A wind turbine blade equipped with root spoilers is analysed using time domain aeroelastic blade element momentum (BEM) simulations to assess the impact of passive devices on the turbine annual energy production (AEP) and lifetime. Previous 2D computational fluid dynamics (CFD) showed a large unsteadiness in aerodynamic coefficients associated with the spoiler, and such behaviour is captured by the OpenFAST simulations when all degrees of freedom are switched off. Once the turbine is fully flexible, a novel way to account for aerofoil-generated unsteadiness in the fatigue calculation is proposed and detailed. The outcome shows that spoilers, on average, can increase the AEP of the turbine. However, the structural impacts on the turbine can be severe if not accounted for initially in the turbine design.

Thanks to a steady rotor size increase over the last decades, the wind energy sector managed to grow. In the onshore wind sector, due to various limitations, the rotor diameter remains constrained, but blades over 60 m long are now common. Larger blades require more attention to detail during the design phase to reduce the cost. The maintenance cost during the turbine lifetime increases too, and a good understanding of the turbine ageing is necessary.

In order to reduce the levellized cost of energy (LCOE), turbine manufacturers had to imagine solutions to increase the energy output of existing turbines. Among such solutions, there are aerodynamic add-ons (AAOs), which are mostly passive devices attached to the blade surface to either lower the acoustic emission or increase the power extraction locally.

With the increasing rotor diameter and hub height, turbine manufacturers are now facing aeroelastic challenges where tower and blades can deform over large distances. Before several extensive measurement campaigns of scaled models in large wind tunnels or in the field were performed (see

ENGIE Green is a French exploiting party of renewable energy sources. During routine maintenance, some cracks have been noted at the blade root near the spoiler installation. The present paper aims at understanding the aerodynamic causes of the structural failure, using state-of-the-art calculation methods and proposing a novel way of predicting the impact of spoiler on the blade structure lifetime. The AEP gain will also be evaluated. Because performing fatigue calculation using CFD would be too computationally expensive, and BEM cannot directly account for aerodynamic unsteadiness, in this paper we propose bridging the gap by utilizing the strengths of both simulation methods. First the methodology to build the aeroelastic model is explained in Sect.

The wind turbine geometry used in the present study was acquired during a scanning campaign on an operating 2 MW turbine (see

Overimposed aerofoil shapes at radial position R6 (

The scanned blade geometry: chord, twist and relative thickness distribution against the normalized radius.

The tools used to perform the spoiler impact assessment are CFD for the polar generation and blade element momentum (BEM) theory for the aerodynamic calculations. BEM is used to calculate associated loads and compute the turbine annual energy production (AEP). The BEM solver used is the AeroDyn module (see

AeroDyn is a well-known tool developed by NREL and has been used in many international or academic projects. A thorough explanation of the BEM theory is available in textbooks such as

The axial induction factor

Then, the inflow angle

The angle of attack,

Read the

Calculate the loads in the rotor plane using

To account for the finite blade span, the Prandtl tip correction factor is calculated.

The initial induction coefficients,

The unsteady BEM equations can be applied: yaw models, dynamic wake model, blade acceleration due to its deflection and tower shadow effect.

A convergence criterion,

After convergence, the local loads (aerofoil level) can be calculated.

Once all elements are converged, the integrated loads (rotor and turbine level) can be computed.

The procedure described relies on steady polar to perform the iterative steps; it is an inherent BEM limitation. However, as highlighted in

Spoiler case

Turbine configurations analysed.

The scan does not give any information on the blade's material, since only the outer skin was measured. Material properties are a crucial element for turbine design, and as part of an academic or wind turbine exploiting party, we do not have access to this information. Therefore, for the rest of the aeroelastic study, the blade and tower mechanical properties will be scaled using the open-source NREL 5 MW turbine (see

Campbell diagram comparison between the NREL reference turbine and the ENGIE Green scaled one. The solid lines show the NREL response, and the dashed lines show the ENGIE Green turbine's response. The dark shaded area illustrates the ENGIE Green's turbine range of operation.

The blade structural properties needed are the edgewise and flapwise local stiffnesses along the radius:

Moreover, the blade and tower modal shapes, necessary OpenFAST inputs, have been recalculated using the scaled mechanical properties. A Campbell diagram illustrates that despite the difference in length and mass, both turbines behave similarly, as desired (see Fig.

A final sanity check was performed on the mass to assess the validity of the scaling. The blade and tower mass where respectively 0.6 % and 1.3 % off compared to the manufacturer's design specifications, which is small enough to be acceptable. Figures

Blade stiffness properties. The blue lines (blue

Tower stiffness properties. The blue lines (blue

The grid independence study and polar generation methodology have already been performed and presented in

CFD-calculated blade section polars defining the BEM model assuming an inflow between 8 and 8.5 m s

The blue dot (blue

The blue dot (blue

Initial BEM simulations showed that high angles of attack can be reached (

The following sections will detail the model set-up used during the aeroelastic simulations. The first goal of the present paper is to determine the maximum aerodynamic potential of spoilers compared to a bare blade, free of any constraints. A second goal is to assess the impact of the spoiler, on the turbine lifetime, when running at maximum power extraction.

The pitch settings for maximum power extraction are unknown. The turbine manufacturers may not recommend maximum power generation pitch settings due to potential noise, stall or load issues. Therefore, using SCADA measurements is not sufficient, and an optimization study is necessary. In order to reduce the optimization space to only a single variable (the pitch settings), we assume that the turbine's rotational speed available thanks to averaged field measurements is optimized and will not vary. Then, a search for the optimum pitch settings was carried out for each wind speed between cut-in (3 m s

Power surface response with varying pitch settings for different wind speeds for the spoiler case using mean aerodynamic polar. The black dotted line is the turbine's optimal pitch settings for maximum power generation.

Optimal pitch settings for both no spoiler and spoiler cases using mean aerodynamic polars.

The blue square (blue

In the first analysis, the turbine is considered rigid (i.e. not flexible) with the hub height 80 m above ground using the standalone AeroDyn module. The aerofoils associated with the CFD-calculated polars precisely define the blade discretization as detailed in Table

The air density in the BEM calculations is considered constant in space and time and is equal to the one used for the CFD polar calculation (

The blue square (blue

The second analysis is a fully flexible turbine with turbulent wind using OpenFAST. The tool TurbSim (see

After running all the turbine configurations, a deep aerodynamic analysis is possible as many sensor outputs are available. For brevity reasons only a small sample of all the available results will be presented. The multiple polar “states” (mean, maxi, mini) allow assessment of the variation around the mean value, giving a measure of unsteadiness. First, the rigid turbine loads, power and AEP are analysed. Secondly, the flexible turbine fatigue impact is analysed.

In Figs.

The blue square (blue

The lift coefficient of the no-spoiler case shows very low values inboard, as expected from very thick aerofoils. After the radial position R7.2 (

The axial induction,

The no-spoiler case show very low induction values at the root of the blade due to the cylinder shape: low lift coefficient and high drag values. The blade's inboard is not efficient to extract energy but the expected load level is consequently low. Where the spoiler is installed the induction increases, and similarly to the lift coefficient the upper band of the variation due to the polar unsteadiness is close to the optimal induction. The average induction level at the spoiler location is close to

Interestingly, for 6, 7 and 8 m s

The local out-of-plane force (

The blue square (blue

The previous figures showed the results at the aerofoil level, and the next phase of the analysis will focus on the integrated values.

The flapwise root bending moment (RBM) is a critical parameter for blade design and is directly linked to

The unsteadiness caused by the spoiler does not seem to be reflected at rotor level, and the coloured area around the mean value is almost nonexistent. Also, because the change is very small, both curves seem overlapped in Fig.

The blue square (blue

The lower RBM value in the spoiler case is explained thanks to the pitch settings, and the same explanation as for the out-of-plane force

The mean power curves for the no-spoiler and spoiler configuration can be plotted (see Fig.

Power curve close-up for the low wind speeds. The blue square (blue

It is to be noted that, interestingly, the power gain of approximately 1 %, for wind speeds up to up to 8 m s

After integrating the mean power curves over a year simulating a wind site condition IEC class II (Weibull shape factor

Spoiler impact on the AEP.

n/a: not applicable.

When using BEM, one cannot use a time-varying description of each angle of attack during the iterative procedure. Using several steady-state polars representing the different possible aerodynamic coefficients allowed for a first estimation of the variation due to the unsteadiness. Analysing the loads or the different aerodynamic metrics (such as presented in Sect.

Detailed AEP gain and variation for all configurations assuming a wind class IEC II.

The total variation in power for each wind speed is found by

Table

Spoiler total AEP variation around the mean value.

Figure

One should note that performing AEP calculation should “smooth out” any variation due to the polar unsteadiness since the power is integrated over a year. Nevertheless, we have chosen to treat each result for each polar state independently in order to define boundaries for the spoiler's potential. The gains presented assume a single turbine operating at maximum power production, as detailed in Sect.

As seen in the previous sections, the rigid modelling shows little AEP benefit of installing the spoiler. However, due to the large increase in the mean local loads and its associated variation introduced by the spoiler, it seems interesting to investigate the damage and fatigue on the turbine. The aeroelastic calculations will be performed by OpenFAST with a fully flexible turbine. The fatigue analysis will focus on the blade only but can be extended to the whole turbine.

A method to account for unsteadiness on a rigid turbine has been presented in Sect.

Plot of the angle of attack versus the vortex shedding frequency for the no spoiler and spoiler cases, with radial position R6 (

OpenFAST time evolution of 0.5 s of the local out-of-plane force (

In

Because of the sampling theorem, the OpenFAST sampling output rate must be at least 2 times higher than the highest VSF. The highest calculated VSF of all sections is approximately 60 Hz. To add safety margin, the OpenFAST output is set to be at 160 Hz, which is equivalent to a time step of

Once all aeroelastic results are available (Fig.

Interpolated OpenFAST results of the local out-of-plane force (

Creation of the intermediate time series by alternating between the different interpolated time series.

An intermediate time series is generated, for each sensor. Again, supposing a periodic variation in the lift and drag coefficients, we assume that the first aerodynamic coefficient “seen” by the aerofoil is from the maximum polar, and it then changes to the mean polar and finally the minimum polar and varies following this cycle for 600 s. Such behaviour leads to the creation of the pink curve in Fig.

One final numerical manipulation is necessary because all intermediate time series created possess different VSF and therefore different

Generation of the final time series using the sampling rate from OpenFAST (

OpenFAST output normal force to the rotor plane for an average horizontal wind speed of 8 m s

This method is repeated for each radial position, each wind speed and all local loads. The results presented in the next sections use the data generated by this method.

Figure

Figure

OpenFAST output normal power spectral density of the normal force to the rotor plane for an average horizontal wind speed of 8 m s

After running in OpenFAST all wind speeds for both turbine configurations and generating the new time series as described in Sect.

The method developed can only account for sectional loads since it relies on vortex shedding frequency. The integrated load such as RBM cannot be associated with any particular VSF.

Lifetime expectancy evolution with respect to out-of-plane local load. The blue square (blue

Lifetime expectancy evolution with respect to the in-plane local load. The blue square (blue

In order to calculate the blade lifetime with the predefined pitch and RPM settings, an ultimate load before rupture for each analysed sensor must be given. Since the material properties are unknown, we used MExtreme (see

Because of the different hypotheses taken, we are only analysing trends and not presenting the direct Mlife results. Therefore a life index (

Life index of the spoiler case.

As suspected in Sect.

To compare the results of the proposed method, Table

Life index of the spoiler case assuming steady polars.

The authors built an aeroelastic BEM model for a commercial 2 MW turbine retrofitted with root spoilers using a 3D blade scan. The turbine was chosen to address a maintenance problem where blades cracked after installing the spoilers. Regarding the AEP gain and rotor integrated load, the spoiler impact is marginal. The AEP increases by a small amount (

A fatigue analysis has been performed using a novel way of capturing the aerodynamic unsteadiness due to the aerofoil's behaviour. It uses 2D CFD flow characteristics (vortex shedding frequency) as well as the results calculated from three different steady polars (maximum, mean and minimum aerodynamic coefficients). The spoiler increases the already locally present unsteadiness and should not be neglected in the turbine's structural design. The spoiler can be detrimental to the turbine lifetime: retrofitting such devices should be done with care, and the turbine's mechanical properties should be re-evaluated prior to installing the spoiler.

A similar conclusion could be drawn for other aerodynamic add-ons generating a similar amount of unsteadiness; however dedicated studies would be necessary. Also, more studies would be required to quantify the impact height and chordwise position of the spoiler.

Finally, the presented method currently relies on 2D assumptions and BEM calculations, and further studies involving 3D CFD are being carried out to assess the vortex shedding behaviour on a rotor.

Blade characteristics.

Hub characteristics.

Nacelle characteristics.

Drivetrain characteristics.

Tower characteristics.

Ultimate loads for various sensors.

Codes and data sets are available on demand.

TP performed the scans post-processing, CFD pre-processing and post-processing, BEM model building, calculations and analysis, and writing of the paper. EG performed CFD verification and helped set up the CFD model. CLB and AF provided feedback from the industrial point of view, and CB helped with proofreading a previous version of the manuscript and physical analysis of the results.

The contact author has declared that none of the authors has any competing interests.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

The authors would like to acknowledge the ANRT (Association Nationale de Recherche Technologique) for their financial support.

This research has been supported by CIFRE (grant no. 2019/1426).

This paper was edited by Alessandro Bianchini and reviewed by two anonymous referees.