Back to the programme printer.gif Print

Delegates are invited to meet and discuss with the poster presenters in this topic directly after the session 'Aerodynamics and rotor design' taking place on Wednesday, 12 March 2014 at 09:00-10:30. The meet-the-authors will take place in the poster area.

Anand Natarajan DTU, Denmark
Anand Natarajan (1) F P Christina Koukoura (1) Allan Vesth (1) Rebeca Rivera Lamata (2) Kenneth Thomsen (3)
(1) DTU, Roskilde, Denmark (2) DONG energy, Gentofte, Denmark (3) Siemens Wind Power, Taastrup, Denmark

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2014 and are published here unedited

Anand Natarajan is a Senior researcher and team manager at DTU Wind Energy, working in the areas of loads prediction on offshore wind turbines, structural design and probabilistic methods. He has a PhD in Aerospace Engineering and 10+ years of experience in wind energy research. Prior to joining DTU, he worked for GE Global Research Center and for Caterpillar Engineering Design center. He is the secretary of the European Technology Platform (TPWIND) Offshore group. He is part of the project management of several Danish and EU funded projects such as the FP7 INNWIND.EU project comprising of 27 organizations.


Extreme design loads calibration of offshore wind turbine blades through real time measurements


Ultimate design load evaluations on offshore wind turbine rotors require loads extrapolation for 50 year return periods
or the use of extreme turbulence input, both of which have significant model and physical uncertainties. It is imperative
to measure loads on a large offshore wind turbine to verify the design envelope. Herein measured loads over one year on
the blades of a 3.6MW offshore wind turbine are used to determine the 1-year extreme load. This measured extreme is
compared to results from stochastic extrapolation based on a limited subset of data whereby the extrapolation is calibrated
to realistic 50 year load magnitudes.


A Siemens 3.6MW offshore wind turbine on an operational wind farm has been fully instrumented for load measurements and is also equipped with a Nacelle mounted LIDAR to measure the mean wind speed and variation at a point in front of the turbine. Each of the turbine blades is equipped with 4 strain gauges, which are mounted 1m from the blade root, for measurement of the flap and edge moments on the blade. Further the drive shaft possesses 4 strain gauges for the measurement of bending moments. Additional strain gauges and accelerometers are installed at 3 different heights along the support structure (4 gauges per height). The loads measured on the turbine are logged as high frequency time series, from which additional statistics is processed. A SCADA system on the turbine transmits ten minute averaged rotational speed, anemometer wind speed, power production, and blade pitch angle and turbine yaw direction. Further SCADA based data from surrounding wind turbines are obtained to correlate measurements such as wind direction and wake effects.
The blade strain gauges are calibrated through a gravity load based experiment during stand still of the turbine. The nacelle mounted LIDAR[1] was calibrated under standard conditions measured at the Høvsøre test site in Denmark. Further wind velocity measurements from the LIDAR on the nacelle are benchmarked with the nacelle mounted cup anemometer during idling of the turbine. During operation of the turbine, wind measurements from both LIDAR and nacelle anemometer are used to determine the mean wind speed and its variation.
The flap and edge moments on all three blades of the turbine are recorded continuously in a database under all operating and stand still conditions from May 2012. The extreme load extrapolation is based on the IEC 61400-1 Ed.3 [2] and the stochastic distribution applied to the tail of the extreme loads data is a Gumbel distribution with a distorted quadratic exponent [3]. The Gumbel distribution with a quadratic exponent is the theoretical solution for asymptotic extreme values [4] and is proven to converge to the 50 year exceedance probabilities for Poisson processes.

Main body of abstract

The blade strain gauge calibration is schematically explained in Fig.1. Two tests were performed where the rotor is slowly rotated through 360 degrees. The first test with a blade pitch angle 0 deg is used for the calibration of the edgewise moment and the second with pitch angle of 90 deg for the calibration of the flap wise moment. The resulting signal (in voltage) is a sinusoidal curve presented in Fig.1, with the weight load taken as unity.
During the blade strain gauge calibration some limitations need to be taken into consideration such as that the calibration is performed in an ideal condition of low wind speed, where the loading due to aerodynamic forces during operation is neglected.
LIDAR wind velocity measurements were compared with the nacelle mounted cup anemometer during idling of the turbine as shown in Fig. 2, which shows that the cup-anemometer and LIDAR measurements are sufficiently correlated when the turbine is not in operation.
The nacelle mounted LIDAR resolution does not allow measurement of turbulence structures, thereby the wind speed standard deviation in Fig. 2 shows a wide scatter with respect to the corresponding standard deviation measured with the anemometer. Therefore, only the mean wind speeds as measured from the LIDAR are used in this study.
A random sample of blade root flap maximum moments as measured on one of the blades of the 3.6MW wind turbine is used as the basis for load extrapolation, where each maximum is over a 10 minute period. Figure 3 compares the normalized blade root maximum moments on blade 1 root for a 10 day period with that from a 100 day period during normal operation of the turbine.
The blade moment normalization is done over a random 1 day extreme blade root flap moment. Thus it can be seen from Fig. 3 that the 100 day extreme flap moment is about 18% higher than the random sampled 1 day extreme moment. Conventional load extrapolation uses simulated data resulting from 12-15 turbulent seeds per mean wind speed and a similar quantity of measured loads is considered herein for extrapolation.
Using the Gumbel distribution with a quadratic exponent for the extreme load probability and the Rayleigh distribution to model the probability of mean wind speed, the probability P that the extreme load Fe exceeds a given level F, is given by Eq. (1),
Wherein a, b, c are the coefficients of the parametric fit to the data at a mean velocity of vi and ni is the number of uncorrelated sampled extreme loads at each mean wind speed. The target 50 year probability of exceedance is 3.8e-07 and the one year probability level is 1.9e-05. In the present case study, 28 random samples of flap loads was considered at each mean wind speed, between 5m/s and 17m/s, since above 17m/s mean wind speed, there are limited measured
samples for this short period. Each sample load was normalized with the one day extreme magnitude that was measured.
Figure 4 displays the result of the load extrapolation performed with Eq. (2) . It can be seen from Fig. 4 , that using this limited measured set, the 50 year extreme flap moment is predicted to be about 30% higher than the 1 day extreme and the 1 year extreme flap moment is predicted to be about 21% higher than the 1 day extreme. Based on this single case and comparing the results in Fig. 4 with that displayed in Fig. 3, it can be seen that the extreme load extrapolation is predicting reasonable trends. However this will be further validated with numerous random samples of flap moments at the blade root for all three blades using a boot strapping technique.


Measured blade root peak flap moments were sampled over different operating conditions of an offshore wind turbine and up to one year of loads measurements are analyzed. The mean wind speed is mainly measured with a calibrated nacelle anemometer and also using a nacelle based LIDAR, when available. The measurements comprise of durations when the instrumented turbine is subject to free wind stream and also when it is in the wake of neighboring turbines. The initial assessment of the extreme one year blade root flap moment based on load extrapolation from a random set of 28 measured values for the flap moment at each mean wind speed indicated that the load extrapolation method employed produced similar results as the measured blade root flap moment over longer periods.
To validate this observation, the statistical variation of the 1 year extreme moment as predicted through extrapolation will be benchmarked with the 1 year extreme measured flap moment by comparison with the mode of the probability distribution of the 1 year extreme moment obtained through boot strapping, that is by repeated sampling of a fixed subset of data used for load extrapolation from a larger population of measured loads. This process will also be repeated for the measured blade root edge moment to determine the 1 year extreme edge moment. Further, since during load extrapolation, the contemporaneous component load magnitudes are lost, another interesting result that will be demonstrated is the contemporaneous measured edge moment corresponding to the one year measured extreme flap moment. The evaluation of these results will also be made for free stream and wake sectors to understand the dependency of extreme loads on wake turbulence.
These results will reveal a validated load extrapolation methodology that can accurately predict the long term extreme blade root loads under realistic offshore operational conditions.

Learning objectives
Extrapolated extremes can possess large uncertainties based on the probability distribution chosen, the number of data points sampled and the interaction of the rotor with the wind. This load case is often design driving for the blade tip deflection and the blade ultimate strains. A key learning from this study is to determine a validated method for load extrapolation, contemporaneous load magnitudes and bounds on the variation of the extrapolated load value based on measurements.

1. Courtney, M ; Wagner, R ; Friis P, T ; Bardon, M ; Davoust, S, Calibrating Nacelle LIDARS, Proceedings of EWEA 2012 - European Wind Energy Conference & Exhibition, EWEA - The European Wind Energy Association, 2012
2. International Standard Wind Turbines—Part 1: Design Requirements IEC 61400–1 Ed. 3, Amendment 1, 2010
3. Natarajan A. and Holley, W.E., Statistical Extreme Load Extrapolation with Quadratic Distortions for Wind Turbines, Journal of Solar Energy Engineering-Transactions of the ASME, 2008, 130 (3): 031017,
4. H.O. Madsen, Krenk, S. & Lind, N.C., Methods of Structural Reliability, Prentice Hall, 1986