Programme

Programme

Thursday 16 March 2017
09:30-10:15 Registration and welcome refreshments in the exhibition and poster area
10:15-10:25 Welcome & opening address by Giles Dickson, CEO, WindEurope
10:25-12:15 Session 1 – Modelling the wind from meso- to planetary scale
12:15-13:30 Lunch in the exhibition and poster area
13:30-15:30 Session 2 – Modelling the wind from micro- to mesoscale
15:30-16:00 Networking break in the exhibition and poster area
16:00-17:45 Session 3 – Models and reality
17:45-18:00 Short break
18:00-19:00 Session 4 – An investor perspective on resource assessment
19:30-23:00 Workshop dinner for all participants
Friday 17 March 2017
08:00-09:00 Welcome refreshments
09:00-11:00 Session 5 – Wakes
11:00-11:45 Networking break in the exhibition and poster area
11:45-13:30 Session 6 – LiDARs
13:30-15:00 Lunch in the exhibition and poster area

Session details

Thursday 16 March 2017, from 10:15-10:25
Welcome and opening address

Giles Dickson, CEO, WindEurope

Thursday 16 March 2017, from 10:25-12:15
Session 1 – Modelling the wind from meso- to planetary scale

Session chair: José Palma, Associate Professor, Engineering Faculty at the University of Porto

Session description: As the trend for more powerful computer power continues, the wind resource evaluation tends to rely more and more on computer modelling, taking advantage of computational techniques and procedures typical of weather forecasting. This session provides a glimpse of the current practice. It will open with a review of global long-term reference data (MERRA-2, by NASA’s Global Modelling and Assimilation Office), followed by a presentation on seasonal prediction of wind power, based on data originated from the ECMWF. The next presentations will cover recent developments in the use of this global analysis by looking at the impact of renewables on the climate and the work in progress within the NEWA project, coupling mesoscale and microscale modelling approaches. The session will end with a short presentation on how these developments can be part of the engineering (IEC) standards and influence the daily life of the engineering and consultant offices in the years to come.

Speakers:

A review of the performance of MERRA-2: the next era of global long-term reference data. Mauricio Pereira
Senior Engineer
DNV GL
Seasonal predictions of wind power generation are now possible Llorenç Lledó
Senior Researcher
Barcelona Supercomputing Center
Global Wind Atlas : preliminary validation and uncertainty assessment Niels G Mortensen
Senior researcher
DTU Wind Energy
Towards a New European Wind Atlas: WRF Sensitivity Experiments and the Mesoscale-to-Microscale Model Chain Andrea Hahmann
Senior Scientist
DTU Wind Energy

Thursday 16 March 2017, from 13:30-15:30
Session 2 – Modelling the wind from micro- to mesoscale

Session chair: Tomas Blodau, Head of Department Wind & Site, Senvion

Session description : From fundamentals of how to extrapolate wind profiles and how to model forests, this session spans to the more abstract questions, how do offshore wecs impact the regional wind resource and where modeling can bring us when we combine meso and micro scales.

Speakers:

A Uniform Approach for Wind Simulations in Forested Areas: Examining the Importance of Data Inputs Peter Enevoldsen
Industrial PhD Student
Siemens Wind Power A/S
Validation of Vertical Wind Shear Methods Circe Triviño
Head of Section
DNV GL
Do turbines dream of modeled winds? Gil Lizcano
R&D Director
Vortex
Short update on the work of the IEC 61400-15 working group Taylor Geer
Member of the IEC 61400-15 working group

Thursday 16 March 2017, from 16:00-17:45
Session 3 – Models and reality

Session chair: Jan Coelingh, Wind resource manager, Nuon (Vattenfall)

Session description: Expectations of reality are input for important decisions, and have to be based on model calculations. The big question then is always: how good is a model in practice, does the reality match the expectation? This session will look at this topic from different angles, both methodological (role of experiments, time-series based modelling) as from the perspective of different users (financiers, operators in cold climates).

Speakers:

Experiments in the New European Wind Atlas Jakob Mann
Professor
DTU Wind Energy
Modelling Wind Farm Energy, Revenue, and Uncertainty on a Time Series Basis Taylor Geer
Service Line Leader – Project Development
DNV GL
Do Uncertainties and Losses in Pre-Finance Wind Farm Energy Assessments Fit to Reality? Kai Moennich
Subject Matter Expert
UL International GmbH – DEWI
Estimating icing losses at proposed wind farms and how to optimise wind farms in cold climate Till Beckford
Engineer, Project Development
DNV GL

Thursday 16 March 2017, from 18:00-19:00
Session 4 – An investor perspective on resource assessment

Session chair: Oisin Brady, Director, Gaoithe Renewable Energy Consulting

Session description: This session will bring together investors and resource assessment experts to explore how resource assessment can best generate value.

Panellists:

Joanna de Montgros
Director
Everoze Partners Ltd
Nicolas Gourvitch
Director
Green Giraffe
Edyta Noworyta
Principal Engineer
European Bank For Reconstruction & Development

Friday 17 March 2017, from 09:00-11:00
Session 5 – Wakes

Session chair: Line Storelvmo Holmberg, Director at Plant Siting & Forecasting, Vestas

Session description: The impact from wakes on the business case certainty of a wind power plant is significant. Despite the effort invested in understanding this field over the last decade wake losses are still an ever returning discussion point and a major contributor to uncertainties in the wind resource assessment process. This session will address both in farm and farm to farm wakes and will focus on the importance of combining measurements, models and operational data to understand the impacts on wakes on the performance of a wind power plant. New measurement techniques and modeling approaches are presented as well as how to utilize wind farm operational data to validate, benchmark and feedback into models and measurement techniques.
This session will increase the understanding of risk mitigation of wake losses as well as identification of opportunities for wind power plant optimization.

Speakers:

Measured power curves in waked flow: Possibilities and limitations Nick Janssen
Data analyst
ROMO Wind
Mesoscale Modelling of Wind Farm Wakes: Implications for large-scale planning Jake Badger
Head of Section
DTU Wind Energy
Wake Modelling in the Time Series Domain Scott Eichelberger
Wind Energy Sales Expert
Vaisala
Comparison of the wake effect and the energy production resulting from a comprehensive annual energy production assessment methodology with SCADA data from an operating offshore wind farm Guillaume Angot
Research engineer
EDF R&D
Improving confidence in wake predictions through operational validations Carla Ribeiro
Head of project development department
DNV GL

Friday 17 March 2017, from 11:45-13:30
Session 6 – LiDARs

Session chair: Andrew Henderson, Senior Lead Windfarm Engineer, DONG Energy

Session description : LiDARs have been used for a number of years now, both onshore and offshore and the current industry focus is achieving greater maturity for a range of technologies, application and techniques. In this session, we will explore how LiDARs can support development of offshore windfarms, focusing on new technologies and reducing uncertainty, from both fixed and floating deployments.

Speakers:

Accuracy of dual-Doppler lidar retrievals of near-shore winds Nikola Vasiljevic
Researcher
DTU Wind Energy
Assessment of repeatability and stability of Lidar performances Paul Mazoyer
Application engineer
Leosphere
Results of the OWA Dublin Bay Scanning LiDAR Trials Alex Clerc
SMART Systems & Innovation Analysis Manager
RES Ltd.
Floating LiDAR uncertainty assessment in an operational context Hugo Herrmann
Offshore Wind Research Engineer
EDF Energy R&D UK Centre
Floating Lidars better accuracy Breanne Gellatly
Director of Business Development & Strategic Alliances
AXYS Technologies

Abstracts

Session 1 – Modelling the wind from meso- to planetary scale

the NEWA model chain

Presenting author: Javier Sanz Rodrigo, CENER

Co-author(s): Roberto A. Chávez Arroyo, CENER
Andrea N. Hahmann, DTU Wind
Stefan Ivanell, Uppsala University

Abstract: The New European Wind Atlas will be based on a mesoscale-to-microscale model chain that will blend long-term wind climatologies with high-resolution site effects. The development of the model-chain follows a verification and validation strategy that will incorporate data from existing experiments, from research and operational wind assessment campaigns, as well as dedicated experiments designed throughout the project. Initial efforts have been directed to reaching consensus in the set-up of mesoscale and microscale codes such that a reference model configuration can be consistently defined as baseline for validation and model improvement activities. Five regions across Europe have been studied with the WRF mesoscale model, following the same sensitivity tests, to determine which settings can be generally applicable in the production of the mesoscale European wind atlas. At microscale, four RANS codes have been tested at the experimental sites, on the assessment of grid dependencies in complex terrain to determine expected numerical errors due to varying spatial resolution and domain dimensions. The GABLS3 diurnal case has been used to benchmark NEWA atmospheric boundary layer (ABL) models dealing with mesoscale forcing and thermal stratification in flat terrain conditions. The Ryningsnäs benchmark has been conducted to test the implementation of forest canopy models embedded in the ABL in simple terrain. These two benchmarks allow NEWA modelers to implement and test fundamental flow cases before analyzing performance in significantly more complex flow cases coming up from NEWA experiments (Hornamossen, Perdigao, Alaiz, Kassel). Current development activities are focused on implementing mesoscale-to-microscale methodologies for AEP assessment and setting up of a validation approach based on tall mast data available from research sites as well as contributions from industry.


the NEWA model chain

Presenting author: Andrea Hahmann, DTU Wind Energy

Co-author(s): Magnus Baltscheffsky, Martin Dörenkämper, Yasemin Ezber, Rogier Floors, Elena Garcia Bustamante, Fidel Gonzalez-Rouco, Xiaoli G. Larsén, Sibel Mentes, Jorge Navarro, Javier Sanz Rodrigo, Alfredo Peña, Tija Sile, Stefan Söderberg, Yurdanur Unal and Björn W

Abstract: As part of the NEWA (New European Wind Atlas) and RUNE (Reducing the Uncertainty of Near-shore Energy) projects a series of mesoscale simulations were conducted to study the sensitivity of the estimates of mean wind speed and annual energy production to different configurations of the WRF model. In RUNE, with focus on coastal resources, we studied the sensitivity to the grid spacing, the PBL scheme and the surface boundary conditions. In NEWA, with focus on onshore resources over the very large European area, we studied the sensitivity to the PBL scheme and the initialization (length and nudging) of the simulations over five European domains with varied wind climates.
The presentation will focus on the validation against measurements for the RUNE experiment and in assessing the sensitivity of simulations to changes in the NEWA runs. The ultimate objective of the simulations is to help make informed decisions on how to setup the mesoscale simulations for wind energy resources in an extensive domain with a large variety of wind climatologies. The presentation also covers what needs to be considered when coupling the mesoscale model output to microscale models.


Towards a New European Wind Atlas: WRF Sensitivity Experiments and the Mesoscale-to-Microscale Model Chain

Presenting author: Andrea Hahmann, DTU Wind Energy

Co-author(s): Andrea N. Hahmann, Javier Sanz Rodrigo, Magnus Baltscheffsky, Roberto A. Chávez, Martin Dörenkämper, Yasemin Ezber, Rogier Floors, Elena Garcia Bustamante, Fidel Gonzalez-Rouco, Stefan Ivanell, Xiaoli Larsén, Sibel Mentes, Jorge Navarro, Alfredo Peña, Tij

Abstract: A mesoscale-to-microscale model chain that will blend long-term wind climatologies with high-resolution site effects will be the basis for the New European Wind Atlas (NEWA). The development of the model-chain follows a verification and validation strategy that will incorporate data from existing experiments, from research and operational wind assessment campaigns, as well as dedicated experiments designed throughout the project. Initial efforts have been directed to reaching consensus in the set-up of mesoscale and microscale codes such that a reference model configuration can be consistently defined as baseline for validation and model improvement activities.
In the mesoscale, as part of the NEWA and RUNE (Reducing the Uncertainty of Near-shore Energy) projects, a series of mesoscale simulations were conducted to study the sensitivity of the estimates of mean wind speed and annual energy production to different configurations of the WRF model. In RUNE, with focus on coastal resources, we studied the sensitivity to the grid spacing, the PBL scheme and the surface boundary conditions. In NEWA, five regions across Europe have been studied, following the same sensitivity tests, to determine which settings can be generally applicable in the production of the mesoscale European wind atlas in an extensive domain with a large variety of wind climatologies. The presentation also covers what needs to be considered when coupling the mesoscale model output to microscale models.
At the microscale, four RANS codes have been tested at the experimental sites, on the assessment of grid dependencies in complex terrain to determine expected numerical errors due to varying spatial resolution and domain dimensions. Two benchmark cases, GABLS3 diurnal case and the Ryningsnäs benchmark has been conducted These two benchmarks allow NEWA modelers to implement and test fundamental flow cases before analyzing performance in significantly more complex flow cases coming up from NEWA experiments (Hörnamosse, Perdigão, Alaiz, Kassel, etc). Current development activities are focused on implementing mesoscale-to-microscale methodologies for AEP assessment and setting up of a validation approach based on tall mast data available from research sites as well as contributions from industry.


Global Wind Atlas : preliminary validation and uncertainty assessment

Presenting author: Niels G Mortensen, DTU Wind Energy

Co-author(s): Jake Badger, DTU Wind Energy
Neil Davis, DTU Wind Energy

Abstract: The Global Wind Atlas was published in 2015 by DTU Wind Energy as part of an international collaboration in the framework of the Clean Energy Ministerial (CEM). The Atlas is meant to address an urgent need to mitigate climate change by sensible renewable energy decisions within wind energy. The objective of the Global Wind Atlas is: to provide global wind resource data accounting for high resolution topographical effects; employ microscale modelling to capture the small scale wind speed variability, which is crucial for better estimates of the total wind resource; use a unified, well-known and documented methodology; ensure transparency about the methodology used; and validate the results in representative selected areas.
The correct usage of the Global Wind Atlas dataset and tools is for aggregation, upscaling analysis and energy integration modelling for energy planners and policy makers. It is not correct to use the data and tools for wind farm siting. Nevertheless, it is of interest for validation purposes to compare Atlas data and predictions to wind measurements in different topographical and climatological settings.
The comparisons presented employ high-quality wind data from Denmark, UK, Egypt, Cape Verde, South Africa, China and Mexico to explore how well the Global Wind Atlas predicts regional and local wind conditions. On average, the Atlas predicts the mean wind speed, power density, annual yield of sample wind turbines and annual yield of sample wind farms very well (1%) in non-complex terrain. In complex terrain, the Atlas seems to overestimate slightly (3-4%). The spreads of predictions are significant: from about 9% for mean wind speed to about 17% for energy yields. Contrasting results for offshore and on-land sites; as well as for flat, hilly and complex terrain sites are presented. Steep terrain and sites associated with pronounced mesoscale flows represent the most challenging conditions.


Seasonal predictions of wind power generation are now possible

Presenting author: Llorenç Lledó, Barcelona Supercomputing Center

Co-author(s): Albert Soret,(Barcelona Supercomputing Center)
Francisco-Javier Doblas-Reyes (Institució Catalana de Recerca i Estudis Avançats)

Abstract: The financial plans for developing a wind farm need to include a section on monthly expected revenues, which have to be estimated from pre-construction wind resource assessment studies. Typically those budgets are precomputed averaging long-term reconstructions of wind speed. The majority of operating wind farms still use those monthly budgets to anticipate revenues from energy sales, or at most use actual generation from the last operating years to refine the climatological monthly budget.
However wind and generation can deviate substantially from climatological averages. When such deviations occur, companies struggle to find out convincing reasons for this deviations: a low technical availability would require action and attention from the company; but the deviations can be due solely to observed winds, and this would have a direct impact on the value of the wind farm and could pose a risk for cash-flow operations. Reanalysis datasets come to help and allow identification of anomalous wind conditions, although they are only available after the period is over -and with some weeks of delay-.
Recent advances in dynamical modelling systems have opened new opportunities for seasonal prediction of wind speed, which allow anticipation of anomalies and extreme events. For certain regions and periods seasonal forecasts do a better job than climatology-based approaches. This forecasts can also be useful for other purposes: planning operations and maintenance activities, grid balance or trading decisions. As part of Copernicus Climate Change Service (C3S), ECMWF will soon deliver seasonal forecasts in real time. Clim4Energy is a proof-of-concept project that aims to tailor those forecasts for its usage in the wind industry, providing capacity factor estimates. In this presentation we will highlight the benefits of this novel methodology and showcase results for specific sites. We will also gain some understanding on what is driving wind anomalies.


Wind Plant Performance Prediction (WP3) Benchmark initiative and an update on the work of the IEC 6400-15 working group (5-10′)

Presenting author: Caleb Phillips, NREL

Co-author(s): Jason Fields, Andrew Clifton, Monte Lunacek, Shawn Sheng, Erik Hale (EDF-RE)

Abstract: The Wind Plant Performance Prediction (WP3) Benchmark initiative is a new joint industry project aimed at performing pre-construction and operational performance reconciliation at scale. This work builds on previous efforts such as the EWEA CREYAP exercises and strives to create a robust operational dataset (currently 12 GW) with which to perform these reconciliations. The project has unprecedented support from all corners of the industry and represents the largest validation and data sharing initiative of its kind. The presentation will walk through the high-level approach as well as offer details on the open source,big data technologies that will be used to perform analysis at scale.
The audience will receive the following key information:
•Overview of Benchmark project including goals, outcomes and how to participate
•Discussion of the technical challenges of pre and post construction analysis reconciliation
•Big data technologies being deployed as part of the project (Python, Spark, Hadoop, and more)


A review of the performance of MERRA-2: the next era of global long-term reference data.

Presenting author: Gemma Daron, DNV GL

Co-author(s):

Abstract: The availability of reliable and consistent long-term reference data is key for defining the long-term wind regime at a proposed wind farm site. Reanalysis products have been widely used within the wind industry for several years, given their global coverage and public availability. The MERRA dataset, in particular, has been shown to provide robust correlations to site measurements across the globe.
In February 2016 the MERRA dataset was discontinued by NASA and succeeded by MERRA version 2, which incorporates a number of stepwise advancements including new and improved satellite observations and an updated assimilation model. This prompted an investigation of the performance of the new dataset as a source of long-term reference data.
DNV GL has undertaken a series of investigations into the consistency of MERRA-2 across the globe and has also compared MERRA and MERRA-2 correlation quality to site data. Both datasets exhibit similar trends and yield similar correlations to site measurements at many locations across the globe. However, there exist locations where the quality of correlations between site measurements and MERRA-2 are significantly worse when compared to the original MERRA dataset. An example of this is seen in parts of Scandinavia, where correlations with MERRA-2 result in significant uncertainty.
In this presentation, we will present our finding on the consistency of MERRA-2 across the world and explore possible reasons for the change in site correlation performance in some regions, with a particular focus on Scandinavia. We shall also present results using alternative reference data, including MERRA-2 data on pressure levels and DNV GL’s Virtual Met Data (VMD) product. We will show examples where these products give an improved representation of the long-term wind regime, providing an alternative to the use of conventional MERRA-2 data products and a reduction in the uncertainty in prediction the long-term wind resource.


Session 2 – Modelling the wind from micro- to mesoscale

A Uniform Approach for Wind Simulations in Forested Areas: Examining the Importance of Data Inputs

Presenting author: Peter Enevoldsen, Siemens Wind Power A/S

Co-author(s):

Abstract: How can the wind industry determine the impact of trees on wind conditions without launching expensive and time consuming measurements of forests?
This research study answers the question by testing and comparing six different public available datasets for roughness information in European forests. The datasets have been examined and compared using the measurements from 22 operating meteorological masts located in different forested areas in Northern Europe. The extensive amount of wind data made it possible to define the impact of each dataset and further to recommend when and how to apply the different approaches. However, the study revealed disappointing results, due to incommensurate and imprecise details of the forest impact. Therefore, the main output of this research study is the development of a uniform approach for estimating the roughness length of coniferous trees in various forest formations, which was developed as a response to the inadequate details of the public available databases. The approach can be applied using public available sources for tree heights (In Europe) by applying 0.3*(tree height-displacement height) for the roughness length and 2*tree height/3 for the displacement height. The proposed approach presented an absolute mean difference of 0.08 m/s when tested against the 22 meteorological masts using a linearized resource assessment model. In comparison, the best of the six pre-defined datasets presented an absolute mean difference of 0.68 m/s. This result was further compared to various CFD approaches, which requires a higher level of details for the forestry information. The results proofed to be better for the proposed methodology in all non-complex sites.


Validation of Vertical Wind Shear Methods

Presenting author: Circe Triviño, DNV GL

Co-author(s): Paul Leask
Till Beckford

Abstract: Wind shear is one of the largest sources of uncertainty in wind farm energy output predictions, especially in developments conducted based on old measurements campaigns often with less than ideal quality. Energy prediction accuracy and associated uncertainty are key to support financial decisions for the wind farm. Both can be improved by gaining knowledge of the vertical extrapolation process during the energy assessment.
There are different methods available to extrapolate wind conditions up to hub height, but there is little consensus in the industry regarding which methods best represent reality and the uncertainty of the prediction. Some effort has been made in the past to identify alternative methods and implement classical methods in different ways to capture the atmospheric effects on wind extrapolation. DNV GL has collected data from over 270 measurement masts distributed globally to produce the largest validation of shear methodologies ever conducted in the industry. Widely used discrete methods based on the power law and the log law and more advanced time-series approaches have been compared to real measured data to assess the quality of each method according to different metrics. Mean wind speed but also frequency distributions extrapolation have been validated in tall masts globally to identify the methods that better behave under different relevant metrics.
The validation also provides information of the size of the errors observed in the extrapolation of wind resource across the sites in the data base and its relation with the different parameters affecting the extrapolation


Articulating models to reach the micro-scale

Presenting author: Alex Montornes, Vortex

Co-author(s): Pau Casso (Vortex)
Gil Lizcano (Vortex)
Pep Moreno (Vortex)

Abstract: The next generation of the wind resource data-sets generated by means of a smooth transition from mesoscale to microscale processes in a seamless modeling chain has come. This new milestone in the atmospheric modeling has been reached by integrating a LES algorithm, broadly used in engineering, within the mesoscale model WRF, widely used in meteorology and, in wind resource assessment particularly. Therefore, high resolution simulations can be produced by explicitly resolving the small eddies represented by fluid dynamics with the rest of the physical processes occuring in the atmosphere (e.g. clouds, land surface, radiation).
During the last two years, Vortex has been researching the way to overcome the main limitations of the cutting-edge NCAR atmospheric model WRF-LES in order to achieve this challenge successfully with a viable computational consumption operativelly. Basically, four difficulties are idientified: i) software efficiency , ii) lateral boundary conditions, iii) Terra-Incognita and iv) performance of the surface layer parameterization.
This contribution will be divided in two parts. First, we will show a full validation of the Vortex-LES product focusing overall on the skills for wind speed and turbulence intensity. The study includes a 1-year of 10-min time-series at around 100 real sites around the world with different met-mast heights, topographic complexity, local weather regimes and climate features. Second, a discussion of the potentialities of this new modeling generation time-series for wind resource assessment as well as manufacturers will be presented.
The conclusions derived from the results show that WRF-LES produces competitive results in terms of metrics (i.e. bias, mean absolute error and correlation) and it characterizes the turbulence patterns associated to the local scale features with a high reliability.


Do turbines dream of modeled winds?

Presenting author: Gil Lizcano, Vortex

Co-author(s): Alex Montornes, Pep Moreno, Pau Casso – Vortex

Abstract: We are in 2017. And this presentation aims to review the limits of time series mesoscale modeling with the technology we have now in 2017. In particular, we see the enormous capacity of new Reanalysis project to improve the large to synoptical conditions all over the World, at the global backend, and we are now able to run almost fully physics models in the time domain at the wind turbine end of the modeling chain.
In this presentation, we will examine the downscaling capacities to localize and refine the information up to windfarm resolving resolution. This task will be carried out using MERRA 2 and the promising new ERA 5 Reanalysis as top drivers and flag mesoscale model WRF with a Large Eddy Simulation module, as downscaling engine.
In practical, we aim to clone observations with model times series that are able to reproduce mean flow conditions with accuracy accepted for EPA , support classification of turbulence and extreme patterns and offer reliable time-dependent long-term information including site specific interannual variability for AEP analysis.
The presentation will cover test cases of applications of 10’ sampling time series output from WRF LES simulations over real sites around the world with different flow complexity, local weather regimes and climate features.
Audience will get a vision of what we can expect in 2017 of the modeling branch of the resource assessment technology and what still are modeling dreams.
(The title of this presentation is obviously an homage to the science fiction novel Do Androids Dream of Electric Sheep? by Philip K. Dick which inspired Blade Runner movie)


On the impact of realistic offshore wind energy exploration scenarios on wind conditions within the German Bight

Presenting author: Martin Doerenkaemper, Fraunhofer IWES

Co-author(s): Gerald Steinfeld (ForWind – Center of Wind Energy Research, University of Oldenburg)
Bernhard Stoevesandt (Fraunhofer IWES, Oldenburg)
Detlev Heinemann (ForWind – Center of Wind Energy Research, University of Oldenburg)

Abstract: Offshore wind energy is a rapidly growing energy source. While in 2011 the globally installed offshore wind capacity was around 4.1 GW, this figure increased to about 12.1 GW by the end of 2015. Out of these, about 11.0 GW were installed within European waters. In Germany, the installed capacity increased from 1.0 GW to 3.2 GW only within the year 2015.
Limited mesoscale wind farm wake studies exist. Past studies were either using idealized or unrealistically large wind farms. Within this presentation, realistic wind farm scenarios taking the as is as well as all planned wind farms within the German Bight into account are simulated using the Weather Research and Forecasting (WRF) Model. Potential further wind farms in a “dormant” phase are not taken into account as it is undecided or unlikely that they will be built in the new future.
The wind farms in this study are simulated utilizing the wind farm parametrization described in Fitch et al, 2012. The study compares three scenarios and their impact on the wind conditions over the German Bight comparing a full year of simulations with a horizontal resolution of 2km. The impact of these wind farms on the wind resource over the German Bight are discussed. In addition, the study also shows how large-scale wind direction changes influence the wake propagation of large wind farm clusters.


Session 3 – Models and reality

Do Uncertainties and Losses in Pre-Finance Wind Farm Energy Assessments Fit to Reality?

Presenting author: Kai Moennich, UL International GmbH – DEWI

Co-author(s): Susanne Horodyvskyy, UL International GmbH – DEWI
Barbara Jimenez, UL International GmbH – DEWI

Abstract: Although the accuracy of energy assessments improved over the years, we still see that on average pre-finance energy assessments rather overestimate than underestimate the later real production figures and, depending on the site’s flow complexity, some projects show large variations from the average in either direction. Additionally, we found a well-fitting P90 value with regards to the number of projects reaching the P90 value, but a mismatch for the P75 energy yield.
A comparison of pre-finance energy assessments with real production data at the Windeurope Summit Hamburg 2016 has been presented, which was focused on the analysis of deviations between real production data and pre-finance energy assessments. At the present work, some selected projects will be analyzed in detail in order to define the reasons of the deviations. Special focus is set on the uncertainties and systematic losses. One approach will be their adjustment according to findings from the analysis of the selected projects. In another, more general approach we will combine losses and uncertainties such that deviations for all P-values are minimized. We will compare these so-derived losses and uncertainties with the ones used in the pre-finance assessment and discuss the effects and reasonability of the different approaches.
The main question to be answered as output from this study will be: Is the current uncertainty and losses assessment reflecting reality or can we minimize deviations for different exceedence levels (P-values) by adjusting uncertainties and losses in the pre-finance assessments? Therefore, the answer will help improving future energy assessments and minimizing project risk.


Estimating icing losses at proposed wind farms and how to optimise wind farms in cold climate

Presenting author: Till Beckford, DNV GL

Co-author(s): Carla Ribeiro (DNV GL)

Abstract: A major challenge for wind energy development in cold climates is to understand the magnitude of energy losses caused by blade icing. The magnitude has been seen to exceed 50% during winter months, and surpass 10% over the course of a year. Such information is paramount to ensure the financial success of individual projects and indeed for the cold climate wind industry as a whole.
To this end, a number of sophisticated models (atmospheric and others) have been developed, but uncertainty remains regarding their accuracy, and validation of these is still limited.
Over the past few years, DNV GL has analysed operational data from over 25 wind farms in the Nordic region, along with meteorological data from over 80 measurement masts. These datasets have been analysed and used to derive empirical methods for predicting icing losses at proposed wind farms. A icing map of Sweden has been derived based on the country’s topography and observed icing losses over different hub height elevations, and a method of correlating pre-construction anemometer icing and post-construction losses has been improved and validated. In addition, the datasets have provided significant insight into the influence of the control strategy of the turbines in the magnitude of the icing losses. Analysis of the data has also suggested the existence of different icing climates across the Nordics and highlighted the significant inter-annual variability of icing events, which can be up to 65% for a site with 5% annual loss.
DNV GL shall present the latest results from the analysis of its Nordic data and the methods used to predict icing losses. Furthermore, using the operational datasets, insights into the driving factors in turbine icing are explored which can be fed back into the layout design process to improve the production of cold climate wind power projects.


Experiments in the New European Wind Atlas

Presenting author: Jakob Mann, DTU Wind Energy

Co-author(s): N. Angelou, 1; J. Arnqvist, 2; D. Callies, 3; E. Cantero, 4; R. Chávez Arroyo, 4; M. Courtney, 1; J. Cuxart, 8; E. Dellwik, 1; J. Gottschall, 3; S. Ivanell, 2; P. Kühn, 3; G. Lea, 1; J. C. Matos, 5, C. M. Veiga Rodrigues, 6; J. M. L. M. Palma, 6,

Abstract: The New European Wind Atlas (NEWA) project will create a freely accessible wind atlas covering Europe and Turkey, develop the model chain to create the atlas, and perform a series of experiments on flow in many different kinds of complex terrain to validate the models. This contribution describes the experiments of which some are completed while others are getting started. All experiments focus on the flow properties that are relevant for wind turbines, so the main focus is the mean flow and the turbulence at heights between 40 and 300 meters. Also extreme winds, wind shear and veer, and diurnal and seasonal variations of the wind are of interest. Common to all the experiments is the use of Doppler lidar systems to supplement and in some cases replace completely meteorological towers. Many of the lidars will be equipped with scan heads that will allow for arbitrary scan patterns by several synchronized systems. Two pilot experiments, one in Portugal and one in Germany, show the value of using multiple synchronized, scanning lidar, both in terms of the accuracy of the measurements and the atmospheric physical processes that can be studied. The largest of the experiments is currently taking place over two steep, parallel ridges in Portugal. It involved 53 meteorological masts and up to 17 scanning lidars among many other instruments. The experimental data for some of the completed campaigns are already used for validation of atmospheric flow models and will by the end of the project be freely available.


Modelling Wind Farm Energy, Revenue, and Uncertainty on a Time Series Basis

Presenting author: Taylor Geer, DNV GL

Co-author(s): Carl Ostridge

Abstract: DNV GL will present a new approach to modelling both wind farm net output and the associated uncertainty on an hourly basis. The model will then be used to better understand the uncertainty around wind farm revenue; additionally, the model can be used to evaluate the impact of various financial products.
The model includes consideration of wakes on a time series basis, a statistical approach to project availability modeling, and refined environmental loss predictions. A stochastic uncertainty model is also used generate many thousands of possible generation scenarios and to quantify the uncertainty associated with the wind farm output on an hourly basis. The time-variance of wind farm output and the price paid for electricity and, in some markets, the interaction between the two, adds a level of complexity to the estimation of project economics. Accurately modeling both wind farm output and the price paid for electricity on an hourly basis is crucial to assessing the level of revenue risk and therefore the economic viability of a project under-development.
The possible financial implications of the increased sophistication of the new modeling techniques will be demonstrated in a number of real-world examples.


Session 4 – Wakes

Measured power curves in waked flow: Possibilities and limitations

Presenting author: Nick Janssen, ROMO Wind

Co-author(s):

Abstract: Though it is well known that turbines under-perform in a wake, they still follow a well-defined power curve. ROMO Wind has shown this by performing power curve measurements on over 40 turbines that are, in different degrees of severity, affected by wakes from other turbines. Measurements were performed using ROMO Wind’s patented iSpin system.
Being able to measure a power curve in a wake gives access to significantly larger amounts of data for power performance studies. Depending on the size of the free sector and the dominant wind direction, it is not uncommon that the amount of valid data points increases with a factor five compared to traditional free-sector based filtering. As statistical uncertainty scales with the square root of the number of data points, this greatly reduces the uncertainty of a measured power curve. Additionally, five times more data means that a power curve can be measured five times faster.
The iSpin system returns a point measurement, i.e. it measures the wind speed at the centre of the rotor swept area. An assumption behind the iSpin measurement is that this point measurement is representative for wind speeds over the entire rotor plane. It was shown that this is the case for most wakes. There exists, however, a definite set of conditions where the abovementioned assumption is not valid. These conditions were determined to be high wind speed conditions in a stable atmosphere and a laminar inflow at the waking turbine.
It was concluded that iSpin power curves in wakes are robust and reliable, abovementioned conditions occur rarely and therefore don’t affect the AEP severely. Studying these conditions can however be extremely useful when learning about the nature of waked flow. The study shows the complexity of a wake and proves that no two wakes are the same.


Mesoscale Modelling of Wind Farm Wakes: Implications for large-scale planning

Presenting author: Jake Badger, DTU Wind Energy

Co-author(s): Patrick J. H. Volker
Andrea N. Hahmann
Hans E. Joergensen

Abstract: There is a pressing need to improve our understanding of wind farm wakes and wind farm wake recovery at the mesoscale (in the 1 km to 100 km range). This is because we must evaluate the influence of wind farm wakes on the environment and neighbouring wind farms. Wind farm wakes can impact neighbouring farms’ energy yield, operations, and turbine loads, and are therefore important in the assessment of wind farm economics. Understanding of wind farms wakes is also relevant to investigate claimed limits to the wind power extraction as investigated in Adams and Keith (2013), Miller et al (2015) and Miller and Kleidon (2016).
Our latest mesoscale wake modelling results show that the influence of wind farm wakes on power production is dependent on the local wind climate, wind turbine technology, installed capacity density and surface roughness at the wind farm site. Consideration of these aspects leads us to a new approach for deciding on wind farm parameters based on mesoscale wake recovery. For wind farms in a high surface roughness setting, very large wind farms with low installed capacity density and placed back-to-back can be the rational choice. For wind farms in a low surface roughness setting, smaller wind farms with higher installed capacity density and separated with turbine free wake recovery zones can be the better option. The presentation will also discuss the influence of trends towards turbines with lower thrust coefficients, and the impact of boundary layer stability on the wind farm wake recovery.
The presentation will give the audience results and outlook of the application of new mesoscale wake modelling that is highly relevant for the planning of long term and large scale exploitation of wind energy in the future.


Wake Modeling in the Time Series Domain

Presenting author: Scott Eichelberger, Vaisala

Co-author(s): Mark Stoelinga (Vaisala)

Matthew Hendrickson (Vaisala)
James McCaa (Vaisala)

Abstract: Wake modeling represents a key input to net energy production estimates, and a significant source of uncertainty in those estimates. Wakes depend on many factors including turbine characteristics, ambient turbulence, wind shear, speed, and direction. Standard wake models attempt to consolidate this variability by applying the wake model to a limited sample from the wind rose, speed histogram, and ambient TI. Another approach is to run a wake model on a many-year time series with hourly granularity, to capture the true distribution of combined factors that affects wakes. This study describes and validates a model that uses this approach. Wakes for each turbine are computed individually and interact in a physically consistent way, precluding the need for posterior models to adjust the effects of wakes from large turbine arrays. The full system has been calibrated using wind farm met and turbine production data under a wide range of conditions, including a broad span of turbulence and stability regimes. This presentation focuses on a specific validation study of the model at a large wind farm on relatively flat terrain, to compare the behavior of time-averaged versus time-varying estimates.


Comparison of the wake effect and the energy production resulting from a comprehensive annual energy production assessment methodology with SCADA data from an operating offshore wind farm

Presenting author: Guillaume Angot, EDF R&D

Co-author(s): Eric Dupont (EDF R&D); Raphaël Bresson (EDF R&D); Laurent Beaudet (EDF R&D)
Sami Barbouchi (EDF R&D UK Centre); Markos Mylonas (EDF R&D UK Centre); Marilia Giannakopoulou (EDF R&D UK Centre)

Abstract: Reducing the uncertainty on the estimation of wake losses is crucial to minimizing the cost of energy, especially for offshore wind.
In this context, EDF-R&D has developed an annual energy production (AEP) assessment methodology based on a chain of mesoscale and full CFD models, in parallel of the operational tools commonly used.
As a first validation case, we present a comparison between simulation results and measurements from an operating wind farm. As the available data set covers only a short period, the goal is not to compute the AEP but to compare the calculated energy production with the SCADA data over this period. The same exercise has been performed on two separate years. The data was filtered so as to perform the comparison on the best quality data.
Regarding the mesocale model, WRF was used. Thanks to a clustering step (using the SOMs method), the number of situations to simulate on the local scale is reduced to 100. Each situation is simulated by the CFD code Code_Saturne developed by EDF R&D.
Results show a good agreement with the measurements. Concerning the net production, the difference between simulations and measurements are -4% for phase 1 and +0.5% for phase 2. The wake measurements show a wake of 11.6% and 11.3% for the two phases, while our simulations indicate 13.1% and 13.5% respectively.
A significant part of the difference between the simulation and the measurements can be attributed to a “stand-by wind turbine effect” (it has been partially quantified) that is not directly related to our methodology.
Finally, the error due to the differences between WRF and the measurements was quantified: error on overall wake effect is significantly reduced to 0.2 and 1.4 points (for the two phases) when adjusting inlet conditions of CFD simulations to measurements. Further future improvements are presented.


Improving confidence in wake predictions through operational validations

Presenting author: Carla Ribeiro, DNV GL

Co-author(s): Marie-Anne Cowan, DNV GL

Abstract: Improved confidence in modelling the wake effect at a wind farm project provides substantial value for project design, project financing, and project operations.
Wakes and wind flow vary substantially depending on the stability of the atmosphere. To accurately capture the flow conditions at a site, a wake or wind flow model must be able to capture these variations and ideally should be validated to give confidence in the model’s ability to predict for a range of conditions. In reality only a limited number of validation datasets are available for such validations, which has exposed the industry to the risk of “over-fit” models which produce erroneous results when extrapolated to new projects
The performance of a wake model is best determined through a comparison of modelled results to measured results on an operational project. The implementation of this process is complicated by variations in project operations and wind flow. DNV GL will propose a comprehensive wake validation procedure to improve the standardization of wake validation results and address the risk of over-fitting by examining the sensitivity of wake models in various conditions. DNV GL will then demonstrate the application of the procedure by applying it to a number of models at multiple operational projects, considering pre-construction measurements and operational data where possible.
The primary result of the work is a comprehensive method of conducting a wake validation, including a sensitivity analysis to ensure results are as broadly applicable as possible. With a standard method of validating wake models and addressing the risk of model “over-fit,” the industry will be a position to substantially increase confidence in wake model results. The secondary result of the work will be a statement of accuracy of a number of industry standard wake models, as well as initial results for next generation wake models.


Session 5 – LiDARs

Results of the OWA Dublin Bay Scanning LiDAR Trials

Presenting author: Michael Stephenson, Carbon Trust

Co-author(s): Peter Stuart, RES; Simon Feeney, RES; Lee Cameron, RES; Alex Clerc, RES
Megan Smith, Carbon Trust
Mathieu Boquet, Leosphere
Keith Barr, LMCT

Abstract: The Offshore Wind Accelerator has carried out a four month trial of two pairs of scanning LiDAR devices : Leosphere and Lockheed Martin. These devices were installed in Dublin Bay and validated against two vertical profiling LiDARs to determine their accuracy and precision. The devices were set up to measure wind speeds in a nominal offshore wind farm within the bay, which stretched to over 10km from the devices at its furthest point. Each device was to measure wind speeds at a range of heights at eight set points offshore and validated against fixed vertical LiDAR. The system suppliers were left to devise their own scanning patterns and algorithms to process the data, and the trial was carried out completely blind with validation by RES as an independent third party.The results show phenomenal accuracy at ranges never tested before in offshore wind (>13km). Results of both single and dual Doppler will be shown, which demonstrate that whilst single Doppler may not be viable for offshore wind, dual Doppler has given very convincing results for both manufacturers.
The results will also show the effect of deploying scanning LiDAR on a potential annual energy production, with uncertainty of the P90/P50 ratio being brought down by between 1-2%. The main conclusion of the trial is that scanning LiDAR can work for offshore wind and can produce very accurate wind speed measurements at very long ranges. There is further work to be done in order to improve some of the algorithms of the devices themselves but for certain sites this technology can have real cost reduction benefits.


Floating Lidars better accuracy

Presenting author: John Slater, Ecoperspective

Co-author(s): Breanne Gellatly
Director of European Operations – AXYS Technologies
Breanne Gellatly [[email protected]]

Abstract: Since the first floating lidars were deployed in 2009 as an R&D project, the technology has come a long way. In 2011 the Carbon Trust set the first standard for validation of the technology which has been followed by a series of validation campaigns. Over the last five years, the technology has not only been site validated but has reached the point where environmental sensitivities like sea state are being eliminated due to the wide pool of evidence supporting the accuracy of the technology in many different conditions. During this presentation we will show evidence from a variety of campaigns which proves the lack of impact of certain environmental conditions leading to reduced uncertainties. We will also discuss the potential to reduce measurement reference uncertainty using an example from our five month validation at the FINO 1 meteorological mast.
This presentation will also explore longer-term applications of floating lidars, including use for power performance testing and as an onsite weather station to feed into operational decision making tools such as weather forecasting and wind farm fatigue forecasting.
The audience this presentation is targeting includes environmental permitting specialists, wind resource engineers, development-stage wind farm investors, foundation designers, O&M planners and asset managers. This presentation will provide to the audience best practice guidelines on how to minimize your data uncertainty through redundancy of sensors and power supplies as well as through validation campaign design. The audience will leave this presentation with a better understanding of 1) what floating lidars can measure beyond wind speed and wind direction 2) how FLIDAR uncertainty has reduced in line with its maturity as a technology and 3) how to design the optimal wind resource assessment campaign using floating lidars.


Assessment of repeatability and stability of Lidar performances

Presenting author: Paul Mazoyer, Leosphere

Co-author(s): Clement HERMASZEWSKI : Leosphere; Florian REBEYRAT: Leosphere

Abstract: Repeatability (different units measure the same) and stability (a given unit measure the same over time) are paramount for wind sensors since it ensures consistency of measurements along projects. The assessment of the repeatability and stability of Lidar over time requires a set of representative datasets to statistically cover the device life. It requires as well as a stable validation procedure which ensure that datasets are obtained with the same environment and criterions.
Each new unit or maintained unit at Leosphere goes through a validation process which includes a several day comparison with a reference Lidar. Comparison issues key performance values such as mean and scatter of winds speed deviations, mean and scatter of winds speed directions or data availability statistics. For 10 years, Leosphere has validated 726 times Lidar on his site. The size of the database is deemed big enough to represent fairly the Lidar characteristics and presentation will show the results of stability and repeatability with this database. For example, a standard deviation of the mean deviation between Lidars and the reference of 0.04 m/s is identified representing less than 0.5% discrepancy at 10 m/s for every Lidars. Besides, validation procedure is presented to validate the consistency of the method.
This presentation will also be put in the light of the study carried out by DTU and DNV GL and presented at EWEA annual event 2014 “How stable are Lidars over time” : a 1% discrepancy is found between calibrations and this is explained by cup variation and environmental variation. The deviation found here is 0.5% and could be of interest as input for project of Lidar calibration without cups.


Accuracy of dual-Doppler lidar retrievals of near-shore winds

Presenting author: Nikola Vasiljevic, DTU Wind Energy

Co-author(s): Michael Courtney

Abstract: In this presentation the accuracy in retrieving horizontal wind speed and wind direction using a dual-Doppler lidar system will be described. First, the line of sight wind speed uncertainty is described followed by the detailed description of the various sources of errors in laser beam pointing with a particular focus on static errors. A methodology for assessing static pointing errors is presented accompanied with results from the method implementation. Afterwards, mathematical relations for the horizontal wind speed and wind direction uncertainties are derived. For the end, the derived mathematical relations are implemented for the uncertainty assessment of the dual-Doppler retrievals of near-shore winds that took place during the RUNE experiment.


Results of the OWA Dublin Bay Scanning LiDAR Trials

Presenting author: Alex Clerc, RES Ltd.

Co-author(s): Peter Stuart, RES; Simon Feeney, RES; Lee Cameron, RES
Michael Stephenson, Carbon Trust; Megan Smith, Carbon Trust
Mathieu Boquet, Leosphere
Keith Barr, LMCT

Abstract: The Offshore Wind Accelerator has carried out a four month trial of two pairs of scanning LiDAR devices : Leosphere and Lockheed Martin. These devices were installed in Dublin Bay and validated against two vertical profiling LiDARs to determine their accuracy and precision. The devices were set up to measure wind speeds in a nominal offshore wind farm within the bay, which stretched to over 10km from the devices at its furthest point. Each device was to measure wind speeds at a range of heights at eight set points offshore and validated against fixed vertical LiDAR. The system suppliers were left to devise their own scanning patterns and algorithms to process the data, and the trial was carried out completely blind with validation by RES as an independent third party.The results show phenomenal accuracy at ranges never tested before in offshore wind (>13km). Results of both single and dual Doppler will be shown, which demonstrate that whilst single Doppler may not be viable for offshore wind, dual Doppler has given very convincing results for both manufacturers.
The results will also show the effect of deploying scanning LiDAR on a potential annual energy production, with uncertainty of the P90/P50 ratio being brought down by between 1-2%. The main conclusion of the trial is that scanning LiDAR can work for offshore wind and can produce very accurate wind speed measurements at very long ranges. There is further work to be done in order to improve some of the algorithms of the devices themselves but for certain sites this technology can have real cost reduction benefits.


Floating Lidars better accuracy

Presenting author: Breanne Gellatly, AXYS Technologies

Co-author(s): John Slater Engineering Director Ecoperspective [email protected]

Abstract: Since the first floating lidars were deployed in 2009 as an R&D project, the technology has come a long way. In 2011 the Carbon Trust set the first standard for validation of the technology which has been followed by a series of validation campaigns. Over the last five years, the technology has not only been site validated but has reached the point where environmental sensitivities like sea state are being eliminated due to the wide pool of evidence supporting the accuracy of the technology in many different conditions. During this presentation we will show evidence from a variety of campaigns which proves the lack of impact of certain environmental conditions leading to reduced uncertainties. We will also discuss the potential to reduce measurement reference uncertainty using an example from our five month validation at the FINO 1 meteorological mast.
This presentation will also explore longer-term applications of floating lidars, including use for power performance testing and as an onsite weather station to feed into operational decision making tools such as weather forecasting and wind farm fatigue forecasting.
The audience this presentation is targeting includes environmental permitting specialists, wind resource engineers, development-stage wind farm investors, foundation designers, O&M planners and asset managers. This presentation will provide to the audience best practice guidelines on how to minimize your data uncertainty through redundancy of sensors and power supplies as well as through validation campaign design. The audience will leave this presentation with a better understanding of 1) what floating lidars can measure beyond wind speed and wind direction 2) how FLIDAR uncertainty has reduced in line with its maturity as a technology and 3) how to design the optimal wind resource assessment campaign using floating lidars.



Floating LiDAR uncertainty assessment in an operational context

Presenting author: Hugo Herrmann, EDF Energy R&D UK Centre

Co-author(s): Cédric Dall’Ozzo (EDF Energies Nouvelles), Camille Bastide (EDF Energies Nouvelles)

Abstract:
Introduction

An AXYS FLIDAR 4M was installed near the Fécamp IEC met mast for four months on EDF Energies Nouvelles’ future offshore wind farm in the French Channel. The measurement uncertainty of the FLIDAR 4M has been compared to an offshore short met mast with a fixed LiDAR (IEC equivalent) at several heights. The traditional accuracy and availability KPIs have been calculated. In addition, EDF R&D have gone a step further to assess mathematically the uncertainty of the FLIDAR during this 4-month measurement campaign.

Approach

To calculate the wind speed uncertainty of the floating LiDAR, a tuned version of the method described in the Annex L of IEC 61400-12-1 Ed.2 has been used. This method calculates the binned wind speed uncertainty of the floating LiDAR based on 10-minute average measurements at heights along the blade swept area.
Main body of abstract

The main objective of the present work is to demonstrate that a wind speed uncertainty below the range 4%-7% stated in the Carbon Trust roadmap for the commercial acceptance of floating LiDAR technology (2013) has been reached on a commercial project using the AXYS FLIDAR 4M floating LIDAR.

In a first part of the presentation, the traditional accuracy and availability KPIs results will be presented. Next, the results of the wind speed uncertainty calculation will be explained.

Conclusion

EDF R&D show how a calculation inspired from the Annex L of IEC 61400-12-1 Ed.2 has resulted in a FLIDAR 4M uncertainty level below 4% and how a large part of floating LiDAR uncertainty comes from the reference itself. To conclude, EDF R&D will explain the need to find alternative ways to validate floating LiDARs (and LiDARs in general), depending less on reference met masts.