DEWI Magazin Nr.

19, August 2001

Open Field Cup Anemometry

Axel Albers, DEWI
Summary
Within the EC co-funded project "Identification of Variables for Site Calibration and Power
Curve Assessment in Complex Terrain" (SiteParIden, contract: JOR3-CT98-0257) large
deviations between cup anemometers of different type under natural conditions have been
identified. Deviations in the order of 2 % have been found to be common even in flat terrain conditions although all compared cup anemometers have been calibrated according
to the latest standards in the wind tunnel. These large deviations are of major significance for the whole wind energy industry because the result of wind turbine power performance tests as well as site assessments are strongly dependent on the type of cup
anemometer in use. The deviations between the cup anemometers have been found to be
influenced by site characteristics. The most important flow and turbulence parameters
influencing the cup anemometer deviations have been identified. From these results follows a first attempt to correct cup anemometer measurements and to classify cup anemometers as well as an advice for further research on the field of cup anemometry.
Introduction

In the SiteParIden project detailed comparisons of cup anemometers of different type under natural conditions have
been performed. Surprisingly large deviations between measurements with different types of cup anemometers have
been found (Fig. 1).

3
2
Difference: v-X - v-Thies [%]

1.

1
0
-1
-2

Risoe P2445b, 30m

Risoe P2445b, 8m
Vector A100, 30m
Vector A100, 8m
Metone 010C-1, 30m
Mierij, 8m
Vaisala WAA151
Friedrichs 4033.1100x, 30m
Friedrichs 4033.1100x, 8m
Thies Compact, 8m

-3

As all compared cup anemometers have

-4
been calibrated in the same wind tunnel
according to the latest standards [1] this
-5
0
5
10
15
20
25
observation must be linked to the turbuv-Thies 4.3303.22.000 [m/s]
lent nature of free air flow (in contrast to
the low turbulence present in wind tun- Fig. 1: Deviations between measurements with various cup anemometers in flat terrain 8 m and 30 m above gorund
nels). The results have been verified at
various sites in flat and complex terrain
up to heights of 83 m above ground (see
reference [2] and [3] for details). The effects on cup anemometers have been found to be site dependent. However, even in flat terrain conditions 83 m above ground significant deviations between the
measurements with different types of cup anemometers in the order of typically 2 % have been observed [3]. This finding is of major interest for the wind energy industry because it implies that wind turbine power performance tests are comparable only
when they are performed with the same type of cup
anemometer and under comparable turbulence characteristics. Furthermore, additional uncertainties for
site assessments are present when the cup anemometer in use for the site assessment does not match
the type of cup anemometer used for the determination of the power curve of the projected wind turbine
type. Since the effect has been discovered the
bodies occupied with the standardisation of power
performance testing aim to overcome the problem.
Basically three different solutions are under discussion [2]:
use of a standardised cup anemometer,
a classification of cup anemometers,
a correction of cup anemometers.

Fig. 2: Test rig for outdoor cup anemometer comparison.

The distance between the cup anemometers is
2.5 m.

53

DEWI Magazin Nr. 19, August 2001

As has been outlined in [3] until lately it was not clear which flow and turbulence parameters and
anemometer properties have to be considered for a proper solution of the problem. Major progress in
this respect has been gained recently from a new large data base of high resolution time series by
DEWI [4], which is subject of this presentation.
2.

Methodology

Two cup anemometers of type Thies 4.3303.22.000 and Ris P2445B have been mounted in parallel
position on a special test rig on top of a 30 m high meteorological mast in flat terrain (Fig. 2). An ultra
sonic anemometer (USA) has been placed in middle of the test rig in order to evaluate turbulence characteristics as well as the vertical flow inclination. The whole test rig is aligned perpendicular to the
actual wind direction by a micro processor controlled drive system. Before anemometers of different
type were compared, detailed tests with two anemometers of the same type mounted on the test rig
have been performed in order to ensure that the two cup anemometer positions are exposed to the
same wind conditions. Furthermore, before the outdoor tests, all anemometers including the USA have
been calibrated in the same wind tunnel according to the MEASNET procedure [1].
A large data base of time series sampled with a frequency of 10 Hz (output rate of USA 4 Hz) has been
collected over a period of 2440.5 h. The aim of this intense data collection was to study which turbulence and flow parameters are significant to explain the differences between the cup anemometer measurements. From the time series about 100 turbulence and flow parameters have been calculated for
each 10 minute period. The ratios between the 10 minute averages of the anemometer measurements
have been analysed in detail on their dependence on turbulence and flow parameters.

1.2

wake sector

free sector

1.15

v-Thies/U
v-Thies/v-horizontal
v-Thies/v-Vector

1.1
Wind Speed Ratio [-]

The Ris P2445B and Thies 4.3303.

22.000 anemometer have been selected
for this investigation because these
anemometers have formerly shown
large deviations, both anemometers are
widely used in Germany and Denmark
respectively and because there is hope
that the anemometers may represent
two groups of anemometers consisting
of the faster anemometers Friedrichs
4033.1100X, the Thies 4.3303.22.000
and the Thies compact series and the
slower anemometers Vector A100, Ris
P2445b and Metone 010C-1.

1.05

0.95

0.9
0

40

80

120

160

200

240

280

320

360

Wind Direction []

A key finding is that the Thies 4.3303.22.

000 anemometer clearly overestimates
even the magnitude of the full three
dimensional wind speed vector (in short
vector wind speed v-vector) in highly turbulent air flow as it is present in wakes of
wind turbines (Fig. 3). Under the same
conditions the Ris P2445 B anemometer still tracks well the horizontal wind
speed component (Fig. 4). Furthermore,
in unperturbed air flow the wind speed
ratio of the Thies anemometer and the
USA is more dependent on site specific
flow parameters than the Ris anemometer (compare Fig 3 and Fig 4 for the
sector 110-350).

54

Fig. 3: Ratio of measurements with a Thies 4.3303.22.000 cup anemometer and the longitudinal wind speed component (U), the horizontal wind speed component (v-horizontal) and the magnitude of
the full 3D-wind speed vector (v-vector) as evaluated with a calibrated ultra sonic anemometer.
1.2
wake sector

free sector

1.15

Wind Speed Ratio [-]

3. Results
3.1 Comparison between Cup Anemometers and Ultra Sonic Anemometer

v-Risoe/U
v-Risoe/v-horizontal
v-Risoe/v-vector

1.1

1.05

0.95

0.9
0

40

80

120

160

200

240

280

320

360

Wind Direction []

Fig. 4: Ratio of measurements with a Ris P2445B cup anemometer and

the longitudinal wind speed component (U), the horizontal wind
speed component (v-horizontal) and the magnitude of the full 3Dwind speed vector (v-vector) as evaluated with a calibrated ultra
sonic anemometer.

DEWI Magazin Nr. 19, August 2001

3.2 Correlation of Time Series

The correlation coefficient between the cup anemometers and the ultra sonic anemometer within 10
minute periods has been investigated. Key findings of this analysis are:
The Ris P2445 B anemometer correlates clearly better with the horizontal flow component
than with v-vector or the longitudinal wind speed component.
The Thies 4.3303.22.000 anemometer has a small tendency to correlate better with v-vector
than with the horizontal flow component. It correlates clearly better with v-vector or the horizontal flow component than with the longitudinal wind speed component.
3.3 Influence of Flow and Turbulence
Variables on Cup Anemometer
Measurements
The influence of flow and turbulence
parameters on the wind speed ratios of
the two cup anemometers and the wind
speed ratios of cup anemometers and the
longitudinal wind speed component, the
horizontal wind speed component and vvector has been investigated following
three different procedures:
The correlation coefficient between the wind speed ratios
based on 10 minute averages
and the flow and turbulence
variables has been evaluated.
Multi variate regression analysis has been performed following the technique described in
[5] with the wind speed ratios
based on 10 minute averages
as dependent variables and the
flow and turbulence variables
as independent variables.
A multi classification technique
as described in [6] has been
applied in order to care for the
co-linearity of the flow and turbulence parameters.
The following results can be drawn in full
consistence from all three evaluation
techniques:
The wind speed ratios with the
Thies 4.3303.22.000 anemometer correlate significantly
with different turbulence variables. In very turbulent air flow
the anemometer measures a
much higher wind speed than
the vector wind speed.

55

DEWI Magazin Nr. 19, August 2001

0.025

1.16
std(x): standard deviation of flow component x within 10 minutes
R(x,y): corellation coefficient between flow component x and y within 10 minutes
L(x): integrale turbulence length scale of flow component x
tensor(xy): turbulent stress between flow component x and y (Reynolds tensor element)
inclination: vertical flow inclination angle averaged over 10 minutes

1.14

v-Risoe/v-horizontal
v-Thies/v-horizontal

1.12
1.1

0.015
Wind Speed Ratio [-]

Significance Coefficient [-]

0.02

v-Thies/v-horizontal, 110-350 (free air flow)

v-Thies/v-horizontal, 350-110, (wake situation)

0.01

v-Risoe/v-horizontal,110-350 (free air flow)

v-Risoe/v-horizontal,350-110, (wake situation)

0.005

1.08
1.06
1.04
1.02
1

0.98

v-vector

inclination

tensor(VW)

tensor(UW)

L(W)

tensor(UV)

L(V)

L(U)

R(V,W)

R(U,V)

R(U,W)

std(V)/U

std(W)/U

std(U)/U

-0.005

Fig. 5: Influence of flow and turbulence variables on wind

speed ratios of cup anemometer measurements and
the horizontal wind speed component. The significance coefficient as gained from multi variate regression
analysis describes the relative change of the wind
speed ratio when the respective flow or turbulence
variable is increased by one standard deviation.

0.96
0.94
0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

std(W)/U [-]

Fig. 6: Influence of the vertical turbulence intensity on wind

speed ratios of cup anemometer measurements and
the horizontal wind speed component.

The wind speed ratios of the Ris P2445 B anemometer and the horizontal wind speed component shows only a very low correlation with all considered flow and turbulence parameters.
Turbulence effects on this anemometer are much smaller than in case of the Thies
4.3303.22.000 anemometer (Fig. 5)
The vertical turbulence intensity is the most significant variable for the explanation between
the differences in the cup anemometer measurements. It has the largest contribution to the
variation of the wind speed ratios (Fig. 5, Fig. 6).
Other significant variables are the flow inclination, the correlation between the lateral wind
speed component and the other two flow components and the integral turbulence length
scale of the vertical flow component (Fig. 5).
The influence of the longitudinal turbulence intensity is much smaller than the influence of the
vertical turbulence intensity.
The most significant flow and turbulence variables are the same for both types of anemometers (Fig. 5).
The air density, air temperature (considered range 5-20C), air pressure and the wind speed
have shown no significant influence on the ratios between the cup anemometers measurements.
4.

Correction of Cup Anemometer Measurements

A correction of cup anemometers is needed to make wind turbine power curve measurements performed with different cup anemometers comparable and to match existing power curve measurements to
the cup anemometer in use for site specific wind speed measurements. The results given in the previous chapter imply that a cup anemometer correction should consider at least the vertical turbulence
intensity. The simplest correction would be a linear correction according to the vertical turbulence intensity. Fig. 7 demonstrates how the variation of the wind speed ratios with the wind direction is reduced
by the wind speed correction.
5.

Classification of Cup Anemometers

A classification of cup anemometers should take into account the sensitivity of the cup anemometers
to flow and turbulence parameters. As a first attempt the sensitivity of cup anemometers on only the
most significant turbulence parameter has been used for the classification of cup anemometers on the
bases of field tests. The sensitivity of the cup anemometers on the vertical turbulence intensity is
expressed by the slope of the regression through the ratios of the cup anemometer measurements and
the horizontal flow component (or v-vector) versus the vertical turbulence intensity. From the slope the
(simplified) measurement error can be calculated by a multiplication with the vertical turbulence intensity. Then the maximum allowable vertical turbulence intensity can be calculated for each class when
the class index is defined as the maximum measurement error in percent. The resulting classification
scheme in Tab. 1 should be considered only as a first attempt to classify cup anemometers by means
of open field measurements. Furthermore, it must be pointed out that except of the Thies

56

DEWI Magazin Nr. 19, August 2001

6.

Conclusions

1.05

free sector

wake sector
1.025
1
v-Risoe/v-Thies [-]

4.3303.22.000 and the Ris P2445B anemometer

the presented slopes of the regressions have been
evaluated from measurements with some shortcomings (see [4]). However, the ranking of the anemometers according to the regression slopes coincides very well with the differences in the cup anemometer measurements as observed in flat terrain
(compare Tab. 1 and Fig. 1). An exception is the
Metone 010C-1 anemometer which has been very
close to the Ris P2445B and the Vector A100
anemometer in the outdoor comparison while its
regression slope is relatively large.

0.975
0.95

not corrected
0.925

linear correction for std(W)/U from free sector

0.9
0.875
0.85
0

40

80

120

160

200

240

280

320

360

Direction []

Fig. 7: Ratios of measurements with a Thies 4.3303.22.000

and a Ris P4225B cup anemometer for the not corrected Thies anemometer and the Thies anemometer corrected for the effect of the vertical turbulence
intensity. The correction has been established from
the free sector 110-350.

Large deviations between measurements with

cup anemometers of different type occur even in
flat terrain and when the anemometers are calibrated uniformly in the same wind tunnel. The
deviations are in the range of some percent.
Wind turbine power curve measurements are strongly effected by the choice of the cup anemometer.
The observed differences in wind speed measurements are dependent on the site specific flow and
turbulence regime.
The vertical turbulence intensity seems to be
Instrument for Horizontal Wind Speed Instrument for Vector Wind Speed
the flow variable with the largest influence
Slope
Maximum Vertical
Class Slope
Maximum Vertical
Class
Turbulence Intensity
Turbulence Intensity
on the differences in the cup anemometer
Thies 4.3303.22.000
Thies 4.3303.22.000
measurements. Among other important flow
0,032
1 0,2215
0,045
1
0,315
0,063
2
0,090
2
variables seem to be the flow inclination, the
0,095
3
0,135
3
correlation between the lateral and the other
0,127
4
0,181
4
0,159
5
0,226
5
flow components and the integral turbulence
Risoe P2445b
Risoe P2445b
length scale of the vertical flow component.
0,196
1 -0,039
-0,255
1
0,0511
0,392
2
-0,511
2
Classifications of cup anemometers and cor0,588
3
-0,766
3
rections of wind speed measurements
0,784
4
-1,021
4
should account for the identified important
0,979
5
-1,276
5
Metone 010C-1 (limited validity)
flow and turbulence parameters.
0,042
1
0,2387
The Ris P2445B cup anemometer seems
0,084
2
0,126
3
to be nearly insensitive against turbulence
0,168
4
effects, at least in flat terrain conditions.
0,209
5
The Thies 4.3303.22.000 cup anemometer
Results from 8m mast (limited validity)
seems to be seriously influenced by the verVector A100 (limited validity)
0,158
1
0,0632
tical turbulence intensity. In very turbulent
0,317
2
flow the anemometer overestimates even
0,475
3
the vector wind speed significantly.
0,633
4
0,792
5
A simple correction algorithm for cup
Mierij (limited validity)
anemometer measurements in flat terrain
0,057
1
0,1743
0,115
2
conditions has been proposed. This correc0,172
3
tion, however, requires a good knowledge
0,230
4
0,287
5
about the vertical turbulence intensity.
Vaisala WAA151 (limited validity)
Although shortcomings of the correction are
0,036
1
0,2792
0,072
2
evident, its application can lead to more
0,107
3
accurate results of wind speed measure0,143
4
0,179
5
ments at least in flat terrain when anemomeFriedrichs 4033.1100x (limited validity)
ters with a high sensitivity on turbulence
0,027
1
0,3699
effects are in use.
0,054
2
0,081
3
A first attempt to classify cup anemometers
0,108
4
according to their sensitivity on the vertical
0,135
5
turbulence intensity, as it is determined by
Tab. 1: Classification of cup anemometers according to their
outdoor measurements, coincides well with
sensitivity on the vertical turbulence intensity

57

DEWI Magazin Nr. 19, August 2001

the observed differences between the cup anemometer measurements.

Future research should emphasise on the verification of the findings regarding the turbulence parameter analysis with other types of cup anemometers and at other locations, especially in complex
terrain. Simulations of turbulence effects on cup anemometers in wind tunnels require a good control over the here identified important turbulence measures.
7.

Acknowledgements

The work has been partially supported by the European Commission under contract number JOR3CT98-0257.
8.

References

[1]
[2]

MEASNET: Measurement Procedure Cup Anemometer Calibrations, 1997

A. Albers, H. Klug, D. Westermann; Outdoor Comparison of Cup Anemometers; Pro-ceedings of
DEWEK 2000
A. Albers, H. Klug, D. Westermann; Cup Anemometry in Wind Engineering, Struggle for Improvement,
DEWI Magazine no. 18, February 2001
A. Albers; Identification of Variables for Site Calibration and Power Curve Assessment in Complex
Terrain, JOR3-CT98-0257, Project Task 6, Relative Power Curve Measurements in Flat terrain, Final
Report, Report-No. JOR3-CT98-0257-0601-DEWI07, June 2001
F. Mouzakis, E. Morfiadakis, P Dellaportas; Parameter Identification on Power Performance of Wind
Turbines Operating at Complex Terrain, 2nd European & African Conference on Wind Engineering
Geneva, Italy, 1997
A. Albers; Comments on Techniques for Multi-Parameter Site Calibration, DEWI Report-No.: JOR3CT98-0257-1198-DEWI01, 1998

[3]
[4]
[5]
[6]

Problematik der Windben

Frank Bttcher, Christoph Renner, Hans-Peter Waldl und Joachim Peinke
Fachbereich Physik, Carl-von-Ossietzky Universitt Oldenburg
1.

Einleitung

Windenergieanlagen (WEA) sind im Betriebszustand unterschiedlichen Belastungen ausgesetzt, die

die Lebensdauer der Anlagen, vor allem durch die wechselnden (instationren) Belastungen, herabsetzen. Neben den deterministischen Anteilen (wie z.B. Schwer- und Kreiselkrfte) sind es insbesondere stochastische Lastanteile, die die mechanische Auslegung von Windkraftanlagen sehr komplex
werden lassen. [1] Die stochastischen mechanischen Belastungen haben ihre Ursache in der schwankenden Windgeschwindigkeit. Sie werden in der Regel in Windturbulenzen und in Windben untergliedert. Der Einflu der Windturbulenz wird oftmals durch den Turbulenzgrad

S=

(1)

wiedergegeben, der das Verhltnis der Standardabweichung des Windgeschwindigkeitsfeldes zu seiner mittleren Geschwindigkeit beschreibt. Die Frage nach der Natur von Windben ist trotz ihrer
scheinbaren Trivialitt bis heute nicht allgemein geklrt und eine strenge Definition des Begriffes findet
sich in der Literatur nicht [3]. Dabei ist die Kenntnis der maximal zu erwartenen Geschwindigkeit einer
Windbe besonders hinsichtlich der Abschtzung von Extrembelastungen von groem Interesse. hnlich relevant und auch schwierig ist die Einschtzung der Hufigkeit von Benereignissen und deren
zeitliche Abfolge [2]. Die Unterscheidung von Ben und Windturbulenzen wird in der Regel folgendermaen formuliert. Whrend die Windturbulenz u(t) als permanente, fluktuierende berlagerung der
mittleren Geschwindigkeit u(t) aufgefat werden kann, stellen Ben nach [3] eine erhebliche Abweichung von der mittleren Windgeschwindigkeit im Bereich von einigen bis einigen zehn Sekunden" dar.
Letztere werden demnach im Gegensatz zur Windturbulenz als einzelne und extreme Ereignisse charakterisiert.
Um zu untersuchen wie Windben vom statistischen Standpunkt aus erfat werden knnen, werden
wir im folgenden zwei Geschwindigkeitszeitreihen auf ihre statistischen Eigenschaften untersuchen:
Einen Freifeld-Windgeschwindigkeits- und einen Labor-Geschwindigkeitsdatensatz. In diesem Artikel
58