Abstracts - WindEurope Workshops
Watch Giles & Pierre Episode 15

Abstracts

Topic 1 – The application of big data and advanced statistical techniques

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.

Analysis and benchmark-oriented evaluation of energy yields with focus on the performance assessment of wind turbines

Presenting author: Philip Görg, Fraunhofer IEE

Abstract

The contribution deals with the topic of analyzing low energetic yields as well as the identification of their different causes and the derivation of possible optimization approaches. For this purpose, a benchmark concept is created which describes the different steps and includes the information required to build a holistic framework which can be used for the performance assessment of a wind turbine (WT).

The contribution compromises the implementation of the suggested concept as well as the deduction of different key indicators which can be used to assess multiple energetic loss causes in terms of their magnitudes. Different key indicators are found which allow to quantitatively assessing the performance, the nacelle misalignment, and the wind measurement system of a WT, respectively. The methodology for deriving these indicators is then applied to the operational data of the Fraunhofer IWES WInD-Pool to determine a critical threshold for each considered loss cause to identify potentials of optimization.
The results show that the majority of all analyzed turbines have a “non-critical performance” in at least 98.50% of their operational time. Moreover, an approximately reverse square correlation between the mean generation ratio and the operational lifetime of all analyzed turbines can be seen. While the turbines performance increases significantly throughout the first year of operation from an initially low level, a continuous decline of the generation ratio can be seen after some operational years..
The usage of status codes from the turbines operating system to identify critical performance events is only partially effective. Most of all data points found to be “critical” cannot be linked to a specific status code (approx. 14%).

Applying the proposed key indicators to identify yaw misalignments of turbine nacelles suggests a threshold of 2.94% (time-based ratio). Furthermore, onshore WT show higher values than offshore WT.

The wind measuring system of all considered WT shows a significantly higher deviation between the measured value and a reference value (wind farm mean) in standstill than in operational times. Additionally, all considered turbine types show a higher tendency to underestimate the wind speed compared to the underlying references during standstill times (median: -0.59m/s).

Using Operational SCADA Data to Optimise Assets

Presenting author: Charles Plumley, Wood

Co-authors: David Robb

Abstract

While performance is often measured by just looking at budgeted outputs, availability and occasionally power curves, a more in depth analysis of the SCADA data can reveal trends and opportunities for optimisation that might otherwise be passed by. Wood’s optimiser data analysis approach involves:

  • Looking at individual event codes to identify common issues across a farm, portfolio or turbine type
  • Monitoring performance in a comparative manner to identify issues at an individual turbine or farm
  • Exploring a wide variety of SCADA signals to identify anomalies and suboptimal designs
  • And predicting the potential for life extension and uprating through comparing site conditions with the wind turbine class

By delving more deeply into the available SCADA data, conclusions can be drawn to highlight where best to focus effort to improve performance. For example, following the data analysis, endemic problems such as pitch or electrical failures can be pursued with the OEM, targeted forestry felling can be justified, controller and hardware improvements can be applied, unnecessary curtailments or operational modes can be removed, and uprating or life extension considered, but only where appropriate.

These techniques have benefited over 1 GW of capacity so far and the lessons learnt and experiences gained are shared here to benefit the wider community.

Visualisation and automation data to decisions

Presenting author: Mark Spring, Lloyd’s Register

Co-authors: Philip Knaute, Matthew Zhang

Abstract

Through working with wind farm owners it has become clear that improvements in performance are hampered by limited access to comprehensive turbine operational and contextual information.  New software tools developed have enabled both an increase in revenue from electricity generated and a reduction in operational costs, due for instance to on-site interventions.  Technical hurdles in the path include pitch angle errors, yaw misalignments, power curtailment, mechanical and electrical losses and production lost during inefficiently scheduled turbine downtime.

Implementations of the approaches described in this paper have demonstrated that seeing a problem clearly has been the first step towards addressing it, diagnosing the causes, mitigating future recurrences and predicting the risk of experiencing each particular mode of failure.  We have succeeded in automating the display of key relationships and combinations of data.  By using domain knowledge from design engineers, technicians, business managers and specialists with relevant experience from outside the wind sector, we have trained the data science tools so as to automate the process of diagnosis and simulation of future operation.  Future risks of failure have been estimated by understanding modes of operation, accumulation of damage, the physics of failure, interactions between environment and turbines, incorporating the influence of maintenance teams, parts suppliers.  This understanding has been incorporated into practices and software.

The process started with a formal FMECA.  Automation of the subsequent steps of data classification, fault diagnosis and failure prediction have relied on novel combinations of proven data science algorithms, face-to-face workshops, user-interface mock-up exercises and condition monitoring systems.

For a system to be resilient it needs to exhibit the following characteristics: plasticity, adaptivity, redundancy, autonomy and self organisation.  This project has demonstrated progress along the path towards this goal for wind power plants.

Catching the context: cross-fleet benchmarking enabled by context-sensitive fingerprinting of SCADA data

Presenting author: Tom Tourwe, Sirris

Co-authors: Elena Tsiporkova, Alessandro Murgia, Mathias Verbeke

Abstract

Wind turbine data, whether it is CMS or SCADA data, is typically analysed using advanced data mining and machine learning algorithms that exploit the time series. Time-sensitive features extracted from hourly, daily, weekly or monthly periods do not reflect explicitly the actual context in which the turbine is operating however. As a consequence, the resulting models are turbine-specific and are difficult to generalise to the whole fleet or portfolio. In this presentation, we will present an alternative approach that is based on characterising the operational context in which the turbines are operating. We will explain how this context can be represented, how each turbine’s performance within the different contexts can be modeled, how this allows to identify groups of turbines that behave similarly, and how ultimately these context-sensitive fingerprints can be exploited to derive context-aware performance and cross-fleet benchmarking.

Power Curve Analysis using high-frequency SCADA Data

Presenting author: Gianmarco Pizza, Nispera AG

Abstract

In the present study, the power curve of a modern multi-megawatt wind turbine is analysed using high-frequency (1 second) SCADA data. Different transients are investigated and described in detail, like e.g. turbine startup, turbine yawing, turbine response to wind speed ramps, and comparison with commonly used 10-minutes averaged data is provided.
Preliminary results have been presented recently at Windenergietage
http://spreewind.de/windenergietage/wp-content/uploads/sites/4/2017/11/26WT0811_F16_1220_Nispera.pdf

WindEurope Annual Event 2023