IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO.

3, SEPTEMBER 2006 717
Simulation Model of Wind Turbine 3p Torque
Oscillations due to Wind Shear and Tower Shadow
Dale S. L. Dolan, Student Member, IEEE, and Peter W. Lehn, Senior Member, IEEE
Abstract—To determine the control structures and possible
power quality issues, the dynamic torque generated by the blades
of a wind turbine must be represented. This paper presents an
analytical formulation of the generated aerodynamic torque of a
three-bladed wind turbine including the effects of wind shear and
tower shadow. The comprehensive model includes turbine-specific
parameters such as radius, height, and tower dimensions, as well
as the site-specific parameter, the wind shear exponent. The model
proves the existence of a 3p pulsation due to wind shear and ex-
plains why it cannot be easily identified in field measurements. The
proportionality constant between the torque and the wind speed is
determined allowing direct aerodynamic torque calculation from
an equivalent wind speed. It is shown that the tower shadow effect
is more dominant than the wind shear effect in determining the dy-
namic torque, although there is a small dc reduction in the torque
oscillation due to wind shear. The model is suitable for real-time
wind turbine simulation or other time domain simulation of wind
turbines in power systems.
Index Terms—Real-time digital simulation, simulation model,
torque oscillations, tower shadow, wind shear, wind turbine.
I. INTRODUCTION
T
ORQUE and power generated by a wind turbine is much
more variable than that produced by more conventional
generators. The sources of these power fluctuations are due
both to stochastic processes that determine the wind speeds at
different times and heights, and to periodic processes. These
periodic processes are largely due to two effects termed wind
shear and tower shadow. The term wind shear is used to de-
scribe the variation of wind speed with height while the term
tower shadow describes the redirection of wind due to the tower
structure. In three-bladed turbines, the most common [1] and
largest [2] periodic power pulsations occur at what is known as
a 3p frequency. This is three times the rotor frequency, or the
same frequency at which the blades pass by the tower. Thus,
even for a constant wind speed at a particular height, a tur-
bine blade would encounter variable wind as it rotates. Torque
pulsations and, therefore power pulsations, are observed due to
the periodic variations of wind speed experienced at different
locations.
Torque oscillations have been noted in several studies. It has
been stated that maximum torque and power were noted when
any individual blade was positioned directly downwards [3],
although Thiringer [1] was unable to certify the dependence of
Manuscript received March 31, 2005. Paper no. TEC-00137-2005.
The authors are with the Department of Electrical and Computer En-
gineering, University of Toronto, Toronto, ON M5S 3G4, Canada (e-mail:
dale.dolan@utoronto.ca; lehn@ecf.utoronto.ca).
Digital Object Identifier 10.1109/TEC.2006.874211
the oscillation on wind shear. It is believed that tower shadow
is also a source of the 3p oscillations observed in wind turbines
although studies [1], [2] are unable to confirm this.
The 3p oscillations are important to model since they could
have wide ranging effects on control systems and power quality.
In systems connected directly to the grid, these torque oscil-
lations would be important to model in terms of grid power
quality. For systems interfaced to the grid through converters,
these torque oscillations would be more important in terms of
converter control. The torque oscillation model would be useful
in studying these effects via a wind turbine simulator or other
dynamic wind turbine modeling tools. Dynamic wind turbine
models are needed to interface with current power system simu-
lation tools like EMTP or PSCAD/EMTDC[4]. Existing models
use either a simple aerodynamic torque representation, or are
excessively complicated and not viable for incorporation into
EMTP-type simulation tools [5].
Several turbine simulators have been created to model the
wind turbine shaft in laboratory studies. Some simulators are
capable of dynamic simulations [6]–[8] while others are only
capable of performing steady-state simulations [9]. The sim-
ulator may only emulate the elements incorporated into the
model. The simplest and most common approach is to use a ba-
sic steady-state torque equation to calculate wind power and use
this to determine the acceleration on the turbine inertia [9]–[11].
Many of the lab simulators reviewed [8]–[13] did not include
the effects of wind shear or tower shadow, making these simu-
lators unsuitable for studying issues that may arise due to these
effects.
In recent literature, dynamic models of wind turbines have
been used where aerodynamic torque was either represented by
steady-state torque curves [14], [15] or by simple sinusoidal os-
cillations [16]. This paper develops a more complete model of
the wind turbine. The formulation involves a torque model for
the three-bladed turbine that includes the effects of wind shear,
and tower shadow. A pragmatic model appropriate for dynamic
wind turbine modeling tools is not available elsewhere that in-
corporates these effects. The formulation will combine and build
upon previous work to develop such a model. Suitable models
for wind shear and tower shadow will be presented that will be
put into a form from which a total wind field over the entire
rotor area may be determined. A method [17] for converting a
wind field into one equivalent wind speed will then be briefly re-
viewed. An equivalent wind speed including contributions from
the hub height wind speed, wind shear and tower shadowwill be
calculated. Finally, a completed normalized torque model will
be presented that is suitable for implementation in a real-time
wind turbine simulator or other time domain simulation.
0885-8969/$20.00 © 2006 IEEE
718 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 3, SEPTEMBER 2006
Fig. 1. Comparison of torque oscillation due to wind shear alone depending
on form of wind shear approximation.
II. WIND SHEAR
Wind speed generally increases with height and this variation
is termed wind shear. Torque pulsations, and therefore power
pulsations, are observed due to the periodic variations of wind
speed seen at different heights. Power and torque oscillate due
to the different wind conditions encountered by each blade as
it rotates through a complete cycle [3]. For instance, a blade
pointing upwards would encounter wind speeds greater than
a blade pointing downwards. During each rotation, the torque
oscillates three times because of each blade passing through
minimum and maximum wind.
It is therefore important to model these wind-shear-induced
3p torque pulsations when studying a wind turbine system. A
common wind shear model, shown as (1), is taken directly from
the literature on wind turbine dynamics [1], [18], [19].
V (z) = V
H
_
z
H
_
α
(1)
For the purpose of this analysis, (1) is converted to a function
of r (radial distance from rotor axis) and θ (azimuthal angle)
giving the following:
V (r, θ) = V
H
_
r cos θ + H
H
_
α
= V
H
[1 + W
s
(r, θ)] (2)
where V
H
is the wind speed at hub height, r is the radial distance
from rotor axis, W
s
is the wind-shear-shape function [18], α is
the empirical wind shear exponent, H is the elevation of rotor
hub, and z is the elevation above ground. The term W
s
(r, θ) is
the disturbance seen in wind speed due to wind shear that is
added to hub height wind speed.
Both Spera [18] and Thresher [19] approximated W
s
(r, θ)
by the second-order-truncated Taylor series expansion shown
as follows:
W
s
(r, θ) ≈ α
_
r
H
_
cos θ +
α(α −1)
2
_
r
H
_
2
cos
2
θ (3)
However, as shown in Fig. 1, the truncated expansion of (3)
eliminates, in three-bladed turbines, the torque oscillations due
to the wind shear when the contributions fromeach of the blades
are summed. This is because when the three blade contributions
are summed, the cos θ term yields a zero contribution while the
cos
2
θ termcontributes only a dc component that adjusts average
wind speed from hub height wind speed to spatial mean wind
speed. This can be seen in Fig. 1, where the resulting torque from
the second order approximation becomes constant, completely
losing the properties of the nonlinear wind shear expression. To
effectively model the 3p effect of wind shear, a cos
3
θ term is
necessary, requiring a third-order-truncated Taylor expansion.
Therefore, to model torque oscillations from wind shear, the
approximation used for W
s
(r, θ) should be as follows.
W
s
(r, θ) ≈ α
_
r
H
_
cos θ +
α(α −1)
2
_
r
H
_
2
cos
2
θ
+
α(α −1)(α −2)
6
_
r
H
_
3
cos
3
θ (4)
III. TOWER SHADOW
The distribution of wind is altered by the presence of the
tower. For upwind rotors, the wind directly in front of the tower
is redirected and thereby reduces the torque at each blade when
in front of the tower. This effect is called tower shadow. The
torque pulsations due to tower shadoware most significant when
a turbine has blades downwind of the tower and wind is blocked
as opposed to redirected [20]. For this reason, the majority
of modern wind turbines have upwind rotors. This paper will
therefore only deal with the tower shadow torque oscillations
in horizontal axis three-bladed upwind rotors. This section will
show theoretically the 3p oscillations caused by tower shadow.
The wind field, only considering tower shadow, is defined as
in (5), where V
H
= hub height wind speed. The termυ
tower
(y, x)
is the disturbance observed in the wind speed due to the tower
shadow that is added to hub height wind speed. Sorensen [17]
modeled tower disturbance using potential flow theory for wind
movement around the tower. Using the reference frames shown
in Fig. 2 yields (6).
V (y, x) = V
H
+ υ
tower
(y, x) (5)
υ
tower
(y, x) = V
0
a
2
y
2
−x
2
(x
2
+ y
2
)
2
(6)
In (6), V
0
is the spatial mean wind speed, a is the tower radius,
y is the lateral distance from the blade to the tower midline, and
x is the distance from the blade origin to the tower midline.
Results for tower radius of 2 mand four different longitudinal
distances between the tower and the blades are shown in Fig. 3.
It can be seen that as expected, the tower shadow effect is
more pronounced when the blades are closer (x smaller) to the
tower.
An alternate tower-shadow-deficit model (7) is developed
in [21] and shown as follows:
υ
tower
(y, x) = −V
0
D

x
(x
2
+ y
2
)
(7)
where D is the tower diameter, y is the lateral distance from
the blade to the tower midline, x is the distance from the blade
DOLAN AND LEHN: SIMULATION MODEL OF WIND TURBINE 3P TORQUE OSCILLATIONS DUE TO WIND SHEAR AND TOWER SHADOW 719
Fig. 2. Dimensions used in tower shadow formula.
Fig. 3. Comparison of tower shadow model (6) with different distances be-
tween the tower and the blades.
origin to the tower midline, and V
0
is the spatial mean wind
speed. Results for this alternate model with a tower radius of
2 m and four different longitudinal distances between the tower
and the blades are shown in Fig. 4.
Comparison of the two models graphically shows that a more
reasonable model is represented by (6), as it models both the
deceleration of the wind flow in front of the tower and the ac-
celeration of the wind flow on each side of the tower. Therefore,
for modeling torque oscillations due to tower shadow, (6) is
preferable and will be used in subsequent model development.
Different reference wind speeds are used in models for the
disturbance due to wind shear and tower shadow. The wind shear
model uses hub height wind speed (V
H
) while the tower shadow
model uses spatial mean wind speed (V
0
). The relationship
between these two wind speeds is formulated in Appendix and
Fig. 4. Comparison of tower shadow model (7) with different distances be-
tween the tower and the blades.
Fig. 5. Variation of m = V
0
/V
H
with α for different R/H ratios.
is summarized in Fig. 5. Most often for time-domain simulation,
only a single wind speed value, V
H
, is available. V
0
would
require calculation from an entire spatial wind field that would
normally be unavailable. Therefore, for all practical purposes, in
the torque oscillation model, tower disturbance will be expressed
in terms of V
H
. Converting (6) from a function of y (lateral
distance) to a function of r (radial distance) and θ (azimuthal
angle) normalized to V
H
yields as follows:
˜ υ
tower
(r, θ, x) = ma
2
r
2
sin
2
(θ) −x
2
(r
2
sin
2
(θ) + x
2
)
2
(8)
where a is the tower radius, r is the radial distance from the
blade to the hub center, θ is the azimuthal angle of the blade,
x is the distance from the blade origin to the tower midline,
and m = [1 +
α(α−1)(R
2
)
8H
2
] as developed in Appendix. It should
be noted that (8) is only valid for 90

≤ θ ≤ 270

as above the
horizontal, tower shadow effects should obviously be absent.
720 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 3, SEPTEMBER 2006
Fig. 6. Comparison of tower shadow at different radii based on a tower with
1.7-m diameter and blades 2.9 m from tower midline.
Fig. 6 shows the variation in the effective tower shadow angle
experienced by different blade elements at varying radial dis-
tances. It is observed that the blade elements closer to the hub
experience tower shadow for a longer period, although the same
wind deficit is seen for all blade elements at an angle of 180

.
IV. DETERMINATION OF TOTAL WIND FIELD—v(t, r, θ)
To determine the total wind field, the results of (4) and (8)
from Sections II and III are combined. The total wind field due
to both tower shadow and wind shear is given as follows.
v(t, r, θ) = V
H
(t)[1 + W
s
(r, θ)][1 + ˜ υ
tower
(r, θ, x)] (9)
v(t, r, θ) = V
H
(t)[1 + W
s
(r, θ) + ˜ υ
tower
(r, θ, x)
+ W
s
(r, θ)˜ υ
tower
(r, θ, x)]. (10)
As W
s
(r, θ)˜ υ
tower
(r, θ, x) would be small compared to other
terms, (11) is a valid approximation of (9). This approach is also
supported in the literature [19].
v(t, r, θ) ≈ V
H
(t)[1 + W
s
(r, θ) + ˜ υ
tower
(r, θ, x)] (11)
The spatially varying wind speed can be calculated using the
total wind field model of (11) or its expanded version as follows.
v(t, r, θ) ≈V
H
(t)
_
1 +α
_
r
H
_
cos θ +
α(α −1)
2
_
r
H
_
2
cos
2
θ
+
α(α−1)(α−2)
6
_
r
H
_
3
cos
3
θ
+
ma
2
(r
2
sin
2
θ −x
2
)
(r
2
sin
2
θ +x
2
)
2
_
. (12)
This total wind field model allows one to determine the wind
speed observed at any particular location in the rotor disk area,
knowing only the turbine parameters, wind shear coefficient and
a single hub height wind speed.
V. EQUIVALENT WIND SPEED FORMULATION BASED ON
EQUIVALENT TORQUE
An effective method for formulating an “equivalent wind
speed” has been developed by Sorensen [17]. The equivalent
wind speed is a representation of the actual spatially varying
wind speed that is defined such that it will give the same aerody-
namic torque. The advantage of this method is that a wind speed
without radial dependence may be used. For clarity and com-
pleteness, Sorensen’s approach is briefly outlined in this section.
The aerodynamic torque produced by a three-bladed wind
turbine immersed in a wind field v(t, r, θ) is given as follows:
T
ae
(t, θ) = 3M(V
0
) +
3

b=1
_
R
r
0
ψ(r)[v(t, r, θ
b
) −V
0
] dr (13)
where T
ae
(t, θ) is the aerodynamic torque, M(V
0
) is the steady-
state blade root moment resulting from spatial mean wind speed
V
0
, R is the radius of the rotor disk, r
0
is the radius at which
blade profile begins, and ψ(r) is the influence coefficient of
the aerodynamic load on the blade root moment. This equation
has been determined through linearization of individual blade
torque dependence on wind speed [17].
An equivalent wind speed v
eq
(t, θ) that does not vary with
the radius is defined which would give the same aerodynamic
torque as the actual spatially varying wind speed. This v
eq
must
be such that
T
ae
(t, θ) = 3M(V
0
) +
3

b=1
_
R
r
0
ψ(r)[v
eq
(t, θ) −V
0
] dr. (14)
Sorensen determined (15) to be the expression for equivalent
wind speed by equating (13) and (14)
v
eq
(t, θ) =
1
3
3

b=1
_
R
r
0
ψ(r)v(t, r, θ
b
)dr
_
R
r
0
ψ(r) dr
(15)
VI. DETERMINATION OF EQUIVALENT WIND SPEED
The total wind field including tower shadow and wind shear
effects will now be converted into one equivalent wind speed.
Three components of this equivalent wind speed will be sep-
arated and solved individually such that the effects from the
hub height wind speed, wind shear, and tower shadow may be
observed separately.
Assuming ψ(r) = kr, and defining n =
r
0
R
and s = 1 −n
2
,
the total wind field (11) may be inserted into (15) to yield (16)
after some initial simplification.
v
eq
(t, θ) =
2V
H
3sR
2
3

b=1
_
R
r
0
_
r +
r
2
α
H
cos θ
b
+
r
3
α(α −1)
2H
2
cos
2
θ
b
+
r
4
α(α−1)(α−2)
6H
3
cos
3
θ
b
+
ma
2
(r
3
sin
2
θ
b
−rx
2
)
(r
2
sin
2
θ
b
+x
2
)
2
_
dr. (16)
This equivalent wind speed will have three components. The
first (v
eq
0
) is due to the hub height wind speed, the second
(v
eq
ws
) is due to the wind shear, and the third (v
eq
ts
) is due to
DOLAN AND LEHN: SIMULATION MODEL OF WIND TURBINE 3P TORQUE OSCILLATIONS DUE TO WIND SHEAR AND TOWER SHADOW 721
Fig. 7. Normalized equivalent wind speed due to tower shadow (v
eq
ts
+
v
eq
0
), wind shear (v
eq
ws
+ v
eq
0
), and combination of wind shear and tower
shadow (v
eq
0
+ v
eq
ts
+ v
eq
ws
).
the tower shadow. Therefore, (16) can be decomposed as (17)
whose components are shown as (18)–(20).
v
eq
(t, θ) = v
eq
0
+ v
eq
ws
+ v
eq
ts
(17)
v
eq
0
=
2V
H
3sR
2
3

b=1
_
R
r
0
[r] dr (18)
v
eq
ws
=
2V
H
3sR
2
3

b=1
_
R
r
0
_
r
2
α
H
cos θ
b
+
r
3
α(α −1)
2H
2
cos
2
θ
b
+
r
4
α(α −1)(α −2)
6H
3
cos
3
θ
b
_
dr (19)
v
eq
ts
=
2V
H
3sR
2
3

b=1
_
R
r
0
_
ma
2
(r
3
sin
2
θ
b
−rx
2
)
(r
2
sin
2
θ
b
+ x
2
)
2
_
dr. (20)
Using the results derived in this section, the normalized
equivalent wind speeds were determined for a turbine with
the following representative specifications: R = 20, H = 40,
α = 0.3, a = 0.85, and x = 2.9. The normalized equivalent
wind speed, caused by the tower shadow and the wind shear
both together and individually, of this configuration are shown
in Fig. 7. It is seen that the effect of the tower shadow is more
dominant than the effect of the wind shear.
A. Solving for v
eq
0
This brief section will calculate the component of the equiv-
alent wind speed that is due to the steady-state hub height wind
speed. As expected and shown by (22), this component is simply
equal to the hub height wind speed, V
H
. It can be seen that this
result is independent of the values of r
0
, n, and s.
v
eq
0
=
2V
H
3sR
2
3

b=1
_
sR
2
2
_
dr (21)
v
eq
0
= V
H
. (22)
B. Solving for v
eq
ws
The component of the equivalent wind speed that is due to
the wind shear is calculated in this section and is given as (28).
Through numerical analysis, it was found that for a conservative
estimate of r
0
= 0.1R, that v
eq
ws
was comparable to the case
where r
0
= 0. Therefore, for the development, r
0
will be taken
as equal to 0 to simplify equations allowing n = 0 and s = 1.
If desired, a true value of r
0
may be used without much more
computational effort.
v
eq
ws
=
2V
H
3R
2
3

b=1
_
R
3
3
α
H
cos θ
b
+
R
4
4
α(α −1)
2H
2
cos
2
θ
b
+
R
5
5
α(α −1)(α −2)
6H
3
cos
3
θ
b
_
. (23)
To further simplify (23), expressions for the sums must be
developed. Using trigonometric identities and the angle defini-
tions shown in (24), these sums are determined and shown in
the form of (25)–(27) as follows.
θ = θ
1
, θ
2
= θ
1
+

3
and θ
3
= θ
1
+

3
(24)
3

b=1
[cos θ
b
] = 0 (25)
3

b=1
[cos
2
θ
b
] =
3
2
(26)
3

b=1
[cos
3
θ
b
] =
3
4
cos 3θ. (27)
We can now substitute (25)–(27) into (23) to yield the fi-
nal expression for equivalent wind speed due to wind shear as
follows:
v
eq
ws
= V
H
_
α(α −1)
8
_
R
H
_
2
+
α(α −1)(α −2)
60
_
R
H
_
3
cos 3θ
_
. (28)
The normalized equivalent wind speed caused by the wind
shear added to equivalent wind speed due to hub height wind
speed (v
eq
ws
+ v
eq
0
) is shown in Fig. 7. It can be observed
that this has a minimum when one blade is pointed directly
downwards but is a relatively small effect (1%). It is also
seen that there is a reduction in the equivalent wind speed due to
wind shear when normalized to V
H
. In this case this depression
is ≈0.5%.
C. Solving for v
eq
ts
The component of the equivalent wind speed that is due to
the tower shadow is calculated and is given in its final form
as (30). The formulation begins by performing the integration
722 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 3, SEPTEMBER 2006
within (20) to yield (29).
v
eq
ts
=
2mV
H
3sR
2
3

b=1
_
a
2
ln (R
2
sin
2
θ
b
+ x
2
)
2 sin
2
θ
b

a
2
ln
_
r
2
0
sin
2
θ
b
+ x
2
_
2 sin
2
θ
b
+
a
2
x
2
sin
2
θ
b
(R
2
sin
2
θ
b
+x
2
)

a
2
x
2
sin
2
θ
b
_
r
2
0
sin
2
θ
b
+x
2
_
_
.
(29)
Numerical evaluation shows that (29) gives nearly identical
results with r
0
= 0.1R and r
0
= 0. Therefore, to further sim-
plify (29), it will be assumed that r
0
= 0 and therefore s = 1.
This allows for the simplification of (29) to (30).
v
eq
ts
=
mV
H
3R
2
3

b=1
_
a
2
sin
2
θ
b
ln
_
R
2
sin
2
θ
b
x
2
+ 1
_

2a
2
R
2
R
2
sin
2
θ
b
+ x
2
_
. (30)
The normalized equivalent wind speed caused by the tower
shadowadded to the equivalent wind speed due to the hub height
wind speed (v
eq
ts
+ v
eq
0
) is shown in Fig. 7. It can be seen that
this has a minimum when a blade is directly downwards and at
≈3%, is much larger than the effect from the wind shear.
VII. EXTRACTION OF FUNCTION ψ(r)
A typical distribution of aerodynamic load can assume that
ψ(r) is proportional to r [17]. However, to use the equivalent
wind speed to calculate torque oscillations the proportionality
constant must be known. This value is not specified in the litera-
ture and thus must be determined. To extract the proportionality
constant we must linearize the classic torque equation’s (31)
dependence on wind speed. Since (13) is itself derived through
linearization, this may be done without additional loss of gener-
ality. We use an operating point around V
0
, since the steady-state
torque depends on the spatial mean wind speed.
T
ae
(t, θ) =
1
2
ρAV
2
R
C
p
(λ)
λ
. (31)
Linearizing (31), we get (32), where V
0
is the spatial mean
wind speed and λ
0
is the tip speed ratio at V
0
.
T
ae
(t, θ) = T
ae
(t, θ)
¸
¸
¸
¸
V =V
0
λ=λ
0
+
∂T
ae
(t, θ)
∂V
¸
¸
¸
¸
V =V
0
λ=λ
0
∆V
=
1
2
ρAV
0
2
R
C
p

0
)
λ
0
+ ρAV
0
R
C
p

0
)
λ
0
∆V. (32)
Defining n =
r
0
R
and s = 1 −n
2
, and with ∆V =
v
eq
(t, θ) −V
0
, (14) may now be transformed to (34).
T
ae
(t, θ) = 3M(V
0
) +
3

b=1
_
R
r
0
kr∆V dr (33)
T
ae
(t) = 3M(V
0
) + 3k
sR
2
2
∆V. (34)
Equating (32) and (34) yields two new important results,
shown as follows.
3M(V
0
) =
1
2
ρAV
2
0
R
C
p

0
)
λ
0
(35)
k =
2ρAV
0
3sR
C
p

0
)
λ
0
. (36)
The first result (35) shows that the addition of the steady-state
blade root moments over the three blades [3M(V
0
)] is equivalent
to the classic torque equation (31) at a particular wind speed. The
second result (36) gives the proportionality constant between
the torque deviation from the steady-state torque and the wind
speed deviation from the average wind speed. This new result
is important since it allows direct calculation of aerodynamic
torque from equivalent wind speed.
VIII. TORQUE OSCILLATIONS
With the three formulations of equivalent wind speed com-
ponents, the overall torque oscillations can now be modeled.
Using the linearized aerodynamic torque relation (34) and
allowing ∆V = v
eq
(t, θ) −V
0
and V
0
= mV
H
, we get the
following results:
T
ae
(t, θ) = 3M(V
0
) +
3ksR
2
2
[v
eq
(t, θ) −V
0
] (37)
T
ae
(t, θ) = 3M(V
0
) +
3ksR
2
2
×[v
eq
ws
+ v
eq
ts
+ V
H
−mV
H
]. (38)
Normalizing (38) to torque at wind speed V
0
, we get the
expression
T
ae
(t, θ) = 1 +
2
mV
H
[v
eq
ws
+ v
eq
ts
+ (1 −m)V
H
]. (39)
Using the end result of the formulation (39), the torque os-
cillations were determined for a turbine with the following rep-
resentative specifications: R = 20, H = 40, α = 0.3, a = 0.85,
and x = 2.9. As an illustration of possible results of the mod-
eling, the normalized torque oscillations due to wind shear and
tower shadow alone, as well as the total torque oscillations of
this configuration are shown in Fig. 8. Again it is observed that
the effects of wind shear on the total aerodynamic torque are
much smaller than those due to tower shadow, although they do
reshape the curve in regions. It is observed that both the oscil-
lations due to wind shear and tower shadow have a minimum
when one blade is pointed directly downwards and a maximum
when one blade is pointing directly upwards. The wind shear
effect is relatively small (1%), while the effect of the tower
shadow is much larger, in this case approximately 6% of the
total aerodynamic torque. It is also seen that there is a small
(≈1%) negative dc offset in the torque oscillation due to wind
shear. This offset seems to disappear in the total torque. This is
due to the normalization by a steady-state torque that occurs at
V
0
. As this dc offset is already contained in the 3M(V
0
) term of
(38), its duplication in the wind shear term is corrected for by
the (1 −m)V
H
term.
Dependence of the total 3p pulsation on wind shear exponent
(α) and dependence of wind-shear-induced 3p pulsation on
DOLAN AND LEHN: SIMULATION MODEL OF WIND TURBINE 3P TORQUE OSCILLATIONS DUE TO WIND SHEAR AND TOWER SHADOW 723
Fig. 8. Resulting normalized aerodynamic torque due to wind shear (1 +
2
mV
H
v
eq
ws
), tower shadow (1 +
2
mV
H
v
eq
ts
), and combination of wind shear
and tower shadow (1 +
2
mV
H
v
eq
ts
+
2
mV
H
v
eq
ws
+
2(1−m)
m
).
Fig. 9 Relative magnitude (per unit mean torque) of 3p pulsation as a function
of wind shear exponent (α).
α is shown in Fig. 9. As observed in the graph, although a
correlation is seen between α and the wind-shear-induced 3p
pulsation, no significant correlation is observed between α and
the total 3p pulsation.
IX. DISCUSSION
The torque model gives two particularly interesting results.
The first is that the tower shadow effects are much more dom-
inant than are the wind shear effects. The second is that the
maximum torque is observed when a blade is pointing directly
upwards. The modeled torque oscillations clearly depend on
turbine parameters R, H, a, and x and site parameter α. The
wind shear component of the oscillations depends on R, H, and
α and the tower shadow component of the oscillations depends
mainly on R, a, and x.
The wind shear exponent obviously has an effect on the torque
oscillations due to wind shear. However, as seen in Fig. 9, there
is not a significant correlation between α and the total 3p pul-
sation. This is due to the wind-shear-induced component being
approximately only 5% of the tower-shadow-induced compo-
nent. This explains why Thiringer was unable to find a good
correlation between α and measured 3p pulsation [1].
In Fig. 9, maximal oscillations occur at a value of α = 0.423,
where the oscillations are approximately 50% larger than those
observed for a typical value of α, such as that used to generate
Fig. 8. As seen in the figure, this would still result in a very small
oscillation. For wind shear, the actual magnitude of Rand H are
not critical as it is only their ratio that has an effect. Typically,
2
5

R
H

2
3
. The higher the ratio, the greater the effect the wind
shear would have as there is a wider range of wind speeds that
the blade experiences in a rotation. For a ratio of
2
3
, the torque
oscillation is five times larger than that observed for a ratio of
2
5
. However, combining the effects of these two parameters to
yield maximal torque oscillations still only amounts to a peak
value of approximately 0.4% of steady-state torque. Although
this torque oscillation is a relatively small one compared to
the tower-shadow-induced oscillations, it is still included in the
pragmatic model. This is done for three reasons. First, the effect
is quite easy to include as it can be represented in a closed form
expression. Second the model also contributes a dc component
that modifies the average torque, and lastly the torque oscillation
reshapes the curve at the peak torque.
For tower shadow, the actual radius of the turbine, indepen-
dent of height, is important. A larger turbine radius results
in both a narrower angle where a torque reduction is seen as
well as a slightly smaller reduction in overall torque. As can
be easily observed by the formula for equivalent wind speed
due to the tower shadow (30), the radius of the tower has a
squared relationship with total torque disturbance. Doubling of
the tower radius will give a fourfold increase in torque distur-
bance. The distance from the tower (x) is also an important fac-
tor. The closer the blades are to the tower, the larger the effect of
the tower shadow.
X. CONCLUSION
Acomprehensive yet pragmatic torque model has been devel-
oped for the three-bladed wind turbine. The model proves the ex-
istence of wind-shear-induced 3p oscillations and demonstrates
that in practice, their presence is masked by the much larger
tower-shadow-induced oscillations. It is determined that max-
imum torque is seen when a blade is pointing directly upwards
for both wind shear and tower shadow effects. The modeled
torque oscillations depend mostly on R, a, and x, as these are
related to tower shadow. Although wind shear causes small 3p
oscillations, it also contributes approximately a 1%dc reduction
in average torque. The proportionality constant between wind
speed variations and torque oscillations is determined, allowing
direct aerodynamic torque calculation from an equivalent wind
speed. This model is a useful representation of the aerodynamic
torque of a wind turbine for use in real-time wind turbine simu-
lators and other dynamic-model-simulation-based applications.
724 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 3, SEPTEMBER 2006
APPENDIX
Arelationship between spatial average wind speed V
0
and hub
height wind speed V
H
is required such that tower shadow and
wind shear formulas can be combined with only one wind speed
term. To calculate spatial average wind speed V
0
, the varying
wind speed from wind shear is integrated over rotor area and
divided by the total rotor area.
V
0
=
1
πR
2
_

0
_
R
0
V
H
[1 + W
s
(r, θ)]r dr dθ (40)
V
0
=
1
πR
2
_

0
_
R
0
V
H
_
1 + α(
r
H
) cos θ
+
α(α −1)
2
_
r
H
_
2
cos
2
θ
+
α(α −1)(α −2)
6
_
r
H
_
3
cos
3
θ
_
r dr dθ (41)
V
0
=
V
H
πR
2
_
R
0
_
2πr +
πα(α −1)r
3
2H
2
_
dr (42)
V
0
=
V
H
πR
2
_

R
2
2
+
πα(α −1)R
4
8H
2
_
(43)
V
0
= V
H
_
1 +
α(α −1)(R
2
)
8H
2
_
= mV
H
. (44)
In (40)–(44), V
H
is the wind speed at hub height, R is the
blade radius, α is the empirical wind shear exponent, and H is
the elevation of rotor hub.
It is shown in Fig. 5 that 0.986 <
V
0
V
H
≤ 1, for
R
H
< 0.67 and
0.1 < α ≤ 1. Therefore, for most cases a simplification that
V
0
= V
H
is justified. For more accuracy (44) can be used.
REFERENCES
[1] T. Thiringer and J.-A. Dahlberg, “Periodic pulsations from a three-bladed
wind turbine,” IEEE Trans. Energy Convers., vol. 16, pp. 128–133, Jun.
2001.
[2] T. Thiringer, “Power quality measurements performed on a low-voltage
grid equipped with two wind turbines,” IEEE Trans. Energy Convers.,
vol. 11, pp. 601–606, Sep. 1996.
[3] S. B. Bayne and M. G. Giesselmann, “Effect of blade passing on a wind
turbine output,” in Proc. Energy Conversion Engineering Conf. Exhib.,
(IECEC) 35th Intersociety, Jul. 2000, vol. 2, pp. 775–781.
[4] J. G. Slootweg, S. W. H. de Haan, H. Polinder, and W. L. Kling, “General
model for representing variable speed wind turbines in power system
dynamics simulations,” IEEE Trans. Power Syst., vol. 18, no. 1, pp. 144–
151, Feb. 2003.
[5] T. Petru and T. Thiringer, “Modeling of wind turbines for power system
studies,” IEEE Trans. Power Syst., vol. 17, no. 4, pp. 1132–1139, Nov.
2002.
[6] D. S. L. Dolan, “Real-time wind turbine emulator suitable for power
quality and dynamic control studies” MASc. thesis, Dept. Electr. Comput.
Eng., Univ. Toronto, Toronto, 2005.
[7] D. S. L. Dolan and P. W. Lehn, “Real-time wind turbine emulator suit-
able for power quality and dynamic control studies,” presented at the Int.
Conf. Power Systems Transients, IPST05, Montreal, Canada, Jun. 19–23
2005.
[8] R. Cardenas, R. Pena, G. M. Asher, and J. C. Clare, “Experimental em-
ulation of wind turbines and flywheels for wind energy applications,”
presented at the EPE Conf., Graz, Austria, Aug. 2001.
[9] L. Chang, R. Doraiswami, T. Boutot, and H. Kojabadi, “Development of
a wind turbine simulator for wind energy conversion systems,” in Proc.
Canadian IEEECCECE2000, vol. 1, Halifax, Canada, 2000, pp. 550–554.
[10] F. A. Farret, R. Gules, and J. Marian, “Micro-turbine simulator based on
speed and torque of a dc motor to drive actually loaded generators,” in
Proc. 1995 First IEEE Int. Caracas Conf. Devices, Circuits Syst., 1995,
pp. 89–93.
[11] P. E. Battaiotto, R. J. Mantz, and P. F. Puleston, “A wind turbine emulator
based on a dual DSP processor system,” Contr. Eng. Pract., vol. 4,
pp. 1261–1266, 1996.
[12] D. Parker, “Computer based real-time simulator for renewable energy
converters,” in Proc. First IEEE International Workshop on Electronic
Design, Test and Applications, Jan. 2002, pp. 280–284.
[13] Y. Matsumoto, H. Umida, and S. Ozaki, “Dynamic simulator of the
mechanical system,” in Proc. IECON, Oct. 1991, vol. 1, pp. 527–
532.
[14] J. B. Ekanayake, L. Holdsworth, W. XueGuang, and N. Jenkins, “Dynamic
modeling of doubly fed induction generator wind turbines,” IEEE Trans.
Power Syst., vol. 18, no. 2, pp. 803–809, May 2003.
[15] D. J. Trudnowski, A. Gentile, J. M. Khan, and E. M. Petritz, “Fixed-
speed wind-generator and wind-park modeling for transient stability stud-
ies,” IEEE Trans. Power Syst., vol. 19, no. 4, pp. 1911–1917, Nov.
2003.
[16] J. Cidras and A. E. Feijoo, “A linear dynamic model for asynchronous
wind turbines with mechanical fluctuations,” IEEE Trans. Power Syst.,
vol. 17, no. 3, pp. 681–687, Aug. 2002.
[17] P. Sorensen, A. D. Hansen, and P. A. C. Rosas, “Wind models for simula-
tion of power fluctuations from wind farms,” J. Wind Eng. Ind. Aerody-
nam., vol. 90, pp. 1381–1402, Dec. 2002.
[18] D. A. Spera, Wind Turbine Technology. New York: ASME Press, 1994.
[19] R. W. Thresher, A. D. Wright, and E. L. Hershberg, “A computer analysis
of wind turbine blade dynamic loads,” ASME J. Solar Energy Eng.,
vol. 108, pp. 17–25, 1986.
[20] E. N. Hinrichsen and P. J. Nolan, “Dynamics and stability of wind turbine
generators,” IEEE Trans. Power App. Syst., vol. 101, pp. 2640–2648,
Aug. 1982.
[21] O. Garcia. (1998). Wind Turbine Dynamic Modelling, [Online]. Available:
www.iit.upco.es/oscar/download/model.ps
Dale S. L. Dolan (S’05) received the B.Sc. (Honors)
degree in biology and B.Ed. degree from the Uni-
versity of Western Ontario, London, ON, Canada, in
1995 and 1997, respectively. He received the BASc.
and MASc. degrees fromthe Department of Electrical
and Computer Engineering, University of Toronto,
Toronto, in 2003 and 2005, respectively, both in elec-
trical engineering. He is currently working toward
the Ph.D. degree at the Department of Electrical and
Computer Engineering, University of Toronto.
His research interests include wind turbine emu-
lation, alternative energy conversion systems, power electronics, and electro-
magnetics.
Peter W. Lehn (S’95–M’99–SM’05) received the
B.Sc. and M.Sc. degrees from the University of Man-
itoba, Winnipeg, MB, Canada, in 1990 and 1992, re-
spectively, both in electrical engineering. He received
the Ph.D. degree in electrical engineering from the
University of Toronto, Toronto, ON, in 1999.
From 1992 to 1994, he was with the Network
Planning Group of Siemens AG, Erlangen, Germany.
Currently, he is working as an Associate Professor at
the University of Toronto.

as shown in Fig. Torque pulsations. For upwind rotors. H is the elevation of rotor hub. Power and torque oscillate due to the different wind conditions encountered by each blade as it rotates through a complete cycle [3]. θ) ≈ α α(α − 1) r 2 r cos θ + cos2 θ H 2 H α(α − 1)(α − 2) r 3 + cos3 θ (4) 6 H III. is defined as in (5). x) = −V0 x D 2π (x2 + y 2 ) (7) where D is the tower diameter. θ)] (2) where VH is the wind speed at hub height. For instance. 1.718 IEEE TRANSACTIONS ON ENERGY CONVERSION. For this reason. VOL. Comparison of torque oscillation due to wind shear alone depending on form of wind shear approximation. x) = VH + υtower (y. where the resulting torque from the second order approximation becomes constant. This effect is called tower shadow. θ) should be as follows. Sorensen [17] modeled tower disturbance using potential flow theory for wind movement around the tower. [19]. the approximation used for Ws (r. 1. x) is the disturbance observed in the wind speed due to the tower shadow that is added to hub height wind speed. shown as (1). To effectively model the 3p effect of wind shear. 3. in three-bladed turbines. the tower shadow effect is more pronounced when the blades are closer (x smaller) to the tower. It can be seen that as expected. a blade pointing upwards would encounter wind speeds greater than a blade pointing downwards. θ) ≈ α cos θ + H 2 H However. y is the lateral distance from the blade to the tower midline. the truncated expansion of (3) eliminates. During each rotation. This section will show theoretically the 3p oscillations caused by tower shadow. requiring a third-order-truncated Taylor expansion. This paper will therefore only deal with the tower shadow torque oscillations in horizontal axis three-bladed upwind rotors. Ws (r. θ) by the second-order-truncated Taylor series expansion shown as follows: r α(α − 1) r 2 cos2 θ (3) Ws (r. only considering tower shadow. α is the empirical wind shear exponent. Both Spera [18] and Thresher [19] approximated Ws (r. y is the lateral distance from the blade to the tower midline. Using the reference frames shown in Fig. The term υtower (y. x is the distance from the blade . V0 is the spatial mean wind speed. the torque oscillates three times because of each blade passing through minimum and maximum wind. It is therefore important to model these wind-shear-induced 3p torque pulsations when studying a wind turbine system. The term Ws (r. This can be seen in Fig. and therefore power pulsations. z α (1) V (z) = VH H For the purpose of this analysis. where VH = hub height wind speed. V (y. the wind directly in front of the tower is redirected and thereby reduces the torque at each blade when in front of the tower. completely losing the properties of the nonlinear wind shear expression. [18]. 2 yields (6). The torque pulsations due to tower shadow are most significant when a turbine has blades downwind of the tower and wind is blocked as opposed to redirected [20]. 1. are observed due to the periodic variations of wind speed seen at different heights. (1) is converted to a function of r (radial distance from rotor axis) and θ (azimuthal angle) giving the following: V (r. WIND SHEAR Wind speed generally increases with height and this variation is termed wind shear. a is the tower radius. A common wind shear model. θ) = VH r cos θ + H H α = VH [1 + Ws (r. is taken directly from the literature on wind turbine dynamics [1]. the torque oscillations due In (6). x) υtower (y. and z is the elevation above ground. NO. 3. θ) is the disturbance seen in wind speed due to wind shear that is added to hub height wind speed. 21. the majority of modern wind turbines have upwind rotors. the cos θ term yields a zero contribution while the cos2 θ term contributes only a dc component that adjusts average wind speed from hub height wind speed to spatial mean wind speed. a cos3 θ term is necessary. SEPTEMBER 2006 to the wind shear when the contributions from each of the blades are summed. and x is the distance from the blade origin to the tower midline. to model torque oscillations from wind shear. This is because when the three blade contributions are summed. Results for tower radius of 2 m and four different longitudinal distances between the tower and the blades are shown in Fig. An alternate tower-shadow-deficit model (7) is developed in [21] and shown as follows: υtower (y. r is the radial distance from rotor axis. II. Ws is the wind-shear-shape function [18]. Therefore. TOWER SHADOW The distribution of wind is altered by the presence of the tower. The wind field. x) = V0 a2 y 2 − x2 (x2 + y 2 )2 (5) (6) Fig.

Comparison of tower shadow model (6) with different distances between the tower and the blades. 5. Therefore.DOLAN AND LEHN: SIMULATION MODEL OF WIND TURBINE 3P TORQUE OSCILLATIONS DUE TO WIND SHEAR AND TOWER SHADOW 719 Fig. θ. (6) is preferable and will be used in subsequent model development. 2. Fig. 4. Results for this alternate model with a tower radius of 2 m and four different longitudinal distances between the tower and the blades are shown in Fig. and V0 is the spatial mean wind speed. . 2 and m = [1 + α (α −1)(R ) ] as developed in Appendix. in the torque oscillation model. Comparison of tower shadow model (7) with different distances between the tower and the blades. VH . θ is the azimuthal angle of the blade. x is the distance from the blade origin to the tower midline. tower shadow effects should obviously be absent. Most often for time-domain simulation. Converting (6) from a function of y (lateral distance) to a function of r (radial distance) and θ (azimuthal angle) normalized to VH yields as follows: υtower (r. The wind shear model uses hub height wind speed (VH ) while the tower shadow model uses spatial mean wind speed (V0 ). The relationship between these two wind speeds is formulated in Appendix and is summarized in Fig. Comparison of the two models graphically shows that a more reasonable model is represented by (6). V0 would require calculation from an entire spatial wind field that would normally be unavailable. Different reference wind speeds are used in models for the disturbance due to wind shear and tower shadow. only a single wind speed value. as it models both the deceleration of the wind flow in front of the tower and the acceleration of the wind flow on each side of the tower. 5. Fig. for all practical purposes. for modeling torque oscillations due to tower shadow. x) = ma2 ˜ r2 sin2 (θ) − x2 (r2 sin2 (θ) + x2 )2 (8) where a is the tower radius. origin to the tower midline. 4. Dimensions used in tower shadow formula. Fig. r is the radial distance from the blade to the hub center. Variation of m = V 0 /V H with α for different R/H ratios. is available. Therefore. It should 8H 2 be noted that (8) is only valid for 90◦ ≤ θ ≤ 270◦ as above the horizontal. 3. tower disturbance will be expressed in terms of VH .

θ) = 2VH 3sR2 + 3 R υ As Ws (r. although the same wind deficit is seen for all blade elements at an angle of 180◦ . DETERMINATION OF TOTAL WIND FIELD—v(t. θ) = VH (t)[1 + Ws (r. knowing only the turbine parameters. r0 is the radius at which blade profile begins. θ)˜tower (r. the results of (4) and (8) from Sections II and III are combined. (14) Sorensen determined (15) to be the expression for equivalent wind speed by equating (13) and (14) 1 veq (t. R the total wind field (11) may be inserted into (15) to yield (16) after some initial simplification. 6. Assuming ψ(r) = kr. r. NO. r. θ) + υtower (r. θ)˜tower (r. 21. (r2 sin2 θb + x2 )2 (16) + This total wind field model allows one to determine the wind speed observed at any particular location in the rotor disk area.9 m from tower midline. θ) that does not vary with the radius is defined which would give the same aerodynamic torque as the actual spatially varying wind speed. and tower shadow may be observed separately. υ (10) where Tae (t. θb ) − V0 ] dr (13) Fig. and defining n = r 0 and s = 1 − n2 . x)] (11) The spatially varying wind speed can be calculated using the total wind field model of (11) or its expanded version as follows. This equivalent wind speed will have three components. 3. veq (t. θ) = 3M (V0 ) + b=1 ψ(r)[veq (t. θ) = 3M (V0 ) + b=1 ψ(r)[v(t. r. This veq must be such that 3 R r0 Tae (t. This equation has been determined through linearization of individual blade torque dependence on wind speed [17]. r. θ. θ) = 3 3 R r0 ψ(r)v(t. The total wind field due to both tower shadow and wind shear is given as follows. x)]. x)] (9) v(t. r. θ) To determine the total wind field. (11) is a valid approximation of (9). θ) = VH (t)[1 + Ws (r. r. This approach is also supported in the literature [19]. Sorensen’s approach is briefly outlined in this section. θ. (r2 sin2 θ + x2 )2 r4 α(α − 1)(α − 2) 3 cos θb 6H 3 ma2 (r3 sin2 θb − rx2 ) dr. θ. The advantage of this method is that a wind speed without radial dependence may be used. It is observed that the blade elements closer to the hub experience tower shadow for a longer period. the second (veqws ) is due to the wind shear. θ) is the aerodynamic torque. θ) is given as follows: 3 R r0 Tae (t. r. The aerodynamic torque produced by a three-bladed wind turbine immersed in a wind field v(t. SEPTEMBER 2006 V. R is the radius of the rotor disk. The first (veq0 ) is due to the hub height wind speed. v(t. θ) − V0 ] dr. M (V0 ) is the steadystate blade root moment resulting from spatial mean wind speed V0 . An equivalent wind speed veq (t. VOL.7-m diameter and blades 2. Fig.720 IEEE TRANSACTIONS ON ENERGY CONVERSION. Comparison of tower shadow at different radii based on a tower with 1. θ)][1 + υtower (r. Three components of this equivalent wind speed will be separated and solved individually such that the effects from the hub height wind speed. 6 shows the variation in the effective tower shadow angle experienced by different blade elements at varying radial distances. θ) + υtower (r. IV. ˜ v(t. θ. r. wind shear. ˜ v(t. x) ˜ + Ws (r. θ) ≈ VH (t)[1 + Ws (r. θb )dr R r0 b=1 ψ(r) dr (15) VI. DETERMINATION OF EQUIVALENT WIND SPEED The total wind field including tower shadow and wind shear effects will now be converted into one equivalent wind speed. and ψ(r) is the influence coefficient of the aerodynamic load on the blade root moment. x) would be small compared to other terms. θ) ≈ VH (t) 1 + α α(α − 1) r cos θ + H 2 r H 3 r H 2 cos2 θ r+ b=1 r 0 r3 α(α − 1) r2 α cos θb + cos2 θb H 2H 2 α(α − 1)(α − 2) + 6 + cos θ (12) 3 ma2 (r2 sin2 θ − x2 ) . EQUIVALENT WIND SPEED FORMULATION BASED ON EQUIVALENT TORQUE An effective method for formulating an “equivalent wind speed” has been developed by Sorensen [17]. wind shear coefficient and a single hub height wind speed. and the third (veqts ) is due to . The equivalent wind speed is a representation of the actual spatially varying wind speed that is defined such that it will give the same aerodynamic torque. For clarity and completeness. θ.

caused by the tower shadow and the wind shear both together and individually. 4 (26) r4 α(α − 1)(α − 2) + cos3 θb dr 6H 3 veqts = 2VH 3sR2 3 R r0 (19) [cos3 θb ] = b=1 (27) b=1 ma2 (r3 sin2 θb − rx2 ) dr. It is seen that the effect of the tower shadow is more dominant than the effect of the wind shear. and combination of wind shear and tower shadow (v eq0 + v eqts + v eqws ).85. Therefore. C. (r2 sin2 θb + x2 )2 (20) We can now substitute (25)–(27) into (23) to yield the final expression for equivalent wind speed due to wind shear as follows: veqws = VH α(α − 1) 8 + R H 2 Using the results derived in this section. The formulation begins by performing the integration . Therefore. it was found that for a conservative estimate of r0 = 0. It can be seen that this result is independent of the values of r0 . α = 0. The component of the equivalent wind speed that is due to the tower shadow is calculated and is given in its final form as (30).5%. 7. H = 40. Normalized equivalent wind speed due to tower shadow (v eqts + v eq0 ). 7. and x = 2. Solving for veqts b=1 sR2 dr 2 (21) (22) veq0 = VH . n. VH .3. expressions for the sums must be developed. veq (t. The normalized equivalent wind speed. a true value of r0 may be used without much more computational effort. 5 6H 3 (23) Fig. A. (16) can be decomposed as (17) whose components are shown as (18)–(20). (28) The normalized equivalent wind speed caused by the wind shear added to equivalent wind speed due to hub height wind speed (veqws + veq0 ) is shown in Fig. a = 0. veqws = 2VH 3R2 3 b=1 R3 α R4 α(α − 1) cos θb + cos2 θb 3 H 4 2H 2 + R5 α(α − 1)(α − 2) cos3 θb . θ2 = θ 1 + 3 the tower shadow. this component is simply equal to the hub height wind speed. In this case this depression is ≈0.9. Using trigonometric identities and the angle definitions shown in (24). of this configuration are shown in Fig. To further simplify (23). the normalized equivalent wind speeds were determined for a turbine with the following representative specifications: R = 20. Through numerical analysis. and s. It can be observed that this has a minimum when one blade is pointed directly downwards but is a relatively small effect ( 1%). r0 will be taken as equal to 0 to simplify equations allowing n = 0 and s = 1. wind shear (v eqws + v eq0 ). that veqws was comparable to the case where r0 = 0. If desired.1R. veq0 2VH = 3sR2 3 α(α − 1)(α − 2) 60 R H 3 cos 3θ . θ) = veq0 + veqws + veqts veq0 = 2VH 3sR2 3 R (17) (18) 3 2π 3 and θ3 = θ1 + 4π 3 (24) (25) [cos θb ] = 0 [r] dr b=1 3 b=1 3 r0 R r0 veqws 2VH = 3sR2 b=1 r α r α(α − 1) cos θb + cos2 θb H 2H 2 2 [cos2 θb ] = b=1 3 3 2 3 cos 3θ. It is also seen that there is a reduction in the equivalent wind speed due to wind shear when normalized to VH . 7. for the development. these sums are determined and shown in the form of (25)–(27) as follows. Solving for veqws The component of the equivalent wind speed that is due to the wind shear is calculated in this section and is given as (28). θ = θ1 . As expected and shown by (22).DOLAN AND LEHN: SIMULATION MODEL OF WIND TURBINE 3P TORQUE OSCILLATIONS DUE TO WIND SHEAR AND TOWER SHADOW 721 B. Solving for veq0 This brief section will calculate the component of the equivalent wind speed that is due to the steady-state hub height wind speed.

VII. this may be done without additional loss of generality. we get (32). It is observed that both the oscillations due to wind shear and tower shadow have a minimum when one blade is pointed directly downwards and a maximum when one blade is pointing directly upwards. k= VIII. where V0 is the spatial mean wind speed and λ0 is the tip speed ratio at V0 . EXTRACTION OF FUNCTION ψ(r) A typical distribution of aerodynamic load can assume that ψ(r) is proportional to r [17]. 1 Cp (λ) . as well as the total torque oscillations of this configuration are shown in Fig. (36) 3sR λ0 The first result (35) shows that the addition of the steady-state blade root moments over the three blades [3M (V0 )] is equivalent to the classic torque equation (31) at a particular wind speed. θ) = 3M (V0 ) + b=1 r0 kr∆V dr sR2 ∆V. (39) mVH Using the end result of the formulation (39). 3. θ) − V0 ] Tae (t. since the steady-state torque depends on the spatial mean wind speed. This allows for the simplification of (29) to (30). Dependence of the total 3p pulsation on wind shear exponent (α) and dependence of wind-shear-induced 3p pulsation on . veqts = − + 2mVH 3sR2 a ln 2 3 b=1 a ln (R sin θb + x ) 2 sin2 θb 2 2 2 2 Equating (32) and (34) yields two new important results. This is due to the normalization by a steady-state torque that occurs at V0 . This value is not specified in the literature and thus must be determined. This offset seems to disappear in the total torque. θ) − V0 and V0 = mVH . As this dc offset is already contained in the 3M (V0 ) term of (38). It is also seen that there is a small (≈1%) negative dc offset in the torque oscillation due to wind shear. and with ∆V = R veq (t. 7. θ) = Tae (t. 2 λ0 λ0 (32) Defining n = r 0 and s = 1 − n2 . H = 40.3. θ) = 1 + [veqws + veqts + (1 − m)VH ]. θ) = 3M (V0 ) + × [veqws 3ksR2 2 + veqts + VH − mVH ]. θ) − V0 . To extract the proportionality constant we must linearize the classic torque equation’s (31) dependence on wind speed. θ) ∂V V =V 0 λ=λ 0 ∆V Cp (λ0 ) Cp (λ0 ) 1 ρAV0 2 R + ρAV0 R ∆V. (30) R2 sin2 θb + x2 2 The normalized equivalent wind speed caused by the tower shadow added to the equivalent wind speed due to the hub height wind speed (veqts + veq0 ) is shown in Fig. veqts = mVH 3R2 3 b=1 R2 sin2 θb a2 ln +1 x2 sin2 θb − 2a R . NO. The wind shear effect is relatively small ( 1%). Using the linearized aerodynamic torque relation (34) and allowing ∆V = veq (t. 8. We use an operating point around V0 . to use the equivalent wind speed to calculate torque oscillations the proportionality constant must be known. θ) = ρAV 2 R 2 λ Linearizing (31). although they do reshape the curve in regions. (31) Tae (t.1R and r0 = 0. the normalized torque oscillations due to wind shear and tower shadow alone. (14) may now be transformed to (34). However. we get the expression 2 Tae (t. its duplication in the wind shear term is corrected for by the (1 − m)VH term. 1 Cp (λ0 ) (35) 3M (V0 ) = ρAV02 R 2 λ0 2ρAV0 Cp (λ0 ) . to further simplify (29).9. Again it is observed that the effects of wind shear on the total aerodynamic torque are much smaller than those due to tower shadow. θ) = V =V 0 λ=λ 0 With the three formulations of equivalent wind speed components. It can be seen that this has a minimum when a blade is directly downwards and at ≈3%. (38) + ∂Tae (t. The second result (36) gives the proportionality constant between the torque deviation from the steady-state torque and the wind speed deviation from the average wind speed. Tae (t. Therefore. a = 0. VOL. we get the following results: 3ksR2 [veq (t. in this case approximately 6% of the total aerodynamic torque. Since (13) is itself derived through linearization. while the effect of the tower shadow is much larger. 21. θ) = 3M (V0 ) + (37) 2 Tae (t. As an illustration of possible results of the modeling. the overall torque oscillations can now be modeled. is much larger than the effect from the wind shear. 3 R Tae (t. α = 0. the torque oscillations were determined for a turbine with the following representative specifications: R = 20.85. TORQUE OSCILLATIONS 2 2 r0 sin2 θb + x2 2 sin2 θb . and x = 2. it will be assumed that r0 = 0 and therefore s = 1. This new result is important since it allows direct calculation of aerodynamic torque from equivalent wind speed. 2 (33) Tae (t) = 3M (V0 ) + 3k (34) Normalizing (38) to torque at wind speed V0 . a2 x2 a2 x2 − 2 2 sin2 θ + x2 2 sin2 θ + x2 ) sin θb (R sin θb r0 b b 2 (29) Numerical evaluation shows that (29) gives nearly identical results with r0 = 0. SEPTEMBER 2006 within (20) to yield (29).722 IEEE TRANSACTIONS ON ENERGY CONVERSION. shown as follows.

the actual magnitude of R and H are not critical as it is only their ratio that has an effect. However. The closer the blades are to the tower. It is determined that maximum torque is seen when a blade is pointing directly upwards for both wind shear and tower shadow effects. the torque 3 oscillation is five times larger than that observed for a ratio of 2 5 . 8. For wind shear. Resulting normalized aerodynamic torque due to wind shear (1 + 2 2 m V v eqws ). This model is a useful representation of the aerodynamic torque of a wind turbine for use in real-time wind turbine simulators and other dynamic-model-simulation-based applications. the actual radius of the turbine. The wind shear component of the oscillations depends on R. tower shadow (1 + m V v eqts ). and x. The second is that the maximum torque is observed when a blade is pointing directly upwards. a. A comprehensive yet pragmatic torque model has been developed for the three-bladed wind turbine. as seen in Fig. H. it is still included in the pragmatic model. the effect is quite easy to include as it can be represented in a closed form expression. a. 9. As seen in the figure. no significant correlation is observed between α and the total 3p pulsation. where the oscillations are approximately 50% larger than those observed for a typical value of α. and α and the tower shadow component of the oscillations depends mainly on R. First. Typically. For tower shadow. Second the model also contributes a dc component that modifies the average torque. DISCUSSION The torque model gives two particularly interesting results. it also contributes approximately a 1% dc reduction in average torque. This is done for three reasons. 9. This explains why Thiringer was unable to find a good correlation between α and measured 3p pulsation [1]. The wind shear exponent obviously has an effect on the torque oscillations due to wind shear. 2 R 2 5 ≤ H ≤ 3 . and lastly the torque oscillation reshapes the curve at the peak torque. The distance from the tower (x) is also an important factor. the larger the effect of the tower shadow. this would still result in a very small oscillation. the greater the effect the wind shear would have as there is a wider range of wind speeds that the blade experiences in a rotation. 8. This is due to the wind-shear-induced component being approximately only 5% of the tower-shadow-induced component.4% of steady-state torque. m Fig. In Fig. However. The modeled torque oscillations depend mostly on R. For a ratio of 2 . The model proves the existence of wind-shear-induced 3p oscillations and demonstrates that in practice.423. such as that used to generate Fig. X. and x and site parameter α. 9. Although this torque oscillation is a relatively small one compared to the tower-shadow-induced oscillations. The first is that the tower shadow effects are much more dominant than are the wind shear effects. 9 Relative magnitude (per unit mean torque) of 3p pulsation as a function of wind shear exponent (α). although a correlation is seen between α and the wind-shear-induced 3p pulsation. independent of height. the radius of the tower has a squared relationship with total torque disturbance. Doubling of the tower radius will give a fourfold increase in torque disturbance. as these are related to tower shadow. A larger turbine radius results in both a narrower angle where a torque reduction is seen as well as a slightly smaller reduction in overall torque. H. a. there is not a significant correlation between α and the total 3p pulsation.DOLAN AND LEHN: SIMULATION MODEL OF WIND TURBINE 3P TORQUE OSCILLATIONS DUE TO WIND SHEAR AND TOWER SHADOW 723 Fig. their presence is masked by the much larger tower-shadow-induced oscillations. The proportionality constant between wind speed variations and torque oscillations is determined. Although wind shear causes small 3p oscillations. As can be easily observed by the formula for equivalent wind speed due to the tower shadow (30). The modeled torque oscillations clearly depend on turbine parameters R. combining the effects of these two parameters to yield maximal torque oscillations still only amounts to a peak value of approximately 0. is important. and combination of wind shear H H and tower shadow (1 + 2 m V H v eqts + 2 m V H v eqws + 2(1−m ) ). maximal oscillations occur at a value of α = 0. The higher the ratio. allowing direct aerodynamic torque calculation from an equivalent wind speed. As observed in the graph. . IX. CONCLUSION α is shown in Fig. and x.

[19] R. 2003. A.. he is working as an Associate Professor at the University of Toronto. Lehn. in 1990 and 1992. α is the empirical wind shear exponent. “Modeling of wind turbines for power system studies.. pp. pp. R. pp. C. “Periodic pulsations from a three-bladed wind turbine. and E. J. Farret.” IEEE Trans. 1995. M. no. Dahlberg. Energy Convers. 527– 532. Currently.. and H is the elevation of rotor hub. Cardenas. Dolan (S’05) received the B. Toronto. Canada. “A computer analysis of wind turbine blade dynamic loads. W. Feb. 1132–1139.” presented at the Int. and P. ON. Available: www. Asher. Jul. Rosas. pp. 1996. V0 R It is shown in Fig. Parker. Test and Applications.” IEEE Trans. 1911–1917.. 3. H. Giesselmann.. 1 πR2 1 πR2 + 2π 0 2π 0 0 0 R R V0 = V0 = VH [1 + Ws (r. 4. [14] J. Pena. vol. 1381–1402. respectively. vol. Toronto. pp. Ozaki. Germany. 4. Power App. N. vol. alternative energy conversion systems. and N.Sc. University of Toronto. “Dynamics and stability of wind turbine generators.724 IEEE TRANSACTIONS ON ENERGY CONVERSION. Energy Conversion Engineering Conf. “General model for representing variable speed wind turbines in power system dynamics simulations. Matsumoto. vol. His research interests include wind turbine emulation.” in Proc. “Power quality measurements performed on a low-voltage grid equipped with two wind turbines. Doraiswami. Electr. 89–93. 90. degrees from the Department of Electrical and Computer Engineering. no. pp. 1.986 < V H ≤ 1.” IEEE Trans. pp.” ASME J. [15] D. Erlangen. 775–781. vol. and M. Solar Energy Eng. Power Syst. 1261–1266. 681–687.iit.ps Dale S. 1994. respectively. vol. Ind. Power Syst. He received the Ph. Wright. A. “Wind models for simulation of power fluctuations from wind farms. [5] T. Dolan and P. [7] D. [21] O. Trudnowski. Canada. “Real-time wind turbine emulator suitable for power quality and dynamic control studies. 2002. both in electrical engineering. 11. [2] T. Canadian IEEE CCECE 2000. “Real-time wind turbine emulator suitable for power quality and dynamic control studies” MASc. Holdsworth. Austria. pp. Ekanayake.Ed. no. 2000. Eng. vol. SEPTEMBER 2006 APPENDIX A relationship between spatial average wind speed V0 and hub height wind speed VH is required such that tower shadow and wind shear formulas can be combined with only one wind speed term. Energy Convers. no. vol. J.. 5 that 0. 17.. in 2003 and 2005. C. Mantz. Gentile. A. (1998). R. Therefore. pp. 18. Feijoo. “Micro-turbine simulator based on speed and torque of a dc motor to drive actually loaded generators. 2001. Exhib. Umida. pp. 601–606. 550–554. London. Clare. (Honors) degree in biology and B.D. 8H 2 V0 = VH 1 + In (40)–(44).” IEEE Trans. in 1999. L. Power Syst. 803–809. L. XueGuang.” J. From 1992 to 1994. “Dynamic modeling of doubly fed induction generator wind turbines.” in Proc.. 2001. Conf. and MASc. Power Syst. Wind Eng. Dec. H. [6] D. and P. D. Syst. B. 128–133. Pract. Toronto. and S. Dept. S. and J. J. NO. ON. M. IECON. A. To calculate spatial average wind speed V0 . 2000. 2640–2648. G. [10] F. 21. for most cases a simplification that V0 = VH is justified. Chang. L.” IEEE Trans. Winnipeg. 18. Bayne and M. Oct.” IEEE Trans. Lehn (S’95–M’99–SM’05) received the B. Power Syst.D.. Aug. Caracas Conf. 1982.” presented at the EPE Conf. L.. [13] Y. Dolan. [3] S. pp. [11] P. [12] D. REFERENCES [1] T. 2. pp.” IEEE Trans. 1991. thesis. F. 3. vol. Petru and T. 2003. 2002. and J. Petritz. respectively.Sc. Power Systems Transients. 16. Thresher. pp. Hershberg. pp. Garcia. degree at the Department of Electrical and Computer Engineering. Aerodynam. A. (IECEC) 35th Intersociety. R. 101. D. He is currently working toward the Ph. [8] R. Aug. [9] L.. 19–23 2005. E. 108. May 2003. Toronto. power electronics. “Effect of blade passing on a wind turbine output. degree from the University of Western Ontario. VH is the wind speed at hub height. both in electrical engineering.. Canada. Spera. Puleston. 1996. Graz. G. Wind Turbine Technology. S. “Experimental emulation of wind turbines and flywheels for wind energy applications. degrees from the University of Manitoba. Sorensen. pp. W. For more accuracy (44) can be used. [Online].Sc. Thiringer. [16] J. “A wind turbine emulator based on a dual DSP processor system.67 and 0. Thiringer and J. 17.-A. and H.” in Proc.upco.. Comput. 2002. and electromagnetics. Sep. Jun. Wind Turbine Dynamic Modelling. no. First IEEE International Workshop on Electronic Design.” Contr. T. “Development of a wind turbine simulator for wind energy conversion systems. Hansen. R is the blade radius. W. Jenkins. vol. “Dynamic simulator of the mechanical system. vol. and E. the varying wind speed from wind shear is integrated over rotor area and divided by the total rotor area. Circuits Syst. θ)]r dr dθ VH 1 + α( r ) cos θ H (40) α(α − 1) r 2 cos2 θ 2 H α(α − 1)(α − 2) r 3 + cos3 θ r dr dθ 6 H V0 = V0 = VH πR2 R (41) (42) (43) (44) 2πr + 0 πα(α − 1)r3 dr 2H 2 πα(α − 1)R4 VH R2 + 2π 2 πR 2 8H 2 α(α − 1)(R2 ) = mVH . 1. MB. VOL. S. L. Thiringer. 1995 First IEEE Int.es/oscar/download/model. H. Hinrichsen and P. Cidras and A. Battaiotto. University of Toronto. Jan. Khan. G.1 < α ≤ 1. 17–25. Polinder. 19. Nov. Halifax.” in Proc. E. 2. . Peter W.. vol. W. “Fixedspeed wind-generator and wind-park modeling for transient stability studies. Canada. Aug. [17] P. [20] E. 4. he was with the Network Planning Group of Siemens AG. Univ. Nov. Nolan. 144– 151. A. Eng. vol. and W. Jun. [4] J. 2002.” in Proc. M. 1. degree in electrical engineering from the University of Toronto. New York: ASME Press. He received the BASc. “A linear dynamic model for asynchronous wind turbines with mechanical fluctuations. J. Montreal. [18] D. vol. IPST05. Devices. pp.. Kling. “Computer based real-time simulator for renewable energy converters. 280–284. L. B.” IEEE Trans. R. 1986. 2005. Kojabadi. for H < 0. Marian. Gules. Slootweg. de Haan. Boutot. in 1995 and 1997.