Topic: Data from operating assets: monitoring and collection methodogies
Health Monitoring Engineer
NEM Solutions (confirmed)
|A.U.R.A: A Novel Condition Based Monitoring System for Wind Turbines|
Rwth Aachen University (confirmed)
Rwth Aachen University (confirmed)
|Results on built-in vibration sensors inside planetary gears|
Junior Engineering Consultant
|SCADA data processing for wind turbine reliability analysis|
Asset Performance Engineer
Natural Power (confirmed)
|Breaking through the OEM filter: A real-world case study on the benefits of independent data collection and monitoring from the wind-farm owner’s perspective|
|Marc Hilbert (on behalf of Dennis Röllinger) Researcher
Rwth Aachen University (confirmed)
|Acoustic Emission as a Tool for Condition Monitoring of Oscillating Pitch Bearings|
|Monitoring of offshore turbines for design and O&M : an overview of the activities of OWI-lab|
Head of wind technology
Sorgenia Green (confirmed)
|From SCADA based to non-dimensional indexes: How to manage your turbines portfolio|
Topic: Post-construction yield analysis
|Nicolai Gayle Nygaard
Senior Wind Energy Analyst
DONG Energy Wind Power (confirmed)
|Benchmarking of wake models for offshore wind farms|
Head of Test & Measurements Group
ReGen Powertech Pvt. Ltd. (confirmed)
|Power Performance Verification Procedures in India|
OST Energy (confirmed)
|Can MERRA data be used to accurately predict offshore wind farm energy production?|
Topic: Real-world power performance
SgurrEnergy Ltd (tbc)
|Observations of the compression zone in front of a wind turbine|
SgurrEnergy Ltd (tbc)
|Offshore power curve tests for onshore costs: a real world case study|
AWS Truepower LLC (confirmed)
|Normalizing Power Curves for Shear and Turbulence|
Head of Product and Technologies
Avent Lidar Technology (confirmed)
|In-depth presentation of turbine to turbine performance optimization projects with turbine mounted Lidar|
Leading Consultant, operations supervision
Etha Wind (tbc)
|Improving yield assessment of wind farms based on operational feedback with focus on icing losses|
|Andrea Nina Eugster
|Challenging accepted methods for post-construction energy yield estimates – lessons learned|
Topic: How to best extend life expectancy?
University of Durham (tbc)
|Wind turbine health: metrics for operational analysis|
Topic: Turbine-by-turbine performance
Head of Product and Technologies
Avent Lidar Technology (confirmed)
|Evaluation of ground and nacelle based LiDAR power curve measurement in complex terrain|
Can MERRA data be used to accurately predict offshore wind farm energy production?
Poster presenter: Richard Abrams, Director, OST Energy
Abstract: We have taken a database of wind farm power production and availability data for ten operational offshore wind farms across northern Europe and compared the normal year results to production estimates, based upon a historical MERRA reanalysis data set of 33 years. The wind farms have been operational spanning the period 2002 to 2014 and are based in the UK, Germany, Holland, and Denmark. The turbine rated power ranges from 1.5 MW to 6 MW. Included in our analysis is a comparison of MERRA data to offshore met. mast data for several point locations, in order to verify the wind climate, and an analysis using the Fuga wake model to derive wake effects. We present the method and findings of our analysis and our determined linear correlation coefficients, including a discussion of individual anomalies and a review of the uncertainties inherent in the approach.
Improving Yield assessments of wind farms absed on operational feedback with focus on icing losses
Poster presenter: Staffan Asplund, Leading consultant, Etha Wind
Abstract: A major inaccuracy in yield assessments in Nordic countries is the icing loss estimate. With new operation supervisory solutions for treatment of big data from wind farms the possibilities to learn from operations increases.
The results from this case study of a 9 turbine Finnish windfarm will contain quantitative and qualitative descriptive information about icing behaviour and impacting parameters. Based on the results, accuracy of yield estimates can be improved. The general usability of the findings from the case study will be implemented in 1-2 other wind farms (different turbines, hub height and locations) to check their general applicability.
In addition a methodology to present operational data in a way supporting easy information feedback to wind farm designers is presented.
The academic and commercial benefit of this work is to decrease uncertainty in yield estimates thus making investments decisions easier.
Tech Lic Staffan Asplund, Leading Consultant, Etha Wind Oy
Dr Margareta Wihersaari , Professor, Åbo Akademi Unversity
Normalizing Power Curves for Shear and Turbulence
Poster presenter: Daniel Bernadett, Chief Engineer, AWS Truepower LLC
Abstract: Wind turbine performance is affected by shear and turbulence. This presents two challenges: predicting performance prior to construction, and optimizing performance during operation. This paper will present two methods for addressing these challenges: a simple matrix method which is independent of time-series calculations, and a more rigorous time-series calculation method which allows more granular insight into plant performance.
A case study will be presented where performance deviations are examined as a function of wind speed, shear, and turbulence. This simple method can be used to make site-specific power loss estimates for pre-construction projects, and can also support a high-level analysis of production data from operational farms.
A more rigorous approach that examines and adjusts a time series of production data for shear and turbulence will also be described. This method directly calculates the contribution to output of wind speeds at multiple heights across the rotor plane using, ideally, observations from remote sensing systems. (Extrapolated tall-tower measurements can also be used, but are likely to be less accurate.) Adjustments are applied to account for the variation of power coefficient as a function of radial position along the blade, as well as the variation of power coefficient as a function of wind speed. Turbulence normalization is applied separately according to the proposed method in IEC 61400-12-1 Edition 2. The proposed method could be used both to provide a more accurate pre-construction energy estimate, as well as to analyze in detail performance deficits in operational wind farms.
The time-series method addresses shortcomings in the rotor equivalent wind speed (REWS), which is included in the proposed IEC 61400-12-1 Edition 2. The REWS predicts a performance benefit in high shear, whereas field observations usually show a deficit. The method described here better matches observations.
Wind turbine health monitoring: the illusion of big data
Poster presenter: Cyril Boussion, PhD Student, TU DELFT
Abstract: Wind turbines are getting bigger, wind parks are going offshore, and the distance to shore is increasing. As a consequence, operation and maintenance costs are increasing and account up to 25 to 30% of the cost of energy. Health monitoring of wind turbines needs to be extended to reduce those costs. Today, only failure-prone components are monitored, by adding specifically designed sensors, and health monitoring systems barely take a system-based approach. Many sensors, already installed, are not used yet: the amount of available data should provide a better and more accurate insight. Most of the investigations try to use all the data to get the most out of them. Since the number of data points is huge, they consider big data analysis techniques: statistical tools, machine learning algorithms, etc. However, most of the approaches neglect the physical properties of the wind turbines and they rather take decisions on a black-box (e.g., neural network) than on knowledge: correlation does not imply causation, and parameters physically related are correlated anyway. Furthermore, all the data may not be needed. This paper demonstrates that smarter ways exist to monitor the health of wind turbines and that they can reduce the operation and maintenance costs: the number of sensors needed is reduced, the relationships between measurements (e.g., physical equations) are included, the redundancies in the sensor network and in the wind farm are strongly considered. As a consequence, the assessments of the health are more accurate, a better insight into the behavior of wind turbines is achieved and the number of false alarm is reduced. This shows that big data is not everything and is not the only option to monitor the health of wind turbines.
Observations of the compression zone in front of a wind turbine
Poster presenter: Peter Clive, Senior Scientist, SgurrEnergy Ltd
Abstract: The determination of the length of the compression (or induction) zone in front of a wind turbine is critical to obtaining accurate power curve tests. Compression zone length defines the minimum distance where a meteorological mast can be installed or lidar measurement acquired for power curve testing.
Conventionally it is recommended in the power curve test standard IEC 61400-12-1 that meteorological masts are installed between 2 and 4 (and preferably 2.5) rotor diameters away from the wind turbine to be tested. However, recent measurements reported in this presentation indicate that this may not be adequate and that measurements in this region may be reading wind speeds affected by the wind turbine compression zone.
A comprehensive measurement campaign has been conducted to investigate this. Scanning lidar units were installed on an offshore wind turbine, recording data in a mode in which the lidar is measuring the incoming wind speed. Two different configurations were used. The first configuration provides a simplified visualisation of the incoming wind speed profile, whereas the second configuration provides a contour plot showing how wind speed varies across the horizontal plane at hub height and along a distance of 1 km in front of the rotor.
In both cases, results show that at 2.5 rotor diameters the compression zone is still evident with wind speed reading varying between 1-3% from the free stream wind speed. In addition, the influence of the compression zone may extend until approximately 3.5 rotor diameters depending on the wind speed.
Therefore, it is possible that current power curve calculation methods are leading to an overestimation of annual energy yield prediction.
Offshore power curve tests for onshore costs: a real world case study
Poster presenter: Peter Clive, Senior Scientist, SgurrEnergy Ltd
Abstract: It is necessary to undertake power curve tests of offshore wind turbines to demonstrate their compliance with the expected levels of power performance on which pre-construction estimates of power production are based. Power curve tests require that the incident wind resource is compared with the corresponding power generated by the wind turbine as a result. The expense of assessing the incident wind resource offshore using offshore met towers has often proved a disincentive and offshore power curve tests have sometimes been neglected as a result.
The draft 2nd edition of the power curve test standard IEC 61400-12-1 allows remote sensing devices such as lidars to be used in the power performance assessment of wind turbines. This offers an opportunity to reduce the risk offshore wind power projects represent by using lower cost lidar based methods. The costs of installing an offshore met tower can be avoided by mounting lidar instrumentation on the wind turbine being tested. However, the draft standard explicitly rules out mounting lidar on the nacelle and requires that ground based methodologies are adopted. This requirement can be fulfilled using scanning lidar installed on the transition piece of the test turbine.
This methodology was successfully trialled at Alpha Ventus Offshore Wind Farm in the German North Sea. This presentation discusses the results. Of the three available methods for assessing the incident wind resource, including met tower and nacelle mounted anemometry, the scanning lidar on the transition piece introduced the lowest uncertainty. The power curve derived using the measurements conformed to within 0.5% of the power performance obtained using met tower based assessments. The cost of the test using the transition piece method was the lowest of all available methods, with the highest degree of compliance with the draft IEC standard of all lidar based methods.
In-depth presentation of turbine to turbine performance optimization projects with turbine mounted Lidar
Poster presenter: Samuel Davoust, Head of Product and Technologies, Avent Lidar Technology
Abstract: SCADA data analysis and on-site inspections are essential to wind turbine performance monitoring and diagnosis. Yet, even with the best practices, turbine wind measurements can prevent from making an accurate diagnosis or remain a source of underperformance. For instance, nacelle anemometer readings may be incorrect and static yaw misalignment cannot always be prevented. A reference wind measurement for each turbine would be an ideal solution to this problematic. We here present a detailed operational case study on turbine to turbine performance verification and optimization using mobile turbine mounted Lidars.
During the course of a 3-month campaign, 10 wind turbines from 3 different manufacturers and installed in complex terrain were analyzed. The turbines were selected based on the O&M performance monitoring reports.
An in depth analysis was performed on each turbine with respect to yaw misalignment detection, operational curve measurement, nacelle anemometer calibration, turbulence intensity analysis. The analysis showed that half of the wind turbines suffered from a significant misalignment of more than 6°, while the other half had an incorrect nacelle anemometer calibration (about 10%). The AEP increase consecutive to the correction of the yaw misalignment is currently being established through relative power curves.
Additionally, other case studies will be introduced to put in perspective these results in other contexts of operations (met. mast instrumented site, flat terrain, different turbines models and owners). Finally, the impact of the yaw optimization on life expectancy will be addressed.
Evaluation of ground and nacelle based LiDAR power curve measurement in complex terrain
Poster presenter: Samuel Davoust, Head of Product and Technologies, Avent Lidar Technology (on behalf of Thomas Burchhart)
Abstract: Assessment of wind turbine performance in complex terrain is a costly and technically challenging task. While ground-based LiDAR is going through IEC standardization for power curve in flat terrain, nacelle based LiDAR is seen as a promising solution to deliver accurate answers in complex terrain with respect to the realized power curve and nacelle anemometer calibration. We here report results obtained in a measurement campaign performed with a Wind Iris nacelle LiDAR installed on a 7.5 MW turbine located in a complex terrain and forested wind farm area. First, the accuracy of the nacelle LiDAR is validated against a ground-based WINDCUBE v2 LiDAR located 290m away (2.3 rotor diameters range). A high quality correlation between the two units is obtained in common measurement volume, which corresponds to the wind sector where the turbine is facing the location of the ground based LiDAR. A power curve is built in this sector, and the impact of shear and turbulence is investigated. Second, the sensitivity of the wind turbine power curve and nacelle anemometer flow correction factor is analyzed with the Wind Iris in the remaining wind sectors. A special care is taken to understand the impact of terrain and turbine blockage effects on the wind flow. The impacts of yaw misalignment and turbulence intensity are also investigated. Different LiDAR and SCADA data analysis methodologies are presented and compared in this case study. The overall conclusion is that nacelle based LiDAR is a serious and competitive tool to investigate wind turbine performance in complex terrain, by providing a wide number of answers and a simplified and cheaper set-up.
Monitoring of offshore turbines for design and O&M : an overview of the activities of OWI-lab
Poster presenter: Christof Devriendt, Scientific co-ordinator, OWI-lab
Abstract: OWI lab is a Belgian R&D initiative for offshore wind energy and infrastructure. One of the main research targets is to develop mid- and long-term monitoring solutions for offshore wind turbines focused on two main purposes.
Design validation and design input for future windfarms e.g.:
- Operational loads on the full turbine and subcomponents: gearbox, generator, etc.
- Loads at the transition piece and stresses in the grouted connection
- Assessing ambient vibration levels in comparison to simulation
- Database handling to study e.g. Wake effects for wind farm lay-out
- Design verification w.r.t. designed resonance frequencies and damping ratios
- Comparison of actual site conditions with predicted data
O&M decisions e.g.
- Turbine to turbine comparison for performance monitoring and fault detection
- Condition monitoring of turbine subcomponents using a mixed statistical/regression-driven strategy
- Remaining fatigue life-assessment using virtual sensors at inaccessible locations
- Structural health monitoring of foundation structures
- The validity of the fleet leader concept for farm monitoring.
In collaboration with our industry partner Parkwind, owner off two offshore wind farms outside the Belgian coast, we are able to validate our research. As such one turbine at Belwind is equipped with a multi physics sensor array and several environmental sensors. Additionally the availability of a subset of the SCADA lead to a huge amount of data.
With two additional turbines at Northwind being equipped as we speak additional challenges with data handling, data availability and data storage are becoming more apparent. To tackle these challenges a collaborative Data-hub for offshore technology and research will be initiated this year.
In our contribution we would like to give an overview of our current research topics, challenges, results and future perspectives.
<strong “gaylenygaard”>Benchmarking of wake models for offshore wind farms
Poster presenter: Nicolai Gayle Nygaard, Senior Wind Energy Analyst, DONG Energy Wind Power
Abstract: The decision to develop and construct a wind power plant is based to a large extent on the capital expenditure and the expected energy yield of the wind farm over time. The latter is a prediction based on analysis of measured wind data, assumptions and modelling. It is therefore of paramount importance for wind power plant developers to thoroughly analyse the post-construction yield of the operating assets in order to challenge the assumptions and improve the modelling. In this way, the bias and uncertainty of future energy yield predictions may be reduced.
A key ingredient in such a re-analysis of yield is the validation of the wake loss prediction, since the wakes represent the principal component of the wind power plant losses. In this presentation we perform a validation of commercial wake models on data from offshore wind farms and highlight the issues and choices involved when analysing SCADA data for benchmarking studies. In our benchmarking trial we compare the commercial software package WindFarmer to operational data from several offshore wind farms. We test the ability of the wake models included in WindFarmer to account for the observed wake losses derived from the SCADA data.
We contrast the performance of the Eddy Viscosity model with that of the PARK model and evaluate whether a large wind farm correction based on added roughness is necessary to account for wake losses in modern offshore wind power plants. In addition, we test if wake model bias can be reduced by averaging over an ensemble consisting of multiple models.
Wind turbine health: metrics for operational analysis
Poster presenter: Jamie Godwin, Postgraduate Researcher, University of Durham
Abstract: Identification of at-risk turbines is essential to reduce maintenance expenditure, increase yield and for advanced operational analytics. In this work, we present a statistically sound methodology based upon empirical data analysis to derive robust condition indices for two major subsystems responsible for a significant quantity of turbine downtime: the pitch system and gearbox.
By identifying at-risk turbines a-priori, existing maintenance resources can be optimised: for instance during periods of intermittent weather where limited maintenance opportunities exist, these opportunities can be exploited by maintaining or inspecting only those turbines which warrant the associated cost based upon their condition. This moves away from traditional “time based” approaches which have been shown to be woefully inadequate offshore.
As such, this work enables a pro-active maintenance strategy for wind farms to be created. As 75% of maintenance expenditure is due to unscheduled downtime, moving away from “fail and fix” methodologies to “predict and prevent” is essential to reducing OPEX costs. Although time based strategies – such as RCM – assist in reducing corrective maintenance, the large quantity of maintenance expenditure spent on intrinsically expensive corrective actions necessitates the need to move to condition based techniques. Similarly, current strategies do not take into account fatigue loading caused by wake effects.
Using freely available techniques, multivariate models are described to assist in the analysis of SCADA data to identify artefacts and quantify health. Using simple statistical techniques, normal behaviours of both the gearbox and pitch system are defined. Normalisation for wind turbine specific characteristics is performed, with a general methodology presented which can be applied to both mechanical and electrical systems to systematically quantify performance.
Employing the derived metrics, component health over time can easily be visualised, providing early warning and diagnosis of wind turbine faults, thus assisting in the prioritisation of maintenance activities including inspection and servicing.
A.U.R.A: A Novel Condition Based Monitoring System for Wind Turbines
Poster presenter: Nagore Guarretxena, Health Monitoring Engineer, NEM Solutions
Abstract: Failure modeling and early detection is a key issue in the wind energy sector in order to avoid important costs derived from unexpected maintenance actions and operations, potentially offshore. To this aim, a novel CBM system that addresses the issue of prognosis and early diagnosis of potential operational failures is proposed. The main process consists of modelling the normal behaviour of the system and then detecting deviations from it. This includes both sudden deviations that need to be detected as close to real-time as possible and slow, evolving degradations that must be taken into account for optimal scheduling of maintenance operation and remaining useful life estimations. In order to illustrate its applicability in a real scenario, historical SCADA data gathered from a set of wind turbines are analyzed and the resulting models are tested in real time. Experimental results show promising advantages over traditional strategies; detecting deviation patterns in the normality models before the anomaly really occurs, and thus avoiding possible damages and costly corrective interventions.
Poster presenter: Marc Hilbert, Researcher, Rwth Aachen University
Abstract: Distributed energy systems such as wind turbines share the properties of (1) having a rising number of similar installed system setups, (2) being installed mostly in remote areas with limited access and (3) needing high system reliability. This makes fault diagnosis and identification (FDI) a crucial but challenging part for operation and maintenance (O&M) of these systems. But because of the large amount of wind turbines there is an increased need for monitoring the overall fleet, additional to individual wind turbines. This work will focus to use condition information of equal components in different machines and under different working conditions, to extract useful information for FDI of those components. A definition for fleet monitoring for FDI will be introduced. It will be shown that by using existing features of the already installed condition monitoring systems and SCADA systems, that by combining these features from different machines, additional FDI information can be gained. Therefore, the focus of data analysis is the fleet information and less only individual systems information. It will be shown that properties of the introduced method can resolve common FDI drawbacks, e.g. setting up alarm thresholds. The method is based on the calculation of selected features from each system in a high dimensional common feature space. A main advantage is the absence of absolute measures for FDI and use of relative measures between components/machines in the fleet. Besides the theoretical approaches, an example using vibration data of bearings will be given. The runs of the bearings were performed with different speed and load and were only stopped by significant degradation. The purpose of this work is to increase system reliability by using existing condition information and, therefore, provide additional information for FDI.
Results on built-in vibration sensors inside planetary gears.
Poster presenter: Marc Hilbert, Researcher, Rwth Aachen University
Abstract: The Internet of Things refers to a network and virtual representations of individual objects (e.g. machine part) in an internet-like structure. This representation enables the comparison of the behavior of the real world object to a virtual counterpart. On the virtual counterpart the reactions of loads or an estimation of remaining life time can be simulated. To realize those industrial visions each object’s needs to have cheap and reliable systems that monitor the object conditions during the full lifetime. Only with these data a mapping of the virtual object will be possible and, therefore, an estimation of the remaining life can be provided. This work will describe a method that covers these challenges on condition monitoring of a wind turbine planetary gear, especially the planet meshing and the planet bearing using a measurement system mounted on the planet gear carrier. The goal is to measure vibrations inside the gearbox on the planet carrier with a standard acceleration sensor and transmit the data to a receiver wireless. Because the sensor and the transducer are installed on the rotating machine part, they can be placed close to the vibration source. Measuring the vibration directly on the planet gear carrier, as presented in this work, enables the use of conventional analysis methods to monitor the bearings of the planets. This will provide more reliable results of the condition of the gear mesh and planetary bearings as at the moment with conventional condition monitoring systems possible. The research and test runs are carried out using state of the art Multi-Megawatt wind turbine gearboxes. Because the research is driven by a university the goal is to share the results and don not develop an independent built-in sensor. Therefore results will provide substantial knowledge for the emerging market of wireless built-in sensors such as SKF InsightTM.
Acoustic Emission as a Tool for Condition Monitoring of Oscillating Pitch Bearings
Poster presenter: Marc Hilbert, Researcher, Rwth Aachen University (on behalf of Dennis Röllinger)
Abstract: Driven by the economic pressure, the manufactures are steadily up scaling their wind turbines. In a quest for cost reduction and lower environmental impact, rotor diameters can be considered as one of the key parameters. One example is the B75 blade made by Siemens with an astonishing 75m total length. But with increasing size of turbine blades comes the need for a more sophisticated load control. Modern Pitch Control Systems with a built-in intelligence make it possible to control the pitch of each blade independently. The so called Individual Pitch Controls (IPC) is quite effective in reducing the fatigue loads but goes hand in hand with a higher stress on the bearings. According to the exemplary report on a collaborative research project “Load reducing control systems for multi-megawatt wind turbines in the offshore sector” the Fraunhofer IWES et al. comes to the conclusion that the bearings of pitch systems undergo a significant increase in fatigue damage which makes a permanent IPC not practical. Design modification of bearings on the one hand and highly sophisticated condition monitoring systems on the other hand could contribute the solution to this challenge. However, conventional condition monitoring systems have their limitations when it comes to the monitoring oscillating bearings. One potential tool for the monitoring of pitch bearings is Acoustic Emission Technology. The high bandwidth and high data sampling rate of modern Acoustic Emission Systems offer the measured information that is needed to estimate the damage during small angular rotation. At the same time this creates huge amounts of data that has to be recorded and processed. In the paper, it will be outlined which burst detection algorithms has been chosen and how the interaction of state-of-the-art hardware with implemented data reduction methods could make Acoustic Emission a tool for condition monitoring of pitch bearings.
Power Performance Verification Procedures in India
Poster presenter: Rathinavel Kumaravel, Head of Test & Measurements Group, ReGen Powertech Pvt. Ltd.
Abstract: Wind industry is growing in terms of having well defined technologies and having well educated investors. The investors are getting seasoned in terms of understanding the different technologies and different concepts of wind turbine generators. They are very keen to know the power performance of the wind turbine in which they have invested, to estimate their rate of returns on investment in setting up of wind farm. In Today’s wind energy practice, power performance verification of wind turbine, post construction of the wind farm, have become more or less a mandatory practice. The industry have reached a level, where the power performance verification is demanded either by the owner of the wind farm or by the lender (financing institution) who funds for setting up of the wind farm.
Though we have power performance measurement standards from International Electrotechnical Commission (IEC 61400-12-1), people around the globe uses different/modified procedure for power performance verification. This research work aims at compiling the different power performance verification procedures followed in Indian wind industry and thereby suggesting the best practice that supplements lesser uncertainty in verifying the power performance of a wind turbine.
SCADA data processing for wind turbine reliability analysis
Poster presenter: Christos Kaidis, Junior Engineering Consultant, MECAL
Abstract: This research project discusses the life-cycle analysis of wind turbines through the processing of operational data (SCADA, Alarm logs and Maintenance logs) from four modern onshore European wind farms. A methodology for SCADA data processing has been developed combining previous research findings and in-house experience followed by statistical analysis of the results. The goal of this research project is to perform reliability analysis of the wind turbines using SCADA data and the relevant Alarm logs. An algorithm has been developed in VBA using the SCADA counters in order to define the downtime events of the wind turbine and the Alarm logs to identify which wind turbine assembly initiated the downtime event. The analysis was performed by dividing the wind turbine into assemblies and the failures events in severity categories. Depending on the failure severity category a different statistical methodology was applied, examining the reliability growth and the applicability of the “bathtub curve” concept for wind turbine reliability analysis. Finally, a methodology for adapting the results of the statistical analysis to site-specific environmental conditions is proposed. The methodology is based on recent findings of other researchers on the impact of wind speed and turbulence intensity on wind turbine reliability.
A better understanding of the inter-annual variation of wind farm energy production
Poster presenter: Carl Ostridge, Engineer, DNV GL
Abstract: Understanding the variability of energy production from year to year is critical to understanding the risks in developing and financing wind farms. A number of factors determine the variability of production, including variable meteorological parameters, such as the wind resource, air density and icing conditions, as well as operational parameters, such as operations and maintenance programs, electrical grid connection reliability and curtailment of electrical output due to grid or operator restrictions.
Better understanding this variability in energy production allows a more accurate assessment of future risk and is therefore important for developers, owners, operators and financers of wind farms. Understanding regional trends of variability also allows better predictions of the benefits of a geographically-diverse portfolio of wind projects.
DNV GL will assess the variability of annual production from wind farms in a number of regions to improve the understanding of IAV. DNV GL will then examine the meteorological and operational factors at the projects to gain insight in to the operational conditions affecting the energy output variability. This investigation will include, but will not be limited to, examining data from available meteorological data sources, including ground-based stations and re-analysis data, as well as operating factors such as availability and environmental conditions. The wind farm energy production variation will be investigated on a regional basis where the available data allow.
Breaking through the OEM filter: A real-world case study on the benefits of independent data collection and monitoring from the wind-farm owner’s perspective
Poster presenter: Setu Pelz, Asset Performance Engineer, Natural Power
Abstract: Natural Power provides independent operational management services to over 35% of the UK’s installed capacity and to over 1300 turbines globally. This case study uses extensive in-house historical operational site data including SCADA data, maintenance data, oil analysis data and component replacement records from multiple on-shore sites totalling several hundred MW in the UK to simulate advanced monthly data analysis during and post the OEM defect liability period.
Market research has shown us that while operational data analysis is gaining significant attention in industry, the vast majority of commercially available asset management products still only provide variations of downtime and availability based reporting. A common wind-farm owner complaint across the industry is the reactive nature of such standard downtime and availability based reporting frameworks and the resulting lack of operational transparency of their assets. Our work explores the application of a multi data-stream collection and monitoring approach using easily accessible real-world operational site data and novel in-house analysis methods to deliver a more comprehensive evaluation of asset operations both in terms of performance and reliability.
The outcomes of the analysis simulation are compared against historical component replacement records and generation data to quantify where possible the potential improvement in wind-farm financial yield. Analysis methods are outlined where possible to allow for plausibility checks.
From SCADA based to non-dimensional indexes: How to manage your turbines portfolio
Poster presenter: Ludovico Terzi, Head of wind technology, Sorgenia Green
Abstract: The proposal demonstrates some applications of SCADA data mining techniques for wind turbine performances analysis, developed by Sorgenia Green in conjunction with Perugia University (Italy).
The philosophy is heavily combining the information of turbine state dynamics with SCADA measurements, for developing a layer of dimensionless indices, independent of the size of the machine and the interface of SCADA control system.
The approach is a climax in the granularity of the analysis, and consequently in the complexity of SCADA data sets post processing.
Using the information contained in the state dynamics, a number map is codified in order to associate to each 10 minute time step a judicious number collecting as much information as possible. Processing the time series of these numbers for each turbine, Malfunctioning Indices are formulated. Their consistency is demonstrated and it is shown how comparison of sample periods against an historical background (one year) is a powerful tool for detecting anomalous operational behaviour. Subsequently, the analysis is specialized to the productive phase, in order to investigate the quality of the power output. In particular, the effect of wake interactions are analysed by the point of view of down performances and mechanical effects.
Wakes cause meandering wind, which the turbine under wake can not follow optimally. Therefore anomalous nacelle blockage and misalignment of nacelle with respect to the wind direction are common symptoms of wake effects. A Stationarity Index and a Misalignment Index are formulated: they quantify how often nacelle is blocked and how often nacelle is misaligned with respect to wind more than a judicious threshold. It is shown that the above Indices are capable to brilliantly codify power degradation due to wakes.
The methods have been successfully tested on a four years data set from two wind farms owned by Sorgenia Green.
(authors: Ludovico Terzi, Davide Astolfi, Francesco Castellani)