Multidisciplinary Design Optimization with HAWTOpt2Multidisciplinary Design Optimization with...

Multidisciplinary Design Optimization with HAWTOpt2

Michael McWilliam, Frederik Zahle

Danish Technical University

Outline

• The HAWTOpt2 Framework

• Conventional Blade Optimization

• Bend-Twist Coupled Blades

• Smart Blades

• Closing statements

2 DTU Wind Energy HAWTOpt2 November 2, 2017

The HAWTOpt2 Framework



HawtOpt2: Tool for Aero-Servo-Elastic Optimization of WindTurbines

HawtOpt2 provides workflows that coupleaeroelastic and structural solverstogether to allow for:

• Simultaneous and fully coupledoptimization of lofted blade shape and thestructural design of a blade.

• Exploration of the many often conflictingobjectives and constraints in a rotor design.

• Detailed tailoring of aerodynamic andstructural properties.

• Constraints on specific fatigue damageloads.

• Placement of natural frequencies.



HawtOpt2: Tool for Aero-Servo-Elastic Optimization of WindTurbines

• Underlying framework: OpenMDAO(written in Python),

• Optimization algorithm: IPOPT interfacedthrough PyOptSparse,

• Geometric parameterisation: FUSED-Wind,

• HAWCStab2/HAWC2 interface:

• AEP,• Frequency placement,• Fatigue in frequency domain,• Reduced/full DLB in time domain,

• BECAS interface:

• Calculation of stiffness properties,• Calculation of material failure andfatigue damage.



Aero-elastic Solvers: HAWCStab2 and HAWC2

• Structural model: geometricallynon-linear Timoshenko finite beamelement model.

• Aerodynamic model: unsteady BEMincluding effects of shed vorticity anddynamic stall and dynamic inflow.

• HAWCStab2: Analytic linearizationaround an aero-structural steady stateignoring gravitational forces.

• HAWCStab2: Fatigue damagecalculated in frequency domain(including effect of ATEFs) based onthe linear model computed byHAWCStab2.

Image from: Sønderby and Hansen, Wind Energy, 2014



Structural Solver: BECAS (BEam Cross section AnalysisSoftware)

• Finite element based tool foranalysis of the stiffness and massproperties of beam cross sections.

• Correctly predicts effects stemmingfrom material anisotropy andinhomogeneity in sections ofarbitrary geometry (e.g., allcoupling terms).

• Detailed stress analysis based onexternally computed extreme loads.



Estimation of Extreme loads using HAWC2

• Fully resolved time simulations of all DLCs (+1000 simulations) is timeconsuming, and inclusion of turbulence is not suited for gradient based MDO,

• A reduced DLB has been developed, which requires <40 cases and omitsturbulence inflow.



Reduced Design Load Basis



The HAWTOpt2 Work-flow

• IPOPT Optimization Algorithm

• Uses OpenMDAO for the glue code

• Cross Section Module (BECAS)

• Steady and Unsteady Aeroelastic Models (HAWCStab2 & HAWC2)



Internal Structure Design Variables

• DP’s control web positions and panel sizes

• Thickness design variables for main load carrying laminates

• Can have fiber angle design variables (not used here)

TRIAX

UNIAX

TRIAX

UNIAXs=−1

s=0.

s=1

DP3

DP5

DP7

DP8DP9

DP10

DP13

DP0

DP4

DP6UNIAX

TRIAX

TRIAX



Optimization Formulation

• Maximize AEP

• Using planform and structural design variables

• Subject to geometric and load constraints

minimizexp,xs,xoper

f({xp, xs xoper,p, w)

subject to g(xp) ≤ 0,

hg(xs) ≤ 0,

hs(xs) ≤ 0,

k({xp, xs}) ≤ 0

Where:

f({xp, xs, xoper},p) =AEP ({0, 0, 0},p)

AEP ({xp, xs, xoper},p)



Design Variables

Parameter # of DVs CommentChord 6 -Twist 5 Root twist fixedRelative thickness 3 Root and tip relative thickness fixedBlade prebend 4 -Blade precone 1 -Blade length 1 -Tip-speed ratio 1 -Trailing edge uniax 2 Pressure/suction sideTrailing edge triax 2 Pressure/suction sideTrailing panel triax 2 Pressure/suction sideSpar cap uniax 4 Pressure/suction sideLeading panel triax 2 Pressure/suction sideLeading edge uniax 2 Pressure/suction sideLeading edge triax 2 Pressure/suction sideDP4 5 Pressure side spar cap position/rear web attachmentDP5 5 Pressure side spar cap position/front web attachmentDP8 5 Suction side spar cap position/front web attachmentDP9 5 Suction side spar cap position/rear web attachmentTotal 65



ConstraintsConstraint Value Commentmax(chord) < 6.2 m Maximum chord limited for transport.max(prebend) < 6.2 m Maximum prebend limited for transport.max(rotor cone angle) > −5 deg -min(relative thickness) > 0.24 Same airfoil series as used on the DTU

10MW RWT.min(material thickness) > 0.0 Ensure FFD splines do not produce neg-

ative thickness.t/wsparcap > 0.08 Basic constraint to avoid spar cap buck-

ling.min(tip tower distance) > ref value DLC1.3 operational tip deflection cannot

exceed that of the DTU 10MW RWT.Blade root flapwise mo-ments (MxBR)

< ref value DLB loads cannot exceed starting point.

Blade root edgewise mo-ments (MyBR)

< ref value DLB loads cannot exceed starting point.

Tower bottom fore-aft mo-ments (MxTB)

< ref value DLB loads cannot exceed that startingpoint.

Rotor torque < ref value Ensure that the rotational speed is highenough below rated to not exceed gener-ator maximum torque.

Blade mass < 1.01 * ref value Limit increase in blade mass to maintainequivalent production costs.

Blade mass moment < 1.01 * ref value Limit increase in blade mass moment tominimize edgewise fatigue.

Lift coefficient @ r/R =[0.5− 1.]

< 1.35 Limit operational lift coefficient to avoidstall for turbulent inflow conditions.

Ultimate strain criteria < 1.0 Aggregated material failure in each sec-tion for 12 load cases.

Flap Angle −15.0 ≥ β ≥ 15.0 Flap angle stays within 15 degree bounds


Conventional Blade Optimization

• Aero-elastic optimization

• Based on the DTU 10MW RWT

• Original DTU 10MW RWT is a CP optimized design

• No sweep or material coupling allowed in the optimization

• All other design variables allowed to vary



Optimized Blade Planform

• Blade length increased to 94.3 m (9.2% increase),

• Slender blade planform, with slightly thicker airfoils,

• Increased prebend to avoid tower clearance constraint.

0 20 40 60 80 100Blade running length [m]

0

1

2

3

4

5

6

7

Cho

rd le

ngth [m

]

BaselineNew design


−15

−10

−5

0

5

10

Blade

twist [de

g.]

BaselineNew design


20

30

40

50

60

70

80

90

100

Relative thickn

ess [%

]

BaselineNew design

0 20 40 60 80 100Blade radial coordiante [m]

7

6

5

4

3

2

1

0

1

Blade out of plane coordinate [-] Baseline

New design



Optimized Blade Internal Structure

0.0 0.2 0.4 0.6 0.8 1.00

20

40

60

80

100

Thickn

ess [m

m]

Trailing edge paneltriaxuniaxbalsauniaxtriax

0.0 0.2 0.4 0.6 0.8 1.00

10

20

30

40

50

60

70

80

Thickn

ess [m

m]

Trailing paneltriaxuniaxbalsauniaxtriax

0.0 0.2 0.4 0.6 0.8 1.00

10

20

30

40

50

60

70

80

Thickn

ess [m

m]

Spar cap

triaxuniaxuniaxtriax

0.0 0.2 0.4 0.6 0.8 1.00

5

10

15

20

25

30

35

40

45

Thickn

ess [m

m]

Leading paneltriaxuniaxbalsauniaxtriax

0.0 0.2 0.4 0.6 0.8 1.00

10

20

30

40

50Th

ickn

ess [m

m]

Leading edge paneltriaxuniaxbalsauniaxtriax

Figure: Material stacking sequence for each region along the optimized blade.



Optimized Lofted Blade

• Internal structure optimized with very few geometric constraints,

• Main laminates placed forward in the cross-section and angled,



Optimized Load Distributions

0 20 40 60 80 100Blade radial coordinate [m]

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Loca

l thr

ust c

oeffi

cien

t [-]

wsp=7.00wsp=8.00wsp=9.00wsp=10.00wsp=11.00


0

1000

2000

3000

4000

5000

6000

7000

8000

9000

Normal force [N/m]



−6

−5

−4

−3

−2

−1

0

1

Blade

torsion [deg

]



0

5

10

15

20

Angle of attack [deg]


Figure: Blade local thrust coefficients and normal forces (upper), and blade torsionand angle of attach (lower) for a range of wind speeds.



A closer look at the structural tailoring

• Bend twist coupling occurs ”hands off” due to the use of a fully coupled MDOapproach,

• Achieved through a forward movement of the shear center and reduction oftorsional stiffness,

• The resulting coupling is referred to as ”shear twist coupling”,

• The angling of the main laminates results in better alignment of the mainloading direction and elastic axis,

• This coupling does not suffer from loss in flapwise stiffness, making it moreattractive that e.g. material coupling.



Power Curve

• Steady state AEP increased by 11.1%,

• Turbulent power curve evaluated at TI=10%, six seeds, AEP gain of 8.7%,

• Aeroelastically tailored blade loses power in turbulent conditions.

0 5 10 15 20 25Wind speed [m/s]

0

2000

4000

6000

8000

10000

12000

Mecha

nical P

ower [kW]

BaselineNew design

4 6 8 10 12 14Wind speed [m/s]

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

Pmek_t/Pmek_s [-]

BaselineNew design

Figure: Power curve for the optimized blade compared to the baseline design forsteady state conditions (dotted lines) and turbulent power curve with 10%reference TI (left), and the ratio between the turbulent and steady state powercurves (right).


Bend-twist Coupled Blades

• Investigated both geometric and material coupled blades

• Based on a 100kW wind turbine with a 10m blade length

• 3 different cases investigated:

• KB1: Reference optimization where torsion is constrained• KB2: Optimization with sweep• KB3: Optimization where fiber angle could vary in the spar cap



Optimal Performance

Quantity KB1 KB2 Diff. KB3 Diff.

AEP [kWh] 405166 412457 1.018 408943 1.009(C=7, k=2)Improvement 15.9% 18.0% 17.0%Blade length [m] 11.064 11.35 1.026 11.231 1.015blade mass [kg] 256.87 257.14 0.999 256.85 0.999TSR [-] 10.084 10.625 1.054 9.779 0.970

• Baseline AEP: 349586 kWh



Optimal Performance

0 5 10 15 20 25Wind speed [m/s]

0

20

40

60

80

100

120

Mech. Power [kW]

ReferenceKB1KB2KB3



Optimal Performance

0 5 10 15 20 25Wind speed [m/s]

0.98

1.00

1.02

1.04

1.06

1.08

1.10M

ech.

Pow

er ra

tio [-

]KB2/KB1KB3/KB1

• Better improvement achieved with geometric coupling

• Optimization limited by tip deflection constraint

• Off-axis fibers reduce stiffness



Optimal Design

0 2 4 6 8 10 12Blade radius [m]

−1.0

−0.8

−0.6

−0.4

−0.2

0.0

0.2

Swee

p [m

]

Blade sweep [m]

KB1KB2



Optimal Design

0 0.2 0.4 0.6 0.8 1Blade Coordinate [-]

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

6

7

8A

ngle

[deg

.]

Reg. 3 Uniax Angle [deg.]

Reg. 9 Uniax Angle [deg.]


Smart Blades

• Investigated the AEP increase for quasi-static flap actuation

• Based on the DTU 10MW RWT

• Optimized both the blade and the flap settings

• 3 different cases investigated:

• DTU 10MW with just the flap• Baseline, blade optimization without the flap• Baseline with the flaps• Combined blade and flap optimization


Smart Blades

Flap Configuration

• Flap only deflected in below rated conditions

• Flap operates in a quasi-steady mode


Smart Blades

Optimal Solution

• Flaps provide a 1% improvement for aero-elastically optimal blades

• Flaps only gives 0.51% improvement for CP optimal blades

Original w/ Flaps Baseline Baseline w/ Flaps Co-Optimization0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

Per

cen

t Im

pro

vem

ent

in A

EP


Smart Blades

Flap Deflection

• Both start at negative angles, increase the decrease with increasing wind speed

• Co-design allowed more aggressive utilization of the flaps

6 7 8 9 10 11 12 13Wind Speed [m/s]

-5

0

5

10

15

Fla

p D

efle

ctio

n [

deg

]

DTU 10 MW RWT w/ Flap Add-on

Baseline Design w/ Flap Add-on

Co-Design


Smart Blades

Power Gain From Flaps

4 5 6 7 8 9 10 11 12 13Wind Speed [m/s]

-20

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

320

Gai

n i

n A

ero

dy

nam

ic P

ow

er [

kW

]

DTU 10 MW RWT w/ Flaps

Baseline w/ Flaps

Co-Design

Co-Design vs. Baseline


Smart Blades

Conclusions

• Wind turbine rotor design is increasingly complicated

• Multiple disciplines in competition• Innovations affect all aspects of the design

• Multidisciplinary design optimization can overcome this complexity

• It can provide better rotor designs• Reveal new innovations• Evaluate new load aleviation technology


Smart Blades

Thank-you for your interest

Comments or Questions?


Multidisciplinary Design Optimization with HAWTOpt2Multidisciplinary Design Optimization with...

Documents

Transcript of Multidisciplinary Design Optimization with HAWTOpt2Multidisciplinary Design Optimization with...