Sie sind auf Seite 1von 14

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/272150167

Load Calculation on Wind Turbines: Validation


of Flex5, alaska/Wind, MSC.Adams and
SIMPACK by means of Field Tests

Conference Paper August 2014


DOI: 10.1115/DETC2014-34670

CITATIONS READS

3 471

4 authors, including:

Roman Rachholz Andreas Mueller


University of Rostock Johannes Kepler University Linz
15 PUBLICATIONS 39 CITATIONS 184 PUBLICATIONS 877 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

MBS/BEM-Co-Simulation for Contact View project

DynAWind View project

All content following this page was uploaded by Jnos Zierath on 18 November 2015.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Proceedings of the ASME 2014 International Design Engineering Technical Conferences &
Computers and Information in Engineering Conference
IDETC/CIE 2014
August 17-August 20, 2014, Buffalo, New York, USA

DETC2014-34670

LOAD CALCULATION ON WIND TURBINES: VALIDATION OF FLEX5,


ALASKA/WIND, MSC.ADAMS AND SIMPACK BY MEANS OF FIELD TESTS


Janos Zierath Roman Rachholz Andreas Muller

W2e Wind to Energy GmbH Christoph Woernle University of Michigan-
18055 Rostock Chair of Technical Dynamics Shanghai Jiao Tong University Joint Institute
Germany University of Rostock Shanghai
Email: jzierath@wind-to-energy.de 18059 Rostock China
Germany Email: andreas.muller@ieee.org
Email: {rachholz, woernle}@uni-rostock.de

ABSTRACT INTRODUCTION
Load calculations on wind turbines are an essential part Load calculations on wind turbines are an essential part of its
of its development. In the preliminary design phase simplified development. In the preliminary design phase simplified multi-
multibody models are used for the estimation of the interface body models are used for the estimation of the interface loads.
loads. The interface loads are used within an iterative develop- The interface loads are used within an iterative development loop
ment loop to design the components of the wind turbine such as to design the components of the wind turbine such as gearbox,
gearbox, blades, tower and so on. Due to the early application of blades, tower and so on. Due to the early application of load
load calculations within the development process, the quality of calculations within the development process, the quality of the
the simulation results has a great influence on the wind turbine simulation results has a great influence on the wind turbine de-
design. sign.
In this contribution the simulation results of the multibody This contribution describes the development of a state-of-
codes alaska/Wind, MSC.Adams and SIMPACK are compared the-art multibody model of a 2.05 MW wind turbine designed by
with measurements obtained from a prototype of a 2.05 MW wind W2e Wind to Energy. The prototype of the wind turbine erected
turbine developed by W2e Wind to Energy. Furthermore, simula- in Tarnow, Mecklenburg-Western Pomerania, Germany, is shown
tion results of the special wind turbine design code Flex5, devel- in Fig. 1.
oped at the Technical University of Denmark Copenhagen, are
taken into account. A statistical and dynamical evaluation of the As it can be seen, the wind turbine is a typical horizontal
simulation and measurement results has been done. Due to the axis design with three blades. It has a double-fed asynchronous
use of the same controller procedures as used on the physical generator. The prototype has a tubular tower and a nominal hub
wind turbine, the wind turbine models show almost the same be- height of 100 m. The nominal rotor diameter of the turbine is
haviour (electrical power, pitch angle, rotor speed) as the wind 93 m. A special drive train concept was developed for the wind
turbine in the field. Differences occur during the evaluation of turbine, see Fig. 1b. Instead of the typical three-point mount-
the interface loads due to the different kinds of wind turbine mod- ing of the drive train which is industrial standard, the rotor is
elling. mounted by a moment bearing well known from tunnel construc-
tion machinery. The gearbox is fixed at the position of its centre
of gravity by a ring mounting consisting of elastic bushings. In
Address
addition the gearbox is also coupled to the rotor by elastic bush-
all correspondence to this author.

1 Copyright
c 2014 by ASME
multibody model with the aerodynamic code and the controller is
presented. Then, a general description of the different simulation
models follows.

Interaction Scheme of the Multibody Program with the


Aerodynamic Code and the Controller
The simplified interaction scheme of the multibody simula-
tion with the aerodynamic code and the controller of the wind
turbine is shown in Fig. 2.

a
position and velocity of generator speed,
blade elements pitch angle

Aerodynamic Code Multibody Code Controller

aerodynamic forces and generator torque,


torques on blade pitch velocity

b
FIGURE 2. SIMPLIFIED INTERACTION SCHEME OF THE
WIND TURBINE MODEL
FIGURE 1. PROTOTYPE OF THE 2.05 MW WIND TURBINE
ERECTED IN TARNOW, MECKLENBURG-WESTERN Pomerania,
GERMANY. a PROTOTYPE. b CAD-MODEL OF THE DRIVE
A discrete interface was developed for the interaction of the
TRAIN
controller with the multibody program. The aim of this interface
is to integrate the same controller software into the multibody
ings. Both elastic mountings behave like an universal joint. As simulation as implemented on the physical wind turbine. The in-
result of the moment bearing and the elastic bushings, no bend- teraction scheme represents a software-in-the-loop principle and
ing moments and shear forces act on the gearbox expanding its was developed in analogy to the hardware-in-the-loop principle
durability. described in [7]. The multibody program contacts the controller
For validiation, different programs are chosen to build up a at discrete time steps and waits until the controller provides the
wind turbine model. Beside the special purpose program Flex5 corresponding output data. Due to the fact that the integrator
the general purpose MSC.Adams 2012 x64 [1], alaska/wind 8.3 of the multibody program generally has a variable step size, the
x64 [2] and SIMPACK 9.3 x64 [3] were chosen to build up interface has to be realised in such a way that the controller is
the wind turbine model. The aerodynamic forces are applied contacted only at prescribed constant time steps. The controller
within MSC.Adams and SIMPACK using the AeroDyn source on the physical wind turbine operates with a cycle time of 10 ms,
code v13.01 developed by NREL [4, 5], whereas alaska/Wind which is also chosen as prescribed time step for the interface.
and Flex5 use their own aerodynamic modules. In contrast to [6], Between the discrete cycle time the output values of the previ-
the aim of this contribution is not to propose one program as a ous time steps are used and kept constant. This interface scheme
reference but to compare all programs with measurements on the does not present any real-time capabilities which is, however, not
prototype. necessary and not realisable for large simulation models.
The multibody model also interacts with the aerodynamic
code. The aerodynamic code used in Flex5 was developed at
THE MULTIBODY MODEL AND ITS INTERACTION WITH Danish Technical University and is integrated directly into the
THE AERODYNAMIC CODE AND THE CONTROLLER multibody code. Within alaska/Wind the aerodynamic code is
The aim of this section is to describe the multibody model also integrated into the multibody code and was developed at the
and its interaction to other codes which are necessary to simu- Institute of Mechatronics in Chemnitz, Germany. The aerody-
late an overall wind turbine model. First, the interaction of the namic code used in SIMPACK and Adams is based on Aero-

2 Copyright
c 2014 by ASME
Dyn provided by the National Renewable Energy Laboratories Parametric Multibody Model of the Wind Turbine in
(NREL). The aerodynamic codes include a calculation based on MSC.Adams
the blade element momentum theory. For this purpose, the blade The multibody model in MSC.Adams is parametrically built
is divided into separate aerodynamic elements. As shown in up. That means, instead of defining the model within the
Fig. 2, the multibody code provides the position and velocity of Adams/View preprocessing environment, the model is created
the blade elements. The aerodynamic code provides the aerody- within the MATLAB environment. The Matlab code generates
namic forces and moments. In addition, the aerodynamic code an Adams command file in the ASCII format, which can be im-
from NREL is extended by the general dynamic wake theory ported by Adams/View. The same principles were also applied
which is based on the acceleration potential theory, see [8]. Fur- for model generation of the high-lift mechanisms of a modern
thermore, SIMPACK provides an interface to the aerodynamic transport aircraft, see [10].
code from ECN based on the lifting line theory, see [9]. For
comparison of the results of the multibody codes used, the blade
element momentum theory is applied within this research work
only.
detailed drive train model

discrete beam blade model


Simple Multibody Model of the Wind Turbine in Flex5
The Flex5 model has a fixed topology comprising overall 28
degrees of freedom as shown in Fig. 3.

Mode Description modal based tower model


Mode 1 1-6 Tower Foundation
7-8 Tower bending thrust direction
Mode 2 9-10 Tower bending side-side
x 11 Tower top yaw FIGURE 4. THE MULTIBODY MODEL OF THE WIND TURBINE
12 Tower top tilt
13 Rotation of main shaft USING MSC.ADAMS
14 Bending of main shaft x-axis
15 Bending of main shaft y-axis
16 1st flapwise mode of blade 1
17 2nd flapwise mode of blade 1
18 1st edgewise mode of blade 1 The Adams model comprises a flexible tower model based
h 19 2nd edgewise mode of blade 1
20 1st flapwise mode of blade 2
on a finite element model, blades built up of discrete beams and
21 2nd flapwise mode of blade 2 a detailed drive train model, see [11]. The discrete beams consist
22 1st edgewise mode of blade 2
23 2nd edgewise mode of blade 2 of lumped mass elements and E ULER-B ERNOULLI beams. Com-
24 1st flapwise mode of blade 3
25 2nd flapwise mode of blade 3 pared to a blade model consisting of flexible bodies, a higher nu-
z
26
27
1st edgewise mode of blade 3
2nd edgewise mode of blade 3
merical stability of discrete beams in MSC.Adams during start-
y
28 Torsion of main shaft up of the wind turbine could be achieved. Furthermore, effects
like centrifugal stiffness are taken into account, and the interface
FIGURE 3. DEGREES OF FREEDOM OF THE FLEX5 WIND loads along the blade can be obtained easily.
TURBINE MODEL As a result, a multibody model with approximately 600 de-
grees of freedom is obtained, see Fig. 4. The higher model depth
of the simulation leads to larger CPU times compared to Flex5.
Within this given topology, the model of a specific wind tur-
The simulation of a ten minute time series with turbulent wind
bine is defined by a fixed set of parameters. A general parameter
conditions on an Intel Core i7-2600 takes about 20 minutes.
file contains, above all, the geometric parameters, the load case
parameters, a tower parameter file, and two parameter files for
the blade containing the mechanical and aerodynamic properties. Structured Multibody Model of the Wind Turbine in
An implementation of another topology of the model is possible alaska/Wind
by a modification of the computer code only. The multibody model within alaska/Wind is structured into
The elasticity of the blade and the tower is represented by submodels. A typical alaska/Wind model comprises submodels
a superposition of the first two modes in the two independent of the foundation, the tower, the nacelle, the yaw drive, the pitch
directions, respectively. The degrees of freedom can be switched drive, the rotor (hub and blades), the drive train, the generator and
off independently within the general parameter file. Due to the the controller. Due to a source-target-concept, the submodels can
small number of degrees of freedom, the calculation time is very be exchanged easily. This is an appropriate way, for example, to
short. A typical run of a ten minute time series with turbulent integrate a detailed drive train model. The alaska/Wind model is
wind conditions on an Intel Core i7-2600 takes about one minute. shown in Fig. 5.

3 Copyright
c 2014 by ASME
time. As a result, the wind turbine model consists of 44 degrees
modal based blade model
of freedom, see Fig. 6.
detailed drive train model

detailed drive train model

modal based blade model


modal based tower model

FIGURE 5. THE MULTIBODY MODEL OF THE WIND TURBINE


USING ALASKA/WIND

modal based tower model


The blades and the tower are generated using the Flex5 in-
put filter of the alaska/Wind Workbench, resulting in a modal
representation of the flexible components. The model com- FIGURE 6. THE MULTIBODY MODEL OF THE WIND TURBINE
prises different types of drive trains. The most simple drive train USING SIMPACK
model consists of two masses only coupled via a torsional spring
damper element. The spring damper element is parametrised
For simulating different load cases time efficiently, an open
according to the first eigenfrequency of the drive train. Also
source code script developed by SIMPACK and refined for the
a detailed drive train model similar to the Adams and Simpack
2.05 MW wind turbine is used. A ten minute time series with
model, see [11], was implemented.
turbulent wind conditions takes approximately 15 minutes on an
The model with the simple drive train has 21 degrees of free-
Intel Core i5-2320.
dom, the detailed drive train model has 44 degrees of freedom.
The simulation of a ten minute time series with turbulent wind
conditions on an Intel Core i7-2600 with the simple drive train
COMPARISON OF SIMULATIONS AND MEASURE-
model takes about 6 minutes, with the detailed drive train model
MENTS
about 13 minutes.
A very important aspect is the experimental validation of
the simulations which are also needed for type certification of
Flexible Multibody Model of the Wind Turbine in SIM- the wind turbine. Therefore, the prototype of the wind turbine is
PACK equipped with numerous measurement sensors, e.g. strain gages
The detailed multibody model of the 2.05 MW wind tur- at the blade root, in the tower or at the low speed shaft, see
bine developed in SIMPACK 8.903b is described in [11]. For Fig. 7a. In addition, a measurement mast equipped with wind
the results shown in this contribution, a revised model built up in vanes and cup anemometers is built up in front of the wind tur-
SIMPACK 9.3 is used. bine, see Fig. 7b.
For comparison with the Flex5 and alaska/wind models, the The aim of this research work is not to establish one of the
tower and the blades are modelled elastically. A detailed drive simulation environments as a reference but to compare equiva-
train model including two planetary gear stages and a spur gear lent models developed by means of the simulation packages with
stage is integrated, comprising torsional stiffness of shafts and the measurement results. The objective is to evaluate the dif-
the contact-stiffness of interacting gears. Blades, tower, and ferent modelling concepts used in the packages for wind turbine
drive train are substructures within the main model. By this, they simulation. In contrast to this method, other research projects
can be easily exchanged in order to simulate different wind tur- validate their program development for wind turbine simulation
bine designs. In order to gain comparable results in accordance by comparison to generally accepted design codes, see [6]. The
to measurements at the wind turbine in the field, the original differences in model generation and build-up of the equations of
controller from the real wind turbine is implemented using an motion lead to difficulties in the direct comparison of the design
interface. The flexible bodies are based on a finite element for- codes. In the authors opinion, it is more meaningful to evaluate
mulation. The tower consists of solids and T IMOSHENKO beam each program by comparing the numerical results with real mea-
elements. Additionally lumped masses are used. The blades surements. Hence, the scatter plots in the next sections compare
are modelled by T IMOSHENKO beam elements. Modal super- the simulations of Flex5, MSC.Adams, alaska/Wind and SIM-
position techniques are used to reduce the number of degrees of PACK with measurements from the prototype of the wind tur-
freedom of the finite element models and to speed up simulation bine.

4 Copyright
c 2014 by ASME
a comparison of the dynamic loads using rainflow counting pro-
cedures, see [12]. The load cycles are estimated for the product
life cycle with respect to the wind distribution in the wind class
GL IIa according to the GL guideline [13].

encapsulated
Strain Gage Electrical Power (Power Curve)
The power curve is important for the economics of a wind
turbine. High earnings especially in the part-load operational
range leads to a fast return of invest which is requested by the
operators of wind farms. The maximum, minimum and mean
electrical power of the Flex5, MSC.Adams, alaska/Wind and
SIMPACK simulations compared to the measurement results are
a shown in Fig. 8.
The diagrams show a very good agreement of simulations
with Flex5, MSC.Adams, alaska/Wind and SIMPACK compared
measurements. The use of the same controller on the prototype
and within the simulations contributes to this result.

Pitch Angle
Measurement The pitch angle shows the functionality of the controller, es-
Mast Wind Turbine
pecially of the pitch controller above rated wind speed. In Fig. 9,
the maximum, minimum and mean pitch angle of blade 1 of the
Flex5, MSC.Adams, alaska/Wind and SIMPACK simulations are
shown in comparison to the measurement results.
b The diagrams show a very good agreement of simulations
with Flex5 compared to measurements. The simulation results
FIGURE 7. MEASUREMENT SETUP OF THE WIND TURBINE. (maximum and mean value) using MSC.Adams,alaska/Wind and
a STRAIN GAGE AT THE BLADE ROOT. b PROTOTYPE OF THE SIMPACK lie above the measured curves. It can be noticed that
2.05 MW WIND TURBINE AND MEASUREMENT MAST the slope of the simulated mean values is slightly steeper com-
pared to the measurements. The use of the same controller on
the prototype and within the simulations contributes to the good
The continuous time series obtained from the measurements agreement of simulation and measurements.
are split up into ten minute time series in analogy to the simulated
time series. The resulting time series are classified with respect
Rotor Speed
to mean wind speed and the turbulence intensity of the wind. To
Another method to check the working principle of the con-
compare simulations and measurements, statistical evaluations
troller is the comparison of the rotor speed nRotor = 30 Rotor / .
of the calculated and measured results are done, as it is difficult
The maximum, minimum and mean rotor speed of the Flex5,
to transfer the wind conditions from the measurement to the sim-
MSC.Adams, alaska/Wind and SIMPACK simulations compared
ulations. The statistical values used for comparison are the min-
to the measurement results are shown in Fig. 10.
imum, maximum, mean value and standard deviation. All calcu-
As it can be seen in Fig. 10, the mean values of simulation
lations and measurements are done with a turbulence intensity of
and measurement agree very well around rated wind speed. The
10 %. For statistical confidence, the calculations are done with
maximum values around rated wind speed lie above the mea-
different wind seeds. A comparison of all measured interface
sured values using MSC.Adams, alaska/Wind and SIMPACK.
loads and operating values is not possible within this paper. For
For the rotor speed the best agreement between simulation and
comparison of the wind turbine behaviour, the measured and cal-
measurement is obtained using the Flex5 code.
culated electrical power, pitch angle and rotor speed are chosen.
To evaluate the simulated loads the bending moments at blade
root and the tilt bending moment at tower base are compared.
The vertical green line within the statistical diagrams denotes the
rated wind speeds. Because the predominant loads are applied on
the blades, the bending moments at the blade root are chosen for

5 Copyright
c 2014 by ASME
Flex5 Flex5
Scatterplot - Maximum-Minimum-Mean Scatterplot - Maximum-Minimum-Mean
2500 25
MAX-Meas MAX-Meas
MIN-Meas MIN-Meas
MEAN-Meas MEAN-Meas
2000 MAX-Sim 20 MAX-Sim
MIN-Sim MIN-Sim
MEAN-Sim MEAN-Sim
1500 15

Theta - []
P - [kW]

1000 10

500 5

0 0
2 4 6 8 10 12 14 16 18 2 4 6 8 10 12 14 16 18
Wind speed - [m/s] Wind speed - [m/s]

MSC.Adams MSC.Adams
Scatterplot - Maximum-Minimum-Mean Scatterplot - Maximum-Minimum-Mean
2500 25
MAX-Meas MAX-Meas
MIN-Meas MIN-Meas
MEAN-Meas MEAN-Meas
2000 MAX-Sim 20 MAX-Sim
MIN-Sim MIN-Sim
MEAN-Sim MEAN-Sim
1500 15
Theta - []
P - [kW]

1000 10

500 5

0 0
2 4 6 8 10 12 14 16 18 2 4 6 8 10 12 14 16 18
Wind speed - [m/s] Wind speed - [m/s]

alaska/Wind alaska/Wind
Scatterplot - Maximum-Minimum-Mean Scatterplot - Maximum-Minimum-Mean
2500 25
MAX-Meas MAX-Meas
MIN-Meas MIN-Meas
MEAN-Meas MEAN-Meas
2000 20
MAX-Sim MAX-Sim
MIN-Sim MIN-Sim
MEAN-Sim MEAN-Sim

1500 15
Theta - []
P - [kW]

1000 10

500 5

0 0
2 4 6 8 10 12 14 16 18 2 4 6 8 10 12 14 16 18
Wind speed - [m/s] Wind speed - [m/s]

SIMPACK SIMPACK
Scatterplot - Maximum-Minimum-Mean
Scatterplot - Maximum-Minimum-Mean 25
2500
MAX-Meas
MAX-Meas
MIN-Meas
MIN-Meas
MEAN-Meas
MEAN-Meas 20
2000 MAX-Sim
MAX-Sim
MIN-Sim
MIN-Sim
MEAN-Sim
MEAN-Sim
15
1500
Theta - []
P - [kW]

10
1000

5
500

0
0 2 4 6 8 10 12 14 16 18
2 4 6 8 10 12 14 16 18
Wind speed - [m/s]
Wind speed - [m/s]

FIGURE 8. MEASUREMENT AND SIMULATIONS WITH FIGURE 9. MEASUREMENT AND SIMULATIONS WITH
FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM- FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM-
PARISON OF THE STATISTICAL VALUES (MAXIMUM, MINI- PARISON OF THE STATISTICAL VALUES (MAXIMUM, MINI-
MUM, MEAN VALUE) OF THE ELECTRICAL POWER MUM, MEAN VALUE) OF THE PITCH ANGLE 1

6 Copyright
c 2014 by ASME
Flex5 Flapwise Bending Moment at the Blade Root
Scatterplot - Maximum-Minimum-Mean
17 The flapwise bending moment at the blade root, see Fig. 11,
16 is caused by the lift of the aerodynamic profile of the blade. The
15
maximum, minimum, and mean flapwise bending moment at the
14
blade root as well as the corresponding standard deviation of
13
the Flex5, MSC.Adams, alaska/Wind and SIMPACK simulations
nRotor - [rpm]

12
compared to the measurement results are shown in Fig. 12.
11

10 MAX-Meas
MIN-Meas
9 MEAN-Meas

8
MAX-Sim
MIN-Sim M flap
MEAN-Sim wind direction
7
2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

MSC.Adams
Scatterplot - Maximum-Minimum-Mean
17

16

15

14

13 FIGURE 11. DEFINITION OF THE FLAPWISE BENDING MO-


nRotor - [rpm]

12 MENT AT THE BLADE ROOT


11

10 MAX-Meas
MIN-Meas
9 MEAN-Meas The direction of the moment Mflap acting on the cut free
MAX-Sim
8 MIN-Sim blade and the wind direction in Fig. 11 indicate that the values
MEAN-Sim
7
2 4 6 8 10 12 14 16 18 of the bending moment are negative. The mean values obtained
Wind speed - [m/s]
alaska/Wind from the simulations with MSC.Adams und SIMPACK with the
Scatterplot - Maximum-Minimum-Mean measurements show a better agreement compared to the Flex5
17
and alaska/Wind simulations. Especially, this is seen at rated
16

15
wind speed, that means around 10 - 11 m/s. However, the abso-
14
lute values from the Flex5 and alaska/Wind simulations exceed
13 those from the measurements, indicating that the Flex5 simula-
nRotor - [rpm]

12 tions lead to conservative load estimations.


11 Beside the statistical evaluation of flapwise bending moment
10
MAX-Meas
MIN-Meas
at the blade root, a dynamic evaluation has been done. Therefore,
9 MEAN-Meas
MAX-Sim
the rainflow matrix is estimated from the time series of the flap-
8 MIN-Sim
MEAN-Sim
wise bending moment. The flapwise bending moment is strongly
7
2 4 6 8 10 12
Wind speed - [m/s]
14 16 18 influenced by the turbulence of the wind. The rainflow matrices
SIMPACK from the simulations are shown in Fig. 13. The rainflow matrices
17
Scatterplot - Maximum-Minimum-Mean show a large number of load cycles at a small load range fora
16 broad range of mean values. The corresponding rainflow matrix
15 of the measurements is presented in Fig. 14.
14 Due to the stochastic behaviour of the wind during mea-
13
surements and the use of different wind simulators during the
nRotor - [rpm]

12
simulations, a qualitative comparison of the rainflow matrices is
11
difficult. Also a quantitative comparison is not appropriate be-
10 MAX-Meas
MIN-Meas cause of the discrete classification during the rainflow counting
9 MEAN-Meas

8
MAX-Sim procedure. To circumvent these problems, the load cycles are
MIN-Sim

7
MEAN-Sim summed over each load range class neglecting their mean val-
2 4 6 8 10 12 14 16 18
Wind speed - [m/s] ues. Subsequently, the load cycles are accumulated in that way
that load cycles with a large load range comprises all load cy-
FIGURE 10. MEASUREMENT AND SIMULATIONS WITH cles with smaller load ranges. The corresponding diagrams of
FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM- the load range vs. accumulated load cycles are shown in Fig. 15.
PARISON OF THE STATISTICAL VALUES (MAXIMUM, MINI- As it can be seen from Fig. 15, the simulations using Flex5
MUM, MEAN VALUE) OF THE ROTOR SPEED show a non-conservative behaviour for large load ranges com-

7 Copyright
c 2014 by ASME
Flex5 Flex5
Scatterplot - Maximum-Minimum-Mean
MBBl1flap - [kNm]

MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

MSC.Adams MSC.Adams
Scatterplot - Maximum-Minimum-Mean
MBBl1flap - [kNm]

MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

alaska/Wind
Scatterplot - Maximum-Minimum-Mean alaska/Wind
MBBl1flap - [kNm]

MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

SIMPACK
Scatterplot - Maximum-Minimum-Mean SIMPACK
MBBl1flap - [kNm]

MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

FIGURE 12. MEASUREMENT AND SIMULATIONS WITH


FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM- FIGURE 13. SIMULATIONS WITH FLEX5, MSC.ADAMS,
PARISON OF THE STATISTICAL VALUES (MAXIMUM, MINI- ALASKA/WIND AND SIMPACK: COMPARISON OF THE DY-
MUM, MEAN VALUE) OF THE FLAPWISE BENDING MOMENT NAMIC VALUES (RAINFLOW MATRIX) OF THE FLAPWISE
AT THE BLADE ROOT BENDING MOMENT AT THE BLADE ROOT
8 Copyright
c 2014 by ASME
Edgewise Bending Moment at the Blade Root
The edgewise bending moment at the blade root, see Fig. 16,
is mainly caused by the dead weight of the blade. Furthermore,
the loads of the dead weight of the blade are superposed by dy-
namical mass effects and aerodynamic loads due to the lift of the
blade. The maximum, minimum, and mean edgewise bending
moment at the blade root of the Flex5, MSC.Adams, alaska/Wind
and SIMPACK simulations compared to measurement results are
shown in Fig. 17.

wind direction
M edge

FIGURE 14. RAINFLOW MATRIX OF MEASUREMENT OF THE


FLAPWISE BENDING MOMENT AT THE BLADE ROOT FIGURE 16. DEFINITION OF THE EDGEWISE BENDING MO-
MENT AT THE BLADE ROOT

Load Ranges vs. accumulated Load Cycles for Comparison


The comparison of simulations and measurement show
some differences in the statistical values. A possible uncertainty
is the calibration of the strain gages at the blade root. The sen-
sors are calibrated by the dead weight and a slow revolution of
Load Range - MBBl1flap - [kNm]

the wind turbine. Typically, the wind turbine coasts freely during
calibration.
Due to the strong influence of the dead weight, the rainflow
matrix has a typical characteristic. The rainflow matrices of the
simulations in Fig. 18 and the rainflow matrix of the measure-
Measurement ment in Fig. 19 show three peaks within all diagrams. The single
Adams
alaska/Wind peak in the higher load range is caused by the first order static
Flex5
SIMPACK
moment of the blade and the revolutions of the wind turbine dur-
3
10 10
4
10
5
10
6
10
7
10
8 ing its life cycle.
Number of accumulated Cycles However, the comparison of the load range vs. the accumu-
lated load cycles in Fig. 20 show a conservative behaviour of all
FIGURE 15. MEASUREMENT AND SIMULATIONS WITH simulations for the edgewise bending moment compared to mea-
FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM- surements. It can be seen that large loads act on the wind turbine
PARISON OF THE DYNAMIC VALUES (LOAD RANGE VS. ACCU- with more than 1 108 load cycles during its product life cycle.
MULATED LOAD CYCLES) OF THE FLAPWISE BENDING MO-
MENT AT THE BLADE ROOT

pared to measurements. In contrast the comparison of simula-


tions using MSC.Adams, alaska/Wind and SIMPACK and mea-
surements show a good agreement or conservative behaviour for
dynamic loads over the whole accumulated load cycle range.
The results obtained using alaska/Wind are the most conserva-
tive ones compared to the other simulation packages.

9 Copyright
c 2014 by ASME
Flex5 Flex5
Scatterplot - Maximum-Minimum-Mean

MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
MBBl1edge - [kNm]

2 4 6 8 10 12 14 16 18
Wind speed - [m/s]
MSC.Adams MSC.Adams
Scatterplot - Maximum-Minimum-Mean

MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
MBBl1edge - [kNm]

2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

alaska/Wind alaska/Wind
Scatterplot - Maximum-Minimum-Mean

MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
MBBl1edge - [kNm]

2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

SIMPACK
Scatterplot - Maximum-Minimum-Mean SIMPACK
MAX-Meas
MIN-Meas
MEAN-Meas
MAX-Sim
MIN-Sim
MEAN-Sim
MBBl1edge - [kNm]

2 4 6 8 10 12 14 16 18
Wind speed - [m/s]

FIGURE 17. MEASUREMENT AND SIMULATIONS WITH


FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM- FIGURE 18. SIMULATIONS WITH FLEX5, MSC.ADAMS,
PARISON OF THE STATISTICAL VALUES (MAXIMUM, MINI- ALASKA/WIND AND SIMPACK: COMPARISON OF THE DY-
MUM, MEAN VALUE) OF THE EDGEWISE BENDING MOMENT NAMIC VALUES (RAINFLOW MATRIX) OF THE EDGEWISE
AT THE BLADE ROOT BENDING MOMENT AT THE BLADE ROOT
10 Copyright
c 2014 by ASME
wind direction
M tilt

FIGURE 21. DEFINITION OF THE TILT BENDING MOMENT


AT THE TOWER BASE

packages provide a very good agreement compared to the mea-


surements. The large influence of the rotor thrust results in same
characteristics of flapwise bending moment and tilt bending mo-
ment at tower base. The best agreement between measurement
and simulation regarding maximum, minimum and mean values
are achieved using alaska/Wind.

FIGURE 19. RAINFLOW MATRIX OF MEASUREMENT OF THE


EDGEWISE BENDING MOMENT AT THE BLADE ROOT CONCLUSION
The present contribution compares the multibody codes
Flex5, MSC.Adams, alaska/wind and SIMPACK with measure-
Load Ranges vs. accumulated Load Cycles for Comparison
ments done on a prototype of 2.05 MW wind turbine developed
by W2e Wind to Energy. The comparisons show in general a
good agreement for all simulation packages. Hereby, the simu-
Load Range - MBBl1edge - [kNm]

lations use the same controller software as implemented on the


physical wind turbine.
The loads at blade root and tower base show some differ-
ences between simulation and measurements. The main loads
act on the rotor of the wind turbine and thus on the blade root.
The best agreements are obtained using the discrete beam model
Measurement
Adams within MSC.Adams, which includes nonlinear effects like large
alaska/Wind
Flex5
deformations and centrifugal stiffness effects. The Flex5-like
SIMPACK blade models in alaska/Wind and SIMPACK produces conser-
4
10 10
5
10
6
10
7
10
8
vative results compared to the measurements. The behaviour of
Number of accumulated Cycles
the alaska/Wind model are the basis of ongoing investigations
FIGURE 20. MEASUREMENT AND SIMULATIONS WITH
within a cooperation of W2e Wind to Energy and the Institute of
FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM-
Mechatronics in Chemnitz, Germany.
PARISON OF THE DYNAMIC VALUES (LOAD RANGE VS. ACCU- An important aspect beyond the results is the usability of
MULATED LOAD CYCLES) OF THE EDGEWISE BENDING MO- the general purpose multibody programs. Within a fixed topol-
MENT AT THE BLADE ROOT ogy model like Flex5 the user parametrises a wind turbine model
only. General purpose multibody programs provide a lot of mod-
elling features but need an extensive knowledge in multibody
Tilt Bending Moment at the Tower Base modelling. The number of model errors and uncertainties may
The tilt bending moment at the tower base, see Fig. 21, is also increase by the number of degrees of freedom of the model.
mainly caused by the thrust of the rotor. Also some dynamic A guided model buildup could support the user. A very good
effects of the mass of the nacelle and the rotor affects the tilt example is the alaska/Wind Workbench. SIMPACK provides an
bending moment at the tower base. The maximum, minimum, example model of the generic NREL wind turbine which can be
and mean tilt bending moment at the tower base of the Flex5, adapted to a specific wind turbine model. The ADWIMO module
MSC.Adams, alaska/Wind and SIMPACK simulations compared for MSC.Adams provided by MSC.Software has not been tested
to measurement results are shown in Fig. 22. so far. Instead, the model buildup and load case generation for
Due to the fact that the wind turbine almost behaves like the Adams wind turbine model has been done by a user-written
a cantilever beam with a single mass at its end, all simulation MATLAB script.

11 Copyright
c 2014 by ASME
Flex5 The most important aspect in the simulation of mechatronic
Scatterplot - Maximum-Minimum-Mean
systems is the integration of the controller. The integration of
the same controller software as on the physical wind turbine in-
creases the quality of the simulation results and helps to avoid
possible errors in the controller design. Among the programs
MBTwrtilt - [kNm]

tested within this study, alaska/Wind is the only program which


provides a free programmable discrete controller interface in
MAX-Meas
C++. MSC.Adams and SIMPACK also provide extension capa-
MIN-Meas
MEAN-Meas
bilities using FORTRAN, but a discrete controller interface has
MAX-Sim
MIN-Sim
to be created by the user. This possibly requires the need of ex-
MEAN-Sim
tensive technical support.
2 4 6 8 10 12 14 16 18
Wind speed - [m/s] All general purpose multibody programs tested have their
MSC.Adams pros and cons. But nevertheless, a transition from specific wind
Scatterplot - Maximum-Minimum-Mean
turbine calculation tools such as Flex5 and BLADED to general
multibody programs is an appropriate way to develop the wind
turbine of the future. It enables a closer look into the dynamical
behaviour of the wind turbine leading to an extensive understand-
MBTwrtilt - [kNm]

ing and an advanced design of wind turbines.

MAX-Meas
MIN-Meas
MEAN-Meas ACKNOWLEDGMENT
MAX-Sim
MIN-Sim The authors would like to thank the German Federal Min-
MEAN-Sim
2 4 6 8 10 12 14 16 18 istry for the Environment, Nature Conservation and Nuclear
Wind speed - [m/s]

alaska/Wind
Safety for supporting this research.
Scatterplot - Maximum-Minimum-Mean

REFERENCES
[1] MSC.S OFTWARE, 2012. Adams 2012.2: Adams 2012 On-
line Help. Tech. rep., MSC Software Corporation, Santa
MBTwrtilt - [kNm]

Ana, California.
[2] I NSTITUTE OF M ECHATRONICS, 2013. alaska 8: Mod-
MAX-Meas
MIN-Meas
elling and Simulation of mechatronic Systems - Users
MEAN-Meas
MAX-Sim Guide. Tech. rep., Institute of Mechatronics, Chemnitz,
MIN-Sim
MEAN-Sim Germany.
2 4 6 8 10 12
Wind speed - [m/s]
14 16 18
[3] S IMPACK, 2013. Simpack v9.3: Documentation to Sim-
SIMPACK pack. Tech. rep., Simpack AG, Gilching, Germany.
Scatterplot - Maximum-Minimum-Mean
[4] L AINO, D. J., and H ANSEN, A. C., 2002. AeroDyn v12.50:
Users Guide. Tech. rep., Windward Engineering (Prepared
for the National Renewable Energy Laboratory), Salt Lake
City.
MBTwrtilt - [kNm]

[5] J ONKMAN, B. J., and J ONKMAN, J. M., 2013. Addendum


to the Users Guides for FAST, A2AD, and AeroDyn Re-
MAX-Meas
leased March 2010 - February 2013. Tech. rep., National
MIN-Meas
MEAN-Meas
Renewable Energy Laboratory, Golden, Colorado.
MAX-Sim
MIN-Sim
[6] TAUBERT, M., C LAUSS, S., F REUDENBERG, H., K EIL, A.,
MEAN-Sim , M., M OSER, W., and W ULF, H. O., 2011. Wind
M ARZ
2 4 6 8 10 12 14 16 18
Wind speed - [m/s] Turbine Design Codes: Eine Validierung von alaska/Wind
mit BLADED, FAST und FLEX5. Tech. rep., Institut fur
FIGURE 22. MEASUREMENT AND SIMULATIONS WITH Mechatronik, Chemnitz.
FLEX5, MSC.ADAMS, ALASKA/WIND AND SIMPACK: COM- [7] W OERNLE, C., K AEHLER, M., R ACHHOLZ, R., H ER -
PARISON OF THE STATISTICAL VALUES (MAXIMUM, MINI- RMANN , S., Z IERATH , J., S OUFFRANT , R., and BADER,
MUM, MEAN VALUE) OF THE TILT BENDING MOMENT AT THE R., 2010. Robot-Based HiL Test of Joint Endoprosthe-
TOWER BASE
12 Copyright
c 2014 by ASME
ses. In Lenarcic, J. (Eds.): Advances in Robot Kinematics:
Analysis and Control, Springer.
[8] S UZUKI, A., 2000. Application of Dynamic Inflow Theory
to Wind Turbine Rotors. PhD thesis, Department of Me-
chanical Engineering, University of Utah, Salt Lake City.
[9] VAN G ARREL , A., 2003. Development of a Wind Turbine
Aerodynamics Simulation Module. Tech. rep., ECN Wind
Energy, Petten, The Netherlands.
[10] Z IERATH, J., W OERNLE, C., and H EYDEN, T., 2009.
Elastic multibody models of transport aircraft high-lift
mechanisms. AIAA Journal of Aircraft, 46(5), pp. 1513
1524.
[11] R ACHHOLZ, R., W OERNLE, C., and Z IERATH, J., 2012.
Dynamics of a Controlled Flexible Multi-body Model of
a 2 MW Wind Turbine. In Proceedings of the 2nd Joint
International Conference on Multibody System Dynamics
- IMSD, Stuttgart.
[12]
K OHLER
, M., J ENNE, S., P OTTER , K., and Z ENNER, H.,
2012. Zahlverfahren und Lastannahme in der Betriebsfes-
tigkeit (German Edition). Springer, Berlin.
[13] G ERMANISCHER L LOYD, 2010. Guideline for the Cer-
tification of Wind Turbines, Edition 2010. Germanischer
Lloyd, Hamburg.

13 Copyright
c 2014 by ASME
View publication stats

Das könnte Ihnen auch gefallen