Want to know what’s going on in the wind energy research community but don’t have time to do a detailed study yourself? Read on for a comprehensive guide to the key research topics at the conference “The Science of Making Torque from Wind” (TORQUE 2018) held by the European Academy of Wind Energy at the Politecnico Milano last week. This article consists a summary of key topics as well as one key “good to know” statement at the end for each of the following six main categories: 1. Aerodynamics and acoustics; 2. Control and monitoring; 3. Design, engineering and new concepts; 4. Measurements and experimental techniques; 5. Modelling and simulation technology; 6. Wind, wakes and turbulence.
Do you belong to a company or research institute looking for support in setting up research programmes in the area of wind energy? Are you looking for tailored technology reports or research reviews? Find out more here:
If you haven’t got time to read the whole article, take a look at these summary graphics for the six categories:
1. Aerodynamics and acoustics
The aerodynamics and acoustics category focused on airfoil simulations, full wind turbine simulations and measurements, the investigation of new wind turbine concepts and noise simulations.
The airfoil simulations showed, for example, that morphing trailing edge flaps can improve airfoil lift by around 15%, that blunt-wavy combined trailing edges can improve the power production by 1-2% by avoiding separation, that active trailing edge flaps can reduce loads by about 4% and that active slats can reduce lift force fluctuations by 20%. However, these studies did not really consider the increased costs of implementing these improvements. Also, improved methods for predicting and understanding transition and dynamic stall and for undertaking Detached Eddy Simulations were looked at.
The full wind turbine simulations and measurements mainly involved the development of computational tools for large wind turbines. This included some interesting comparisons of aerodynamic measurements using surface pressure sensors on an operating 2 MW wind turbine with simulations as well as results from the EU project AVATAR, in which 10 MW wind turbines were aerodynamically modelled for the first time. This showed that engineering design tools based on Blade Element Momentum (BEM) methods or Vortex Methods can over-predict fatigue loads by up to 15%. Additionally, there was quite a focus Computational Fluid Dynamics (CFD) of yawed wind turbines, showing that simulations do not yet compare particularly well to experiments, and a detailed numerical Fluid-Structure Interaction study on the NREL 5 MW wind turbine. One study even looked at developing a framework for applying the Lattice Boltzmann Method to wind energy applications as an alternative to conventional CFD, in which the Boltzmann gas equation is calculated at molecular level, instead of the Navier-Stokes equations. Very little work has been done on this in the research community but it may be interesting in the future, due to the potential improved performance for complex flows and its ability in parallel computing.
New wind turbine concepts included measuring an improvement in lift coefficient of 0.3 and a reduction of loads using lift control with fluidic jets on a model wind turbine, simulations showing that double-rotor vertical axis wind turbines are capable of producing more power than two alone-standing turbines under certain conditions and that ducted wind turbines can lead to significant power improvements. However, these improvements must be compared to the expected increase in costs before the potential of these ideas can be properly assessed.
In the area of noise, topics included comparing simulations and measurements of noise emission from wind turbines in wakes, showing that turbulent inflow noise is increased in the wake but that the overall noise in the wake can be reduced due to the velocity deficit, as well as comparison of some wind turbine noise emission models coupled to BEM aerodynamics, showing a number of discrepancies.
Good to know: high-fidelity CFD and measurements can capture the physics of 10 MW-scale wind turbines; however, engineering models still need improvement.
2. Control and monitoring
The control and monitoring category focused on wake steering, wind farm control in general, failures and condition monitoring, load reduction and fatigue as well as wind turbine control.
The investigation of wake steering was the hottest topic in this category, mainly involving the development of control strategies for increasing wind farm power and reducing loads by steering the wakes via yawing the wind turbine based on LiDAR measurements and on engineering models. The method appears to be promising; however, it lacks validation and implementation on real wind farms. Other topics of wind farm control were also looked into, including the development of novel real-time estimation solutions for a medium-fidelity dynamical wind farm model, which allows the dynamics of changing atmospheric conditions and flow dynamics to be integrated into a wind farm controller, as well as the development of an active power controller with the help of Large Eddy Simulations (LES) and a sparse sensor algorithm, reducing computational cost.
The key topics in load reduction and fatigue were the integration of LiDAR measurements in control systems and Individual Pitch Control for load reduction, both involving the development of promising algorithms using computational models.
In the area of failures and condition monitoring, analysis methods including disturbance observer and phased-locked loop, genetic algorithms, order tracking and machine learning, deep autoencoders and data fusion as well as the statistical evaluation of SCADA data were the main focus. Additionally, a new monitoring system for wind parks using passive radar was introduced, showing that it can be used to detect component faults.
For single wind turbine control, the Delft research controller was introduced – an open-source and community-driven baseline controller that can be used by research groups to develop and compare new control algorithms. Other topics were a SCADA data mining method for quantifying the improvement of aerodynamic retrofits, Gaussian process machine learning for estimating the maximum power coefficient and drivetrain losses and neural network modelling of the aerodynamic coefficients of wind turbines, all important in the wind turbine design process.
Good to know: control strategies have been developed to increase wind farm power and reduce loads by steering the wakes via yawing the wind turbine based on LiDAR measurements and on engineering models. The strategies appear to be promising; however, they currently lack sufficient validation and implementation on real wind farms.
3. Design, engineering and new concepts
In the category of design, engineering and new concepts, the main research topics were blade design, wind turbine concepts, wind farm design and drivetrain design.
Blade design included structural and aerodynamic design. In the area of structures, the importance of a realistic representation of core materials which takes into account the resin uptake in sandwich constructions of wind turbine blades was highlighted and the ONE SHOT BLADE® technology was introduced for a 50 m blade. This technology allows production of a wind turbine blade in a single infusion and without any bonding processes, saving significant production time. Additionally, the design of bend-twist coupled blades was improved in order to reduce the loads and at the same time avoiding power loss at lower wind speeds. In the area of aerodynamics, it was shown that blade tip extensions with a winglet can increase the power production by 2.6% and that flutter aerodynamic instabilities can be reduced on 100 m blades by the correct choice and placement of materials.
Interesting wind turbine concepts that were investigated were a 20 MW reference wind turbine, a two-bladed 10 MW rotor with teetering hub, two concepts for the integration of hydro-pneumatic energy storage in floating wind turbines, and an alternative concept for a high-altitude wind power generator. Additionally, it was shown via an investigation into the dynamic behaviour of a 3 MW wind turbine that passive structural control techniques are essential for modern wind turbine design. In general, it should be noted that only a small about of radical innovation seems to be present at the moment in the wind energy research community.
Wind farm design topics included a method for reducing multi-modality in the wind farm layout optimisation problem, showing that wind farm layout optimisation can be improved by factors of between 1% and 10%. A numerical prediction of normal and extreme waves at the Fukushima offshore site showed that using a computational domain covering whole the Pacific Ocean improves the prediction accuracy of significant wave height and wave period. Furthermore, an improved model describing the joint probability distribution of wind speed, wave height and wave period is shown to agree well with measurements.
On the topic of drivetrain design, main bearings and blade bearings were examined. The load distribution in a three-row roller-type rotor blade bearing was examined, showing that fatigue lifetime can be optimised through careful design. Additionally, a review paper showed that, as wind turbines are getting larger and more flexible, more attention to main bearings needs to be made and industry practices are needed that enable sufficient model exchange or interfaces and thus effective exploitation of the benefits of a simulation model throughout the turbine life cycle.
Good to know: initial wind turbine designs on the order of 10-20 MW have been investigated; however, due to the increased elasticity of the components, design tools require further improvement and a more holistic approach – as well as in some cases some new industry standards.
4. Measurements and experimental techniques
In the category of measurements and experimental techniques, the main research topics were wind measurements and aerodynamics, power measurements, loads, structures and vibrations and wind tunnel tests.
On the topic of wind measurements and aerodynamics, LiDAR measurements were done in wind turbine wakes in several projects, showing that wind speed deficits and wake propagation paths can be quantified using two LiDARs in a co-planar set-up in order to retrieve horizontal and vertical wind speeds. It was also shown that axial velocity error of wake measurements can be reduced on average from 7.4% to 2.8% by combining a SpinnerLidar set-up with simulated LiDAR measurements. Additionally, the New European Wind Atlas, a joint research effort from eight European countries, co-funded under the ERANET Plus Program, was introduced. An offshore wind atlas extending 100 km from the European coasts is foreseen within the project, based on mesoscale modelling and various observational datasets. Part of the project involves carrying out large scale field experiments at a high spatial and temporal resolution, providing a significant upgrade to the experimental databases currently available.
Power measurements included the calculation of the power output loss based on thermographic measurement of the leading edge condition combined with simulations based on the Blade Element Momentum method, showing in one test case that a contamination level of up to 90.4% leads to a decrease in annual energy production of between 4.7% and 2.7%, depending on the wind conditions, as well as some studies of power curve measurement methods and uncertainties. Additionally, one study of the performance of a soft kite wind generation device was presented. I am fairly sceptical regarding kite energy, mainly because a whole lot of research and development seems to be going on without anyone seeming to really have much of an idea of the potential energy production or efficiency of the devices.
There was a large interest in the area of loads, structure and vibrations, including interesting techniques such as in-blade load sensing using trailing-edge flaps, showing how the lift coefficient varies with flap angle, and sub-soil strain measurements on an operational offshore wind turbine for design validation and fatigue assessment, showing that it is possible to check design assumptions and identify main fatigue contributions using this indirect, cost-effective measurement technique. Additionally, layouts of low cost sensors for Operational Modal Analysis for the identification of early-stage damage have been tested in order to reduce the investment in monitoring equipment, showing that acceptable results can be obtained with just one low cost well-designed bi-axial sensor, if it is placed in an adequate tower section.
Wind tunnel tests have always been an important topic in wind energy research, because they allow the dynamic behaviour of wind turbines to be investigated under known input conditions. The main topics of interest were investigating the effects of inflow turbulence and dynamic inflow effects on wind turbine behaviour, allowing control strategies to be optimised, as well as some detailed studies on the unsteady aerodynamics using techniques such as Particle Image Velocimetry and Infrared thermography, providing a basis for the improvement of design tools.
Good to know: effective wind turbine loads and vibration monitoring is still very expensive; however, techniques are being developed in order to reduce the costs of monitoring systems, which is vital for the reduction of operating costs in the future.
5. Modelling and simulation technology
The main topics in the category of modelling and simulation technology were aerodynamic and aeroelastic simulations, structure and load modelling, wind turbine and wind farm modelling as well as modelling failures and maintenance.
Aerodynamic and aeroelastic simulations formed a large part of this category, mainly involved Large Eddy Simulations (LES) of wind farms and aero-elastic modelling for special load cases. LES is more accurate that the traditional Reynolds-Averaged turbulence modelling method but are also more computationally expensive. The studies included a comparison to the new mid-fidelity wake simulation software from the National Renewable Energy Laboratory in the USA (NREL), FAST.Farm, showing that it reasonably accurately predicts the performance of individual turbines, wake meandering behaviour and averaged wake-deficit effects. I was very impressed by a comparison between LES and wind tunnel measurements of one hundred static turbine models, showing that experiments with a naturally developing boundary layer and a clean flow above provide well-defined conditions that can be simulated accurately in LES and therefore are useful for the validation of wind farm models. Also interesting was the coupling of LES to a mesoscale Weather Research and Forecasting model, concluding that numerical results agree well with meteorological data from the met tower in the test case examined. The aero-elastic modelling for special load cases included analysis of membrane blades, of swept blades, of oscillating trailing edge flaps as well as of blades during extreme deflection events.
The key subjects in the topic of structure and load modelling were a stability analysis of wind turbines with bend-twist coupled blades, load validation of aero-elastic simulations with measurements performed on a 850 kW wind turbine and wind turbine site-specific load estimations using artificial neural networks calibrated by means of high-fidelity load simulations. This showed that a feedforward neural network with two hidden layers and 11 neurons per layer, trained with the Levenberg Marquardt backpropagation algorithm, is able to estimate blade root flapwise damage-equivalent loads more accurately and faster than a polynomial chaos expansion (PCE) trained on the same data set.
Wind turbine and wind farm modelling included high-resolution periodic mode shapes identification for wind turbines, showing that the first low-damped modes of the system could be identified with good accuracy, as well as an experimental analysis of the force transmission in a rotor bearing support system, enabling a detailed analysis of the mechanical interactions between the main bearing, the gearbox and the structural components. Another interesting study involved analysing a database of wind farm performance data compared to the original wind resource assessment predictions and trying to understand the resulting bias. I found this particularly interesting as in my experience this appears to be quite a problem for wind energy project investors, especially because the actual performance tends to be lower than predicted. Further work is, however, required in order to understand the reasons for this bias.
There was also some interest in the area of modelling failures and maintenance, due to the increased need for advanced maintenance strategies in the wind energy industry. This included training Bayesian belief networks based on failure records, technology specific covariates, as well as measurements of the environmental and operational conditions at the site, achieving a very good accuracy in predicting component failures. Actually I was surprised that there were not more papers on this topic, as it is becoming increasingly important in the industry as operators are tending to do Operations & Maintenance themselves in order to reduce costs.
Good to know: Large Eddy Simulations are becoming standard in the research community for simulating atmospheric boundary layer and wake effects, and the new tool from NREL, FAST.Farm, appears to be a reasonably accurate lower-fidelity tool. There is a need, however, for a decision-making tool to help companies assess when to use which level of fidelity.
6. Wind, wakes and turbulence
This category of the conference was probably the largest and most important. The main part of this was modelling and simulations, despite there being an entire separate topic at the conference called Modelling and Simulation Technology (see above). Other themes in this section were wind tunnel tests, field measurements and wind resource assessments.
In the area of modelling and simulations, there were many Large Eddy Simulation (LES) studies, which were similar to the ones presented in the modelling and simulation technology section. Several simulations in complex terrain showed that the main flow features can be captured despite the relatively simple set-up by comparisons to field measurements; others looked at the validation of the actuator line model and the wake meandering model within an LES framework. Also interesting was a data-driven machine learning method for wake modelling, required by operators in estimating the reserve power required by the most recent grid codes, and the investigation of a wind farm layout called vertical staggering, in which wind turbines of different heights are installed in order to reduce wake effects. This showed that vertical staggering can increase the power production in the entrance region of a wind farm, but may increase the power fluctuations and have negative impacts for the wind turbine manufacturer.
Wind tunnel tests included an investigation of the influence of ground roughness on the wake of a yawed wind turbine, showing that the wake is less deflected with increasing terrain roughness, and of the interaction between free-stream turbulence and blade tip vortices, showing that the strength of the vortices is reduced depending on the turbulence levels. Another study measured the impact of gusty inflow on an adaptive camber airfoil using Particle Image Velocimetry, finding that the coupled leading and trailing edge allows lift peaks to be reduced significantly for suddenly occurring changes in angle of attack.
Several field measurements of wake effects were undertaken, including the wake behind an offshore wind farm observed with dual-doppler radars, finding that the wake region extends at least 17 km downstream of the wind farm and that the observations agreed well with two wake models, provided a coastal gradient is accounted for. An important study, in my opinion, was the investigation of wake characteristics between research-scale and full-scale wind turbines, showing that wakes from different-scale turbines expand similarly when not limited by the ground, but that the meandering magnitude is not easily translatable across scales. However, the far wakes of different-sized rotors at the same absolute height are scalable. This should be accounted for in the interpretation of research results.
There was a particular focus on the topic of wind resource assessment, including introduction of an annual energy production estimation methodology for onshore wind farms over complex terrain using a Reynolds-Averaged Navier-Stokes model with actuator discs an, a steady state CFD approach for stratified flows and a comparison of meso-micro methodologies. In this work, a hierarchy of methodologies that rely on coupling mesoscale and microscale models was evaluated as a trade-off between accuracy and computational cost. This goes back to my comment in Section 5 regarding the need for a decision-making tool to help companies assess when to use which level of fidelity.
Good to know: Large Eddy Simulation have established themselves as a standard for modelling wake effects; however, more work is required on the combined modelling of wake effects with complex terrain and weather effects.
The conference proceedings are freely available here: http://iopscience.iop.org/volume/1742-6596/1037
Did you like this article? Do you belong to a company or research institute looking for support in setting up research programmes in the area of wind energy? Are you looking for tailored technology reports or research reviews? Get in touch here: