Abstracts

Supporters

Exhibitors

Sponsors

Deutsche Wind GuardDeutsche Wind GuardLLoyds RegisterULUptake

Media Partners

RENews

SESSION 1: Deriving value from Big Data

Power Curve Evaluation Based on Limited Wind Speed Range

Presenting author: Axel Albers, Deutsche WindGuard

Abstract

Wind turbine power curve tests must cover a wind speed range from below cut-in wind speed to clearly above rated wind speed according to the current testing standards IEC 61400-12-1 and IEC 6140-12-2. A procedure has been developed which allows reducing this wind speed range to only about 50% of rated wind speed to 80% of rated wind speed (roughly 6 m/s to 10 m/s), while still the power curve for the entire wind speed can be re-constructed with reasonable accuracy. The procedure makes use of the turbulence normalisation procedure as given in Annex M of IEC 61400-12-1 with certain modifications and has been successfully tested on the basis of a set of power curve tests.

The developed procedure constitutes an important milestone for power curve tests on the basis of new wind measurement technologies like airborne sensors or scanning LiDARs. It is further useful in combination with traditional measurement techniques where only a small wind direction sector is applicable for the power curve test or for reducing the measurement period.


The EME – Power Forecasting Improvement Research Project

Presenting author: Joana Mendes, UK Met Office

Co-authors: Paul Smith (ECMWF); Leo Hume-Wright (UK Met Office); Hui Yang (Envision Energy Shanghai)

Abstract

Project Description: The EME – Power Forecasting Improvement research project currently consists of a multidisciplinary alliance (Envision Energy, UK Met Office, ECMWF, Wind Europe & Aarhus University). Its aims to demonstrate the reduction in error possible in renewable power forecasting through the use of novel wind and solar forecast data and upgrades to current ‘best practive’ methodologies. This research alliance will demonstrate the value of ensemble and probabilistic forecasting and further enhance the use of meteorological forecasts in the renewable energies sector. The expected impact of the research project is a limitation of the risks related to grid instability, unpredictable energy prices which will ultimately benefit European citizens by decreasing the levelized cost of energy produced by wind and solar technologies.

Learning Outputs: The oral presentation will target the initial outputs from the EME – Power Forecasting Improvement research by introducing the following findings:

  • A comparative study of forecasting methods for weather conditions at 50 renewable energy sites (40 wind farms and 10 solar farms) spread all over China.
  • Revealing numerical weather prediction (NWP) models’ abilities to determine future energy output from renewables.
  • Examine the impact of NWP for various locations, in order to suggest how and which local specifications affects the forecasting accuracy

The output is expected to inform the audience about state-of-the-art methods, and further design future research requirements for weather forecasting. The oral presentation will also inform the audience about the second phase of the EME – Power Forecasting Improvement research project. The second phase will reveal how improved weather forecasting methods can assist the energy industry in limiting the risks related to grid instability, unpredictable energy prices and ultimately benefit European citizens by decreasing the levelized cost of energy produced by wind and solar technologies


Combining IoT and Big Data Analytics to improve Wind Farm Performance

Presenting author: Bruno Pinto, Sereema

Co-authors:

Abstract

Digitalization brings a load of revolutionary ingredients to the wind industry, thus completely transforming the way wind farms optimization is at work. For industry in general, one of the key elements dwells on the emergence of IoT. With a huge load of data already coming from the SCADA, will these technologies benefit the wind industry or just add more useless data to the stack? In other words, is combining IoT data with SCADA data a truly worthy idea?

To answer this, Big Data Analytics were applied to heterogeneous data sources such as traditional SCADA, met mast and Lidar data on the one hand and IoT sensors data on the other hand. The volume and velocity of data, transferred directly from the data acquisition feed, ranges from a frequency of 10 minutes for the SCADA averages up to several hundreds Hz for the IoT sensors. The amount of data collected and streamed nears 1 000 values per second and per wind turbine. By adapting cloud-based algorithms to the wind turbine specificities, this approach brings scalable results in terms of predictive maintenance, early default detection and performance optimisation.

Applied to a fleet of 50 wind turbines, each equipped with IoT sensors, this approach provided tailored KPI enabling easy focus on underperforming turbines.

As another example, an improvement of the wind sector management of 4 operating wind turbines was obtained resulting in a 5% AEP increase.


Value from free-text maintenance records: converting wind farm work orders into quantifiable, actionable information using text mining

Presenting author: Erik Salo, University of Strathclyde

Co-authors: Dr David McMillan

Prof Richard Connor

Abstract

The aim of this project is to demonstrate how data and text mining techniques can help wind farm operators to extract unique, quantifiable, site- and asset-specific maintenance information from historic work orders. Understanding how maintenance efforts have been distributed in the past can help develop a more evidence-based maintenance strategy for the future in terms of labour intensity, budgeting and logistics of spare parts. However, work order records – where significant information is entered by a human in the form of free text – can present a particularly complex data source for analysis.

Our approach introduces a novel combination of machine learning techniques supported by a database of domain vocabulary and expert judgement. Significant focus is on term recognition, aided by spelling error correction and semantic matching of synonyms and abbreviations. Task descriptions can thereby be classified by meaning, not just the words present. In the first instance this creates a frequency distribution of all the different tasks carried out. Categorical data can then be extracted about maintenance of different functional locations and subsystems, as well as the occurrence of different failure modes.

Data from major onshore wind farms in Scotland was used to test our approach against undertaking a similar analysis manually. Potential savings were identified on the order of weeks of effort, or £-9k in labour cost per wind farm, in addition to the benefits of an improved maintenance strategy.

The remaining challenges mainly lie in increasing accuracy and reducing operator input. These are being addressed by our continued research, but also provide opportunities for collaboration and standardisation across the wind energy industry to maximise the value of data.

SESSION 2: Enhancing turbine performance using data

From real wind to actual power output

Presenting author: Samuel Davoust, GE Renewable Energy

Co-authors: Thomas Fric, Minesh Shah

Abstract
  • Wide variation of wind characteristics influencing power performance occur across sites, and also within a site. These characteristics include air density, full shear profile, turbulence intensity (TI), wind veer and upflow.
  • Power curves can be restricted to be valid to specific wind conditions, which are commonly named the Inner Range. However, it is frequent for wind conditions to be outside of the Inner range, called Outer range.
  • To reduce power performance uncertainties, power curve should continue to evolve to increase accuracy and the range of validity. We conduct a case study to evaluate three approaches:
    1. Adding an Outer Range power curve to account for typical performance in abnormal conditions
    2. Extend the Inner Range by predicting a site-specific annual average power curve from wind resource assessment analysis
    3. Translate the Inner Range PC to any other wind condition using an adjusted wind profile and turbulence intensity corrections adapted from the IEC-6100-12-1 Edition 2 [2]
  • These power curves publication approaches are evaluated in terms of accuracy and complexity for pre-construction annual energy production (AEP) assessment and for power curve field testing.

Monitoring O&M data from 11,443 turbines worldwide: approaches, challenges and lessons learned

Presenting author: Peng Hu, Longyuan Beijing Wind Power Engineering Technology Company

Co-authors: Xu his, Li Shaowu, Liu Ruihua

(Longyuan Power Group Corporation)

Abstract

Longyuan power Group Corporation is the largest wind power operator in the world, now running 11,443 WTGs in China, South Africa and Canada, with capacity over 17GW, in 289 wind farms, totaling 93 types from 23 OEMs. Longyuan has started its data analysis from vibration and lubrication monitoring in 2009, and started performance analysis based on SCADA data in 2011.

Gradually Longyuan is now building a big data platform to face the challenges and trying to dig gold from the mine of data. We are now illustrating the performance and reliability of every wind turbine, analyze power performance, composition of the loss of power, causes of faults and highlight the most frequent faults and faulty or degraded WTGs.

In the integration and analysis of the O&M data, we met many challenges: the variety of the OEMs, the different WTG components and configuration, the low manual data quality, and the huge data.

We’ve got up to nearly 70 types of WTG, the faults or states are different in different types, the number of the states are 200~400 each. As a support to raise the accuracy of the fault data, Longyuan has built a special deviation response mechanism to collect the fault data: the Longyuan O&M system collects the 10min data from wind farms, the local operator of the wind farm would review the fault reports and response to the O&M system the deviation of the states of the WTG, these data accumulates to 3 million texts every year. At the very beginning, these manual reports are dealt by experts, now based on the experts’ experience, we’ve developed a Text Information Classification method to deal with the manual reports automatically, which saved manpower and time cost greatly.

Longyuan brought the Production-based availability into wind farm efficiency assessment method. From WTGS user’s view, combined with the ongoing power performance measurements, met mast data, 10 min data and the standardized fault data, Longyuan is able to divide the loss of power into classes. In China, the Grid plays an important part in the power system, it decides the curtailment ratio and plan the O&M of the transmission line. The O&M of the northern wind farms are also impacted greatly by the icing lines in the wind farm. To distinguish the impact elements and its influence, we made further detailed category.

China is imbalanced in economy and power distribution. The northern farms are greatly impacted by the curtailment, while the southern farms are impacted less, which brought great confusion in the energy availabilities, we developed new algorithm to evaluate the availabilities excluding the curtailment, in order to offer a same base line for different turbines in different regions. Furthermore, we brought in new calculation methods for base potential energy production, using the best-performance turbines in the farm as the base line, trying to compare the wind farms which doesn’t have ideal met masts and topographic features.

The efficiency analysis data and reports are promoted to the in-field management, every season the power curve census help every wind farm to find the degraded wind turbines. The degraded data was labelled in the analysis process, we use the labelled data to carry out further analysis, with the models of gearbox temperature, wind wine, etc. to confirm the cause of the degradation, or to forecast and prevent the degradation.


Presenting author: Jesus Navarro, DNV GL

Co-authors: Carlos Albero; Amilcar Zambrano

Abstract

PURPOSE OF THE WORK

Wind energy has been racing in the last 30 years in a constant push for innovative technology, more reliable, and more efficient wind turbines. Despite this history has been a success, the wind sector has also faced the challenge of demanding sites, new technologies and a maturity trend in which not minor hurdles have been jumped. The analysis of all these events, has made the sector strong, and more reliable, but best practices should be followed to make the most of these episodes. In DNV GL experience in operational wind farms and specifically in the area of the Root Cause Analysis, it is clear that a solid methodology allows us to have more reliable and solid conclusions that would allow us to move forward providing the sector with a reputation that will stand the test of the most demanding conditions.

We will present a couple of real examples of RCA analysis and conclusions, with a double target:

  1. Presenting the typical structure of a detailed RCA after a detected failure;
  2. Considering 1 as a lesson learnt to prevent a possible future failure, taking also into account the possible information and knowledge limitations detected in 1 to solve/mitigate these limitations in the future.

All this knowledge regarding the failures analysis and how to:

  1. Study and
  2. Solve these issues is even more important in the current context of very new technologies and, on the other hand, the possible life extension of old wind turbines.

This will be depicted through a business case of a real RCA performed by DNV GL in the context of our work.

APPROACH

An RCA process normally consists in 5 steps (according IEC 62740): Initiation, establishing facts, analysis, validation and presentation of results. The RCA is considered a dynamical systematic investigation that will receive feedback in some of its steps.

DNV GL implements the BSCAT methodology (Barrier-based SCAT (Systematic Cause Analysis Technique) in the RCA analysis, in order to use a systematic process and also to give a clear and graphic representation of the analysis and results.

Wind turbines are quite complex systems, and catastrophic failures can depend on several factors. Therefore, the analysis of such as a failure should be conducted by a multidisciplinary team and with an appropriate knowledge.

A RCA is typically focused on cause, responsibilities and mitigation measurements, amongst other. But the whole process should be seen as an opportunity. If a catastrophic failure is appropriately analysed could give loads of positive outcomes:

Improved manufacturing skills

Improved O&M procedures and operations

More reliability in the fleet of wind turbines and also:KNOWLEDGE.

As we have already mentioned, this knowledge can be used not just to prevent the same failure at the same wind farm in the future. This knowledge is a very valuable tool to be used as part of the wind farms operation, enhancing wind turbines performance.

DNV GL has acquired a significant experience in RCA analysis, increasing its knowledge in operational wind farms. This deep knowledge is used to support wind farms stakeholders.


Production gain assessment due to improved dynamic yaw control settings

Presenting author: Nicolas Quievy, ENGIE

Co-authors: Marc Van Caillie

Thomas Museur

Paul Poncet

Nicolas Girard

Abstract

Wind turbines need to face the wind in order to maximize the amount of produced energy and decrease the fatigue on the rotor. When it is not the case, there is an error called yaw misalignment. This misalignment can be static, i.e. permanent misalignment of the rotor towards any given constant wind direction, or dynamic, i.e. the rotor does not track efficiently the continuous varying wind direction. In the latter, the yaw drive is reacting too slow or too fast. Therefore an optimum between production increase and mechanical loading of the yaw drive has to be found. It is recognized in the wind industry that improvement of the dynamic yaw control can produce a gain in production. A potential improvement was detected in ENGIE fleet. For this purpose, a retro-engineering of the yaw controller was performed. The model was constructed and validated using 1 Hz turbine data. The model was used to determine new optimized control settings in order to increase the production by keeping constraints on the mechanical loading of the yaw drive. The optimized settings were introduced in the software controller of a turbine end November 2017. The quantification of the relative gain in production is ongoing.

SESSION 3: Lowering operational costs

Wind Energy Yield Assessment, A look-back analysis on ENGIE plants

Presenting author: Simon Courret, ENGIE

Co-authors: Olivier Moreau

 

Abstract

A robust and sustainable wind energy industry requires a predictable return on investment. However wind energy plants are underperforming compared to what was initially predicted for commitment decision. What are the reasons behind: an inaccurate energy yield assessment (EYA)? the wind is not blowing this year (again)? turbines do not perform well? etc. In this context, ENGIE made a look back analysis on the EYA of 99 wind farms. This investigation aimed improving the accuracy of the EYA for new projects and the assumptions on losses, unavailabilities, etc. made in the business plans. The actual production over an observation period was corrected by taking account of losses due to (un)controllable unavailability, turbine underperformance and wind resource anomaly. The corrected production was then compared to the estimated production at maximum capability over the reference period. The difference constitutes the error related to the EYA assessment. The results showed an overestimation of the EYA in the past. With the integration of new EYA specifications in the internal business case guidelines the gap was reduced. Understanding the potential causes of the gaps by taking account of factors impacting the EYA (topography, roughness, wake effect, etc.) was attempted. Complex relief, high roughness, low ratio between measurement and hub height tend to increase the gap. However, even a high amount of plants were considered, the diversity of situations and practices makes the identification of errors very challenging. The developed look-back analysis is regularly used today by ENGIE and is continuously integrating new wind energy plants to make it more reliable and profitable.


Guideline on Operational Data Assessment

Presenting author: Martin Strack, Deutsche WindGuard Consulting GmbH

Abstract

The assessment of operational losses and determination of energetical availability plays a major role in evaluation of status and performance of an asset. In Germany, for wind turbines operating according to the new feed-in-law (EEG 2017), such evaluation of operational data has become even a legal requirement, to be performed after 5, 10 and 15 years of operation in order to identify and quantify different energy yield losses. As certain losses can lead to an adjustment of the feed-in tariff or even to a repayment of revenues, this assessment can have a major impact on the balance of an asset.

For this reason, an expert group from experienced participants from wind farm operators, wind turbine suppliers and consultants has been founded, working on the requirements and methodology for such assessment and preparing a guideline which was published in January 2018 (FGW TR10). The author of this abstract has played a major role in the expert group, by leading the subgroup which worked out a methodology for energy loss quantification, which should accomplish this task in reproducible, robust and accurate way. The defined methodology adopts a big data approach the author has been developing on basis of extensive experiences gained in operational data assessment for on- and offshore projects [1][2][3][4].

The proposed presentation will explain the general achievements of the TR10 guideline as well as the methodological approach and application experience of the loss quantification procedure. The first step of the guideline is the definition if the scope of data and information to be provided, and a scheme to derive the relevant turbine status from turbine specific status and error log information. When interpreting the turbine status information, the definitions from IEC 61400-26-1 are applied in order to consider international standardisation and ease the generalisation of the procedure.

For explanation of the methodology for energy loss quantification, the findings from measurement campaigns and operational data evaluations at the largest wind energy project planned in Europe [1][3] will be briefly shown, complemented by a summary of results from extensive offshore operational data assessments and measurement campaigns [2][4] which address especially important questions on usability of nacelle anemometer data for this purpose. It will be shown how these experiences have been used for designing a procedure for energy yield loss calculation on basis of a big data approach of SCADA- and external meteorological data.

The developed guideline and defined calculation procedure may be the most comprehensive and precise guideline on requirements and methodology for operational data assessment available today. So, as learning objective, the proposed talk may provide valuable guidance and triggers for the own jobs of the delegates.

References:

[1] Martin Strack et.al.: Big data approach of wind resource and operational data analysis in cold climate. Presentation held at WindEurope Summit 2016, Hamburg, Germany.

[2] Martin Strack: Quantification of grid related energy yield losses: Experience gained in German offshore projects. Contribution to WindEurope Summit 2016, Hamburg, Germany.

[3] Martin Strack et.al.: Big data approach of wind resource and operational data analysis. Presentation held at German Wind Energy Conference (DEWEK) 2017, Bremen, Germany.

[4] Martin Strack: Quantification of grid related energy yield losses: Experience gained in German offshore projects. Presentation held at German Wind Energy Conference (DEWEK) 2017, Bremen, Germany.


Towards Revenue Assessment

Presenting author: Henrik Sundgård Pedersen, EMD International

Co-authors: Wiebke Langreder, EMD International A/S

Abstract

The later years have seen a move from time based availability towards energetic availability in order to find the most appropriate form of addressing losses covered by contractual agreements between the owner and the OEM.

Even though energetic availability obviously is a more conclusively linked to production numbers, it might not necessary be easily translatable to lost revenue in a spot market, particularly in mature markets with a high penetration of wind energy generation. With focus on maintaining a high energetic availability, maintenance will usually take place in low wind periods. On the other side operation failures will often occur at high wind speeds thus in periods with lower spot market prices. This connection between production and spot market prices, together with compensated downtimes, such as utility deregulation, create a rather complicated picture.

We examined for a number of cases the dependence between operational losses, time and energy based availability in connection with spot market prices and the resulting impact on the revenue stream.

For this purpose, SCADA data of a number of projects has been analyzed. Losses due to availability and other losses have been quantified. If the turbine production is sold on the spot market, or on a market with variable tariffs, the loss time series can be combined with the spot marked time series resulting in a price tag on the lost production. This results in a comparison of time based, energetic and revenue availability of the wind turbine.


Improving offshore wind farm performance through novel access strategies

Presenting author: Fernando Sevilla Montoya, DNV GL

Abstract

Due to the challenges of accessing offshore wind turbines, an optimal O&M strategy is critical in achieving a high level of wind farm performance. One key area of the O&M strategy is the crew transportation and turbine access solution for the project.

In order to increase wind farm accessibility, improve project performance and reduce operational costs, crew transportation and turbine access solutions are scrutinised during the operational phase of the project. The choice of an access strategy is highly site specific, depending predominantly on the distance to the site, number of turbines and site conditions. Current access strategies, such as the deployment of standard catamaran Crew Transfer Vessels, can achieve accessibilities at offshore projects in the North Sea in the range of 50 to 70%.

As technology has evolved, new vessel technology and methods of access have emerged and are now available for operational projects, enabling potential benefits to accessibility and hence project performance, that were not previously attainable. Based on detailed time-domain and Monte Carlo simulations, DNV GL has scrutinised the benefits to project availability and O&M costs that novel access strategies, such as recently-trialled  vessels (Surface Effect Ship), helicopter support and use of Service Operations Vessels (SOV), could represent.

For this study, DNV GL has performed computer simulations that recreate all relevant real operational activities required for three operational offshore wind projects and estimated their current availability and operational expenditure.  DNV GL has then changed the O&M strategy via implementation of novel vessel and access technologies, enabling detailed analysis of the benefits that these technologies could unlock in offshore wind farm availability and operational expenditure.

The results of this work show that a significant increase in wind turbine availability in the range of 1-2% can be achieved through changes to current O&M strategies.

SESSION 4: Back to the future: from post-construction yield analysis to life extension

Surrogate Modelling of load and power output variation in wind farms

Presenting author: Nikolay Dimitrov, DTU

Co-authors: Anand Natarajan

Abstract

The wake-induced load effects experienced by wind turbines within a wind farm are important considerations for structural design and can impose restrictions on the wind farm layout or the turbine design choices. Wake-induced load analysis is a central part of wind farm planning and computation of the expected fatigue life of wind turbines.  . In the present study, a procedure for mapping wake-induced fatigue damage equivalent loads to a computationally efficient surrogate model approximation is defined and demonstrated. The dynamic loads are simulated using an aeroelastic software with the Dynamic Wake Meandering (DWM) model [1], [2]. Using the surrogate mapping function, the load variation can be efficiently estimated for a wind farm with arbitrary layout. We will also discuss the possible applications of the load map for addressing different challenges in wind farm design and operation. These are the following:

1) Wind farm layout optimization. The load mapping procedure provides quick estimation of loads as well as their analytical derivatives, which enables the use of gradient-based optimization techniques for various objectives such as to maximize net efficiency from the farm while minimizing fatigue damage.

2) Estimation of expected turbine fatigue lifetime. The rate of fatigue degradation can be computed for individual turbines within a wind farm based on their specific exposure to ambient conditions and wake-added turbulence.

3) Structural reliability analysis. Given the continuous, differentiable nature of the surrogate model, it can be readily implemented in structural reliability analysis which can estimate the site-dependent reliability for individual turbines.

4) Optimizing operation and maintenance. The individual wind turbine load and reliability predictions can be used for optimal maintenance scheduling based on estimated degradation rates.

In our presentation, we describe the load mapping procedure using an example with a specific wind farm, and we outline the framework of the applications defined above.  References

[1] C. Galinos, N. Dimitrov, T. J. Larsen, A. Natarajan, and K. S. Hansen (2016). Mapping wind farm loads and power production – a case study on horns rev 1. Journal of Physics: Conference Series, 753(3), 032010.

[2] G. C. Larsen, H. A. Madsen, K. Thomsen, and T. J. Larsen (2008). Wake meandering: a pragmatic approach. Wind energy, 11(4), 377–395.


Reconciling the past with the present: the impact of developments in reanalysis data on pre- and post-construction yield analysis

Presenting author: Nathan Hill, Lloyd’s Register

Co-authors: Matthew Zhang,

Nathan Hill and

Tim Crutchley

Abstract

Pre-construction assessments, by their very nature use the best practice models several years before any wind farm becomes operational. During that time the data, models and tools available develop at a rapid pace. This study examines in detail the impact of changes in the available reanalysis models over the last decade – and the impact on pre- and post-construction yield assessments.

In some regions analysts are observing changing predictions by as much as 3 % of the project energy yield on an identical like for like comparison – purely due to different versions of the reanalysis models used (the change from MERRA to MERRA-2 for example). This has significant impacts for the viability and value of these projects which is a great concern to all stakeholders in a wind farm.

The study focusses on four primary reanalysis sources, MERRA, MERRA-2, ERA-Interim and ERA 5. A comparison is made with known consistent reference sources across the globe in order to benchmark each source’s performance and further understand how this performs over time.

The results and conclusions made enable an increased level of understanding of the accuracy of each of the reanalysis data sources – and how bias varies over time and geographical location. Project owners, investors and consultants working on those projects will be able to better benchmark their asset performance based on the presented insights into long-term wind trends.


Leveraging Envision Energy’s EnOS IoT platform towards automated post-construction yield analysis for benchmarking and improving the accuracy of the Greenwich Systems yield predictions

Presenting author: Gregory Oxley, Envision Energy

Co-authors: Fatma Demet Ulker, Senior Researcher, Envision Energy

Xing Sun, Data Science Analyst, Envision Energy

Jianzhi Li, Data Science Analyst, Envision Energy

Abstract

EnOS is a smart, scalable and open-source platform that is enabling the IoT for energy; the Energy Internet. Through monitoring, real-time computing and data analysis it provides surveillance and control for how, where and when wind turbines generate power.

As the IoT platform which provides the foundation for all of Envision’s energy applications, EnOS hosts the Greenwich Systems platform; an internal wind resource assessment tool providing the full micro-siting workflow with a state-of-the-art microscale CFD engine. Mounted on EnOS in concert with Envision’s WindOS surveillance and monitoring application a post-evaluation benchmarking comparison is in progress, with post-construction yield analyses being compared with pre-construction yield predictions.

The goal of these benchmarking activities is, firstly, to set a baseline AEP accuracy distribution of the Greenwich Systems platform; guiding the evaluation of methodology improvement hypotheses and further refining best practices. Secondly, these activities will serve to better understand and quantify the uncertainties involved in the various stages of the micro-siting workflow and how these uncertainties propagate to AEP prediction uncertainty.

In this work, the architecture of the benchmarking system will be presented, with a focus on:

initial benchmarking results across 23 sites;

the methods used to obtain post-construction yield estimates suitable for comparison with the original predictions;

global and regional sensitivity analyses to determine the sources of uncertainties that yield high variation and over/under prediction of AEP;

best practices across different farm classes to inform our future micro-siting efforts.


Lifetime extension reports and the redesign of O&M concepts for aging turbines

Presenting author: Philipp Stukenbrock, 8.2 Consulting AG

Abstract

Differences between the design loads and the actual loads on-site can lead to the possibility of operating the wind turbine past its design life. Based on more than 300 lifetime extension (LTE) assessments the first results and challenges will be outlined.

Utilizing an aero elastic simulation the individual overall lifetime is calculated per main component. In most cases you can extend the lifetime of your asset and then what next. Even with a permit to prolong the operation you will have to clarify further economic question before you continue to operate. You have to show your understanding of the corresponding effects on the O&M concept for old wind farms.

We will outline the key point and discuss this from a practical experience from the field. This offers a real benefit in comparison to the research topics for LTE and the theoretical papers currently in discussion.

In more detail we will discuss the following:

Each WEC has an individual lifetime which is affected by the on-site wind conditions. Using an analytical approach, the lifetime of each WEC main component can be calculated. How does that work and is it convenient for the local authorities?

Consequently, the weak points of a WEC can be determined and the risk of damage caused by fatigue can be reduced. Knowing the overall lifetime serves as a basis for reliable organizational and financial decisions. How can you include that in your O&M concept?

Moreover, the analysis helps to estimate financial and structural risks when the conditions on-site change e.g. in the case of new WECs built in the direct vicinity of existing WECs. The new wind turbines influence the turbulences on-site and cause a wake effect. This might lead to lower energy yields of the existing wind turbines and even influence the structural safety in a negative way. The analysis offers a direct comparison between the two settings and we will show several example of real projects to discuss the participants or on a panel.

In addition, the results of the simulation can be used to give advice on how to operate a WEC more safely and adjust it to the individual on-site conditions.

Also, the individual maintenance plan can be adjusted according to the lifetime of the main components and repairs of critical components can be planned in a long-term schedule. During regular inspections it is possible to check the weak points, e.g. the rotor blade connection, more closely and notice anomalies at an early stage. To extend the lifetime of the wind turbine it is possible to:  –  adjust the operation mode and/or renew the affected component.

During the discussion with wind farm owners and operators and project execution of hundreds of lifetime extension assessments, the following challenges appear to be the most relevant:

  • Collection of data (retrofits, tracked repairs and other rather old documentations)
  • Inconsistent calculation approach within the industry and heterogeneous expectation by local bodies
  • Expected earnings after 2020 (Germany) without subsidies and the uncertainty of spot market price

We will outline in a few slides what do you need to consider in order to operate without subsidies. This will include an example showing a different way of producing power coupled with the electricity market and your gains on lifetime if you shut down your turbine based on low electricity prices.

In summary, each WEC has its individual lifetime which can be analyzed based on the on-site wind conditions. Thus, the analysis enables operators, project developers and investors to plan individually and with a high reliability. However, without outlining an operation strategy for aging wind farms, the capability to prolong the life of your turbine is just a first step for a financial decision to continue operation.


An online digital twin for real time calculation of remaining life

Presenting author: Francesco Vanni, DNV GL

Co-authors: Thomas van Delft, Michael Wilkinson

Abstract

The trend of decreasing cost of energy means that wind farm owners and operators are constantly seeking new ways to improve the performance of their wind turbines, reduce downtime and maintenance costs, increase operational efficiency and extend the life of their assets, thereby reducing levelised cost of energy. These challenges can be met using on-line digital twins that take operational data and undertake calculations in real time.

This presentation will focus on a fatigue lifetime estimator that makes use of 10-minute SCADA signals to estimate the site conditions experienced by the turbines across the wind farm. The effect on the cyclic loading of different structural components is calculated via a loads database generated using a full aeroelastic model of the wind turbine. The loads database captures the loading associated with different modes of operation as well as with downtime. Uncertainty is calculated and propagated via sensitivity analysis. This approach allows the algorithm to run with a generic limited data set (and higher levels of uncertainty) or to make use of more complex signals (accelerometers, blade loads, LiDAR) to reduce uncertainty. The resulting estimate of accumulated fatigue is translated into an equivalent “age”. The system is currently deployed on several operating wind farms and is being used to monitor the ageing of turbines to identify opportunities to extend life or increase power rating, and prioritise inspections using a risk based approach. Some case studies from the real world experience will be presented.

SESSION 5: Innovations in operations & hybrid systems

Machine Learning for automated detection of wind farm underperformance

Presenting author: Jon Collins, DNV GL

Co-authors: Giacomo Rossito, Michael Wilkinson

Abstract

Successful identification of turbine intermittent underperformance has historically relied on analyst-led review and manual “flagging”. This can be enhanced with good time series analysis tools, but can still be a laborious and time-consuming activity. In recent times there has been interest in automation of underperformance flagging. The authors have investigated a range of techniques to achieve this objective from basic parameter algorithms to advanced statistical and machine learning techniques. A number of case studies will be presented to compare these methods and demonstrate the relative effectiveness of machine learning.

For example, a case study will be presented that demonstrates that by feeding a machine learning algorithm with all the underperformance flags on 3 turbines in a wind farm it was capable of effectively flagging the fourth turbine on the farm to a 95% level of accuracy.

Following the comparison of methods, the authors have now operationalised the most promising techniques by deploying these on a real time online digital twin environment. This is currently active on over 20 wind farms and some selected real cases of underperformance will be presented to demonstrate the effectiveness.


Examining the business incentives for investments in coupled wind – storage systems

Presenting author: Peter Enevoldsen, Envision Energy

Co-authors: Peng Hou (DTU); Joshua Eichman (NREL)

Abstract

Varying electricity prices is one of the consequences of the global electrification and transition towards societies powered by renewables. This study examines business incentives for integrated renewable energy power systems by revealing the investment potential of coupling wind farms with the following storage technologies, in order to stabilize the fluctuations in the energy market:

Hydrogen (Electrolyzer) with and without fuel cells

Compressed Air Energy Storage

Lithium-ion batteries

Redox flow batteries

The research design is inspired by the optimization framework proposed by Hou et al. (2017) which couples the sequential quadratic programming and the adaptive particle swarm optimization.

The output of this research is a strategy for applying optimization methods for a power system consisting of each of the targeted storage technologies and wind turbines, which due to the stochastic nature wind provides an excellent case study. The case studies will be based on an onshore wind farm with an installed capacity of 12 MW, and an offshore wind farm with an installed capacity of 96 MW.  It will therefore be possible to rank each of the storage technologies to determine the most profitable storage method throughout the lifetime of an on and offshore wind farm. This research furthermore investigates the trade-offs between exporting hydrogen directly or using it as a storage medium to re-generate electricity at a period when it is more valuable. That period is determined by forecasting of electricity prices and grid demands which have been established using historical data from the Danish and German electricity markets in the period 2014-2017 where several hours introduced negative electricity prices. The number of hours with peak electricity prices is expected to increase with the increasing capacity of installed renewables, why it is sought to reveal whether strategic optimization methods can exploit such peaks. Therefore, the following scenarios are examined to provide an overview of the future energy systems.

Scenario 1: Wind turbine performance without storage, in order to benchmark the other approaches

Scenario 2: Storage with electricity export during times with high electricity prices in the electricity markets

Scenario 3: Storage where electricity is exported to the grid at the wholesale rate or/and used in an electrolyzer to produce and export hydrogen.

The learning objectives provided to the audience are the following:

Introduction of various wind-storage systems

Ranking of storage solutions for on -and offshore wind power

Business incentives for storage scenarios

Strategic optimization method for operating on – and offshore wind farms

ROI and LCOE changes for each scenario tested


SCADA analysis might be the best return on investment you ever get

Presenting author: Matthew Colls, Prevailing

Co-authors: Neil Atkinson

Abstract

Experience shows that performance can be improved at virtually every wind farm.  Amazingly most of these issues can be detected through careful analysis of SCADA data alone.  No expensive LiDAR devices, no secondary SCADA systems, no power performance masts, and at much, much lower cost.  This data driven approach opens the door to portfolio wide performance optimisation, squeezing every drop of return from the assets.  The yield improvement from just one of the turbines in a project in a single year will often repay the analysis investment.  The rest is for the owner.

Through acquisition and refinance analysis, Prevailing has analysed SCADA data from hundreds of turbines.  A grand tour of configuration and control issues will be presented, uncovered through SCADA data analysis alone, and previously unknown to the project owners.