER

Co-author(s): Fernando Borbón (CENER)
Javier Sanz (CENER)
Daniel Paredes (Iberdrola)
Almudena Garcia (UPNA)

Abstract: The objective of this work is to develop a methodology for atmospheric stability characterization to be used in connection with wind resource assessment campaigns and wind farm design tools.
Research met-masts at Alaiz (CENER’s Test Site next to Pamplona) and Wind farm mast at an Iberdrola location near Cuenca (IBR) are used as benchmarks to establish procedures and methodologies that can systematically be applied to sites with a similar or lower level of instrumentation.
Several stability parameters have been studied, based on the Richardson and Froude numbers and the Obukhov length, to determine suitable methods for the categorization of wind conditions (wind speed, wind shear, turbulence intensity) in terms of atmospheric stability. The methods are examined based on their theoretical background, implementation complexity, instrumentation requirements and practical use in connection with wind energy applications.

This presentation summarizes results related to onshore met-masts IBR, in relatively simple terrain, and Alaiz-MP5 in complex terrain. Special focus is given to explaining the post-processing methodologies to derive stability from raw data considering fast-response sonic anemometers as well as slow-response mast instruments.

BigDataEurope Platform for Wind Energy Applications

Presenting author: Dimitri Foussekis, C.R.E.S.

Co-author(s): Fragiskos Mouzakis ([email protected])

Abstract: Big Data technology addresses the galloping need for creation of new knowledge and innovative solutions based on the management, analysis and interlinking of voluminous, variable and fast generated data, which cannot be handled efficiently with conventional tools. BigDataEurope (BDE) project embraced this rapidly evolving technology and developed an open platform. Several pilots have been developed, addressing various societal challenges, including wind energy.
The Big Data Europe Integrator Platform (BDI) is an open-source platform based on Docker, today’s virtualization technique of choice. The base Docker platform of BDI can work on a local development machine, or scale up to hundreds of nodes connected in a swarm. Components such as Apache Spark, Hadoop HDFS and Apache Flink can be built into a pipeline through a simple graphical UI.
In the field of Wind Energy, the first pilot case developed, addressed the research on system monitoring of a Wind Turbine. The case regards the data acquisition and analysis of voluminous data streams delivered from a network of FPGA-based data acquisition modules covering the condition and operational monitoring of the structure and the drive train. The data streams exceed 15GB per hour and the collected data base is subject to periodical analysis for procedure development. Apache SPARK and HDFS along with specific analysis modules were used for the data management and processing.
Another related pilot was developed regarding the climatic dynamic downscaling, for weather forecasting, using publicly available global climate data (ECMWF or ESGF) and the WRF mesoscale model. It consists of a Docker platform ready to be deployed to any HPC infrastructure, providing forecasting according to user requirements. The extension of the pilot by implementing CFD tools for site assessment purposes is pursued.

Research on the Effect of Wind Parameters on the Wind Energy and Economic Characteristics of Wind farm

Presenting author: KEONHOON KIM, Korea Institute of Energy Research

Co-author(s):

Abstract: The installed capacity of wind turbines is growing and enlarging by the governmental support program in the world. But, its growing grades are strongly affected by the economic characteristics in each circumstance and the economic characteristics also must be influenced by several major parameters like an average wind speed, scale and shape parameters for wind speed distribution, geographical configuration and the construction cost, etc. In general, the wind speed parameters like C and k are considered as the most effective and important parameters for the economic characteristics of wind farm. Thus, it must be important to understand the effect of wind parameters on the constructions for the wind farm.
In this presentation, the effect of these main parameters on the economic characteristics of wind farm are analyzed and discussed. For the understanding on effect, the comparative calculations like sensitivity analysis are analyzed with several selected wind parameters. As a result, the scale parameter related on the annual average wind speed is a most effective parameter on the wind energy and economic characteristics of wind farm construction, necessarily. In addition, it is also valuable to understand how much errors in wind speed parameters during measurement and/or analysis process probably affect on the wind potential energy and economic status of wind farm.

Understanding the behaviour of a wind farm under real conditions: a complete study case

Presenting author: Bruno Pinto, Sereema

Co-author(s): Thierry Ferrand, Kevin Michel, Jérôme Imbert

Abstract: An important cause of underperformance in wind turbines is their lack of adaptation to the real and local conditions. Wind turbine’s settings and control systems are parameterized to optimise their functioning under reference conditions. These conditions can strongly differ from those experienced on-site and therefore lead to sub-optimal performances.
We demonstrate how we can optimize the power output, the availability and the maintenance schedule through the understanding of the specific behaviour of each wind turbine. To do so, a study case of 6 operating wind turbines is presented. The results were obtained by:
1 – equipping each turbine with a connected smart-sensing device embedded with multiple sensors such as accelerometers, gyroscopes and compass;
2 – applying specially developed analysis algorithms to the acquired data;
3 – adding a comparative layer for wind turbines under similar local conditions.
Multiple results were obtained and made available to the wind farm owner and operator: an online and continuous monitoring of the rotor balance status of each wind turbine; overall and comparative vibration analysis of the structures; sectorized analysis of the vibrations levels quantifying the impact of particular working conditions, such as wake sectors and curtailments.
Adapting wind turbines to the local conditions is the next step towards wind farm optimisation. Our results show that simple strategies and analysis can permit a gain on both the power output and life expectancy of wind turbines.

Topic 5 – Wakes

Wind Farm Performance Analysis from scada information : why a model approach is required ?

Presenting author: Alexis DUTRIEUX, ATM-PRO S.P.R.L.

Co-author(s): Dr. Alexis DUTRIEUX, ATM-PRO S.P.R.L., Rue Saint-André, 7, BE-1400 Nivelles, Email: [email protected]
Dr. Laurent RAKOTO, Maintenance Partners Wallonie S.A., Rue des Gerboises, 1, BE-5100 Naninne, Email: [email protected]
Mr P

Abstract: Wind farm performance analysis is a key step for long-term sustainability and acceleration of wind project return on investment. One key element is the evaluation of wake effect impacts on wind turbine output power. Such wake effect evaluation is performed according a sector wise analysis based on wind direction. The wind direction is calculated from nacelle position and relative wind direction measured from the wind vane. The nacelle position is calculated from the “north offset” and an increment given by the encoder. This nacelle position may be altered due to missing or extra counts from the incremental encoder. This results in erroneous direction corrupting the wake effect analysis.
In order to reconcile wind direction with real world, MAESTRO Wind, a meso-gamma meteorological model, is used in “CLIMATOLOGY” mode to reproduce time series of data at wind turbines. This enables to generate a well-referenced wind direction and to correct observed wind direction and nacelle position. Subsequently a consistent wake effect analysis can be performed on the whole wind farm.
Such an approach has been tested on an existing wind farm. The analysis allowed to better understand the performance issues of the wind farm and the identification of improvements in terms of pro-active maintenance with respect to fatigue and load mitigation.
The aim is to improve “problem” detection and the timing of the required actions to be taken if the nacelle is not in the correct position with respect to wind direction. The meteorological model run in “FORECAST” mode can therefore be very helpful to implement such KPI to survey wind farm performance and detect in real time nacelle misalignment.
The perspective is to replicate the approach on other wind farms and develop a procedure that enables to extract both wind direction / nacelle position, but also nacelle misalignment if any exists.

Validating Wake Effect in an Onshore Complex Low Latitude Wind Farm

Presenting author: Paulo Henrique Valente Campos, Casa dos Ventos

Co-author(s): Gabriel Mendes Oliveira Lima
Casa dos Ventos,Project Engineer

Abstract: This work aims to answer some important questions regarding wake loss measurement impacts in a complex site, located in the inland northeast of Brazil, where Casa dos Ventos developed, built and is currently operating over 100 turbines. Several met masts used in the pre-construction phase are still measuring in the operating phase, some of them located downstream and others upstream of the prevailing wind direction, relative to the layout.
Additionally, operational data extracted from the SCADA system gathering important variables like: BOP availability, energy production by Wind Turbine, nacelle anemometers wind speeds, and other important data.
Combining all information and using some methodologies like:
Direct analysis of met masts data, comparing long-term statistics (long-term wind speed, turbulence intensity and turbulence spectrum density function) of pre-construction phase with the post-construction phase
statistics.
Comparison between predicted and realized production by wind turbine, correcting data with other losses like electrical losses, wind curtailment, underperformance, and unavailability in order to properly access the wake effects.
One important result is the observed turbulence acting on the farm in contrast with the calculated pre-construction turbulence. This analysis shows value because turbulence is the one of principal effects, which produces wearing on the machines, so influencing directly in the machine lifetime and O&M costs.
The final intention of this work is to compare the calculated pre-construction wake and the measured wake, enhancing knowledge about this effect in the site and providing better predictions of AEP (Annual Energy Production).

Wake predictions using different RANS turbulence models on full wind turbine rotor CFD simulations and actuator disk model compared with wind tunnel measurements

Presenting author: Nikolaos Stergiannis, Vrije Universiteit Brussels (VUB)

Co-author(s): Mark C. Runacres (Vrije Universiteit Brussel, Belgium) ; Jeroen van Beeck (von Karman Institute for Fluid Dynamics)

Abstract: The development of large-scale wind energy projects has created the demand for increasingly accurate and efficient models that limit a project’s uncertainties and risk. Wake effects are of great importance and together with the complex terrain effects are employed to the optimization process of wind farms. Despite a growing body of research, there are still many open questions and challenges to overcome. To investigate the physics involved in a single wake, full wind turbine rotor simulations at high spatial resolution are compared with a simplified actuator disk model that is used by the industry for wind farm applications.
The steady state 3-D Reynolds-Averaged Navier Stokes (RANS) equations are solved in the open-source package OpenFOAM, using different two-equation turbulence models. For the full rotor computational fluid dynamics (CFD) simulations, the multiple reference frame (MRF) approach was used to model the rotation of the blades. For the simplified cases, the actuator disk model was used with the experimentally measured thrust (Ct) and pressure (Cp) coefficient values. The performance of each modelling approach is compared against wind tunnel wake measurements of the 4th blind test organized by NOWITECH and NORCOWE in 2015.