Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

ARTICLE IN PRESS

Journal of Wind Engineering


and Industrial Aerodynamics 92 (2004) 1265–1279
www.elsevier.com/locate/jweia

The starting and low wind speed behaviour of a


small horizontal axis wind turbine
A.K. Wright, D.H. Wood
School of Engineering, University of Newcastle, University Drive, Callaghan, NSW 2308, Australia
Received 17 February 2004; received in revised form 6 August 2004; accepted 12 August 2004
Available online 18 September 2004

Abstract

In order to extract the maximum possible power, it is important that the blades of
small wind turbines start rotating at the lowest possible wind speed. The starting performance
of a three-bladed, 2 m diameter horizontal axis wind turbine was measured in field tests,
and compared with calculations employing a quasi-steady blade element analysis. Accurate
predictions of rotor acceleration were made for a large range of wind speeds, using a
combination of interpolated aerofoil data and generic equations for lift and drag
at high angles of incidence. Also, significantly different values for the wind speeds at
which the turbine rotor starts and ceases to rotate were determined, indicating limitations in
the traditional method of describing starting performance with a single ‘cut-in’ wind speed
based on 10-min averages of wind speed and turbine power. The blade element calculations
suggest that most of the starting torque is generated near the hub, whereas most power-
producing torque comes from the tip region. The significance of these results for blade design
is discussed.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: Small wind turbine; Starting; Field test; Blade element method

Corresponding author. Tel.: +61-2-4921-6210; fax: +61-2-4921-7050.


E-mail address: andrew.k.wright@studentmail.newcastle.edu.au (A.K. Wright).

0167-6105/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2004.08.003
ARTICLE IN PRESS
1266 A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279

Nomenclature

Cl, Cd aerofoil lift and drag coefficients


I mass moment of inertia of the rotor (kgm2)
Ts starting time (s)
t time (s)
U wind speed (m/s)
Uc cut-in wind speed (m/s)
Us starting wind speed (m/s)
a angle of attack
O blade angular velocity (rpm)
Oa average blade angular velocity during a starting sequence (rpm)
Re Reynolds number

1. Introduction

Small wind turbines need to be affordable, reliable and almost maintenance free
for the average person to consider installing one. This often means a sacrifice of
optimal performance for simplicity in design and operation. Thus, rather than using
the generator as a motor to start and accelerate the rotor when the wind is strong
enough to begin producing power, small wind turbines rely solely on the torque
produced by the wind acting on the blades. Furthermore, small wind turbines
are often located where the generated power is required, and not necessarily where
the wind resource is best. In these low or unsteady wind conditions slow starting
reduces the total energy generated. Also, a stationary wind turbine fuels the
perception of wind energy as an unreliable energy source.
Only a small amount of published work on wind turbine starting has been
found. Ebert and Wood [1] and Mayer et al. [2] presented starting sequences from
experiments conducted using two separate machines but with the same 5 kW
rated blades. In both studies, the wind speed (U) was in the range 5–8 m/s, which
was considerably higher than Uc, the cut-in wind speed of the turbine, conventionally
defined as the lowest speed at which power is produced. Both references describe
30–50 s ‘idling’ periods of slow rotation, with little acceleration after the initial
start. In this condition the blades experience high angles of attack (a), are stalled,
and are generating only a small amount of torque. When a eventually decreases
sufficiently to produce high lift:drag ratios, the rotor accelerates rapidly. Very
few small turbines have pitch adjustment, so the angles of attack during starting
are usually high and unfavourable. The starting sequences for varying pitch in
Mayer et al. [2] showed that increasing pitch (i.e., reducing a) decreases the idling
period. The blades used presently are pitched by 51 relative to the blades of Ebert
and Wood [1].
ARTICLE IN PRESS
A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279 1267

Mayer et al. [2] simulated rotor acceleration with an adaptation of blade element
analysis, also used in our analysis and described shortly. Their set of experiments
included various pitch angles, and better agreement with rotor acceleration
predictions was found at higher pitch, and hence lower angles of attack. A
suggested reason for this was the large uncertainty in lift and drag values for high a;
and very low Reynolds number, Re. Also highlighted in their analysis was the need
to accurately characterise the resistive torque of the turbine. This comprises
the frictional torque in the gearbox and drivetrain as well as windage and
cogging torque of the generator. Clausen and Wood [3] observed that overcoming
resistive torque generally becomes more difficult as turbine size decreases,
which is one reason that ‘micro’ turbines used on yachts often have a large number
of blades.
Wood [4] estimated the conventional cut-in wind speed, Uc, of a 600 W and a 5 kW
turbine. He used blade element theory to predict the aerodynamic torque generated
on the stationary blades by assuming there was no kinetic energy absorbed from the
wind. Cut-in was assumed when the aerodynamic torque balanced the resistive
torque. He found that Uc was over-estimated by the calculations. It will be shown
that the actual starting behaviour of a turbine is significantly more complex than
assumed in Wood’s [4] analysis and that Uc is not necessarily a good measure of the
starting behaviour.
The main aim of this paper is to improve the analyses already referred to, and to
compare the results with measurements from a turbine whose resistive torque was
characterised much better than previously. In addition, the experiments cover a wide
range of wind speeds, from below cut-in to above rated speed. It will be shown that
the starting behaviour is remarkably well predicted by the blade element analysis
using a combination of interpolated aerofoil data and generic expressions for
aerofoil lift and drag at high angles of incidence.

2. Experiment

The three-bladed, 2 m diameter wind turbine (Fig. 1) had a permanent magnet


generator, and was designed to produce 600 W at a rotor speed (O) of 700 rpm and
U=10 m/s. The blade design for this turbine is described by Wood and Robotham
[5]. For testing, the turbine was mounted at a height of 8 m above the roof of a four
store building, with trees of similar height to the turbine in the near vicinity; pictures
of the test site can be found at: www.wind.newcastle.edu.au. The unsteadiness of the
wind at this site caused the turbine rotor to stop quite frequently during lulls in the
wind, and subsequently generate a starting sequence during a subsequent gust. U
was measured with a Synchrotac 710–1960 cup anemometer, and wind direction was
measured with a Met One Instruments 020C wind direction sensor. Both were
mounted 1.5 m below the turbine rotor axis with 1.2 m horizontal separation between
the anemometer and wind vane (Fig. 1). The uncertainty of the calibrated
anemometer is estimated to be about 0.1 m/s at 10 m/s, and the anemometer’s
dynamic response relative to the rotor is sufficiently rapid to have negligible effect on
ARTICLE IN PRESS
1268 A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279

Fig. 1. Field testing: the wind turbine, wind vane, and cup anemometer.

calculated results. O was measured to a precision of 2 rpm by an optical sensor and


reflective strips mounted on the rotating generator casing, with the sensor output
updated every 1/8th of a second. Turbine yaw direction was measured with an
optical encoder, and all measurements were sampled at 200 Hz.
The permanent magnet generator output was rectified and fed into a 24 V battery.
The battery placed no load on the generator until OE250 rpm was reached, at which
point the generator voltage was sufficient to begin charging the battery. We will
assume that the beginning of power extraction completes the starting sequence. This
arrangement, whereby the turbine does not generate power until the blades reach a
substantial angular velocity, is common for small turbines, and indicates the need to
distinguish between the Uc and the wind speed at which the blades start rotating. The
generator had a cogging torque of 0.36 Nm when stationary, and a constant resistive
torque of 0.24 Nm when rotating. These values were determined by winding the rotor
shaft with a line, attaching a series of known masses, and then recording the
acceleration of the rotor as the masses were dropped. The turbine’s rotor moment of
inertia (I) was 0.43 kgm2 with the main contribution coming from the blades, and
was determined at the same time as the resistive torque. This value of I was
consistent with that obtained from a finite element analysis of the blade for
structural design purposes. The major advantage of the present turbine over that
used by Ebert and Wood [1] and Mayer et al. [2] is the accurate characterisation of
the resistive torque and the moment of inertia.
ARTICLE IN PRESS
A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279 1269

3. Calculations

Small wind turbine starting involves three complex areas of aerodynamics;


unsteadiness, high angles of attack, and very low Reynolds numbers, but it is
important that the analysis be no more complicated than traditional blade element
theory used to determine the extracted power. Ebert and Wood [1] concluded that
the effects of unsteadiness were small by determining from their experiments
measures of unsteadiness such as pitch rates and reduced frequencies. Our present
turbine is significantly smaller (blade length of 1 m compared to 2.5 m), accelerates
more rapidly, and is hence more susceptible to unsteady effects. A model for
correcting aerofoil lift coefficients for dynamic stall is described by Rasmussen et al.
[6] in terms of da/dt and d2a/dt2. Unfortunately the model is only validated for
ao251, and the model constants are dependent on aerofoil type. Its application is of
little quantitative benefit, but rough comparison shows that dynamic effects at low a
are only significant for da/dt values higher than those found during our tests.
Unsteady effects were neglected and a quasi-steady analysis was used in a similar
manner to Mayer et al. [2]. Conventional blade element momentum analysis was
modified to assume no flow deceleration through the blades, and the net torque on
the rotor equated to the product of rotational moment of inertia and angular
acceleration. In other words, the aerodynamic torque is assumed to accelerate the
blades, rather than extract kinetic energy from the wind. The resulting ordinary
differential equation was integrated using the standard 4th order Runge–Kutta
method to determine the angular velocity as a function of time. Net torque is the
difference between the calculated aerodynamic torque and the cogging torque of
0.36 Nm or 0.24 Nm depending on whether the rotor is stationary or rotating.
The results of Mayer et al. [2] highlighted the uncertainty in characterising lift and
drag characteristics of aerofoils at high a and low Re, with little data available at
appropriate Reynolds numbers. A renewed attempt was made at constructing lift
and drag curves from existing published data. For our aerofoil (SD7062) the
published lift and drag data (Lyon et al. [7]) is restricted to ao201 and Re X6  104.
This data was combined with the results of Ostowari and Naik [8] for the NACA
4412 for 201oao1801, at Re=2.5  105, the lowest value tested. However the
applicability of this composite lift and drag data was questionable, as during a
typical start Re varies with time and radial location between about 104 and 105. In
this range, aerofoil performance at low a is significantly affected by the behaviour of
a flow separation bubble that covers much of the aerofoil surface, e.g., Refs. [9,10],
so lift and drag prediction is difficult and very sensitive to the flow conditions.
A simple alternative to the composite lift and drag data was considered. Following
Meyer and Kroger [11], Wood [4] assumed that lift and drag coefficients (Cl and Cd
respectively) for any aerofoil at high a can be approximated by the ‘flat plate’
equations
C l ¼ 2sin a cos a (1)

C d ¼ 2sin2 a: (2)
ARTICLE IN PRESS
1270 A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279

2.5
(a) : aerofoil data
2
lift / drag coefficent

(b) : Eqs. (1) & (2) Cd


1.5

1
Cl
0.5

0
0 15 30 45 60 75 90
angle of attack (deg)

Fig. 2. Lift and drag coefficient: composite aerofoil data and Eqs. (1) and (2).

Wood [4] showed these equations to be reasonably good fits to data from three sets
of aerofoils for a4451, which is the range of most interest for starting. Eqs. (1) and
(2) underestimate typical high lift aerofoil behaviour for ao451, so we would expect
better rotor performance than predicted when a is in this range. Eqs. (1) and (2) have
been used in a second set of predictions, for which rotor behaviour depends only on
the pitch and twist and chord distribution of the blade, but not on the aerofoil profile
of the blade sections. Calculated rotor speeds using the composite aerofoil lift and
drag data shall be designated (a), predictions using Eqs. (1) and (2) designated (b),
and a third set of predictions, found by averaging the Cl and Cd values of the two
methods designated (c). A comparison of the Cl and Cd curves is shown in Fig. 2.

4. Results and discussion

4.1. Starting wind speed

Field test data were analysed to determine the wind speed at which the blades
first rotate (Us). Fig. 3 shows 506 values of Us plotted against dU/dt. Us is
the instantaneous wind speed when rotor motion is first detected, and dU/dt is the
rate of change of wind speed calculated over a 2 s period centred on the start time.
The vast majority of starts occur when U is increasing, which is not unexpected, but
other than this Us shows no obvious correlation with dU/dt. Similarly, no correlation
with turbine yaw or yaw rate was observed. It was assumed that measured
instantaneous values (typically during a gust) accurately represent the actual
conditions at the rotor, however the somewhat limited response of the rotor speed
sensor, anemometer and wind vane possibly obscures any effect of dU/dt, yaw angle
and yaw rate. The average Us was 4.6 m/s, compared to a predicted value of 4.7 m/s
using the composite aerofoil data, and 5.6 m/s using Eqs. (1) and (2). It is important
ARTICLE IN PRESS
A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279 1271

3.5

3.0

2.5

2.0

1.5
dU/dt

1.0

0.5

0.0

-0.5

-1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
starting wind speed, Us (m/s)

Fig. 3. Measured starting wind speed (Us), and the rate of change of wind speed (dU/dt) when the rotor
first begins to rotate.

to note that once the blades have started rotating, there can still be a considerable
time before they reach the O at which power is first produced, which is usually set by
the control system and is independent of wind speed. The complex relationship
between Us and Uc can only be understood by considering the full starting sequences.

4.2. Starting sequences

Data sets in which the rotor accelerated from rest up to 250 rpm, which we define
as a successful ‘start’, were selected from about 70 h of field test data, yielding 160
starting sequences. The average start time, Ts, was 28 s, the minimum length 9 s and
the maximum 101 s. Fig. 4 shows that Ts depends strongly on the wind speed.
Predicted rotor speed calculations were produced for each starting sequence,
with two starts selected to illustrate some of the important features of a high and a
low U start.

4.3. High wind start

Fig. 5 illustrates data from a start during wind speeds exceeding the turbine’s rated
speed, and predicted rotor speeds that are typical of the results in general. During
this start, the average yaw error, the angle between the rotor axis and the wind
direction, was 131. In this case and in general, yaw behaviour during the starting
sequences had no obvious effect on the accuracy of predicted rotor speed, so no
attempt was made to correct for yaw error.
The predicted rotor speed (a) using the composite aerofoil data over-estimates the
rotor acceleration. The Re during the start (Fig. 6) is significantly lower than the Re
of the aerofoil data for high angles of attack used in the interpolation (Re=250 000),
ARTICLE IN PRESS
1272 A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279

120

100

80
start length (s)

60

40

20

0
3 4 5 6 7 8 9 10
average wind speed (m/s)

Fig. 4. Starting time and the average wind speed of the sequence.

14 winds peed 250

actual rpm
12 predicted rpm (a) 200
predicted rpm (b)

rotor speed (rpm)


wind speed (m/s)

predicted rpm (c)


10 150

8 100

6 50

4 0
0 2 4 6 8 10
time (s)

Fig. 5. High wind speed starting sequence.

particularly at the blade tip. Poorer performance at lower Re is not unexpected.


Also, a reduces rapidly during this short starting sequence (Fig. 7), and any dynamic
stall effects would reduce the actual rotor acceleration. However, since over-
estimation of rotor acceleration was not limited to starts with rapid changes in a;
inadequate aerofoil data at high a and low Re is probably the largest contributing
factor to the difference.
Conversely, predicted rotor speed (b) using generic Eqs. (1) and (2) under-
estimates the rotor acceleration, while prediction (c), using average Cl and Cd data
from the two methods results in an accurate estimate of rotor acceleration.
ARTICLE IN PRESS
A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279 1273

1.25E+05

100000
Reynolds number

75000

50000

25000 blade root

blade tip
0
0 2 4 6 8 10
time (s)

Fig. 6. Prediction (c) Reynolds number variation for high wind speed starting sequence.

90
blade root
80

blade tip
70
angle of attack (deg)

60

50

40

30

20

10
0 2 4 6 8 10
time (s)

Fig. 7. Prediction (c) angle of attack variation for high wind speed starting sequence.

Fig. 8 shows that during the initial stages of the start, most of the torque is
generated by the blade root (the first of 20 blade elements, 36.75 mm in width,
beginning 0.24 m from the rotor axis). At about t=6 s, a at the blade tip drops below
451 as the rotor speed increases above 100 rpm, and the proportion of torque
generated at the blade tip begins to rise.

4.4. Low wind start

Fig. 9 shows a starting sequence at low U for which Ts is much longer than at
higher U. The turbine starts in a gust of about 6 m/s, but the rotor idles at low speed
ARTICLE IN PRESS
1274 A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279

0.2
blade root
0.175

blade tip
0.15
torque (Nm)

0.125

0.1

0.075

0.05

0.025

0
0 2 4 6 8 10
time (s)

Fig. 8. Prediction (c) torque generated by inner and outer blade elements for high wind speed start.

10 wind speed 250

actual rpm
predicted rpm (a) 200
8
predicted rpm (b)

rotor speed (rpm)


wind speed (m/s)

predicted rpm (c)


6 150

4 100

2 50

0 0
0 5 10 15 20 25 30 35 40 45 50
time (s)

Fig. 9. Low wind speed starting sequence.

for the next 25 s as U varies between 3 and 6 m/s. This is an example of the idling
period referred to in the Introduction. Eventually, a 6 m/s gust at t=30 s accelerates
the rotor beyond 250 rpm, despite a subsequent drop in U to below 5 m/s. The idling
period is relatively short in duration, and only evident in low wind starts. The 5 kW
turbines of Ebert and Wood [1] and Mayer et al. [2] showed a similar or greater
idling time even during high wind starts. The present turbine’s increased blade pitch
and lower I are the probable explanations for the reduction. The identification of the
idling period is helped by the analysis of Wood [4] who showed that the aerodynamic
ARTICLE IN PRESS
A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279 1275

torque and hence angular acceleration of the stationary blades depends on U2, see
Eq. (9), with the constant of proportionality dependent on the blade geometry,
including the pitch. Thus O is initially linear in t for constant wind speed. The idling
period begins when either the blade geometry and angular velocity cause the torque
to reduce from this value, or if U decreases as the blades accelerate. We note that if
the aerodynamic torque for a given blade depends only on U2 and O; as required by
the quasi-steady analysis using Eqs. (1) and (2), then Ts U2 which correlates the
data in Fig. 4.
Actual rotor speed is again bounded by the predictions of methods (a) and (b),
with the average (c) giving good agreement for most of the sequence. Prediction (b)
predicts quite accurately the initial stage of the starting sequence, however rotor
speed fails to rise out of the high a idling period in such a low wind speed. During the
idling period the Re is very low (Fig. 10), a is higher than 401 (Fig. 11), and the hub
region generates more torque than the blade tip (Fig. 12).
The accuracy of the blade element calculations in Figs. 5 and 9, particularly
prediction (c), suggests that the calculated torque variation along the blade shown in
Figs. 8 and 12, indicating that the hub region generates most of the low-speed
torque, is genuine. Since the tip region is responsible for the power-producing
torque, we conclude that blades may be optimised for starting performance through
the shape (pitch, twist, and chord) of the hub region and for maximum power
through the shape and aerofoil section for the tip.

4.5. General low wind speed behaviour

It is reasonable to conclude from the results presented so far that dO/dt for
Oo250 rpm can be calculated quite accurately using the quasi-steady analysis and by

60000

50000
Reynolds number

40000

30000

20000

blade root
10000
blade tip
0
0 5 10 15 20 25 30 35 40 45 50
time (s)

Fig. 10. Prediction (c) Reynolds number variation for low wind speed start.
ARTICLE IN PRESS
1276 A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279

90

80
blade root

70 blade tip
angle of attack (deg)

60

50

40

30

20

10

0
0 5 10 15 20 25 30 35 40 45 50
time (s)
Fig. 11. Prediction (c) angle of attack variation for low wind speed start.

0.05
blade root

0.04 blade tip


torque (Nm)

0.03

0.02

0.01

0
0 5 10 15 20 25 30 35 40 45 50
time (s)

Fig. 12. Prediction (c) torque generated by inner and outer blade elements for low wind speed start.

averaging the composite aerofoil data with the generic lift and drag Eqs. (1) and (2).
It follows that for each U and O, it is possible to determine whether the rotor is
accelerating, steady, or decelerating. This was tested using a matrix comprised of a
range of initial O(0–250 rpm), and a range of constant U (0-10 m/s). The subsequent
O was then simulated for 15 s at constant U. If the resulting O was higher than
250 rpm the simulation was deemed invalid, otherwise the average dO/dt for the time
ARTICLE IN PRESS
A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279 1277

500
steady
450
accelerating
400
decelerating
350
rotor speed (rpm)

300

250

200
(ii)
150
(iii)
100

50
(i)
0
2 3 4 5 6 7 8
wind speed (m/s)

Fig. 13. Turbine low wind speed behaviour.

period was calculated and scaled by dividing by the average rotor speed Oa : Each
calculated Oa value was allocated to one of three categories:
(i) dO/dt/Oao0.02 decelerating
(ii) 0.02odO/dt/Oao0.02 steady
(iii) dO/dt/Oa40.02 accelerating

The solid lines in Fig. 13 are the calculated boundaries between the U and O states:
(i) decelerating, (ii) steady, and (iii) accelerating rotor. It must be noted that the
division between the steady and unsteady zones is somewhat arbitrary.
For comparison, field test data were processed to produce 15 s data sets of U and
O for which U varied by less than 20%. Using the categorisation described in the
previous paragraph, O in each data set was identified as steady, accelerating or
decelerating and represented by an appropriate symbol in Fig. 13. The ‘steady’ data
in Fig. 13 forms the upper limit of O at any U. There is a clear relationship between
the maximum O and U, which is fixed by the particular control system, a load
provided by constant battery voltage. Below this line the majority of data points are
in their expected zone, (i), (ii) or (ii). The starting wind speed Us is clearly between 4
and 5 m/s, but the ‘cut-out’ wind speed is possibly 2.5 m/s or lower. The phenomenon
of a relatively low U and high O producing a low a and better than predicted
performance is a possible explanation of the grouping of ‘accelerating’ points
between 150–250 rpm and 3–4 m/s.
Fig. 14 is a simplified version of Fig. 13, showing the shape of the calculated steady
rotor speed curve of the current wind turbine at low U. It highlights the importance
of reducing the Us (5.1 m/s in this case) to maximise power capture in lighter winds.
ARTICLE IN PRESS
1278 A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279

500

400
rotor s peed (rpm)

area of rotor deceleration

300

200
area of rotor acceleration

100

0
2 3 4 5 6 7 8
wind speed (m/s)

Fig. 14. Predicted steady rotor speed curve.

Note that the flat line between 4.2 and 5.1 m/s represents the difference of 0.12 Nm
between the static (0.36 Nm) and dynamic (0.24 Nm) cogging torque, and indicates
the high sensitivity of starting performance to cogging torque.
In practice, Uc is determined, as is every other point on the turbine’s power curve,
from 10-min averages of the U and turbine power output. The results in Figs. 13 and
14 imply that Uc can be significantly less than the starting wind speed, Us, in
contradiction of the assumptions used in Wood [4]. A more realistic, but still simple
assessment of a turbine’s starting capabilities would be to determine Us (=5.1 m/s
from Fig. 14, or 4.6 m/s from the data, Fig. 3) and the ‘cut-out’ wind speed as the
lowest U at which power is produced during an interval comparable to the starting
time. For the present turbine the cut-out speed is about 2.5 m/s. It is likely that Uc
will lie between cut-out and Us, and may well be site-dependent.

5. Conclusions

Data from the field testing of a three-bladed 600 W wind turbine was analysed to
identify various characteristics of its starting performance. The blades started
rotating at a wind speed of 4.6 m/s on average, but this varied between 2.5 and 7 m/s,
and generally coincided with increasing wind speed. Predictions of rotor acceleration
up to a power producing level were made using a quasi-steady adaptation of blade
element analysis employing composite lift and drag data and also generic equations
for high angles of attack. Given the uncertainty associated with lift and drag data at
high angles of attack and low Reynolds numbers, these predictions compared well
with 160 measured occurrences of rotor acceleration over a large range of wind
speeds. The acceleration and deceleration of the rotor at speeds below its controlled
maximum speed, and for a range of wind speeds was calculated and compared with
data. Excellent agreement was found.
This method of analysis of the low wind speed performance of a small wind
turbine has potential as a relatively simple design tool. The simplicity of the generic
ARTICLE IN PRESS
A.K. Wright, D.H. Wood / J. Wind Eng. Ind. Aerodyn. 92 (2004) 1265–1279 1279

equations for lift and drag has particular appeal, and is probably a good starting
point for lift and drag approximation for many blades. Starting is not usually the
primary concern of the designer, however a simple method of predicting a turbine’s
starting performance is useful, particularly if siting turbines in low or unsteady
winds. The method described and tested here is suitable for this purpose.
Furthermore, most starting torque is generated near the hub, and most power-
extracting torque comes from the tip region, so that it should be possible to optimise
starting performance while maintaining good power performance.

Acknowledgements

This work was performed with the support of a University of Newcastle Research
Scholarship and the Australian Co-operative Research Centre for Renewable
Energy. Equipment was purchased with the assistance of a University Research
Management Committee grant. The authors acknowledge Ian Miller’s significant
contribution to the wind turbine instrumentation and data acquisition.

References

[1] P.R. Ebert, D.H. Wood, Observations of the starting behaviour of a small horizontal-axis wind
turbine, Renew. Energy 12 (3) (1997) 245–257.
[2] C. Mayer, M.E. Bechly, M. Hampsey, D.H. Wood, The starting behaviour of a small horizontal-axis
wind turbine, Renew. Energy 22 (2001) 411–417.
[3] P.D. Clausen, D.H. Wood, Recent advances in small wind turbine technology, Wind Eng. 24 (3)
(2000) 189–201.
[4] D.H. Wood, A blade element estimation of the cut-in wind speed of a small turbine, Wind Eng. 25 (4)
(2001) 249–255.
[5] D.H. Wood, T. Robotham, Design and testing of high performance blades for a 600 watt horizontal-
axis wind turbine, Proceedings of the first Australian Wind Energy Conference, Newcastle, June
28–30, 1999, pp. 91–96.
[6] F. Rasmussen, J.T. Petersen, H.A. Madsen, Dynamic stall and aerodynamic damping, J. Solar
Energy Eng. 121 (1999) 150–155.
[7] C.A. Lyon, A.P. Broeren, P. Giguere, A. Gopalarathnam, M.S. Selig, Summary of Low-Speed Airfoil
Data, Soartech Publications, Virginia Beach, 1997.
[8] C. Ostowari, D. Naik, Post stall studies of untwisted varying aspect ratio blades with an NACA 4415
airfoil section—part 1, Wind Eng. 8 (3) (1984) 176–194.
[9] H. Matsumiya, T. Kogaki, K. Kieda, M. Iida, K. Waseda, On the nature of a flow field around an
airfoil with a separation bubble, numerical and experimental approaches, Proceedings of the
European Wind Energy Conference, Copenhagen Denmark, 2001, pp. 510–513.
[10] T.J. Mueller, J.D. DeLaurier, Aerodynamics of small vehicles, Ann. Rev. Fluid Mech. 35 (2003)
89–111.
[11] C.J. Meyer, D.G. Kroger, Numerical simulation of the flow field in the vicinity of an axial flow fan,
Int. J. Numer. Meth. in Fluid. 36 (2001) 947–969.

You might also like