Programme outline

Thursday 14 April 2016
08:00-08:45 Registration and welcome coffee in the exhibition and poster area
08:45-09:05 Welcome and opening addresses

Session 1 – Extracting value from operating data

11:00-11:45 Refreshments break in the exhibition and poster area – Poster session
11:45-13:30 Session 2 – Understanding failures, detecting faults
13:30-15:00 Lunch, followed by coffee in the exhibition and poster area
15:00-17:00 Session 3 – Using data for advanced performance modelling
17:00-18:30 Drinks reception in the exhibition and poster area – Poster session
19:30-23:00 Workshop dinner for all participants
Friday 15 April 2016
08:00-09:00 Welcome refreshments
09:00-11:00 Session 4 – Reporting and analysis from operating assets
11:00-11:45 Refreshments break in the exhibition and poster area
11:45-13:30 Session 5 – Innovation in operations
13:30-15:00 Lunch, followed by coffee in the exhibition and poster area

Extra session (independent from EWEA workshop)

After the EWEA workshop, on Friday 15 April, from 15.00 to 18.00 the EU project ClusterDesign held a separate workshop at the Bilbao Exhibition Centre, where they presented progress on the development of a ToolBox for the integrated design and control of clusters of offshore wind farms.

For more information, please visit:


Detailed programme

Thursday 14 April 2016 – 8:45 – 9:05

Welcome address by Arantxa Tapia, Councillor for Economic Development and Competitiveness, Basque Government

Opening address: Christian Jourdain, Services Marketing Director, Gamesa


Thursday 14 April 2016 – 09:05 – 11:00
Session 1: Extracting value from operating data

Session chairs: Mike Anderson, Group Technical Director, RES Ltd / Christian Jourdain, Head of Marketing & Communication, Services, Gamesa

Session description: Big data has been on every lips for some years now. Consulting service companies, OEM, and owners have been collecting data and analyzing assets to detect turbines’ failures or low performance at earlier stages, and finally develop prognosis tools and systems. However, we are still far from it as turbines’ failures and performance are multi-factors. During this session utilities and main OEMs shared their vision and results on big data applied to operating wind farms.


Valeri Voev, Head of Risk Management and Data Analytics, Siemens Wind Power (confirmed)
David Vernooy, Global Digital Wind Initiative, GE Renewable Energy (confirmed)
Mark Zagar, Plant Siting and Forecasting Specialist, Vestas (confirmed)
Carles C. G., Fleet Performance Analyst, Vattenfall Wind Power (confirmed)
Followed by Q&A
First Discussant: Adrijan Ribaric, Head of Industrial Internet Systems, Sentient Science (confirmed)


Thursday 14 April 2016 – 11:45 – 13:30
Session 2: Understanding failures, detecting faults

Session chair: Lars Landberg, Director, Strategic Research & Innovation, Renewables, DNV GL

Session description: This second session focused on faults and failures and how to understand and detect these from very large amounts of data. While detecting faults and failures is valuable, being able to predict them with some lead time is even more so.

The session therefore discussed the use of various techniques to detect and predict faults and failures, including machine learning. Data from operating wind farms used in all the applications.


Utilizing wind-turbine failure and operating data for root-cause analysis Katharina Fischer
Senior Scientist
Fraunhofer IWES (confirmed)
High-dimensional data analysis for early fault detection in wind turbines Esmaeil S. Nadimi
Associate Professor of Electrical Engineering
University of Southern Denmark (confirmed)
Big Data Analytics for SCADA: Machine Learning Models for Fault Detection and Turbine Performance Elizabeth Traiger
Senior Researcher
DNV GL (confirmed)
Turbine performance: lessons learned by Leosphere through hundreds of projects Paul Mazoyer
Application Engineer
Leosphere (confirmed)
Wind Turbine Major Components Failure Predicting Based on SCADA Data Analysis Li Shaowu
China Longyuan Power Group Corp. Ltd. (confirmed)

Thursday 14 April 2016 – 15:00-17:00
Session 3: Using data for advanced performance modelling

Session chair: Hans Ejsing Jørgensen, head of meteorology, DTU Wind


Can we evaluate performance changes based on SCADA power curves? Anthony Crockford
Technical Director
Arista (confirmed)
Performance analysis on turbine yaw misalignment with higher data sampling rate than 10min statistics Niko Mittelmeier
Wind Farm Performance Expert
Senvion GmbH (confirmed)
Benchmarking pitch system reliability and reducing Cost of Energy through advanced design Prasad Padman
Global Head of Marketing
Moog Inc. (confirmed)
Results from the Offshore Wind Accelerator (OWA) Power Curve Validation using LiDAR Project
Alex Clerc
Technical Manager
RES (confirmed)
Use of Higher Frequency data for Turbine Performance Optimisation Michael Wilkinson
Service Line Leader, Asset Operations & Management
DNV GL (confirmed)


Friday 15 April 2016 – 09:10-11:00
Session 4: Reporting and analysis from operating assets

Session chair: Sebastian Mertens, Projects and Services Leader (Onshore Wind – EMEA), GE

Session description: In this session, participants heard about new methods to identify the origin and nature of production losses, illustrated with operating data from a variety of operating sites in several countries, in Europe, Africa, Asia and the Americas. Two presentations had a specific focus on operations in cold climate areas, with a respective focus on the detection of icing losses and on the assessment of maintenance productivity factors. Participants could also learn about the incremental impact of asset portfolio diversification on assets performance.


Next-generation wind portfolio strategy: how important is diversification? Scott Eichelberger
Wind Energy Offering Manager
Vaisala (confirmed)
Methods for detection of icing losses in SCADA data Staffan Asplund
Managing Consultant
Etha Wind Oy (confirmed)
Remodeling power productions time series to minimize post-construction uncertainties Gil Lizcano
R&D Director
Vortex (confirmed)
Energy yield reconciliation in monthly O&M reports Claire Puttock
Technical Analyst
RES Ltd. (confirmed)
A Study of Maintenance Key Performance Indicators for the European Offshore Wind Farms in Cold Climate Regions Mahmood Shafiee
Lecturer in Engineering Risk, Reliability and Maintenance
Cranfield University (confirmed)

Friday 15 April 2016 – 11:45-13:30
Session 5: Innovation in operations

Session chair: Rüdiger Knauf, Chief Technology Officer, Siemens Windpower and Renewables

Session description: This last session, focused on innovative techniques and approaches to wind farm operations. It included concrete examples of how to best manage the life of  assets by using a lifecycle analysis of critical components or using smart sensors to perform rotor balance analyses. The session also explored new ways of measuring performance offshore and how the use of drones could impact the way we think about wind farm operations. Participants were also be provided with an overview of existing frameworks and practices from various countries that have already experienced the decommissioning of wind farms on a large scale, including lessons on how to minimize the environmental impacts of decommissioning. Moreover the session included a look at emerging ways of operating wind assets in maturing markets and related infrastructure options.


Multi-WTG performance assessment offshore, using a single scanning Doppler Lidar Rémi Gandoin
Measurements Engineer
DONG Energy Wind Power
Infrastructures to enable wind and solar projects to actively participate in balancing European grid systems Keir Harman
Head of Asset Operations and Management
The drones are coming! Lars Landberg
Director, Strategic Research and Innovation
DNV GL (confirmed)
Applying smart sensors technology to improve the analysis of operating wind turbines Bruno Pinto
R&D and Data Analysis Manager
SEREEMA (confirmed)
Reliability-centred maintenance: cost-effective techniques to minimize turbine failure Jurgita Simaityte
Senior Data Analyst
Nordex Energy GmbH (confirmed)
Decommissioning of wind farms ensuring low environmental impact Liselotte Aldén
University Lecturer, Earth Science
Uppsala University (confirmed)



Session 2 – Understanding failures, detecting faults

Utilizing wind-turbine failure and operating data for root-cause analysis

Presenting Author: Katharina Fischer, Senior Scientist, Fraunhofer IWES

Abstract: In order to achieve a further reduction of the cost of wind energy, enhancing the reliability of wind turbines plays a key role. For that purpose, it is important to systematically make use of experience and data gained on existing turbine fleets in order to direct and support the development of technical improvements: Field data cannot only reveal the main sources of unreliability and maintenance cost among the turbine components. It can also provide valuable information to shed light on the root causes and mechanisms leading to failure. Such knowledge is essential as a basis for working out effective countermeasures for the existing fleets as well as for new turbine or component developments.

In this presentation, the utilization of field data for the purpose of root-cause analysis and reliability improvement is demonstrated with examples from an ongoing project on power-converter reliability, in which 16 companies – among them wind-turbine and converter manufacturers, operators and maintenance service providers – have joined forces with Fraunhofer IWES and academia in order to identify the causes and mechanisms underlying the frequent converter failures in wind turbines: Failure data derived from maintenance and spare-part records is used for weak-point analysis. The failure observations are examined with respect to possible temporal or spatial patterns. A combined analysis of failure data and operating (10min SCADA) data is carried out to assess if there are systematic differences in the operating histories of turbines with high and turbines with low converter failure rates. All results are evaluated with respect to potential indications of failure causes and mechanisms.

In contrast to the classical reliability-analysis methods that typically require age or usage information of the components of interest, the approaches presented here are based on failure rates and can therefore also be applied also in cases in which such information is missing.

High-dimensional data analysis for early fault detection in wind turbines

Presenting Author: Esmaeil S. Nadimi, Associate Professor, University of Southern Denmark

Co-authors: Jurgen Herp, Victoria Blanes Vidal

Abstract: In this study, high-dimensional data analysis methods to address the problem of fault detection of wind turbines are proposed. Fault detection requires continuous monitoring and processing of Big Data generated by the wind turbines and recorded by the supervisory control and data acquisition (SCADA) system. In this paper, random matrix theory and nonparametric kernel distribution are used to model the oscillations of the nacelles of wind turbines, from a Danish wind farm, prior to the occurrence of a major fault in one of the turbines. We establish universality by referring to the asymptotic distribution of the empirical spectral density (ESD) of the sample covariance matrix deviating from the Marchenko – Pastur (MP) law almost surely due to the occurrence of the fault. We establish a nonparametric estimate of the empirical spectral density of the sample covariance matrix with a given kernel function and a bandwidth parameter. We further analyze the empirical eigenvalue density function of the sample covariance matrix and compare with the MP law prior to the fault, during the fault and after the fault has occurred. The results of this study show that deviation from the MP universality law with probability close to one can be seen prior to the occurrence of the fault. To the best of our knowledge, this study is the first one that demonstrates the applications of high-dimensional signal processing techniques such as random matrix theory towards fault detection in wind turbines.

Big Data Analytics for SCADA: Machine Learning Models for Fault Detection and Turbine Performance

Presenting Author: Elizabeth Traiger, Senior Researcher, DNV GL

Abstract: Machine learning techniques are well-suited to operational turbine performance and fault detection given the intricacy and abundance of existing monitoring data. Complexity in the intra-turbine flow dynamics within a farm make rendering exact system modelling outside the realm of possibility. Statistically based data analytics including neural networks and machine learning algorithms overcome this limitation. Recent advances in computing have made the application of computationally intensive statistical techniques commonplace within the engineering community. We introduce distributed machine learning methods and provide two machine learning models applications. Using historical wind farm measurements from SCADA data, we develop a non-physical statistical model of the system to predict component failures. A supervised learning algorithm will be utilized to analyze the wealth of operational data recognizing conditional state patterns. This information coupled with additional expert knowledge of turbine component limits forms the basis of a statistical model to identify future failures. Also, a machine learning model predicting turbine performance over a range of climatic conditions is presented. An unsupervised learning algorithm has been trained on power performance data from a collection of independent tests. Key turbine performance drivers are identified. Model validations of the turbine performance model show improved predictive accuracy over traditional matrix classifier methods.

Turbine performance: lessons learned by Leosphere through hundreds of projects

Presenting Author: Paul Mazoyer, Application Engineer, Leosphere

Co-authors: Paul Mazoyer, Guillaume Coubard-Millet, Matthieu Boquet, Florian Rebeyrat

Abstract: Boosting turbine performances relies on both getting a reliable picture of actual turbine performance and optimizing conversion of wind energy by the wind turbine. 4 years of campaigns are analyzed containing a large span of operators and a large span of turbine type and permits to outputs general statistic that leaded to two lessons: First lesson is that the nacelle-mounted Lidar Wind Iris addresses a large variety of applications regarding the performance optimization and monitoring. Second lesson is that analysis often revealed unexpected results.

The statistics of yaw misalignment measured are presented: it shows that 52% suffers from yaw misalignment above 4° inducing important long term losses of AEP and 40% of the turbine have a nacelle transfer function shifted by more than 8% inducing misinterpretation of the turbine performances.

Power curve measurements with Wind Iris are performed before and after yaw misalignment correction to assess the gain of correction. A real-life example is presented. The analysis also show a typical nacelle transfer function analysis. The yaw misalignment value can generally be computed in a short period (nearly 5 days) but for complex flow, a deeper analysis of wind data is necessary to identify the behavior of yaw misalignment. A case study is presented using an innovative algorithm to study complex behavior of the yaw misalignment. As of finding the best yaw misalignment for a turbine, a power coefficient method is proposed which permits to identify the yaw misalignment leading to the best wind energy conversion.

Results show that the under-performances is come sometimes due to a bias from nacelle anemoter or that some turbine ‘s best yaw misalignment can be surprising.

Wind Turbine Major Components Failure Predicting Based on SCADA Data Analysis

Presenting Author: Li Shaowu, Engineer, China Longyuan Power Group Corporation Limited

Co-authors: Xu Jia(Jay), Liu Ruihua, Zhang Zijun, Wang Long, Long Huan

Absract: Longyuan power wind energy is China’s largest wind power operator, in 2009, Longyuan began to develop vibration test and oil test technology, in 2011, began to develop performance analysis technology based on SCADA data, in 2014, began to develop monitoring and predicting technology of the turbine main components based on SCADA data. In recent years, Longyuan encountered several serious damage such as gearbox or blade broken, the damage brought great economic losses. In order to find out the reason and to prevent such accidents from happening again, we investigated a lot through traditional methods, vibration signals, oil and scada data analysis, but all the methods have failed. The traditional methods can’t predict or diagnose these problems. Therefore, we tried to use other method to solve such problems. This time, we mainly introduce two monitoring model based on SCADA data, one is for gearbox, the other is for blade.

A deep neural network (DNN) is trained with the data of normal gearboxes to predict its performance. The developed DNN model is next tested with data of the normal and abnormal gearboxes. The abnormal behavior of the gearbox can be detected by the statistical process control charts via the fitting error. The capacity of the monitoring model for detecting the abnormal behavior of gearbox is validated by two gearbox failure cases collected from wind farms located in China.

A Deep Autoencoder (DA) based model is proposed to extract a derived indicator, the reconstruction error (RE), for monitoring blade failures from SCADA data. To effectively detect RE shifts in online monitoring, the exponentially weighted moving average (EWMA) control chart is considered. The effectiveness of the proposed monitoring approach is validated by five blade failure cases collected from wind farms located in five regions of China.

The computational results prove that the proposed monitoring approach is able to identify incoming gearbox or blade failures.

Session 3 – Using data for advanced performance modelling

Can we evaluate performance changes based on SCADA power curves?

Presenting Author: Anthony Crockford, Technical Director, Arista

Co-authors: Philippe Cambron, PhD Candidate in Wind Energy, École de Technologie Supérieure; Jean Grassin, O&M Manager, JP Energie Environnement; Francis Pelletier, President, Arista

Abstract: A common method for performance evaluation is to monitor the power curve based on nacelle wind speeds and other SCADA data. The advantage is that the data is collected over the full lifetime of the wind farm (without needing a met mast), although there can be major challenges to arrive at an accurate result.

We will present three case studies from wind farms in Canada and France, which highlight the issues with nacelle-based power curves:

1) A 100+ MW wind farm has been in operation for over 5 years, without major modifications, yet the power curves calculated for each year of data show a variability of ±1-2% in terms of the associated energy yield. Since the differences are not attributed to real changes in the wind turbine operation, these results indicate the flaws of a power-curve based comparison of wind turbine performance.

2) The nacelle wind speed measurements are found to be significantly impacted, for two wind farms undergoing wind turbine modifications (installation of new anemometers and vortex generators). Again, a power curve comparison before and after those installations is affected, since it is difficult to separate the effects on power and wind speed.

3) Finally, we will demonstrate the quantification of performance gain for a modern upgrade package for a wind farm, in terms of relative changes in production changes compared to control wind turbines. The method does not consider the wind speed measurement, so it is insensitive to the previously identified issues with nacelle anemometry.

Performance analysis on turbine yaw misalignment with higher data sampling rate than 10min statistics

Presenting Author: Niko Mittelmeier, Wind Farm Performance Expert, Senvion GmbH

Co-authors: Tomas Blodau, Martin Kühn

Abstract: The correct alignment into the wind is a key requirement for optimal turbine performance. The wind industry standard data collection for turbine performance analysis are 10min statistics. The purpose of this investigation is on the one hand to identify the best data sampling rate for turbine performance comparison and on the other hand to identify the best turbine alignment. To answer these questions, a field test was set up. SCADA data from two neighboring turbines has been sampled with 1Hz resolution. One turbine has been forced to operate for two weeks with +5° and then for another two weeks with -5° yaw misalignment. Having the second turbine as reference, the best sampling rate to identify the performance changes due to the forced yaw misalignment was found.

Benchmarking pitch system reliability and reducing Cost of Energy through advanced design

Presenting Author: Prasad Padman, Global Marketing Director, Moog Inc.

Co-authors: Francesco Vanni, DNV GL; Erika Echavarria, DNV GL; Michael Wilkinson, DNV GL

Abstract: Improving turbine reliability and uptime is one of the critical aspects of reducing wind generation’s Levelised Cost of Energy (LCoE) and remaining competitive in today’s renewable market.

One of the challenges faced by turbine manufacturers and sub-system suppliers lies in understanding where the efforts to improve reliability will result in the largest LCoE benefits. Limited access to operational field data means that the impact of improved design on LCoE cannot be easily quantified.

By combining benchmarking analysis of large sets of operational data with Cost of Energy modelling, DNV GL and Moog calculated average values of pitch system failure rate and associated turbine downtime, and determined what reduction in LCoE can be achieved by improving the current reliability levels.

The paper presents the quantitative results of the study and explains how operational data from over 6GW of installed wind farm capacity (exceeding 5 million turbine days), collected from a range of sources, were aggregated and analysed to produce a benchmark of pitch system failure rates and understand their variation across different geographic regions, turbine sizes and types of actuators (electric and hydraulic).

It then describes how Turbine.Architect and other cost modelling tools developed by DNV GL were used to study the influence of pitch system reliability on turbine and wind farm capital and operational costs (CapEx and OpEx), energy capture and finally Cost of Energy.

Finally, the paper discusses how a new generation of pitch systems, based on an improved design of drive electronics, motors, and back-up power assemblies could significantly lower maintenance costs, improve turbine reliability and reduce LCoE.

Results from the Offshore Wind Accelerator (OWA) Power Curve Validation using LiDAR Project

Presenting Author: Alex Clerc, Technical Manager, RES

Co-authors: Peter Stuart (RES), Lee Cameron (RES), Simon Feeney (RES), Ian Couchman (FNC)

Abstract: The OWA Power Curve Validation using LiDAR project aims to assess the effectives of performing LiDAR based power curve measurement offshore. A total of 10 LiDAR power curve analyses will be presented spanning multiple LiDAR technologies; Nacelle LiDAR, Scanning LiDAR and Floating LiDAR. Three of the LiDAR datasets have concurrent meteorological masts available for comparison (of which two are fully IEC compliant). The study represents a unique overview of the current state of the art in power curve measurement with LiDAR.

Conclusions will be drawn regarding the capability of LiDAR to perform contractual power curve measurements in the absence of expensive offshore met masts. The implications for the uncertainty assigned to Offshore LiDAR based power curve measurements will be discussed and recommendations on LiDAR deployment configuration will also be made.

The LiDAR data is further explored to enable improvements in turbine performance modelling. For each power curve test a parametric study of the dataset is conducted to characterise the sensitivity of power deviation against each signal in the dataset (air density, turbulence, shear, inflow angle, etc.). This reveals the parameters which have the most significant relationship with real world power deviations. These observed deviations are compared to theoretical expectations from advanced turbine performance models and the effectiveness of industry standard correction methods (IEC air density and turbulence corrections, REWS correction, inflow angle correction) is evaluated. This insight into the performance of turbines in a broad variety of wind conditions provides valuable input for model development.

Use of Higher Frequency data for Turbine Performance Optimisation

Presenting Author: Michael Wilkinson, Service Line Leader, Asset Operations & Management, DNV GL

Co-authors: Keir Harman, Barbara Savini, Francesco Vanni

Abstract: Analysis of 10-minute SCADA data is now established as an important tool for the optimisation of operational wind turbine performance. Many SCADA systems also generate higher frequency data at around 1 Hz but due to the high volume these data have historically been ignored and archived or even discarded. However, with increasing computing power, and decreasing storage costs, analysis of these data is becoming increasingly viable. The authors will demonstrate, through the use of case studies, the benefit of looking in more detail at these higher frequency data. Insight can be gained into the performance of the turbine: for example the nacelle anemometer wind speed signals can be used to measure turbulence and assess how high and low turbulence affects real-world turbine performance. Through the analysis of the correlation between signals such as wind speed, generator speed, pitch angle and/or electrical power, it is possible to observe high level controller behaviour such as: closed-loop pitch and torque controller relationships; above- and below-rated power transitions; and supervisory controller actions.

SCADA-based condition monitoring with 10-minute data, although valuable, is often limited to temperature analysis. Using 1 second data unlocks a range of additional analysis possibilities, for example monitoring first tower frequency deviation over time to assess structural health and estimating fatigue loading which provides valuable data for turbine life extension.

With 10-minute data, the effectiveness of a turbine’s yaw strategy can be difficult to assess. However with higher frequency data it can be demonstrated that more frequent yaw-manoeuvres leads to increased energy capture.

The authors will present real case study examples from analysis of higher frequency data.


Session 4 – Reporting and analysis from operating assets

Next Generation Wind Portfolio Strategy: How Important Is Diversification?

Presenting Author: Scott Eichelberger, Wind Energy Offering Manager, Vaisala

Abstract: As installed wind capacity increases, companies find themselves exposed to large portfolios of operational wind projects with many choices as to what markets and regions to expand to next. For some stakeholders, this raises a concern as to how much organizational effort should be spent in diversifying a portfolio. For anyone that owns securities, the concept of portfolio diversity is clear. To date most wind industry strategy focuses on optimizing the cross-section of locations with the strongest wind resource, the easiest to build sites, and the highest power prices. Rarely has the question of incremental impact to the existing portfolio been taken seriously. Using multiple reanalysis datasets as the basis for analysis, this presentation discusses how much can actually be done about this problem. How diverse are the typical development regions? Is negative correlation achievable? Is there a portfolio size at which there are few options to increase portfolio diversity? These questions and more will be discussed using real world scenarios and quantitative analysis of the variability and covariance of resource across the major wind regions. The analysis consists of simulating portfolios of varying geographical diversity and size. A portfolio economic model is developed so that results can be discussed in terms of economic impact.


Methods for Detection of Icing Losses in Scada Data

Presenting Author: Staffan Asplund, Managing Consultant, Etha Wind Oy

Co-authors: Christian Granlund Etha Wind Oy, Teppo Hilakivi Puhuri Oy

Abstract: Estimating the icing losses is an essential part of post-construction yield analysis in northern Europe. The production losses during harsh icing conditions are usually detected and flagged by the Scada system, however, significant amounts of icing losses are often left undetected by the system as long as the turbine is not completely stopped. Alternative methods are thus required in order to estimate the total icing losses of a wind farm.

The methods used in this study were all based on the Scada data of the specific wind parks, and the methods were tested on two wind parks (with different turbine types) in Finland. No other data than the Scada data and the contractual power curve of the turbines were used. The main challenge in this study was to find a method that with high accuracy differs between icing losses and other types of sub-optimal production. This was achieved by doing an in-depth analysis of the performance during icing conditions and comparing the production behavior during icing conditions with the production behavior during periods when sub-optimal production occurred, but icing losses could be ruled out. Based on this, icing occurrence could be described mathematically and data containing icing losses could be successfully identified.

The yearly production losses was estimated and the consistency of the methods were checked by comparing the results for the two wind parks, as they were operated differently from each other in icing conditions. When a satisfying ice loss detection method have been defined and an estimate for the Scada period have been calculated, it would also be of interest to estimate the long-term normalized icing losses. This can be done by using a neural network model developed earlier by Etha Wind for this purpose.

Remodeling power productions time series to minimize post-construction uncertainties

Presenting Author: Gil Lizcano R&D Director Vortex

Co-authors: Abel Tortosa, Vortex; Pau Casso, Vortex

Abstract: A challenging application for high resolution wind modeling technology is to translate wind conditions hindcast into multi-year power production time series.

Modeled power time series can be used to analysis windfarm sensitivity to interannual variability, to map ramps occurrences, to quantify extreme and hostile environmental conditions like icing and to assess windfarm reactivity to all these events. Such information is very valuable to define windfarm timeline scenarios to support decision making process and to minimize uncertainties in the real windfarm operation.

Train & predict procedure is employed to derive power time series allowing different degrees of modeling complexity. We opted in this work for a dynamic wind to power approach based on a heuristic (or bigdata) remodel of a large set of hindcast atmospheric parameters and windfarm operation and production outputs. This approach can more effectively discriminate different meteorological conditions and its windfarm responses.

This works presents analysis of wind and power hourly time series spanned over 20 years for a selection of sites in Europe, North America, Brazil, South Africa and India. Validation against observed data is computed to provide a reference for bias and uncertainty thresholds.

The work illustrates the sensitivity of annual and monthly power fluctuations to resource conditions variability for different real project in each regional context. Results are compared to AEP P50 and P90 yield pre-construction estimations as reference.

Selection of different wind conditions anomalous periods are carried out to quantify and to visualize out-of-normal windfarm operations and the probabilities of occurrences in a 20 year frame. Different test cases are shown for extreme conditions. Ramps occurrences are mapped as baseline information for windfarm and grid operation. All results are summarized in terms of windfarm asset management needs.

The data employed in the analysis have been kindly provided by different windfarms developers and owners.

Energy yield reconciliation in monthly O&M reports

Presenting Author: Claire Puttock, Technical Analyst, RES

Co-authors: Alex Clerc, Peter Stuart, Lee Cameron

Abstract: This presentation seeks to explain a method for energy yield reconciliation which can be quickly performed on a periodic basis (e.g. monthly) to give a full view of an asset’s energy yield.

Through analysis of the period’s SCADA, historic performance and historic reference data it is possible to give a detailed breakdown of the reasons that a month’s production is different to budget (“production variance”). Production variance can be fully broken down in to the following categories:

  • Operational losses
    • Downtime
      • Turbine
      • BoP
      • Grid
      • Environmental
      • Normal operations
    • Running losses
      • Curtailment
      • Icing
      • Sub optimal operation
  • Energy resource variation
    • The MWh effect of wind speed variation from budgeted long term wind speed is calculated every month
  • Budget error
    • Industry standard post construction analysis techniques can be used to reforecast a more certain budget every month. If the Owner’s budget is different to the more realistic reforecast budget it will contribute to production variance.

Furthermore, given that industry standard post construction analysis generally results in a linear model of production as a function of reference wind speed, the presentation will explain a procedure to derive an analogous linear model from the pre construction energy yield assessment, allowing a rolling comparison of the site’s operational yield with pre construction expectations to be easily reported.

Using MERRA in the UK and France it is normally possible to reconcile production variance to within 5% in a particular month and within 1% over the course of a year. Some residual error is expected due to missing or inaccurate SCADA and noise in site’s relationship to reference data.

A full breakdown of production variance gives many benefits including:

  • improves communication between the Operator and the Owner
  • ensures relevant KPI’s can be defined for O&M activity
  • empowers Owner to make more informed financial decisions about the asset

A Study of Maintenance Key Performance Indicators for the European Offshore Wind Farms in Cold Climate Regions

Presenting Author: Mahmood Shafiee, Lecture in Engineering Risk, Reliability and Maintenance, Cranfield University

Abstract: Substantial investments have been made in recent years for wind energy development in cold climate regions. A large number of wind turbines are planned to be built in the near future at sites with severe climatic conditions. The European Wind Energy Association (EWEA) has forecasted that between 45 and 50 gigawatts of wind energy will be built in cold climates by 2017 [1]. Even though cold regions have a high wind potential, the installed wind turbines are exposed to low temperatures outside the standard operational limit, leading to icing of structures. Icing phenomena may dramatically shorten the life expectancy of wind turbines and increase the risk of premature failures if sufficient (effective) maintenance is not undertaken within a time period. Therefore, an efficient management of inspection, maintenance and repair programmes for wind energy assets operating in cold climate areas is crucial to maximize reliability, minimize breakdowns, and reduce long-term costs [2].

In this paper, a number of maintenance performance indicators for offshore wind farms in cold climates are defined through survey of experts in the field and the published literature. Some indicators identified include: Mean time between failures (MTBF), Mean time to repair (MTTR), Mean time between repairs (MTBR), number of preventive maintenance work orders divided by corrective maintenance work orders, preventive maintenance labor hours divided by emergency labor hours, spare parts usage, maintenance crew efficiency (work hours completed on schedule per estimated time), work order discipline (part of labor work accounted for on work orders), annual maintenance costs, etc. Then, a quantitative framework is developed to evaluate and monitor the productivity of inspection, maintenance and repair programmes for cold climate wind turbines. The presented framework is applied to several European offshore wind farms built in cold climate areas and their corresponding maintenance performance indicators are calculated. Finally, the results are compared with the productivity of inspection, maintenance and repair programmes for offshore wind farms located in normal climatic conditions. A sensitivity analysis is also conducted to identify the factors having the greatest impact on maintenance productivity of offshore wind farms in cold regions. Our results indicate that the poor accessibility to cold regions, unavailability of means of transportation, and lack of trained technicians are the main factors negatively influencing the maintenance productivity of wind farms in cold climate regions.

Session 5 – Innovation in operations

Decommissioning of wind farms – ensuring low environmental impact

Presenting Author: Liselotte Aldén, University lecturer, Uppsala University

Co-authors: Andrew Barney

Abstract: The installed capacity of wind power has increased substantially in the last few years. In 2014 over 50 GW were added bringing the total capacity to 370 GW, this is remarkable when considering that total capacity was just 14 GW in 1999. As wind turbines have lifetimes between 15 to 25 years the demand for decommissioning will grow in a similar way to the increase of installed capacity, meaning a time lag of 15 to 25 years. It is becoming clear this will be a challenge.

Decommissioning already is happening on a large scale in countries that were front runners in using wind power such as Denmark and Germany. A primary challenge when decommissioning wind farms is ensuring as small environmental impact as possible. To address this challenge extensive knowledge and practice is required.

In many countries a security bond is required during the permitting of wind farms. This security bond is intended to ensure that the wind farm will be decommissioned and the location be restored.

This article will discuss and recommend how wind farms can be decommissioned with low environment impact and compare the practices in several of countries. Wind power decommissioning laws, regulations, permits, history, activity costs and the disposal and restoration options in Sweden and around the world will be analyzed and compared.

Special focus will be placed on the amount of financial security required, the degree of restoration and its environmental impact and cost implications.

Reliability Centered Maintenance: cost-effective techniques to minimize turbine failure

Presenting Author: Jurgita Simaityte, Senior Data Analyst, Nordex Energy GmbH

Co-authors: Thomas Zedler

Abstract: The main challenges of the wind farms asset owner or service provider is not only to keep turbines on operation, but also to optimize the efficiency of the fleet by the means of reliability, availability and the related costs. Here a complex process arises: sustain the present reliability and establish cost-effective maintenance strategies while, at the same time, ensuring knowledge transfer from lessons-learned in the field back to “design for reliability” stage.

In this presentation the author will showcase an example of lifecycle analysis of critical components and its input into overall system reliability model. Important considerations here also are given to the uncertainty of the results and its influence to the decision making process. The Reliability Centred Maintenance concept and its realization will be demonstrated from the service providers and operators perspective and how it can be applied to establish optimal solutions and which main benefits can be delivered.

Multi-WTG performance assessment offshore, using a single scanning Doppler Lidar

Presenting Author: Rémi Gandoin, Measurements engineer, DONG Energy Wind Power

Co-authors: Benny Svardal (CMR), Valerie Kumer (University of Bergen), Raghavendra Krishna Murthy (Leosphere)

Abstract: In this paper we describe the setup and results of a scanning LiDAR measurement campaign at an offshore park, aiming at performing multiple power curves and characterizing the wind flow variation in the vicinity of the wind farm. The Scanning LiDAR is located on an offshore substation, 1.8 km at the West of the Anholt wind farm. The lidar has a nominal range of 3.5km, and scans towards the first row of wind turbines, measuring the inflow and wakes of three wind turbines. A vertical profiling lidar, also located on the substation, provides additional wind speed and direction measurements, and a wave buoy provides information about the water temperature and sea state. The data are processed in order to reconstruct velocity measurements at 2.5 rotor diameters in front of the turbines, allowing for multiple and concurrent power curve analysis. At the same time, the dataset also provide some novel insight into wake effects and turbulence outside and within the wake of a wind turbine.

The drones are coming!

Presenting Author: Lars Landberg, Director, Strategic Research & Innovation, Renewables, DNV GL

Co-authors: Elizabeth Traiger, DNV GL

Abstract: Drones have clearly become the “new black” within the application of robotics to all kinds of problems. Wind Energy is no exception, and in this presentation a quick introduction to state-of-the-art drones will be given, and after that the application of drones to inspect wind turbines and wind farm sites (potential and operating) will be discussed.

Real life examples of both applications will be given, and issues like accuracy of positioning, quality of images, treatment of the resulting data will be discussed. Also examples of practical issues will be given, like battery life, insurance, aviation regulations and so on. A brief discussion on the economics around the use of drones will also be given.

At the end some perspectives on the future application of drones and robotics will be given, these include flying BVLoS (Beyond Visual Line of Sight), autonomous operation and the potential of using AI in connection with robotics in wind energy.

Applying smart sensors technology to improve the analysis of operating wind turbines

Presenting Author: Bruno Pinto, R&D and Data Analysis Manager, SEREEMA

Co-authors: Jérôme Imbert, [email protected]

Abstract: Smart sensors technology allows to easily add capacities to operating wind turbines in order to optimize their maintenance and increase their AEP in a cost effective way. Smart sensors rely on the Internet of Thing infrastructure to provide added value and ready to use information to the End User so that the correct actions can be determined and conducted.

To exemplify the potential of this new approach to the diagnosis of operating wind turbines, we present a study case of rotor balance monitoring using a connected device placed on the wind turbine. The benefits of using IoT technologies for the rotor balance studies when compared to the classic measurements procedures are, among others: a continuous surveillance versus a punctual measurement, the capacity to determine and select the perfect conditions for the analysis and the ability to provide highly integrated information as well as detailed data when needed.

Four wind turbines (1.7 MW rated power) were equipped with the connected-smart sensing device with embedded accelerometers and firmware adapted to the rotor balance analysis. The continuous surveillance permits the definition of a baseline behavior and threshold levels adapted to each turbine. The adapted algorithms provide the identification of rotor imbalances (aerodynamic and masse imbalances) and their evolution with time. The results are based on 10 minutes periods with the perfect conditions for the analysis. On average over one hundred 10 minutes periods are studied per month and per wind turbine to provide a high confidence level diagnosis of the rotor balance.

The final result of this process is an automated, on-line diagnosis of the rotor balance of all equipped wind turbines. The ready to use information on the status of the wind turbines is accessible through the internet for operators and maintenance teams.

Infrastructures to enable wind and solar projects to actively participate in balancing European grid systems.

Presenting Author: Keir Harman, Head of Asset Operations and Management, DNV GL

Abstract: With the rapid increase of renewable energy penetrating our grid systems there is a strong demand for all projects to operate more like conventional generating plants, in a more visible, predictable and controllable way. This talk will focus on the need and opportunities on European grid systems for renewables to offer balancing services and other ancillary services. For example, rapid curtailment, reactive power export control, fast frequency response and even spinning reserve. It will look at how such services must be underpinned with accurate forecasting, reliable data and real-time communications with the system operator. A particular focus on the capability of wind farms and how they can operate in a dynamic way to offer these services will be presented. The potential for aggregation of projects and technologies including how they may be coupled with storage is also considered. One conclusion being that participation in this way will open up a revenue stream to project owners that counters the current diminishing income from government support mechanisms as renewable markets mature.