Energy 109 (2016) 294e309

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Comparative studies on control systems for a two-blade variablespeed wind turbine with a speed exclusion zone
Jian Yang a, Dongran Song a, b, Mi Dong a, *, Sifan Chen b, Libing Zou b, Josep M. Guerrero c
a

School of Information Science and Engineering, Central South University, Changsha, PR China
China Ming Yang Wind Power Group Co., Ltd., Zhongshan, PR China
c
Department of Energy Technology, Aalborg University, Denmark
b

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 2 November 2015
Received in revised form
17 March 2016
Accepted 24 April 2016

To avoid the coincidence between the tower nature frequency and rotational excitation frequency, a SEZ
(speed exclusion zone) must be built for a two-blade wind turbine with a full rated converter. According
to the literature, two methods of SEZ-crossing could be adopted. However, none of them have been
studied in industrial applications, and their performance remains unclear. Moreover, strategies on power
regulation operation are not covered. To fully investigate them, this paper develops two control systems
for a two-blade WT (wind turbines) with a SEZ. Because control systems play vital roles in determining
the performance of the WT, this paper focuses on comparative studies on their operation strategies and
performance. In these strategies, optimal designs are introduced to improve existing SEZ algorithms.
Moreover, to perform power regulation outside the SEZ, two operation modes are divided in the proposed down power regulation solutions. The developed control systems performance is conrmed by
simulations and eld tests. Two control systems present similar capabilities of power production and
SEZ-bridging. Nevertheless, at the cost of signicantly increased tower loads, one captures 1% more
energy than the other. Overall consideration must be made for the control system selection for a WT with
a SEZ.
2016 Elsevier Ltd. All rights reserved.

Keywords:
Two-blade variable speed wind turbine
Control system
Speed exclusion zone
Tower resonance
Power capture
Tower loads

1. Introduction
A wind turbine system is a system that converts mechanical
energy obtained from wind into electrical energy through a
generator. It can be categorized by types of generators used, power
control methods, constant- or variable-speed operations, and
methods of interconnection with the grid [1]. To ensure high performance while minimizing costs, new solutions are developed
constantly for WT (wind turbines) (). Fundamental changes have
been addressed, such as continuously variable transmissions [2,3]
and new sensing technologies [4,5]. Meanwhile, advanced control
algorithms have been widely studied, such as soft computing
techniques [6,7] and sustainable control [8]. Despite the development of good concepts in recent years, engineering and science
challenges still exist.
Modern high power WTs are typically designed in a variablespeed type, capturing wind energy and reducing the mechanic

* Corresponding author.
E-mail address: mi.dong@csu.edu.cn (M. Dong).
http://dx.doi.org/10.1016/j.energy.2016.04.106
0360-5442/ 2016 Elsevier Ltd. All rights reserved.

loads effectively. However, a wide speed operation region allows

the resonance between rotor rotary frequency and natural frequencies of other structural components. To tackle underlying
problems, some methodologies are applied during the design
phase, including natural characteristic calculations and potential
resonance analyses [9]. Considerations include not only the certain
gap reserved among the natural frequencies of the blades, tower
and driver train but also the avoidance of coincidences among
natural frequencies and external resonance force [10]. It is recommended that the eigen-frequency of the rotor blade be outside
a 12% range of the rotational frequency of the WT and the lowest
mode frequency of the tower be kept outside ranges dened as
10% of the rotor frequency and 10% of the blade passing frequency [11]. In practical applications, the tower resonance is
dangerous because it results in the vibration of the whole WT set.
For a three-blade WT, it is possible to move the natural frequency
to the region between 1 P and 3 P by redesigning the tower's
thickness and radius. However, this approach does not work for a
two-blade WT because changing the tower's natural frequency to
be lower than 1 P or higher than 2 P will greatly increase the cost.
Therefore, to prevent the WT from operating in the SEZ (speed

J. Yang et al. / Energy 109 (2016) 294e309

Nomenclature

qset,qm
wA,wB,wC, wD
wb,wc
wo
wr p
wr pl , wr ph
wr t
wr tl ,wr th
wr m
Topt
Pset

the pitch angle set-point and the measured pitch

angle.
four speed points at optimal tip speed section.
the lower and upper speed boundaries of the
speed exclusion zone.
the critical speed of a two-blade wind turbine.
the speed reference of the pitch controller
wr p in low power mode and high power mode
the speed reference of the PI torque controller
wr t in low power mode and high power mode
the measured rotor speed
the optimal generator torque
the power command from wind farm controller

exclusion zone), the only feasible way is to redesign the control

system.
Control algorithms for a WT with a SEZ are described in
previous works [12e17]. Among these works, two control approaches can be distinguished. The rst one, recorded in [12], is
based on the torque control with a conventional lookup table.
The second one, proposed in [13e15], is developed based on a
proportional integral (PI) torque control method. In both of
them, a certain speed region, including the critical speed and its
vicinity, is built up to form the SEZ. Differences between them
are the means of establishing and bridging over the SEZ. The rst
approach is to create an ambiguous function between rotor
speed and generator torque, so that the generator can accelerate
to cross the SEZ, through an unbalanced relation between the
aerodynamic torque and demanded generator torque. The second is to gradually adjust the speed reference from one xed
speed boundary to another. Despite the two approaches available, studies about their applications in real wind turbines are
few. As far as we know, only in [16] are different widths of SEZs,
based on the second approach, investigated and validated on a
1.3 kW test rig. In addition, in [17], the rst approach is
employed for the design of a two-bladed WT's control system. In
the wind energy industry, control strategy validation through
eld trials is vital and irreplaceable. Based on eld trials and
related data analysis, for the control approach applied in [17],
two drawbacks are exposed: i) the experimental turbine fails to
cross over the SEZ under certain wind conditions; ii) the power
capture performance is unsatisfactory. Therefore, optimization
techniques must be further investigated. Moreover, the performance of available control approaches is not studied in the
literature, which is vital for WT designers and owners to select a
control system for a WT with a SEZ.
The control strategies discussed above are utilized only to
maximize power production while maintaining the desired rotor
speed and avoiding equipment overloads [18]. Currently, wind
farms are required to play roles similar to those of conventional
power plants in power systems [19]. As a result, WTs are commanded to regulate power according to the power set-points set by
central control systems of wind farms. Thus, these WTs must
perform three power generation tasks: power optimization, power
limitation, and power regulation. These three tasks are fullled in a
certain operation region, constrained by the rotor speed. In the case
of a WT without a SEZ, it is necessary only to limit the rotor speed to
the speed reference by the pitch controller under the power limitation. To date, many studies have focused on generic WTs,

295

Prated, Pm
Pset b

the rated power and the measured electrical power

the power set-point to the boost converter
controller
Pl
the power set-point from the lookup torque
controller
Pl l , Pl h
Pl in low power mode and high power mode
PB, PC
the power set-points at rotor speeds wB and wC
PE, PF
the upper and lower power limits at the speed
boundaries wb and wc
Pl1, Pl2, Pl3
three power limits at the speed boundary wc
Ph1, Ph2, Ph3 three power limits at the speed boundary wb
ttask,tcross
time of control system task and set time to cross the
SEZ
Hs, Hm,Hl
three hysteresis time
Mx, My, Mz the rolling, nodding and yawing moments

especially those with doubly fed induction generators [20e24]. For

a WT with a SEZ, specic control strategies must be studied, which
are required to perform power generation tasks while maintaining
the rotor speed outside the SEZ. However, there is no literature on
such strategies.
The objective of this work is to perform comparative studies of
control systems for a two-bladed WT with a SEZ. Starting from
available methods, this paper develops two control systems to
perform power generation tasks while bypassing the SEZ. For both of
them, three operation strategies are discussed, including power
optimization, power limitation and power regulation. In such strategies, optimal designs are introduced to improve existing SEZ algorithms and solve their problems. Moreover, to perform power
regulation outside SEZ, simple yet effective down power regulation
solutions are presented. The control strategies are veried through
simulations and eld tests. Their performance is evaluated according
to International Electro-technical Commission (IEC) standards.

2. Studied two-blade WT
2.1. Basic information
The studied WT is a two-blade 3.0 MW super compact drive
machine. It is manufactured by China Ming Yang Wind Power
Company, and its specications are shown in Table 1.
The WT has a super compact structure, and its main body
consists of two parts: the energy conversion system and its supporting tubular steel tower. The energy conversion system diagram
is shown in Fig. 1, including a blade rotor, a low-ratio gearbox, a

Table 1
Specications of the studied WT.
Parameters

Value

Rotor diameter
Number of rotor blades
Rated electrical power
Rotor speed range
Nominal rotor speed
Rated wind speed
Rotor moment of inertia

110 m
2
3000 kW
6.0e21.0 rpm
16.2 rpm
12.2 m/s

Generator moment of inertia

2:1
103 kg$m2

23.94
3 m/s
20 m/s

Gearbox ratio
Cut-in wind speed
Cut-out wind speed

1:5
107 kg$m2

296

J. Yang et al. / Energy 109 (2016) 294e309

Fig. 1. Energy conversion system diagram of the studied WT.

PMSG (permanent magnet synchronous generator) and a full-scale

power converter (consisting of diode rectiers, DC-Boost converters and grid inverters).
2.2. Characteristic curves of the studied WT
By using the Bladed software application [25], the characteristic
curves of the studied machine are obtained. Curves of the aerodynamic power coefcient (Cp) vs. the TSR (tip speed ratio), and
thrust coefcient (Ct) vs. TSR are shown in Fig. 2. Conventionally,
the pitch angle and TSR for maximum Cp acquisition are called the
optimal pitch angle and optimal TSR, respectively. Fig. 2 shows that,
for the studied WT, the maximum Cp is 0.454, and the corresponding optimal pitch angle and optimal TSR are 0 and 10.5,
respectively. Meanwhile, the optimal pitch angle is changing in the
range of 1 to 1 along with the TSR variation in the scope of 8e12.
In addition, Ct increases with decreasing pitch angle when the TSR
is a constant. According to [26], tower loads are proportional to the
thrust coefcient. Therefore, to reduce tower loads, it is benecial
to maintain a large pitch angle and a lower TSR.
Fig. 3 shows the Campbell diagram of the studied machine, in
which the coupled modes are functions of rotor speeds. At a rotor
speed of approximately 10 rpm, the blade passing frequency 2 P
crosses the frequencies of the lowest two tower modes in the stationary frame. To avoid excessive excitation of these modes, a SEZ must
be set up, which is handled by the control system studied in this work.
2.3. Control system architecture of the studied WT
The control system of a modern WT is usually divided into two
levels: the generator control and the WT control. These two control
levels are characterized by different bandwidths [22]. For the
studied turbine, a unied control architecture is adopted, running a
WT and ensuring energy injection from power converters into the

electricity network at maximum efciency [27]. Fig. 4 illustrates the

architecture, in which a Siemens IPC P320 is the control unit. Based
on the Pronet protocol, the power converter and other major
components are controlled by one unique controller within two
task periods of 250 ms and 10 ms, respectively. With this unied
architecture, relations and constraints among different control
levels become clear. Therefore, it turns out to be quite convenient to
implement control algorithms for the WT.

3. Operation strategies of the studied WT

Considering the power generation system, there are three
operation tasks for modern WTs [24]:
 Limiting the output power to the rated power for high wind
speeds (power limitation);
 Maximizing the power extracted from the wind for a wide range
of wind speeds (power optimization);
 Adjusting both active and reactive powers to set-points ordered
by the wind farm control system (known as power regulation
operation or deloaded operation or de-rating operation).
When these three tasks are executed, the rotor speed must be
maintained in the predened range. Otherwise, the machine would
suffer from overload. For a generic WT with no SEZs, its rotor speed
is controlled within a continuous operation zone limited by the cutin speed and the rated speed. However, for a WT with a SEZ, its
rotor speed not only is constrained by the cut-in and rated speeds
but also must be held away from the critical speed. Separated by the
SEZ, there are two operation zones: a low speed zone and a high
speed zone. Therefore, the existence of the SEZ affects such WTs
power optimization and power regulation operations. For the sake
of simplicity, SEZ-related torque control and pitch control strategies
will be discussed, whereas other controls unaffected by the SEZ are

Fig. 2. The aerodynamic power coefcient and thrust coefcient curves of the studied WT.

J. Yang et al. / Energy 109 (2016) 294e309

297

Fig. 3. Campbell diagram of the studied WT.

Fig. 4. Control system architecture of the studied WT.

neglected. For the studied WT, the pitch controller is used to control
a hydraulic pitch system and the torque controller is used to control
the DC-Boost converter. The common operation strategies
employed are summarized as follows.
 In the power limitation condition, the operation strategy for the
studied WT is mainly in charge of the pitch controller. The rotor
speed is controlled to be the rated value by the pitch controller,
and the generator torque is limited to the rated value by the DCBoost converter controller. As illustrated in Fig. 5, the pitch
controller contains three main parts: a PD controller and two
fuzzy logic units. Regarding the PD controller, its input is the
error between the reference wr p and the feedback wr m , and its
output is the set value of pitch speed to the hydraulic proportional valves. Two fuzzy logic units, FC1 and FC2, are designed

for the pitch bias determination and over-speed problem prevention, respectively [28].
 In the power optimization condition, the torque controller is
responsible for the optimized operation, and the pitch angle is
maintained at its optimal value by the pitch controller. In this case,
the rotor speed is controlled by the torque controllerdnot only to
track the optimal TSR outside the SEZ but also to cross over the SEZ.
 In the power regulation condition, the operation strategy requires cooperation between the pitch controller and the torque
controller. According to [24], three control strategies are available for DFIG WTs with no SEZs. Recall that down power regulation mainly involves the scheduling power and rotor speed
set-points; these control strategies can also be employed by
the control system of PMSG WTs. However, a special down power strategy must be developed for a WT with a SEZ.

Pset _ b
Gain
scheduling

FC1

wr _ p -

PD
+

wr _ m

L1
+

dt

set

Com +

FC2
Fig. 5. The structure diagram of the pitch controller.

L2
+

298

J. Yang et al. / Energy 109 (2016) 294e309

4. Control systems of the studied WT

As mentioned previously, there are two different control approaches for a WT with a SEZ in previous works. Based on them,
two control systems (denoted as Control Systems 1 and 2) are
developed for the studied WT. Control System 1 is based specifically on [17], whereas Control System 2 is based on [16]. Meanwhile, optimal techniques are presented to improve conventional
SEZ-crossing methods. Furthermore, power regulation strategies
are proposed to full power output adjustment.

4.1. Control System 1

4.1.1. Structure of Control System 1
The structure of Control System 1 is illustrated in Fig. 6,
including four main parts: the pitch controller, the speed reference
unit, the DC-Boost converter controller and the power set-point
unit.
Based on Fig. 6, the operation strategies are summarized as
follows:
 Power limitation strategy: The reference wr p for the pitch
controller is the rated value, and the power set-point Pset b for
the DC-Boost converter controller is calculated based on the full
powerespeed curve and wr m .
 Power optimization strategy: The pitch angle is maintained at its
optimal value by the pitch controller, and the rotor speed is
adjusted by the torque control strategy explained as follows.
 Power regulation strategy: The down power regulation strategy
is divided into low power mode and high power mode. Both are
determined by the power command Pset from the wind farm
controller and the power division point PE, which corresponds to
the upper power limit at the lower speed boundary of the SEZ.
When Pset > PE, the WT operates in high power mode: wr p takes
wr ph , which is generated from the high powerespeed curve and
Pset. Meanwhile, Pset b takes Pl h , which is derived from the full
powerespeed curve and wr m . When Pset  PE, the WT operates
in low power mode: wr p takes wr pl generated from the low
powerespeed curve and Pset, whereas Pset b takes Pl l derived
from the low powerespeed curve and wr m .

4.1.2. Optimized torque control scheme in Control System 1

The torque control scheme is illustrated in Fig. 7, including three
parts: the power set-point unit, the bias unit, and the DC-Boost
converter controller.
The DC-Boost converter controller controls the generator torque. A PI controller is employed to control the Boost converter
current, the set-point Iset b of which is obtained by dividing the
power set-point Pset b by the rectier's output voltage Um b . FC1, a
fuzzy logic unit, is used to decouple the pitch controller and the
torque controller [28]. The power set-point unit determines the
powererotor speed lookup table, which includes normal points
predened according to the WT's aerodynamic data and special
points related to the SEZ. In this work, eight pairs of powererotor
speed points are shown in Table 2. In [17], we proved that for a twoblade WT, proper widths of the SEZ and its neighbouring zones can
be 10%. Here, the SEZ is preset to 9e11 rpm, its two neighbouring
zones are dened as 8.2e9 rpm and 11e11.9 rpm, and the upper
and lower power limits at two speed boundaries of the SEZ are 18%
and 2%, respectively.
To enhance the SEZ-bridging capability under different wind
conditions, a hysteresis technique is presented to replace the predened powererotor speed points within the SEZ. As illustrated in
Fig. 8, the technique is described as follows: when the rotor speed is
increased above the lower speed boundarywb(9.0 rpm), the power
set-point Pset b is decreased with a certain rate to the end point
PF(2.0%); when the rotor speed is decreased below the upper speed
boundarywc(11.0 rpm), the power set-point is increased with a
certain rate to the end point PE(18.0%).
4.2. Control System 2
4.2.1. Structure of Control System 2
Similar to Control System 1, Control System 2 also contains four
main parts: the pitch controller, the speed reference unit, the DCBoost converter controller and the power set-point unit. Its structure is illustrated in Fig. 9.
Based on Fig. 9, the operation strategies for the control system
are as follows:
 Power limitation strategy: both speed references of the pitch
controller and the PI torque controller are the rated value. As a

Fig. 6. Structure of Control System 1.

J. Yang et al. / Energy 109 (2016) 294e309

299

Fig. 7. The torque control scheme in Control System 1.

Table 2
Powererotor speed lookup table.
Measured value of rotor speed (rpm)

result, the rotor speed and generator torque are maintained at

their rated value by the pitch controller and the torque
controller, respectively.
 Power optimization strategy: the pitch angle is kept at its
optimal value by the pitch controller, and the rotor speed is
controlled by the PI torque control strategy.
 Power regulation strategy: wr p of the pitch controller and Pset b
of the Boost converter controller are derived based on two power modes as mentioned earlier. When Pset > Ph3, the WT
operates in high power mode: wr p takes wr ph , which is
calculated from the high powerespeed curve and Pset, and wr t
takes wr th , which is obtained from the full powerespeed curve
and wr m . When Pset  Ph3, the WT operates in low power mode:
wr p takes wr pl , which is calculated from the low powerespeed
curve and Pset, whereas wr t takes wr tl , which is obtained from
the low powerespeed curve and wr m . Meanwhile, Pset b is
derived from the output of the PI torque controller and wr m .

Power set-point (100%)

6.0
8.2
9.0
11.0
11.9
13.7
15.0
16.2

0.0
8.0
18.0
2.0
17.0
35.0
48.0

100.0

4.2.2. Optimized torque control strategy in Control System 2

The torque control strategy is illustrated in Fig. 10 and also
contains three parts: the power set-point unit, the bias unit, and the
boost converter controller. The Boost converter controller and the
bias unit are the same as those in Control System 1. The power setpoint unit refers to a PI torque controller and a mode selection unit.
The design of the PI controller is a routine with the assistance of
Bladed. Here, it is worth noting that the controller gains are dened

Fig. 8. Powererotor speed curve in Control System 1.

Select mode
1: Low power mode
2: High power mode

Pset
Ph3

Pset
Ph3
Pset
Ph3

Pset

1
wr _ p

Low powerspeed curve

Pitch
controller

wr _ ph

Pset > Ph3

2
High powerspeed curve

Pset

wr _ tl
Tlim it

Ph3
wr _ m

Pset
Ph3

wr _ pl

Ph3

1
wr _ t
Tlim it

Low powerspeed curve

wr _ th
Pset > Ph3

Tlim it

Full power-speed
curve
Fig. 9. Structure of Control System 2.

PI torque
controller

Pset _ b

Boost
converter
controller

300

J. Yang et al. / Energy 109 (2016) 294e309

Fig. 10. The torque control scheme in Control System 2.

wA  wb, wc  wD and wb  wc, respectively. Its value changes, when

a mode transition (WT_lowhigh_transition/WT_highlow_transition)
is triggered by the variation of Pm.The mode transition is determined by the time duration of the compared result between Pm and
the predened power limit. To cross the SEZ under various winds, a
variable transition technique is employed. In this technique, conditions for WT_lowhigh_transition and WT_highlow_transition are
summarized in Table 3. Predened are several parametersdnamely, six power limits (Ph3,Ph2,Ph1,Pl3,Pl2 and Pl1), three hysteresis times (Hl,Hm andHs), and a crossing time tcross with two
values. For the studied WT, their values are given in Table 4.

in terms of generator torque with respect to the high speed shaft. Its
parameters are given as kp 8300.0[Nm/(rad/s)], k i 1300.0[Nm/
(rad/s)], and the gain scheduling factor is 1.5. In addition, the
optimal generator torque Topt is calculated as Topt kw2r m [29]. For
the studied WT, k 14322[Nm/(rad/s)2].
Calculating the speed reference and torque limits for the PI
torque controller, the mode selection unit is in charge of the SEZ
algorithm. To carry out the comparison to Control System 1, the SEZ
with same range of 9e11 rpm is preset. Based on the PI torque
controller, the powererotor speed characteristic curve of the WT is
shown in Fig. 11.
In the mode selection unit, three modes are dened according to
the WT's operation in different rotor speed ranges. In Fig. 11, three
operation modes, named low speed mode, high speed mode and
SEZ mode, correspond to rotor speed ranges of wA  wb, wc  wD
and wb  wc, respectively. The speed reference wr t and torque limit
Tlimit for the PI torque controller in these modes are calculated by
the algorithm described in pseudo code as follows:

As illustrated in Figs. 8 and 11, there are two powererotor speed

characteristic curves for the WT with two control systems. In view
of the WT performance, which is dependent on its powererotor
speed characteristic curve, control systems impacts on the per-

In the pseudo code above, WT_speed_mode_ag is determined

by measured rotor speed wr m and measured electrical power Pm. It
takes one of three valuesdnamely, WT_lowspeed_mode, WT_high
speed_mode and WT_TEZ_mode, based on the location of wr m in

formance in terms of power capture and tower loads can be

assessed.
Considering power capture, the performance of the WT with
Control System 1 is inferior to that with Control System 2. On one

4.3. Assessment of two control systems

J. Yang et al. / Energy 109 (2016) 294e309

301

Although a basic assessment has been obtained based on analyses of the operation principles of two control systems, it is
indispensable to perform a detailed performance comparison
through nonlinear simulations and eld tests, which is important to
give designers the condence to choose a suitable controller for a
WT with a SEZ.
5. Performance comparisons of two control systems
5.1. Comparative study based on simulation

Fig. 11. Powererotor speed curve in Control System 2.

Table 3
Transition conditions in variable transition technique.
Transition type

Condition

WT_lowhigh_transition

T(Pm
T(Pm
T(Pm
T(Pm
T(Pm
T(Pm

WT_highlow_transition

>
>
>
<
<
<

Crossing
time (tcross)

Ph3) > Hs
Ph2) > Hm
Ph1) > Hl
Pl3) > Hs
Pl2) > Hm
Pl1) > Hl

ts
tl
tl
ts
tl
tl

side, Control System 2 obtains better TSR-tracking than the Control

System does at four speed points (wA, wb, wc and wD). On the other
side, two optimal tip speed sections (wA  wb andwc  wD) are
better handled by the PI torque control strategy in Control System 2
than by eight powererotor speed points dened in the lookup table
of Control System 1. Regarding these two sides, the WT with Control System 2 would produce more power. However, the evaluation
is established on the static energy balance theory, which is valid
only on the premise that the WT rotor has a small inertia of
moment and the winds change slowly.
Regarding tower loads, the performance of the WT with Control
System 1 outweighs that with Control System 2. This deduction is
based on two aspects. For the rst aspect, tower loads are affected
by the WT's operation points outside the SEZ. According to the
analyses of the resonance problem discussed in [17], tower vibration amplitude decreases with increasing difference from the critical speed. Therefore, tower loads can be determined by the degree
by which the rotor speed converges to the critical speed. As illustrated in Figs. 8 and 11, Control System 2 possibly operates the WT
at the speed boundary of the SEZ, whereas Control System 1works
in neighbouring zones of the SEZ. In this aspect, Control System 1
produces fewer tower loads than Control System 2. For the second
aspect, tower loads increase with increasing instances of SEZcrossing. The crossing instances with Control System 2 are more
abundant than those with Control System 1 because the condition
for SEZ-crossing is easier to satisfy in Control System 2. When the
WT with Control System 1 can operate in neighbouring zones, that
with Control System 2 will work at the speed boundary of the SEZ.
Therefore, the WT with Control System 2 produces more tower
loads than that with Control System 1.

In this section, two works are performed through detailed

simulations with Bladed: control algorithm validation and performance evaluation in terms of structure loads and power production. To enhance the power capture capability, the control
algorithm in Control System 2 is further improved by adjusting the
optimal pitch angle. The details are as follows: the measured rotor
speed and electrical power are used to examine the TSR, and pitch
angles are adjusted to the optimal value based on the calculated
TSR. The correlation between the optimal pitch and the TSR can be
obtained by checking the Cp curves shown in Fig. 2. Therefore, three
controllers are developed as the external dynamic library.
Controller 1, Controller 2 and Controller 3 refer to control algorithm
1 (in Control System 1), control algorithm 2and updated control
algorithm 2 (in Control System 2), respectively. In view of the fact
that the simulation running time is shorter than the WT's real
operation time, the hysteresis times employed by Controllers 2 and
3 are shortened to 60 s, 10 s and 1 s in simulations.
5.1.1. Validation of the proposed control algorithms
Regarding SEZ-related controls, two operation scenarios, power
optimization operation and power regulation operation, are
considered. To validate the effectiveness of the controllers, 13
simulation tests are implemented, which are preset by the two
scenarios with single point history and 3D turbulent winds. The
single point history winds are set to step winds from 3 to 12 m/s,
and 3D turbulent winds are dened with 6 m/s mean wind speed of
three typical turbulence intensities (14%, 16%, and 18%). In this
work, for the sake of simplicity, only two representative simulation
results are shown; one is based on the power optimization case,
and the other is the power regulation case with winds of 16% turbulence intensity. Among numerous simulation data obtained from
Bladed, six signals are shown: wind speed, rotor speed, output
electrical power, pitch angle, and nacelle sideeside and foreeaft
accelerations. The simulation results with Controller 1, Controller 2,
and Controller 3 are plotted in black, red and green, respectively.
The simulation results of power optimization are illustrated in
Fig. 12a. It is clear that all three controllers succeed in bridging the
SEZ. However, their differences are obvious. First, the instances of
SEZ-crossing are not the same. Three instances occur for Controllers
2 and 3, whereas there is only one for Controller 1. Second, before
and after SEZ-crossing, Controllers 2 and 3 maintain rotor speeds
nearer the speed boundaries of the SEZ than Controller 1. Third,
except for several points, the WT's nacelle accelerations with
Controller 1 are slightly smaller than those from other two controllers. These differences impact power capture and tower loads,
which will be numerically presented in the next section.

Table 4
Parameters for the studied WT.
Parameter

Hs

Hm

Hl

Ph3

Ph2

Ph1

Pl3

Pl2

Pl1

tl

ts

Value

3s

30 s

300 s

540 kW

440 kW

410 kW

200 kW

325 kW

350 kW

15 s

10 s

302

J. Yang et al. / Energy 109 (2016) 294e309

Fig. 12. Simulation results among three controllers: (a) at power optimization case and (b) at deloaded case.

In the down power regulation case, the power regulation demand is set to 450 kW before 290 s and increased to 550 kW at
290 s with a ramping rate of 50 kW/s. The simulation results
illustrated in Fig. 12b it show that all three controllers succeed in
following power commands while bypassing the SEZ. Four differences are distinguishable. First, the SEZ-crossing instances are
different. Three instances occur for Controllers 2 and 3, whereas

there is only one for Controller 1. Second, before and after SEZcrossing, the rotor speeds with Controllers 2 and 3 are upheld
tightly to the speed boundaries of the SEZ, whereas that with
Controller 1 is locate in the SEZ's neighbouring zones. Third, both
the nacelle foreeaft and sideeside acceleration amplitudes with
Controller 1 are obviously smaller than those with the other two
controllers. Finally, pitch actions behave differently when the

J. Yang et al. / Energy 109 (2016) 294e309

Table 5
Summarized numerical results from Fig. 12a.

303

Table 7
The DELs of four components with SN4.

Controller

Mx (MNm)

My (MNm)

Mz (MNm)

Averaged power (MW)

Component

Mx (kNm)

My (kNm)

Mz (kNm)

1
2
3

3.757
7.413
7.361

7.298
8.917
8.899

1.103
1.066
1.066

0.502
0.509
0.510

Blade root (steel)

Blade root (GRP)
Hub (steel)
Yaw bearing (steel)
Tower bottom (steel)

5640.79
5591.17
393.36
452.34
5003.96

2281.65
4204.39
2491.80
2472.33
9983.03

57.40
68.64
2491.84
2482.58
2482.45

output power reaches the power demand. These differences have

direct impacts on the WT's performance, which will be presented in
the next section.
5.1.2. Performance comparisons with simulation results
For performance comparisons, three simulation results are
presented. The rst two from the discussed simulation cases are
used for preliminary evaluation. The third is taken for detailed
comparisons, which is obtained from a complete set of simulations
according to the design requirement of the IEC standard [30].
To preliminarily evaluate the WT's performance with different
controllers, the averaging process function provided by Bladed is
used to calculate the averaged power and averaged tower moments. The numerical results from Fig. 12a and b are summarized in
Tables 5 and 6, respectively. The results of the power optimization
case in Table 5 show that the averaged power production is similar
between Controllers 2 and 3, as are the tower moments. Compared
with Controller 2, there is a slight power output increase for
Controller 3. This result proves that only the pitch angles differ in
the two controllers. By comparing the results between Controller 1
and Controllers 2/3, obvious discrepancies are found. Controller 2
increases the averaged power by approximately 1.6% but doubles
the tower Mx moments and increases the My moments by more
than 20%. By checking the results of the deloaded case in Table 6, it
is found that Controllers 2 and 3 are similar in producing power and
tower moments. This ts with the fact that the trajectories of rotor
speed and pitch angle in Fig. 12b are almost overlapped in the two
controllers. Compared with Controller 1, Controller 2 increases the
power output by more than 3.1% yet increases the tower Mx and My
moments by more than 85% and 22%, respectively. Because pitch
actions under the power regulation operation directly affect the
aerodynamic thrust and thus the tower moments, these comparative results are different from those in Table 5.
In accordance with the IEC standard [30], a complete set of
simulation series is performed to calculate the design loads, which
is essential to evaluate the controller impact on the loads before
carrying out the eld testing. In the simulation series, different
winds are dened based on the analysis of wind resource measurement at the wind farm site where the studied machines are
deployed: the annual average wind speed at hub height is 6.42 m/s,
and the characteristic turbulence intensity at 15 m/s is 12%. Because
the same pitch control algorithms and supervisory control strategies are performed in all three controllers, fatigue loads rather than
extreme loads are mainly affected. Therefore, performance comparisons are conducted on fatigue loads and power production.
To understand component loads, the coordinate system for load
outputs should be dened. The coordinate systems of Bladed are
given in the Appendix. With the coordinate systems, the damage

Table 6
Summarized numerical results from Fig. 12b.
Controller

Mx (MNm)

My (MNm)

Mz (MNm)

Averaged power (MW)

1
2
3

4.723
8.640
8.592

9.115
10.880
10.812

0.805
0.973
0.972

0.381
0.393
0.393

equivalent loads (DELs) are calculated based on the assumption

that the WT's lifetime is 20 years and the press cycle time is
1.0E 08. By using a Wohler exponent of 4 for steel and 10 for the
glass reinforced plastic (GRP), the DELs of four components (steel
blade root, GRP blade root, hub, yaw bearing, and tower bottom)
with Controller 1 are shown in Table 7. By treating its results as the
baseline, the comparative results of Controllers 2/3 are presented in
Fig. 13 (the GRP blade root DELs of the three controllers are almost
equal and thus are not included). The comparative results are
summarized as follows:
 Tower bottom DELs: the Mx DEL is increased by nearly 60%, and
the My DEL is increased by more than 10%.
 Other DELs: no signicant change is found; that is, only the
increments of the My DELs of the blade roots reach 5%, whereas
the others are less than 3.5%.
By comparing the DELs between Controllers 2 and 3, it is found
that the related DELs are very similar to each other. Only an
increment of 2% for the Mz DEL of the blade root is produced by
Controller 3, whereas other differences are less than 1%. The
optimal pitch angle adjustment applied to Controller 3 accounts for
the increased Mz DEL of the blade root.
To observe the contributions of different wind speeds to the
tower's DELs, the tower bottom Mx and My DELs of design load case
(DLC) 1.2 are shown in Fig. 14. At wind speeds of 4 m/s and 6 m/s,
the tower bottom Mx and My DELs with Controllers 2 and 3 almost
double those with Controller 1. The reason is that the rotor speeds
with Controllers 2 and 3 at low winds are limited to the speed
boundaries of the SEZ. At wind speeds of 8 m/s and 10 m/s, the
tower bottom Mx DELs of the three controllers are almost equal,
whereas the tower bottom My DELs of Controllers 2 and 3 are
higher. This is because the rotor speeds with the other two controllers have reached the rated speed, but that with Controller 1 has
not. Therefore, a larger thrust is produced by higher TSRs. Above the
rated winds, slight differences among tower DELs are shown, which
are affected by torque demand differences in turbulent winds.
Based on the simulation results of DLC 1.2, the averaged power
at different wind speeds is calculated. As shown in Fig. 15, results
from Controllers 2 and 3 are compared with the baselinedthat is,
the result of Controller 1. It is clear that different averaged power is
produced by the three controllers. Compared with Controller 1, the
other two controllers increase the power production at wind
speeds of 8 m/s, 10 m/s and 12 m/s but decrease the power production at other wind speeds. The increased power production at
medium wind speed is caused by the optimal TSR tracked by these
two controllers. The decreased power production by less than 0.3%
above the rated winds can be explained by the power loss model,
which is determined by the rotor speed and generated power. The
lower power production at 4 m/s and 6 m/s seems to contradict the
results shown in Table 5. However, it is reasonable when considering the inuence of different turbulence intensities.
To assess the overall power production performance of the three
controllers, the AEP (annual energy production) is calculated based
on the averaged power at DLC 1.2 and the wind characteristic on the

304

J. Yang et al. / Energy 109 (2016) 294e309

Fig. 13. DEL comparisons of four components among three controllers.

Fig. 14. Comparisons of tower bottom Mx and My DELs at DLC 1.2 among three controllers.

Fig. 15. Averaged electrical power comparison at DLC 1.2 among three controllers.

wind farm site. The AEP with Controller 1 is 6716.47 MWh, whereas
that of Controllers 2 and 3 is slightly higher with results of
6762.78 MWh and 6764.19 MWh, respectively. Thus, there is a 0.7%
difference in AEP, which must pay for a 10% increase in the tower
bottom DELs.
5.2. Comparative study through eld tests
After validation through simulations, the control algorithms are
transferred into the programmable logic controller (PLC) program
and then integrated into the control systems of the studied WT. The

eld testing site is located in a wind farm on the coast of southern

China, in which there are ten 3 MW two-blade WTs and seven 2 MW
three-blade WTs. Before the testing, the control systems of the 3 MW
WTs employ the lookup table torque control algorithm. To carry out
eld tests, two of the ten machines, named N15 and N16, are chosen as
testing objectives because their locations and their power production
performance are quite similar. Control System 1 is used to update N16,
and Control System 2 is tested on N15. Because Controller 3 produces
more power than Controller 2 in simulations, its updated algorithm is
adopted in Control System 2 for testing. The eld tests were carried
out in June 2015 for a duration of three weeks.

J. Yang et al. / Energy 109 (2016) 294e309

305

Fig. 16. Cross over curves of SEZ on eld testing for: (a) N15 with Control System 2 and (b) N16 with Control System 1.

5.2.1. Field testing results

In the eld testing, the control systems are tested in different
wind conditions under normal grid operations. Although power
regulation strategy is developed in the control system, this function
is inactivated during the tests because in that wind farm, there is no
such requirement to date.
Because different SEZ algorithms are employed by two control
systems, the results of SEZ-crossing recorded in a 10 ms period are
shown in Fig. 16a and b.
It can be observed that when the SEZ is crossed over, the output
power varies signicantly. The peaks of the output power are
750 kW and 540 kW for N15 and N16, respectively. Meanwhile, the

crossovers of the SEZ occur at different wind speeds: near 5.5 m/s
for N15 and 4.5 m/s for N16. In addition, both nacelle foreeaft and
sideeside accelerations increase with more transitions between
two speed zones. The different acceleration amplitudes could be
the results of varying winds experienced by the whole rotor.
To further illustrate the different behaviour of the two control
systems, another eld testing result recorded for one day (24 h) is
presented in Fig. 17. Because the result is with a 10 s sampling
period, nacelle acceleration signals are excluded. It is very clear
that the rotor speed trajectories and SEZ-crossing instances are
different for N15 and N16, whereas wind conditions are surprisingly similar.

306

J. Yang et al. / Energy 109 (2016) 294e309

Fig. 17. Field testing curves on one typical day (black curves for N15 and red curves for N16). (For interpretation of the references to color in this gure legend, the reader is referred
to the web version of this article.)

5.2.2. Statistics analysis of eld testing data

Because load measurement devices are not equipped in testing
WTs, only power performance evaluation is conducted by referring
to the IEC standard [31]. The data collection is performed between
10 July and 10 August and recorded with 10 min averaged values.
Four measurable data points (wind speed, rotor speed, output power, and pitch angle) are collected and form a valid dataset after
removing corrupted data. Based on the valid dataset, 10 min averaged TSRs are computed from the instantaneous TSRs calculated at
each sampling point from the measured wind and rotor speeds.
Four characteristic curves of N15 and N16, including rotor speed

wind speed, pitch angleewind speed, TSRewind speed, and powerewind speed, are illustrated in Fig. 18. The former three characteristic curves are quite different, whereas the powerewind speed
curves are similar. This shows that one obvious SEZ ranges from 9 to
11 rpm, and the pitch angle of N16 is maintained at 3 , whereas the
pitch angle of N15 varies in the area of 2e4 . For the testing WTs,
the pitch angle of 3 is the optimal pitch angle (the same as the
0 illustrated in Fig. 2). TSRs of N15 are maintained near the optimal
value of 10.5 in the wind speed range of 4e5 m/s and 7e9 m/s,
whereas N16's TSRs are not constant in the whole wind speed
range. Meanwhile, TSRs of N15 and N16 are distributed in different

Fig. 18. Characteristic curve comparisons between N15 and N16.

J. Yang et al. / Energy 109 (2016) 294e309

Fig. 19. Averaged output power comparison between N15 and N16.

ranges. The TSRs of N15 are scattered between 9.0 and 11.5 at low
winds of 4e5 m/s and between 9.8 and 11.2 at high winds of
7e9 m/s. By comparison, the TSRs of N16 are more concentrated. It
means that the dynamic tracking TSR capability of N15 with Control
System 2 is inferior to that of N16.
To numerically compare the power capture performance of the
two control systems, the averaged output power of N15 and N16
are calculated. By setting the averaged power of Control System 1
as the baseline, comparative results are shown in Fig. 19. It is
obvious that N15 outputs more power below rated winds except at
the wind speed of 7 m/s. This result is consistent with the
TSRewind speed characteristic curve (shown in Fig. 18): at a 7 m/s
wind speed, the TSRs of N15 and N16 are near the optimal value of
10.5, whereas those of N16 are much denser. Compared with the
simulation results, more power is obviously produced by N15 in
the low wind range (3e5 m/s), whereas the power increasing
trend is similar in the high wind range (8e12 m/s). These differences can be explained by different time lengths and the inuence
of different turbulence, especially in low winds. Again, AEPs of N15
and N16 are calculated based on the eld testing results, which are
5763.1 MWh and 5695.8 MWh, respectively. It is proved that N15
with Control System 2 produces more power than N16 with
Control System 1. However, the AEP obtained from eld testing
results is less, approximately 15%, than that obtained by simulation, for which the possible reasons could be the wake loss and
model tolerance.
6. Conclusions
This paper presents a comparative study on two control systems
for a two-blade WT with a SEZ, which is built to avoid tower
resonance. The SEZ of the studied WT is set up and bridged by an
appropriate torque control, performed through a Boost converter
controller at power optimization operation in collaboration with
the blade pitch control at power regulation operation.
In this paper, two control systems (Control Systems 1 and 2)
are developed based on existing torque control strategies, in which
three operation strategies have been performed. At power optimization operation, Control System 1 employs a conventional
lookup table torque control strategy, whereas Control System 2
uses a PI torque controller. To guarantee successful SEZ-crossing
under different wind conditions, a hysteresis technique and a
variable transition technique are performed in Control Systems 1
and 2, respectively. For power limitation operation, the two control systems use the same pitch angle controller. Regarding both

307

the power regulation and the SEZ to be handled at deloaded

operation, two power operation modes are divided based on the
comparative result between the upper power limit of the SEZ and
the power regulation command. In this way, the WT operates in
the low speed range with low power command and in the full
speed range with high power command. As a result, the WT can
produce maximal power while maintaining its rotor speed outside
the SEZ.
Based on analyses of their operation principles, the impact of
control systems on the WT performance is assessed: Control System
2 would produce more power at the cost of increased tower loads
compared with Control System 1. The assessment is further veried
through simulations and eld tests. For general operation cases
without down power regulation, detailed simulation tests are fullled according to the design requirement of IEC-64100. The simulation results illustrate the capability of developed control systems
to perform the discussed tasks. Meanwhile, the simulation results
show that, on the one hand, fatigue loads caused by Control System 2
are surely larger than those of Control System 1: increased DELs on
other components are less than 6%, but raised tower DELs are signicant, representing more than a 60% improvement and 10% increase for tower Mx DEL and My DEL, respectively; on the other
hand, 0.7% greater power production is obtained by Control System 2
compared with Control System 1. The detailed numerical results
have shown that the increased DELs are mainly contributed by a
wind speed range corresponding to the SEZ. Following the simulation tests, eld testing is implemented to validate the control systems and compare power production performance. The eld testing
results show that both control systems are capable of controlling the
WT to build up and cross over the SEZ. Again, it has been demonstrated that energy capture performance is enhanced by Control
System 2. According to a comparison of the results between simulations, an increased AEP of 1.1% is achieved by Control System 2.
The simulation results also reveal that, at power regulation
operation, Control System 2 produces more power than Control
System 1 at the cost of increased tower loads. However, in this
circumstance, there is a risk of frequent SEZ-crossings when the
power regulation command is switched between high power and
low power modes. Therefore, the WT would suffer from high tower
loads. In this case, it is necessary to design a proper wind farm
controller to send proper power commands to each WT with the
SEZ. Meanwhile, deliberate evaluation strategies are necessary to
carry out thorough comparisons because no applicable evaluation
standard is available to follow. These aspects would be the subject
of future publications.
Acknowledgements
This work is supported by the National Natural Science Foundation of China under Grant 61573384 and the National High
Technology Research and Development Program (863 Program) of
China under Grant 2015AA050604. This work is also nancially
supported by the Project of Innovation-driven Plan in Central South
University, No. 2015CX007 and the Fundamental Research Funds
for the Central Universities of Central South University under Grant
2015zzts050.
Appendix
The coordinate systems for load outputs in this study are
dened by Bladed. They are based on the GL convention and are
shown in following gures.

308

J. Yang et al. / Energy 109 (2016) 294e309

Fig. 20. Coordinate systems for load outputs.

References
[1] Marques J, Hey H. A survey on variable-speed wind turbine system. Proc Braz
Conf Electron Power 2003;1:732e8.
[2] Petkovic D, Cojbasic Z, Nikolic V, Shamshirband S, Kiah MLM, Anuar NB, et al.
Adaptive neuro-fuzzy maximal power extraction of wind turbine with
continuously variable transmission. Energy 2014;64:868e74.
[3] Shamshirband S, Petkovic D, Amini A, Anuar BN, Nikolic V, Cojbasic Z, et al.
Support vector regression methodology for wind turbine reaction torque
prediction with power-split hydrostatic continuous variable transmission.
Energy 2014;67:623e30.
[4] Newsom RK, Berg LK, Shaw WJ, Fischer ML. Turbine-scale wind eld measurements using dual-Doppler lidar. Wind Energy 2015;18(2):219e35.
[5] Pena A, Hasager CB, Gryning SE, Courtney M, Antoniou I, Mikkelsen T. Offshore
wind proling using light detection and ranging measurements. Wind Energy
2009;12(2):105e24.
[6] Nikolic V, Shamshirband S, Petkovic D, Mohammadi K, Cojbasic Z,
Altameem TA, et al. Wind wake inuence estimation on energy production of
wind farm by adaptive neuro-fuzzy methodology. Energy 2015;80:361e72.
[7] Shamshirband S, Petkovic D, Saboohi H, Anuar BN, Inayat I, Akib S, et al. Wind
turbine power coefcient estimation by soft computing methodologies:
comparative study. Energy Convers Manag 2014;81:520e6.
[8] Kanev S, Engelen TV, Engels W, Wei XK, Dong JF, Verhaegen M. Sustainable
control. 2012. http://www.ecn.nl/docs/library/report/2012/e12028.pdf.

[9] Yao XJ, Liu YM, Liu GD, XING ZX, Bao JQ. Vibration analysis and online condition monitoring technology for large wind turbine. J Shenyang Univ Technol
2008;29(6):627e32.
[10] Shan GK, Wang XD, Yao XJ, Zhang CC. Stability analysis on MW wind turbine.
Acta Energiae Solaris Sin 2008;29(7):786e91.
[11] Veritas N. Guidelines for design of wind turbines. Det Norske Veritas: Wind
Energy Department, Ris National Laboratory; 2002.
[12] Schaak P, Corten GP, Hooft ELV. Crossing resonance rotor speeds of wind
turbines. In: Proc. EWEC, Madrid, Spain; 2003.
[13] Bossanyi EA. The design of closed loop controllers for wind turbines. Wind
Energy 2000;3:149e63.
[14] Bossanyi EA. Wind turbine control for load reduction. Wind Energy 2003;6:
229e44.
[15] Bossanyi EA. Controller for 5MW reference turbine. 2009. http://www.
upwind.eu/.
[16] Licari J, Ekanayake JB, Jenkins N. Investigation of a speed exclusion zone to
prevent tower resonance in variable-speed wind turbines. IEEE Trans Sustain
Energy 2013;4:977e84.
[17] Song DR, Yang J, Dong M, Yan Q, Zhang B. Control strategy to avoid tower
resonance for two-blade variable-speed wind turbine. J Vib Shock 2015;34:
90e8.
[18] Miller NW, Sanchez-Gasca JJ, Price WW, Delmerico RW. Dynamic modeling of
GE 1.5 and 3.6 MW wind turbine-generators for stability simulations. In: IEEE
2003 power engineering society general meeting; 2003. p. 1977e83.

J. Yang et al. / Energy 109 (2016) 294e309

[19] Jauch C, Matevosyan J, Ackermann T, Bolik S. International comparison of
requirements for connection of wind turbines to power systems. Wind Energy
2005;8:295e306.
[20] Rodriguez-Amenedo JL, Arnalte S, Burgos JC. Automatic generation control of
a wind farm with variable speed wind turbines. IEEE Trans Energy Convers
2002;17(2):279e84.
[21] Hansen AD, Sorensen P, Iov F, Blaabjeg F. Centralised power control of wind
farm with doubly fed induction generators. Renew Energy 2006;31(7):
935e51.
[22] Sorensen P, Hansen AD, Iov F, Blaabjeg F, Donovan MH. Wind farm models
and control strategies. Roskilde, Denmark: Ris National Laboratory; 2005.
Technical Report.
[23] De Almeida RG, Castronuovo ED, Pecas Lopes JA. Optimum generation control
in wind parks when carrying out system operator requests. IEEE Trans Power
Syst 2006;21(2):718e25.
[24] Fernandez LM, Garcia CA, Jurado F. Comparative study on the performance of
control systems for doubly fed induction generator (DFIG) wind turbines
operating with power regulation. Energy 2008;33(9):1438e52.

309

[25] GH bladed user manual. Garrad Hassan and Partners Ltd; 2009.
[26] Soleimanzadeh M, Wisniewski R. Controller design for a wind farm, considering both power and load aspects. Mechatronics 2011;21:720e7.
[27] Gort S, Doran HD, Weber K, Norbert H. Communication aspects of wind turbine control-architecture redesign. In: International conference on power
engineering. Energy and Electrical Drives; 2011.
[28] Yang J, Song DR, Han H, Tong PS, Zhou L. The integrated control of fuzzy logic
and model-based approach for variable-speed wind turbine. Turkish J Electr
Eng Comput Sci 2015;23(6):1715e34.
[29] Wang N, Johnson KE, Wright AD. Comparison of strategies for enhancing
energy capture and reducing loads using LIDAR and feed-forward control.
IEEE Trans Control Syst Technol 2013;21:1129e42.
[30] International Electro-technical Commission. IEC 61400-1 International standard, wind turbines e Part 1: design requirements. 3rd ed. Geneva,
Switzerland: IEC; 2005.
[31] International Electro-technical Commission. IEC 61400-12-1 International
Standard, Wind Turbines ePart 12-1: power performance measurements of
electricity producing wind turbines. 1st ed. Geneva, Switzerland: IEC; 2005.