Report

Teaser, summary, work performed and final results

Periodic Reporting for period 2 - WINDMIL (Smart Monitoring, Inspection and Life-Cycle Assessment of Wind Turbines)

Teaser

In recent years, Europe has been faced with the problem of excessive energy consumption, closing in on exhaustion of available energy resources. In view of this, the notions of sustainability and resilience have become paramount in resource management and policy-making. Within...

Summary

In recent years, Europe has been faced with the problem of excessive energy consumption, closing in on exhaustion of available energy resources. In view of this, the notions of sustainability and resilience have become paramount in resource management and policy-making. Within such a context, the European Union (EU) has reported that a mandatory target of 20% for renewable energy\'s share in the EU by 2020 was established, with wind energy supplying 14%. Furthermore, the Energy Roadmap 2050 predicts that wind energy could supply between 31.6% and 48.7% of Europe\'s electricity. Wind energy production is well on its way to meet these forecasts currently accounting for almost half (43%) of the global generating capacity of renewable energy sources. Yet, despite the level of refinement that has been reached in the past few years concerning the production of these systems, their life-cycle management is still at its infancy as evidenced via the short life-span, currently set at a mere 25 years, and the lack of efficient operation and maintenance (Ο&Μ) schemes. The cost of the latter may in fact rise up to 25-30% of the total levelized cost per kWh produced over the lifetime of a wind turbine (WT) or 75-90% of the investment costs.

With a number of WTs currently reaching their design span, stakeholders and policy makers are convinced of the necessity for reliable life-cycle assessment methodologies. However, existing tools have not yet caught up with the maturity of the WT technology, leaving visual inspection and offline non-destructive evaluation methods as the norm. Most current installations feature Supervisory Control And Data Acquisition (SCADA) systems, which however fail to provide information on structural performance, including damage, fatigue and deterioration. Structural Health Monitoring (SHM) systems that have been proposed for improving safety and minimizing downtime and maintenance costs, currently rely either on visual inspection and non-destructive evaluation methods that are implemented on a periodic basis, or on very dense sensor layouts that require very careful configuration for automated monitoring. In contrast to these practices, we herein propose a flexible, permanent “protection suit” for WTs, i.e., a long-term monitoring solution featuring low-cost sensors, able to follow the structure from cradle-to-grave and to offer real-time feedback on structural condition.

Scope and Objectives
We propose an innovative Life-Cycle Assessment (LCA) framework for the main structural parts of WTs (blades, tower and foundation) relying on the use of monitoring data for decision-making during the entire life-cycle of the structure. We aim to combine easily deployed, low-cost sensor technology, with next-generation data processing methodologies for maximum feature extraction, with the goal of smart condition assessment and optimal maintenance planning for WTs.
Within such a context, the overarching objective of the project lies in maximizing of WTs’ efficiency by a) decreasing O&M costs; b) providing guidelines for more efficient yet more economical future designs c) enhancing the economic viability of a powerful renewable energy source; and d) providing infrastructure operators with a means for reliably estimating risk, therefore avoiding overestimation and excessive insurance policies. In materializing this goal, the project aims:
1. To develop appropriate performance assessment methodologies, which are able to account for the lack of precise input (loading) information as well as environmental and modeling uncertainties, toward the assessment of structural performance across two temporal scales, namely the short- and long-term one. The former refers to the handling of sudden anomalies typically linked to extreme events (strong winds, waves, earthquake), while the latter refers to deterioration processes that evolve across a lengthier time span and which form an adverse factor for extending the life-cycle of thes

Work performed

The methodological tools that form the core of the WINDMIL project are organized along two main interacting tracks, in accordance with the original description of the project proposal, namely the forward (physics-based) and the inverse (data-driven) track.
In the forward track, adequate computational models are compiled, able to account for the uncertainties relating to diverse loading (wind, wave, wake effects, etc), site-specific conditions (e.g. soil-structure-foundation interaction), as well as necessary modeling simplifications. The inverse track, on the other end, deals (i) with the extraction and handling of information obtained via monitoring (through appropriate sensor deployments) and (ii) with the derivation of data-driven system representations. The latter includes parametric and nonparametric representations, as well as time-invariant and time-variant models for tracking the evolution of the WT dynamics. The novel methods developed as part of WINDMIL are fueled by the synergy of the forward and inverse engineering, building on the fusion of data with models. We call such schemes, coupling physics with data, hybrid models. We have so far developed new methods, along all three tracks, i.e., in terms of forward, inverse and hybrid modeling.

The main tasks accomplished to-date, are in line with the originally envisioned Work Packages (WPs) and include:

1. Forward Engineering - WT Simulation Models (WP1)
Surrogate Modelling and Uncertainty Quantification
We develop multi-type models (physics-based, data-driven or hybrid) that are able to simulate the WT in terms of its components or as a whole and, more importantly, can do this in the required computational time. For the case of diagnostics for example, this assessment has to be delivered in real-time, or near real-time. The latter is not a straightforward task, if one considers that the physics-based analysis of composite structures, such as wind turbine blades in particular, is typically accomplished by means of refined Finite Element models, and is thus costly. Computational homogenization offers a means for incorporating micro-scale effects, while solving at the macro-scale, thereby accelerating computation. Nevertheless, it is still computationally demanding and the effect of geometric uncertainties on the effective homogenized material properties are hard to analyse with a limited number of Monte-Carlo samples given a high dimensional set of random parameters. In a first exploration (Mylonas et al., 2017/UNCECOMP) we have tackled the challenge of computational time by developing Polynomial Chaos Expansion (PCE) surrogate models of the effective homogenized anisotropic stiffness tensor. This tool offers an opportunity to take physics into account in a computationally efficient and scalable manner for modeling complex phenomena tied to damage accumulation on WT blades.

Wake analysis for Structural Health Monitoring (SHM)
Wind turbines structures are described by complex dynamics operating under a wide range of environmental and operational conditions. Amongst these, the varying nature of the wind excitation forms perhaps the main driver for the variability of the induced dynamics. In this sense, the features of the dynamic response of an up-wind wind turbine are expected to be differentiated with respect to that of a similar wind turbine receiving the slower, meandering and more turbulent wind stream of a wake. An important objective of structural health monitoring (SHM) is the life-cycle assessment and the calculation of the remaining useful life-time of structural components, which in turn requires an accurate estimate of the loading pattern on those components. Ideally one would assess fatigue loads on all wind turbines components’ in the farm by direct and comprehensive measurements. When direct measurements through sensors on all turbines are not possible, one can rely on forward physics-based numerical simulations (of various fidelities). This, though,

Final results

Powered by the methods described above, which lie beyond the state-of-the-art, we merge knowledge from the structural engineering, mechanical engineering and energy fields, and bring them together in a holistic decision support tool (main WINDMIL deliverable) that is invariably of cross-disciplinary nature. It is important to stress once again, that this approach to O&M of WTs is new and has been so far missing. The current culture focuses on mechanical components such as the gearbox, completely ignoring the structure, which according to the norms features a prohibitively short nominal life, and which naturally experiences faults and damages that call for expensive inspections and repairs. Structural testing is to this date limited to the prototype testing of WTs, and is not found as a culture for WTs that are in operation. Using the methods and tools described above, we hope to change this status quo and bring intelligent and robust diagnostics into the wind energy sector, thereby redefining the state-of-the-art.

Quite importantly, we have in parallel initiated a number of activities for dissemination of the technology to the practice and industry. This effort is primarily driven by instrumentation projects on actual operating turbines and Wind Farms.
Starting from scaled models, we have built a small-scale wind turbine, which will be used in the laboratory and will serve as a benchmark for linear and non-linear vibrations and dynamics tests, system identification, active and passive control, shake table and hybrid tests.
Moving onto a full-scale system, we are in the process of deploying a dense network of strain and accelerations based sensors on a 7kW, 13m rotor diameter wind turbine (variable speed & variable pitch), located in Switzerland, in collaboration with Aventa AG (http://www.aventa.ch/)
In addition, in early 2019 we will be deploying our sensors on several multi-megawatts wind turbines in a wind farm located in Greece in collaboration with the OSMOS group (https://www.osmos-group.com/en). Hence, the three wind turbines (at three different physical scales) will constitute a test bed to our aforementioned research. This deployment on a farm level will allow us to study the derivation of data-driven indicators for the interaction of turbines (e.g. wake effects) and for studying the influence of site effects on performance.

Apart from those activities, which necessitate our involvement for the installation of sensors, we have been further working on the processing of existing data, obtained through our collaboration with manufacturers and public operators of wind farms.
Our collaboration with Siemens Gamesa has ensured access to data and models from a prototype turbine (SWT-2.3-108) in Boulder, Colorado, where the developed methods for system identification and diagnostics are currently verified. A similar type of verification is offered on data provided by our collaborators in the OWI Lab (Belgium) and is reported in Noppe et al. (2018/ISMA).
Our collaboration with the public operators Vattenfall (Sweden) and Ørsted (former DONG Energy in Denmark) has offered us with access to SCADA data of the Lillgrund Wind Farm (48 turbines) and the Anholt offshore Wind Farm (111 turbines), respectively. This data is particularly valuable for the Decision Support framework we develop, which is built to operate on a farm level (Abdallah et al. 2018/ESREL)

The knowledge produced thus far, as part of this project, has been disseminated in a number of publications appearing in journals and peer-reviewed conference proceedings, as well as in keynote talks by the PI (Prof. Eleni Chatzi), as reported in the Project Output, and on the project webpage (http://www.chatzi.ibk.ethz.ch/erc-stg-windmil.html)

We are currently working on extending the concept of decision tree learning to Object-Oriented decision tree learning:
- where the WT is viewed as a multi-layered system of objects (e.g. structure, controller, actuator, etc.) define

Website & more info

More info: http://www.chatzi.ibk.ethz.ch/erc-stg-windmil.html.