Renewable Energy 62 (2014) 506e515

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Multi-objective optimisation of horizontal axis wind turbine structure and energy production using aerofoil and blade properties as design variables
Gunter Reinald Fischer a, *, Timoleon Kipouros b, Anthony Mark Savill b
a b

Nordex Energy GmbH, 22419 Hamburg, Germany Craneld University, Craneld MK43 0AL, United Kingdom

a r t i c l e i n f o
Article history: Received 6 March 2013 Accepted 12 August 2013 Available online 7 September 2013 Keywords: Bernoulli beam Horizontal axis wind turbine Multi-objective optimisation Parallel coordinates Tabu search Wind turbine blade design

a b s t r a c t
The design of wind turbine blades is a true multi-objective engineering task. The aerodynamic effectiveness of the turbine needs to be balanced with the system loads introduced by the rotor. Moreover the problem is not dependent on a single geometric property, but besides other parameters on a combination of aerofoil family and various blade functions. The aim of this paper is therefore to present a tool which can help designers to get a deeper insight into the complexity of the design space and to nd a blade design which is likely to have a low cost of energy. For the research we use a Computational Blade Optimisation and Load Deation Tool (CoBOLDT) to investigate the three extreme point designs obtained from a multi-objective optimisation of turbine thrust, annual energy production as well as mass for a horizontal axis wind turbine blade. The optimisation algorithm utilised is based on Multi-Objective Tabu Search which constitutes the core of CoBOLDT. The methodology is capable to parametrise the spanning aerofoils with two-dimensional Free Form Deformation and blade functions with two tangentially connected cubic splines. After geometry generation we use a panel code to create aerofoil polars and a stationary Blade Element Momentum code to evaluate turbine performance. Finally, the obtained loads are fed into a structural layout module to estimate the mass and stiffness of the current blade by means of a fully stressed design. For the presented test case we chose post optimisation analysis with parallel coordinates to reveal geometrical features of the extreme point designs and to select a compromise design from the Pareto set. The research revealed that a blade with a feasible laminate layout can be obtained, that can increase the energy capture and lower steady state systems loads. The reduced aerofoil camber and an increased L/D-ratio could be identied as the main drivers. This statement could not be made with other tools of the research community before. 2013 Elsevier Ltd. All rights reserved.

1. Introduction Regardless of the recent economic challenges, global wind turbine installation can record healthy growth gures year by year. Despite this encouraging development, competition between companies in the sector is intensied and the prot margins are shrinking. The development of new, more reliable and efcient turbines is one way to answer this competitive pressure. Therefore, tools like the Computational Blade Optimisation and Load Deation Tool (CoBOLDT) can help to gain new insights into a highly constrained and complex design space represented by wind turbine rotor blade development.

* Corresponding author. E-mail address: gscher@nordex-online.com (G.R. Fischer). 0960-1481/$ e see front matter 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.renene.2013.08.009

For the presented research paper three objective functions are minimize, that are proportional to the cost of energy (CoE) of a wind turbine investment over its life time. This approach has been taken because the direct calculation of CoE is partly ambitious and many cost factors cannot be inuenced by the turbine manufacturer but are depended on the site where the wind turbine is erected. To evaluate all cost factors Fingersh et al. [1] proposed a wind turbine scaling and cost model, in which statistical data of thousands of turbines is used to derive empirical formulations for the costs of the main wind turbine components. However, the physics behind a certain turbine design is not reected. We are approaching the topic from a manufacturer perspective and focus on the reduction of wind turbine thrust and blade mass as proxy of turbine capital cost and the improvement of Annual Energy Production (AEP). For this purpose we concentrate on the geometrical blade properties, namely the shape of the aerofoils and the blade

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

507

Nomenclature AEP annual energy production AoA angle of attack BEM blade element momentum CFD computational uid dynamics Cl lift coefcient CoBOLDT Computational Blade Optimisation and Load Deation Tool CoE cost of energy DoF degrees of freedom EPD extreme point design FEM nite element method GA genetic algorithm GFRP glass bre reinforced plastics HAWT horizontal axis wind turbine IEC international electrotechnical commission MOTS Multi-Objective Tabu Search 2D-FFD two-dimensional Free Form Deformation

functions, since they represent turbine parameters with high impact on the chosen objective functions. The paper is subdivided into six sections. After this introduction and the methodology and research outline in chapter two, the construction and functional details of the optimisation package CoBOLDT are presented in chapter three. In chapter four a design case using the Nordex Rotor Blade NR41 [2] as initial design is shown and the optimisation setup is stated. A detailed post optimisation analysis of several points in the Pareto set will follow in chapter ve which contains a parallel coordinates analysis to relate geometrical features to the considered performance metrics. This analysis is supposed to help the designer to concentrate on only those parameters that are important for the design problem. The conclusion in chapter six summarizes the ndings and discusses possible future developments. 2. Methodology and research outline In the eld of wind turbine blade optimisation the choice of the optimisation technique is particularly challenging since traditional gradient-based algorithms are often unable to cope with numerical, mathematical or physical noise of the processed signals. In some cases the solution can be to evaluate and improve the gradient quality by curve approximation [3]. Another way to circumvent the problem of establishing suitable gradients can be the use of metaheuristic optimisation approaches. Especially Genetic Algorithms have been in the focus of many recently published articles [4e8]. A Paper from Benini and Toffolo [9] shows how a multi-objective GA can solve the trade-off between structural requirements in the root section versus the aerodynamic requirements at the middle and tip section of the blade. The design case uses a stall regulated wind turbine with chord and twist parametrised with Bezier curves and the tip speed and tip speed ratio as scalar parameters. The structural mass is estimated with an I-beam for which the skin thickness is calculated with glass bre reinforced plastics (GFRP) and loads from a steady Blade Element Momentum (BEM) simulation. This structural model originally proposed by Fuglsang and Madsen [10] has also been used by other researchers in this eld [11,12]. However in the current paper the core of the optimisation package CoBOLDT is based on the Multi-Objective Tabu Search (MOTS) as described by Jaeggi et al. [13]. It is a fast and efcient algorithm with a powerful local searching scheme which has been proven to successfully solve multi-dimensional aerodynamic

optimisation problems [14e18]. Since MOTS was designed to solve large problems it is able to run in a parallel computing environment on a 64 bit Linux operating system. The memory management and the selection of new parameters is the task of the master code, whereas the evaluation of each design will be performed by a number of slaves. We prefer this type of optimisation over the mentioned examples of GAs as the small modication from one design to the other in MOTS is perfectly suited for parameters with high impact on the optimisation objectives. In terms of the aerodynamic performance evaluation of the rotor three different methods with varying delity and associated computational cost are most common. The slowest option is Computational Fluid Dynamics (CFD) which is therefore mainly used to verify rotor performance after the optimisation [17] or to investigate wind turbine sub models like the blade tip [19]. Secondly potential ow theory methods like in Abedi et al. [20] can be used but are also quite slow and usually require knowledge of the wind turbine wake. The third and most prevalent method is the BEM method, which is very fast a reasonably accurate [3,17,17] and is employed for this research. The automated structural layout with a Bernoulli beam model has been chosen due to its fast computational speed opposed to the Finite Element Method (FEM) presented by Jureczko et al. [7] or Chen et al. [21]. The model takes advantage of the real laminate properties of the structure and can provide a more realistic mass and stiffness prediction opposed to the structural model of Benini and Toffolo [9] or Young and Wu [11]. The polar data needed for the BEM is created with Xfoil [22] like shown in other publications in the eld [6,23]. The paper combines these theories and applies them in multidisciplinary, multi-objective optimisation to an industrial scale wind turbine to achieve the afore mentioned research objective. Opposed to other design system presented in the literature a simultaneous relative parameterisation of aerofoils and the blade functions allows to exploit the effects of both design areas without being limited to one of them. The structural model at the same time should guarantee the structural feasibility, which is considered another novelty. The design strategy can hence give a more holistic indication of what kind of aerofoils properties are needed at which span wise position and how important the actual blade functions are. However the tool cannot replace the designer as in the absence of a loads base turbine cost model the selection of a compromise design is to some extend always subjective.

Fig. 1. Workow of CoBOLDT.

508

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515 Table 1 Target functions and hard constraints. Target functions TFAEP AEPini /AEP TFMass mass/massini TFThrust thrust/thrustini Hard constraints Steady chord reduction outboard of chordmax Steady absolute thickness reduction Steady relative thickness reduction Maximum chord < 4 m Maximum tip speed < 74 m/s No exceedance of stall angle

3. Description of the CoBOLDT The current optimisation package CoBOLDT features seven modules that provide the functionality for optimisation, parameterisation, geometry generation, performance evaluation and structural layout as presented in the ow chart in Fig. 1. Before the optimisation process can be initialised the user needs to specify at least two target functions that are minimised as well as for every variable the initial value, initial step size and minimum and maximum ranges. After the optimisation process is invoked a design vector is created and fed into the rst parameterisation module. Rather than an absolute parameterisation of blade functions we implemented a relative parameterisation that depends on normalised initial blade functions. The advantage of the relative parameterisation is that more complex features, like sudden changes in the curvature at the hub or the tip of the blade, can be captured with a lower number of design variables. The new function is still similar to the initial blade function but simultaneously a reduction of the design space dimensionality is achieved. The drawback, that not every blade can be reproduced with this technique is compensated by the possibility to replace the initial design for a restart run. For each function four variables are specied which control the shape of two tangentially connected cubic splines. The rst two specify the X and Y position of the spline connection point whilst the remaining two dene the slope at the connection point (a1) and the end point at the tip (a2). The slope at the root side is set to zero to prevent changes of the blade root diameter. If the spline curve has values below zero they are subtracted from the initial blade function or added if values are positive. An example of the method and the effect on the blade geometry is depicted in Fig. 2. After several geometrical constraint checks (see Table 1) have successfully been passed to secure transportation and turbine noise level, the perturbed blade functions are accepted and the aerofoil parameterisation routine starts. The spanning aerofoil family that is used to create the blade is normalised to a chord length of one and comprises 8 proles ranging from 100 percent relative thickness at the circular root section to the thinnest aerofoil at the tip. To reshape the aerofoils two-dimensional Free Form Deformation (2DFFD) is used which is a method originating from computer graphics that has received a lot of attention in aerospace parameterisation [24e27].

As opposed to aerofoil generation with Bezier Splines [5,6,28,29] it has been integrated due to the relative nature of parameterisation and the capability to parametrise a design with a small number of parameters whilst preserving the topology characteristics of the baseline shape. The blade root and tip are not supposed to be changed and therefore aerofoils at this location are excluded in the presented approach. For every prole eight control points are dened as shown in Fig. 3. The number of control points can be adjusted as appropriate to t the parameterisation aim. The rst and the last control point on the suction and pressure side cannot be distorted to maintain the leading and trailing edge positions and prevent a change of the chord length. The remaining control points can each be moved with two degrees of freedom along and perpendicular to the chord. To reduce the order of dimensionality we have chosen to use the same set of parameters for every aerofoil since parameters for control point movement perpendicular to the chord are scaled with the relative thickness. Hence perturbations of thick proles are larger than those for thin proles and smooth generation of the loft surface without the problem of long-waved buckling is ensured. Following the second parameterisation module the nal blade is built from the blade functions and the aerofoils in the geometry generation module of CoBOLDT. The resulting geometrical blade description list and the triangulated surface are the basis for the static BEM and the structural calculation respectively. In the last stage of this module, when all corrections of the aerofoil shape have been done to secure manufacturability of the whole blade (e.g. correction for minimal trailing edge thicknesses), equidistant sections are considered to evaluate the aerodynamic performance. For this purpose, Xfoil [22] was chosen, because an initial exploration conrmed its fast execution and accurate prediction of aerodynamic coefcients for thin aerofoils. The calculation is

Fig. 2. Spline parameterisation of blade chord.

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

509

Fig. 3. 2D-FFD parameterisation of aerofoils.

Fig. 4. 360 degree lift coefcients at Re 3 Mio. for the NR41.

limited to proles thinner than 35% relative thickness. The code was extended to include the sequential calculation of the aerofoil properties from 10 to 20 angle of attack (AoA) at a xed conservative Reynolds number of 3 million. The decision to x the Reynolds number has been taken because at this point in time the information about the rotational speed, to calculate a Reynolds number for a given blade, is not yet available. Each polar is then calculated twice with one design case using the natural transition and the other run being tripped at 1% chord on the pressure and the suction side. The actual polar is established with a user-dened mixing ratio (currently 50:50) of the two cases to account for surface roughness due to dirt and surface imperfections from production. The mixing of polars in this case is preferred over the use of the parameter Ncrit in Xfoil as there is no information on how to relate turbulence magnitude to Ncrit values for long scale atmospheric turbulence. Within the Xfoil code an additional section was appended which takes care of the extrapolation of the 10 to 20 AoA polar to the full 360 polar needed by the BEM code. The algorithm is an industry standard approach and was adopted from the visual basic code Aerofoil Prep published by NREL [30]. In principle it has 2 components, which are the 360 extrapolation developed by Viterna and Janetzke [31] and Tangler and Kocurek

[32] and the 3D stall correction by Du and Selig [33] and Eggers et al. [34] for lift and drag coefcients. Several publications describe the successful application of this methodology [5,6,23]. For the 360 extrapolation, the linear part of the lift curve needs to be considered that will serve as input for a trigonometric function tting for lift, drag and moment coefcient. The actual extrapolation uses the information of the thin plate theory [31] with the associated limitations for thicker aerofoils. The 3D stall correction depends on the radial position, the chord length, the rotational speed and the wind velocity of each prole on the rotor which is estimated from the turbine operation at maximum tip speed. It can be shown that the inboard proles exhibit a stronger three dimensional behaviour than the proles near the blade tip. This is due to strong coriolis and centrifugal forces as well as a big variation of dynamic pressure in this area [35], that effectively delay the stall and provide for higher lift and drag coefcients. An example of the resulting 360 lift coefcient (Cl) plots for ve selected radial stations from 40 to 100 percent relative blade length (r/R) is shown in Fig. 4. In the next module the actual evaluation of the turbine performance is conducted using the static BEM theory [36] whereas the method of calculating the axial and rotational induction factors has been adopted from Hansen [37]. The sequence of calculation and more details of the setup can be found in Fischer et al. [17]. The results of the aerodynamic module are the value of the annual energy production along with the rotor thrust at rated power. Moreover, section moments at discrete radii in apwise and edgewise directions are derived and prepared for use in the last module of CoBOLDT that calculates the mass and the stiffness of the current blade design. The mentioned mass and the stiffness estimation module are completely automated. In addition to the section moments as external loads it needs the blade contour at discrete radial stations

Table 2 CoBOLDT Parameters. Parameter x-chord y-chord a1-chord a2-chord x-thick y-thick a1-thick a2-thick 3-X 3-Y 4-X 4-Y 5-X 5-Y 6-X 6-Y Initial value 0.5 0 0 0 0.5 0 0 0 0 0 0 0 0 0 0 0 Initial step size 0.005 0.002 0.002 0.005 0.005 0.002 0.002 0.005 0.002 0.002 0.002 0.002 0.002 0.002 0.002 0.002 Min range 0.1 0.2 0.2 0.5 0.1 0.2 0.2 0.5 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 Max range 0.9 0.2 0.2 0.5 0.9 0.2 0.2 0.5 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1

Fig. 5. Visualisation of a blade section in the structural model.

510

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

the internal structural sizing, which will add (strain ratio > 1) or remove layers (strain ratio < 1) in the sizing zone until the ratio is equal to one and therefore a minimum weight is obtained. The initial layout data are then updated with new thicknesses to reach a faster convergence next time the structural sizing module is started. The blade mass as the nal target function is then provided back to the global optimisation routine. In the current version of CoBOLDT no parameters from the structural layout (e.g. the position or width of the girders) are dened as parameters for the global optimisation, however the option is considered for a future release of CoBOLDT. 4. Design case NR41 with CoBOLDT The design case chosen to demonstrate the optimisation capabilities of CoBOLDT is based on a generic blade (NR41) for an 82 m diameter wind turbine. The blade manually designed for the specications of a 1.5 MW pitch controlled wind turbine and has been described in a paper by Fischer et al. [13]. The initial aerofoil series as been designed by Claas-Hinrik Rohardt at the Deutsches Zentrums fr Luft- und Raumfahrt in Braunschweig and is parametrised with 8 variables, namely 3-X, 3-Y, 4-X, 4-Y, 5-X, 5-Y, 6-X and 6-Y (see Fig. 3). The parameters are similar for each blade section to keep aerofoils compatibility. The functions for chord and the absolute thickness will be altered using four parameters for each. Table 2 explains the parameterisation setup of the master code according to Figs. 2 and 3 in more detail. The steady reduction of chord as well as of the absolute and relative thickness is required by the utilized BEM code to distribute the spanning prole correctly. The geometrical requirement of the maximum chord being below 4 m stems from the road transportation for onshore blades underneath bridges, whereas the last two constraints are needed to guarantee a certain noise emission in normal operation. During optimisation the blade design is checked to satisfy all the hard constraints. The target functions are AEP at an IEC Class 3 wind site with a probability of wind following a Rayleigh

Fig. 6. Optimisation history.

as well as an initial internal structure denition and initial laminate layout of the load carrying structure. The initial laminate layout contains the information which zones should be automatically sized (e.g. main girder and aft girder see Fig. 5) and which should stay untouched. To get the internal loads the blade is modelled as a Bernoulli beam with the actual laminate layout and blade geometry serving as input to calculate the beam stiffnesses in ap and edgewise directions corresponding to Librescu and Song [38]. The strains are evaluated on the blade skin at discrete radii and circumference positions out of the external loads and stiffnesses in apwise and edgewise directions [38]. In the next step, for each discrete position of the skin the calculated strains are divided by a previously dened allowed design strain. The strain ratio is used for

Fig. 7. Objective function space.

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

511

Fig. 8. Full dataset in parallel coordinates [42].

distribution (mean wind 7.5 m/s), the blade mass and the thrust of the rotor. The maximum root bending is not considered since it heavily correlates with the blade mass. A summary of the optimisation objective functions is shown in Table 1. The run is performed on a 32-core Linux cluster with one slave initialised on every 2.4 GHz core. After 78,773 evaluated blade designs and 2717 optimisation iterations the run was manually stopped. The optimisation took 84 h of wall clock time. 5. Post optimisation analysis of design case NR41 The optimisation strategies (see Ref. [13]) over the number of optimisation iterations are presented in Fig. 6. It shows that 23

intensication moves (green cross) were carried out which lead to the exploration of ve regions with fast optimisation progress and hence low number of unsuccessful iterations (red cross). The other two available search strategies, namely diversication and step size reduction, were not performed at all. This is mainly due to two reasons that the optimisation process was not nished by the time it was manually stopped. The optimiser was still able to nd enough Pareto solution and the unsuccessful iteration counter (red cross) never reached the threshold necessary to perform a diversication or step size reduction move. The obtained Pareto surface according to the objectives in Table 1 is shown in Fig. 7 and proves the efciency of MOTS to explore complex and fragmented objective function spaces. Especially the relation of AEP and thrust is worth

Fig. 9. Extreme point designs in parallel coordinates [42].

512

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

mentioning in that context, as the projection reveals a discontinuity in the concave area of the projected Pareto front, which could hardly be explored by gradient-based optimisers. The same discontinuity can also be found in the AEP and mass related Pareto set projection, even though the front exhibits a rather convex behaviour. The extreme point design for each objective function has been investigated further to link the geometrical properties of the design dened by the parameters to the physics that lead to improved performance. A powerful tool in the post optimisation analysis is the application of parallel coordinates in the design space as published by Kipouros et al. [39] and Inselberg [40,41]. The complete dataset with 516 Pareto points is used as the basis for the analysis. It was segmented into three clusters using the Euclidean distance as a metric in the parameter space as visualised in Fig. 8. The method builds the hierarchy starting from the individual blades by merging clusters that are geometrically similar, until the required number of clusters is obtained. The AEP discontinuity mentioned in the previous section is clearly visible and is formed mainly by the red and the blue cluster. To nd the blade designs with the biggest improvement for each of the three target functions, namely the extreme point designs (EPDs), we use the slider to highlight a small cluster (up to approx. 5) of very similar extreme point design in Fig. 9. The AEP EPDs are marked in dark green as well the mass and thrust EPDs in orange and blue respectively. Table 3 underneath shows the designs with the lowest value for each objective function. The comparison in Fig. 9 reveals that: 1. Mass and thrust EPD exhibit a high correlation of parameters, except for x-chord, y-chord, y-thick and 6-y 2. The biggest difference between mass and thrust EPD occurs for y-thick, which translates to a bigger absolute thickness of the bending beam for the design with lower mass 3. Parameters y-thick of all EPDs only differ by one step size. This points to a poor choice of the initial step size for this variable. 4. Huge difference for most parameters between AEP EPD and the mass and thrust EPDs 5. Bigger chord (y-chord) is associated with a high axial induction factor and therefore a high AEP In terms of the design driving variables for the prole parameterisation, 3-y, 4-x, 5-y and 6-x can be identied as most inuential. Since all parametrised aerofoils are changed with the same parameters, a closer look to one of the spanning aerofoils reveals the design trend for the whole aerofoil family. Fig. 10 shows that from the initial aerofoil designs all EPDs show a considerable reduction in aerofoil camber with the associated lower downstream ow angle. Again, AEP as opposed to mass and thrust can be identied with some unique properties. Firstly, the AEP EPD exhibits the lowest camber of all aerofoils. Secondly, the pressure

TFThrust TFMass TFAEP 6-Y 6-X 5-Y 5-X 4-Y 4-X 3-Y a2-thick a1-thick x-thick a2-chord a1-chord y-chord x-chord Table 3 Extreme point and compromise design parameters. y-thick 3-X

AEP Mass Thrust Comp

0.2650 0.3750 0.3350 0.2550

0.002 0.008 0.006 0.002

0.0880 0.038 0.038 0.094

0.0400 0.1100 0.1150 0.0450

0.6000 0.6000 0.6050 0.6050

0.002 0.0000 0.002 0.002

0.0080 0.008 0.010 0.0100

0.110 0.095 0.085 0.100

0.006 0.006 0.006 0.004

0.048 0.026 0.028 0.048

0.004 0.0100 0.0100 0.004

0.056 0.046 0.048 0.056

0.0060 0.0060 0.0020 0.0000

0.008 0.014 0.014 0.010

0.054 0.032 0.030 0.054

0.052 0.040 0.048 0.050

0.9919 0.9978 0.9978 0.9927

1.0028 0.9909 0.9963 0.9993

0.9777 0.9632 0.9573 0.9641

Fig. 10. Aerofoil design trend.

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

513

Fig. 11. Lift and drag coefcients (r/R 70%).

recovery region at the suction side shows a comparably at shape, which proved in a subsequent CFD calculation (not included in the paper) to reduce the tendency for stall at high angles of attack (AoA). To calculate the polars the blade is cut at specied radial sections and therefore the relative thickness can change. In Fig. 11 the aerodynamic coefcients that correspond to proles at (r/R 70%) are considered, which translate to a relative thicknesses of 18.6% (Initial), 20.9% (AEP), 22.3% (Mass) and 20.8% (Thrust). It is obvious that the thin 18.6% prole exhibits very low drag at low AoA. Besides this advantage the other designs e although they are much thicker e can show a 10% better L/D-ratio in the design range of 6e 8 AoA. Fig. 12, which shows the L/D distribution at partial loading with a wind velocity of 7.5 m/s is underlining the superiority of the new prole designs. Fig. 13 depicts the AoA distribution at ve distinct operational wind speeds for the compromise design. For 7e 8 m/s the turbine is still at partial loading with constant tip speed ratio and variable tip speed. The local AoAs are below the value at which maximum L/D is obtained (thick red curve) and well below the local stall angles (thick black curve). At 9 m/s the maximum tip speed of 74 m/s is reached, the turbine increases the counter moment at the generator to increase power output further and the local angle of attack rises. At 10 m/s the turbine is just before rated power and here the local AoA coincides with the curve for the highest L/D. At 11 m/s the rated power of 1500 kW is fed into the grid and the turbine is in pitching mode to regulate the power output the this value.

The comparison in terms of the objective functions can be found in Table 3. The increase in L/D for the prole family translates to a higher AEP value of the AEP EPD (0.8%) and the mass and thrust EPDs (0.2%). Additionally, the thrust of the thrust EPD and mass EPD is considerably reduced by 4.3% and 3.7%. The difference in blade mass between initial and mass EPD is found to be 0.9%. If the normalised variance (absolute variance over the mean) of the design variables in the Pareto set is considered (see Fig. 14) the relative impact of parameters on the design problem can be determined. In this case the normalised variation is used to allow for comparison of parameters with different ranges. From the bar chart it is obvious that the variance of x-chord, a1-chord, a1-thick and 3-x accounts for a major proportion of the total variability. Nevertheless, y-thick which was found to have a big inuence on mass, has a very low normalised variance. The reason for this problem is the big initial step size. Only 3 distinct values are present in the Pareto range for the parameter (see Fig. 8) and the optimiser had no chance to reduce its step size due to the cancellation of the optimisation run prior to the rst step size reduction. At this point another approach to step size reduction with MOTS should be suggested. 1. After a predened number of optimisation iterations, when a good approximation of the Pareto set is available, determine maximum and minimum values (Pareto range) of each parameter in the Pareto set.

Fig. 12. L/D distribution at 7.5 m/s.

514

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

Fig. 13. AoA distribution for 7e11 m/s for the compromise design.

2. Reduce or increase the step size to t a predened number of designs in the Pareto range. 3. After the application of this method the normalised variation of each parameter really reects its relative importance (as the variance is not skewed by different step sizes) and parameters with low importance might be used to reduce the dimensionality of the optimisation problem. The described analysis of the design driving parameters and their inuence might be a good way to pro-actively design blades to comply with the rules found. However in a commercial context the question of choosing a good compromise design is of high importance. For this purpose it is sensible to pick the design with the lowest cost of energy e as a composite target e after the multiobjective optimisation. The total cost analysis considers the changes of component cost due to the different load level as well as the income from the energy produced. Since the actual cost model is condential it cannot be disclosed with this paper. However from Table 3 it can be noted that the compromise design inherits most of its features from AEP EPD.

6. Conclusions This paper describes the effect of changes in blade and aerofoil geometry on turbine performance for a horizontal axis wind turbine blade using the Computational Blade Optimisation and Load Deation Tool. The design case that is presented is based on an industrial scale wind turbine with a rotor diameter of 82 m and a rated power of 1.5 MW. The three target functions that are considered are maximisation of the annual energy production, minimisation of the blade mass as well as minimization of the rotor thrust. They have been chosen since there are turbine properties in the control of the turbine designer with a high impact on cost of energy. In terms of the parameterisation 16 variables have been used to alter the blade geometry relative to the datum design. Eight parameters are dened to change the blade chord and the blade absolute thickness (4 parameters each) whereas 8 parameters serve as input for the aerofoil parameterisation. Post optimisation analysis reveals that for the compromise design all objectives could be improved. The geometrical design drivers that lead to the improvements could be studied with a selection of three extreme point designs chosen from the associated clusters in a parallel coordinate analysis. It was found that: 1. AEP extreme point design is very different from the extreme point designs of mass and thrust 2. Reduction in camber of the aerofoils lead to an enhanced L/D-ratio in the design range 3. Reduction in blade chord resulted in a lower blade thrust 4. Increased relative prole thickness allowed for the reduction of blade mass 5. Close before rated power the compromise design showed an AoA distribution that allows to operate the aerofoils at the local maximum L/D-ratio Finally it has been shown that the normalised variation of parameters in the Pareto set (to nd the most energetic parameters) is skewed due to the initial step size value of parameters prior to the rst step size reduction. To mitigate this effect a novel approach has been suggested which scales the parameters to a set Pareto range after a certain number of optimisation steps. To further improve the CoBOLDT optimisation package the implementation of a multi-dimensional surrogate model and the

Fig. 14. Normalised variance of the Pareto Set.

G.R. Fischer et al. / Renewable Energy 62 (2014) 506e515

515

development of a restart model based on relative parameterisation is envisaged. Also the stationary BEM solver is due to be replaced with a transient solver to account for aero-elastic effects during turbine operation. References
[1] Fingersh L, Hand M, Laxson A. Wind turbine design cost and scaling model. Technical Report NREL/TP-500e40566. Golden, CO: National Renewable Energy Laboratory; 2006. [2] Rochholz H. Technischer Bericht e Aerodynamische Auslegung des Rotorblattes NR41. Unpublished Nordex Communication; 2009. [3] Hjort S, Dixon K, Gineste M, Olsen SA. Fast prototype blade design. Wind Engineering 2009;33:321e34. [4] Maki K, Sbragio R, Vlahopoulos N. System design of a wind turbine using a multi-level optimization approach. Renewable Energy 2012;43:101e10. [5] Vesel RW. Optimization of a wind turbine rotor with variable airfoil shape via a genetic algorithm. Masters thesis. Aeronautical and Astronautical Engineering. Ohio State University; 2009. [6] Xuan H, Weimin Z, Xiao L, Jieping L. Aerodynamic and aeroacoustic optimization of wind turbine blade by a genetic algorithm. In: 46th AIAA aerospace sciences meeting and exhibit, 7e10 January 2008, Reno, Nevada, AIAA 20081331 2008. p. 1e12. [7] Jureczko M, Pawlak M, Mezyk A. Optimisation of wind turbine blades. Journal of Materials Processing Technology 2005;167:463e71. [8] Belessis MA, Stamos DG, Voutsinas SG. Investigation of thecapabilities of a genetic optimization algorithm in designing wind tur-bine rotors, European Union Wind Energy Conference and Exhibition,Goteborg, Sweden, May 20-24 (1996). [9] Benini E, Toffolo A. Optimal design of horizontal-axis wind turbines using blade-element theory and evolutionary computation. ASME Journal of Solar Energy Engineering 2002;124:357e63. [10] Fuglsang PL, Madsen HA. A design study of a 1 mw stall-regulated rotor. Ris-R799(EN). Roskilde, Denmark: Ris National Laboratory; 1995. [11] Young W-B, Wu W-H. Optimization of the skin thickness distribution in the composite wind turbine blade. In: International conference on uid power and mechatronics (FPM) 2011. p. 62e6. [12] Gigure P, Selig MS, Tangler JL. Blade design trade-offs using low-lift airfoils for stall-regulated hawts. In: ASME/AIAA wind energy symposium, Reno, Nevada, January 11e14, 1999 1999. [13] Jaeggi DM, Parks GT, Kipouros T, Clarkson PJ. The development of a multiobjective tabu search algorithm for continuous optimisation problems. European Journal of Operational Research 2008;185:1192e212. [14] Kipouros T, Molinari M, Dawes WN, Parks GT, Savill AM, Jenkins KW. An investigation of the potential for enhancing the computational turbomachinery design cycle using surrogate models and high performance parallelisation. In: Proceedings of ASME Turbo Expo 2007 2007. [15] Kipouros T, Ghisu T, Parks GT, Savill AM. Using post-analyses of optimisation processes as an active computational design tool. In: ICCES7; 2008. p. 151e7. [16] Kipouros T, Jaeggi DM, Dawes WN, Parks GT, Savill AM, Clarkson PJ. Insight into high-quality aerodynamic design spaces through multi-objective optimization. Computer Modeling in Engineering and Sciences 2008;37:1e23. [17] Fischer GR, Kipouros T, Savill AM. Multi-objective shape optimisation for horizontal-axis wind turbine blades. In: 53rd AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics and materials conference, 23e26 April 2012, Honolulu, Hawaii, AIAA 2012-1353 2012. p. 1e10. [18] Ghisu T, Parks GT, Jarrett JP, Clarkson PJ. Integrated design optimisation of gas turbine compression systems. In: 49th AIAA/ASME/ASCE/AHS/ASC structures,

[19] [20] [21] [22]

[23]

[24] [25] [26] [27]

[28]

[29] [30] [31]

[32]

[33]

[34]

[35] [36] [37] [38] [39]

[40] [41] [42]

structural dynamics, and materials conference, 7e10 April 2008, Schaumburg, IL, AIAA 2008-1979 2008. p. 1e20. Gaunaa M, Rthor P, Srensen NN, Dssing M. A computationally efcient algorithm for the aerodynamic response of non-straight blades; 2011. Abedi H, Davidson L, Voutsinas S. Vortex method application for aerodynamic loads on rotor blades. In: EWEA conference 2013. Chen J, Wang Q, Shen WZ, Pang X, Li S, Guo X. Structural optimization study of composite wind turbine blade. Materials & Design 2013;46:247e55. Drela M. Xfoil: an analysis and design system for low reynolds number. In: Conference proceedings on low Reynolds number aerodynamics, 5e7 June 1989, Notre Dame, Indiana 1989. Kenway G, Martins JR. Aerostructural shape optimization of wind turbine blades considering site-specic winds. In: 12th AIAA/ISSMO multidisciplinary analysis and optimization conference, 10e12 September 2008, Victoria, British Columbia Canada, AIAA 2008-6025 2008. p. 1e12. Sederberg TW, Parry S. Free-form deformation of solid geometric models, SIGGRAPH. Association of Computing Machinery 1986;20:151e60. Samareh JA. Aerodynamic shape optimisation based on free-form deformation. AIAA 2004-4630; 2004. p. 1e13. Samareh JA. Novel multidisciplinary shape parameterization approach. Journal of Aircraft 2001;38:1015e24. Samareh JA. Survey of shape parameterisation techniques for high-delity multidisciplinary shape optimisation. The American Institute of Aeronautics and Astronautics Journal 2001;39:877e84. Doosttalab M, Frommann O. Effects of the design parameters on the multidisciplinary optimization of atback airfoils for large wind turbine blades. In: 13th AIAA/ISSMO multidisciplinary analysis optimization conference, 13e15 September 2010, Fort worth, Texas 2010. Grasso F. Usage of numerical optimisation in wind turbine airfoil design. Journal of Aircraft 2011;48:248e55. Hansen C. Airfoilprep e nwtc computer-aided engineering tools; 2012. Viterna LA, Janetzke DC. Theoretical and experimental power from large horizontal-axis wind turbines. TM-82944, Technical Report. NASA Lewis Research Center; 1982. Tangler J, Kocurek JD. Wind turbine post-stall airfoil performance characteristics guidelines for blade-element momentum methods. In: 43rd AIAAAerospace sciences meeting and exhibit, 10-13 January 2005, Reno, Nevada, AAIA 2005-0591 2005. p. 1e10. Du Z, Selig MS. A 3-d stall-delay model for horizontal axis wind turbine performance prediction. In: ASME wind energy symposium, 36th AIAA aerospace sciences meeting and exhibit, AIAA 1998-0021 1998. Eggers AJ, Chaney K, Digumarthi R. An assessment of approximate modeling of aerodynamic loads on the uae rotor. In: 41st aerospace sciences meeting and exhibit, 6e9 January 2003, Reno, Nevada, AIAA 20030868 2003. p. 1e10. Schreck SJ, Srensen NN, Robinson MC. Aerodynamic structures and processes in rotationally augmented ow elds. Wind Energy 2007;10:159e78. Glauert H. Airplane propellers. In: Durand WF, editor. Aerodynamic theory. Springer; 1963. Hansen MO. Aerodynamics of wind turbines. Earthscan; 2008. Librescu L, Song O. Thin-walled composite beams e theory and application. Springer; 2006. Kipouros T, Jaeggi DM, Dawes WN, Parks GT, Savill AM, Clarkson PJ. Biobjective design optimization for axial compressors using tabu search. AIAA Journal 2008;46:701e11. Inselberg A. Parallel coordinates: visual multidimensional geometry and its applications. Springer; 2008. Inselberg A. The plane with parallel coordinates. Visual Computer 1985;1: 69e91. Hinterberger H. Visulab 4.2 (software); 2005.