Validation of uncertainty in IEC damage calculations

based on measurements from alpha ventus

DeepWind 2016

January 21st, 2016

Kolja Müller, Po Wen Cheng

Stuttgart Wind Energy (SWE), University of Stuttgart, Germany

Content of presentation

„Can assumptions of environmental conditions in the design process

adequately represent real loads?“

→ IEC 61400-03 DLC 1.2, load variation

• Research at alpha ventus, turbine, measurements & simulation model

• Applied procedure

• Measurement selection

• IEC assumptions

• Statistical evaluation

• Conclusions

2

Source: DOTI (www.alpha-ventus.de, 21.12.2015)

Motivation gut, eher

noch ausbauen, was

das Ziel der

Zu ausführlich

Die Motivation auf Folie

3 ist gut, könnte aber

noch ausgebaut werden.

Der eine zentrale Satz

oben ist für deine

Argumentation sehr

wichtig. Vielleicht kannst

du für den Zuhörer

erklären wie das

Standardvorgehen nach

IEC ist. Darauf baut sich

dann dein roter Faden

auf, dass du diese

Annahmen mit

Messdaten/Simulation

überprüfen möchtest.

Offshore test field alpha ventus (North Sea)

3

> 100 sensors since 2011

• SCADA

• Loads

• Accelerations

• Environmental conditions

• Corrosion

Statistical & high resolution

(50 Hz) data available online

5 MW

FATB

MresTT

MM

,

,- tower top resulting bending moment

- tower base fore-aft bending moment

B. Load-

reducing

control and

load

monitoring

C. Design

conditions for

future wind

turbines

OWEA LOADS project

A. Load

analysis and

probabilistic

load

description

Applied Simulation Method

4

Wind turbine model:

Tool: Flex5 (28 DOF)

Dynamics: nonlinear elastic multi-body system (MBS)

with modal shape functions

Aero: BEM theory with correction models

Control: pitch and torque

4

[Sourc

e: D

OT

I]

Coupled integrated approach, dynamically linked at tower base

D. K

aufe

r et al. “

Valid

ati

on

of

an

In

teg

rate

d S

imu

lati

on

Meth

od

wit

h H

igh

Reso

luti

on

Lo

ad

Measu

rem

en

ts o

f th

e O

ffsh

ore

Win

d T

urb

ine R

Ep

ow

er

5M

at

Alp

ha

Ven

tus ."

Jo

urn

al o

f O

cean

an

d W

ind

En

erg

y,

Vol. 1

, N

o.

1

Substructure and foundation model

Tool: Poseidon (n DOF)

Dynamics: FE model

Elements: Bernoulli beams and force elements

Hydro: irreg. sea states, Morison equation

Validated for equivalent environmental conditions

Variation of measured loads can be represented with simulations

Applied procedure for validation of fatigue load

variation implied in IEC design assumptions

5

Selected measurements

Environmental parameters (IEC DLC 1.2, Hindcast data)

Selection criteria

DEL & damage over lifetime

simulation

Monte Carlo / Bootstrap

Damage statistics

DEL & damage over lifetime

Damage statistics

Comparison

Monte Carlo / Bootstrap

Simulation results (high resolution)

All measurements

Equal

format of

datasets

Environmental

data (mean

values)

Load measure-

ments (high

resolution)

presented

Comparisons

Sim.

Meas.

Selection of measurements

Turbine status

- Only power production

- Only free flow conditions

- No curtailment periods

Quality of measurements

- High resolution data available for

considered sensors

- No fault conditions of sensors

- No outliers (Palmgren Miner Rule applied)

April 2011 - January 2012

(10 months of measurements)

N

210°-270°

6

Simulation input: IEC DLC 1.2 environmental

conditions

Applied IEC simplifications (DLC 1.2):

• Environmental conditions with dependence

on wind speed and wind direction

•

• = 90th percentile

• = 50th percentile

• Constant values

•

• Azimuth error

• Water depth

• Marine growth

• Wind-wave-misalignment

• Soil conditions

),(,, vfTHTI Ps

TI

Ps TH ,

14.0

P

s

T

H

TI -Turbulence intensity

- wave height

- wave period

v - wind speed

- wind direction

- wind shear 7

Assumptions of IEC DLC 1.2 environmental

conditions

Assumptions of IEC DLC 1.2 environmental

conditions

8

Applied IEC simplifications (DLC 1.2):

• Environmental conditions with dependence

on wind speed

•

• = 90th percentile

• = 50th percentile

• Constant values

•

• Azimuth error

• water depth

• Marine growth

• wind-wave-misalignment

• soil conditions

),(,, vfTHTI Ps

TI

Ps TH ,

14.0

100 simulations

Applied procedure for validation of fatigue load

variance implied in IEC design assumptions

9

Selected measurements

Environmental parameters (IEC DLC 1.2, Hindcast data)

Selection criteria

DEL & damage over lifetime

simulation

Monte Carlo / Bootstrap

Damage statistics

DEL & damage over lifetime

Damage statistics

Comparison

Monte Carlo / Bootstrap

Simulation results (high resolution)

All measurements

Equal

format of

datasets

Environmental

data (mean

values)

Load measure-

ments (high

resolution)

presented

Comparisons

Sim.

Meas.

Calculation of lifetime damage

10

DEL of time

series

Damage of time

series

Occurence

probability of

wind bin

Damage over

lifetime

Load time series simulation

m

(m = 4)

(Nref= 2e6)

Wir raten dir nur Formeln

zu zeigen, die du auch

erklärst. Sonst ist der

Zuhörer schnell

abgelenkt.

Formeln in Formelfolie

überführen

… …

repeat i=1..r..10.000

Damage statistics:

Monte Carlo & Bootstrap evaluation

11

Da

vis

on

, An

tho

ny C

hri

sto

ph

er,

an

d D

avid

Vic

tor

Hin

kle

y. B

oo

tstr

ap

me

tho

ds a

nd

oth

eir a

pp

lica

tio

n. V

ol. 1

. C

am

bri

dg

e u

niv

ers

ity p

ress, 1

99

7.

draw n=1..100

samples

return samples to bin

Ddataset

nj

j

iji DD1

… …

repeat i=1..r..10.000

Damage statistics:

Monte Carlo & Bootstrap evaluation

12

Da

vis

on

, An

tho

ny C

hri

sto

ph

er,

an

d D

avid

Vic

tor

Hin

kle

y. B

oo

tstr

ap

me

tho

ds a

nd

oth

eir a

pp

lica

tio

n. V

ol. 1

. C

am

bri

dg

e u

niv

ers

ity p

ress, 1

99

7.

draw n=1..100

samples

return samples to bin

Ddataset

nj

j

iji DD1

Results (1): variation of damage

13

measure

ments

sim

ula

tions

Boxplo

ts s

how

ing m

edia

n, 1,5

,95 a

nd 9

9 p

erc

entile

s

]/[ smvhub]/[ smvhub

]/[ smvhub ]/[ smvhub

Blue

Error

Legende ist nicht

klar

#16

die Skalierung der

Achsen in den

Plots rechts bei

epsilon

machst, kann man

die Unterschiede

besser sehen.

FATBM , MresTTM ,

Results (1): variation of damage

14

measure

ments

sim

ula

tions

Boxplo

ts s

how

ing m

edia

n, 1,5

,95 a

nd 9

9 p

erc

entile

s

]/[ smvhub]/[ smvhub

]/[ smvhub ]/[ smvhub

Blue

Error

Legende ist nicht

klar

#16

die Skalierung der

Achsen in den

Plots rechts bei

epsilon

machst, kann man

die Unterschiede

besser sehen.

FATBM , MresTTM ,

Results (2): validation of variation of damage

15

]/[ smvhub

Normalized damage variation from bin median damage

[%],%medD

Legende: „

„

das erkennt man dann

besser

Wir raten dir nur

Formeln zu zeigen, die

du auch erklärst. Sonst

ist der Zuhörer schnell

abgelenkt.

]/[ smvhub

no clear consistency of difference observable

largest differences observable for regions around peak damage

FATBM , MresTTM ,

ref

D

med

D

D

med ,%- median damage

- % deviation of damage from median damage

- reference damage [%]

)()(,%

ref

hubmedhubD

D

vDvDmed

Conclusion & outlook

Methodology for validation of variation of damage by design assumptions

- Measurement selection

- Monte Carlo and Bootstrap methods

- Comparison of percentiles

→ Significant variation of loads from simulations observable

→ Difference between measurements and simulation varies

→ Calculation of probability of exceedence possible and could be relevant

Variation of damage cannot be captured by IEC design assumptions

-> Goals of the IEC fatigue evaluation regarding load variation?

a) Strictly conservative

b) Match variation of loads experienced in real environment

16

Acknowledgement

Thank you for your attention

This research is part of the RAVE projects OWEA -

“Verification of offshore wind turbines” and OWEA

Loads.

It is funded by the Federal Ministry for the Economic

Affairs and Energy (BMWi).

17

[wikipedia.org/wiki/Windmessmast]

Full scale validation of numerical models

18

Load

measurements

Environmental

measurements

plausibility checks

& calibration

Simulation Comparison 1. Time domain (rotor-/nacelle

rotation)

2. Frequency domain (natural

frequencies)

3. Statistics (min, mean, max,

stddev)

Validated simulation model

(equivalent environmental

conditions)

Selected events

1. Obtain

measurements

2. Obtain results 3. Compare results Aerodynamic

loads

Tower

drag

Hydro-

Dyn.

loads

soil-pile

interact

D. K

aufe

r et al.,

ISO

PE

2013

Simulation

Measurements Variation of measured loads can be represented with

simulations when considering variance of environmental

conditions

Validation of load variation

19

Can variance of loads be represented by simulations?

Simulation study considering variation of • Wind speed

• Turbulence intensity

• Wind shear

• Wave height

• Wave period

based on 5 year Fino1 data

Variation of measured loads can be represented with simulations when

considering variance of environmental conditions

Tower base fore aft bending moment

DEL – damage equivalent load

Damage bootstrap evaluation (tower bottom)

20

[%]%,sim

[%]%,mes

s

mvhub 13

s

mvhub 13

][n

][n

)[%]10,6(%

)[%]10,6(%

]/[ smvhub

]/[ smvhub

Boxplo

ts s

how

ing m

edia

n, 1,5

,95 a

nd 9

9 p

erc

entile

s

Bootstrap Rate of change

measure

ments

sim

ula

tio

ns

Du betrachtest

„Variation“ in den Plots,

das ist sehr schwer zu

verstehen. („Validation

of

Was ist

das nicht alles

durchschaut. Vielleicht

musst du da mehr Zeit

investieren, bzw. sehr

klar vorstellen.

Formeln unten in

Formelfolie überführen

Bootstrap evaluation (tower top)

21 21

Bootstrap Rate of change

measure

ments

sim

ula

tio

ns