Trabajo de Investigación Formativa

TERMODINAMICA I
UNIVERIDAD NACIONAL DE TRUJILLO
UNIVERSIDAD NACIONAL DE TRUJILLO

FACULTAD DE INGENIERÍA QUÍMICA
ESCUELA ACADÉMICO PROFESIONAL DE INGENIERÍA QUÍMICA
CURSO: TERMODINAMICA I
DOCENTE:
MS. ROSA NOMBERTO TORRES
INTEGRANTES:
➢ GOMEZ PONTE FIORELLA
➢ AVALOS YUPANQUI GILMER
➢ RODRIGUEZ RODRIGUEZ JHON
➢ VELASQUEZ ACEVEDO ALISSON
➢ MARCOS ALEJANDRO LUCIO RIVAS
CICLO: V
AÑO:
2020
TERMODINAMICA I
UNIVERIDAD NACIONAL DE TRUJILLO
LINKOGRAFIAS EN INGLES
Kumar, Atul; Ullac, Muhammad; Pandey, Bishwajeet. (07 de 2018). Wind Energy: A Review
Paper . Obtenido de
https://www.researchgate.net/publication/326340286_Wind_Energy_A_Review_Paper
Kumar, Raja; Singh, S; Srivastava, R; Saket, R. (14 de 09 de 2019). Dynamic

reluctance air gap modeling and experimental evaluation of. Obtenido de
https://www.sciencedirect.com/science/article/pii/S2090447919301261
Li, Xuxia; Hu, Yingying; Wang, Yao; Deng, Jiaojiao. (13 de 09 de 2020). Cluster Wind
Power Uncertainty Model and Operation. Obtenido de
https://iopscience.iop.org/article/10.1088/1742-6596/1449/1/012077
Lui, Q. (17 de 07 de 2018). Optimization of Wind Fan Selection and Operation.

Obtenido de https://iopscience.iop.org/article/10.1088/1755-
1315/170/4/042017/pdf
Smith, Craig; Barthelmie, R; Pryor, S. (16 de 07 de 2013). In situ observations of the

influence of a. Obtenido de large onshore wind farm on near-surface:
https://iopscience.iop.org/article/10.1088/1748-9326/8/3/034006
Surana, Kavita; Doblinger, Claudia; Diaz, Laura; Hultman, Nathan. (31 de 08 de 2020).
Effects of technology complexity on the emergence and evolution of wind
industry manufacturing locations along global value chains. Obtenido de
https://www.nature.com/articles/s41560-020-00685-6
Thresher, R. (14 de 04 de 2018). Taking advantage of modern turbines. Obtenido de

https://www.nature.com/articles/s41560-018-0168-2
Yang, Jingjing; Liu, Xiaoliang; Dong, pentao; Huang, Mingjie. (09 de 02 de 2018).
Influence of wind power grid connection on power quality and reactive power
compensation scheme. Obtenido de
https://iopscience.iop.org/article/10.1088/1742-6596/1087/4/042084/pdf
Journal of Physics: Conference Series
PAPER • OPEN ACCESS
Influence of wind power grid connection on power quality and reactive

power compensation scheme
To cite this article: Jingjing Yang et al 2018 J. Phys.: Conf. Ser. 1087 042084
View the article online for updates and enhancements.
This content was downloaded from IP address 181.64.57.90 on 13/09/2020 at 16:27

First International Conference on Advanced Algorithms and Control Engineering IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1087 (2018)
1234567890 ‘’“” 042084 doi:10.1088/1742-6596/1087/4/042084
Influence of wind power grid connection on power quality and

reactive power compensation scheme
Jingjing Yang1，Song Wang2*, Yan Tian2, Ruining He2, Pengtao Dong1, Mingjie
Huang1, and Xiaoliang Liu1
1
Weifang Power Supply Company,State Grid Shandong Electric Power Company,
Weifang, China
2
School of Mechanical, Electrical & Information Engineering, Shandong University,
Weihai, China
*Corresponding author e-mail: wszbs3@sdu.edu.cn
Abstract. In order to explore the influence of wind power grid connection on power quality
and its control scheme, this paper builds a wind power generation system model in PSCAD
software, and studies the changes of active power, reactive power, DC bus voltage and
harmonics of wind farm when wind speed changes according to the models of basic wind, gust
wind, gradual change wind and random wind. And compare the electrical energy indexes
before and after the installation of reactive power compensation devices SVC and STATCOM
to verify the control effect of the reactive power compensation devices on the electrical energy
quality.
1. Introduction
The large-scale development of wind power is one of the most important means for all countries in the
world to deal with the increasingly urgent environmental deterioration and the strategic demand for
energy security. The installed capacity of wind power in our country is also increasing year by year.
However, due to the randomness of wind power, the output of the wind farm cannot be controlled.
After large-scale wind power being connected to the grid, the level of power grid indicators such as
voltage deviation, voltage fluctuation, flicker and harmonics will obviously drop [1]. In order to
ensure that the power quality meets the requirements, it is necessary to analyze the power quality,
judge various factors affecting it, and improve the power quality.
In this paper, basic wind, gust wind, gradual change wind and random wind are analyzed and
simulated in theory, and explore the output of wind power grid-connected system when wind speed
changes. The wind power grid-connected system model is built on the PSCAD software platform, and
the control effects of reactive power compensation control devices SVC and STATCOM on the power
quality are analyzed respectively, which verifies the feasibility of adding reactive power compensation
devices to improve the power quality and reactive power compensation [2].
2. Wind speed model and its influence on grid connection
2.1. Wind speed model

The simulated air volume usually consists of four wind speed models, namely, basic wind model, gust
model, gradual change wind model and random wind model, which is single or combined.
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
1234567890 ‘’“” 042084 doi:10.1088/1742-6596/1087/4/042084
Basic wind model. The basic wind model is usually established by Weibull distribution (Weibull)
parameters obtained from wind speed measurements at the location of the wind farm. The
mathematical expectation of Weibull distribution is
 1
v = Aτ 1 +  , (1)
 K
where v is basic wind, m/s; A is scale parameters of bull distribution; τ 1 + 1  is gamma parameter.
 K
When consider the simulation of seconds, the basic wind can be set to a constant, which is set to 11
m/s in this model.
Gust model. Gusts can usually be used to express sudden changes in wind speed. The expression of
gusts vWG is as follows.
0, t < t1G
 (2)
vWG = vCOS , t1G < t < t1G + tG
0, t > t + t
 1G G
MaxG   t t1G  
vcos = 1 − cos 2 − π  . (3)
2   tG tG  
Where, MaxG 、 tG 、 t1G are respectively expressed as gust maximum, acting time and starting time.
Fig.1 depicts a simulated gust with a basic wind speed of 11 m/s, a maximum of 5 m/s, an acting
time of 1s and a starting time of 3s.
As can be seen from Fig.1, the wind speed produced a gust of wind with a maximum speed of 5
m/s from 11 m/s at 3 seconds. The maximum wind speed reached 16 m/s and then returned to 11 m/s
at 4 seconds.
Figure 1. Gust Figure 2. Gradient wind Figure 3. Random wind model

model model
Gradient wind model. The gradual change wind model is usually used to reflect the gradual change
of wind speed. Fig. 2 shows a simulated gradual change wind with a basic wind speed of 11 m/s, a
maximum of 5 m/s, a duration of 2s, and a starting time of 3s.
As can be seen from Fig. 2, the wind speed changed from 11 m/s to 11 m/s at 3 seconds, with a
maximum speed of 5 m/s. The maximum wind speed reached 16 m/s and then returned to 11 m/s at 5s.
Random wind model. The random characteristics of random wind speed changes can be expressed
by random noise wind speed components. Fig.3 is a simulated random wind speed sequence with a
basic wind speed of 11 m/s, a noise amplitude control parameter of 0.05 rad/s, a surface resistance
coefficient of 0.0192, and a disturbance range of 600 m.
2.2. Influence of wind speed change on grid connection

During this transient simulation process, the wind speed will slowly change around the nominal speed
of 11 m/s, and the active power, reactive power, DC bus voltage and harmonic changes of the wind
farm are observed in Fig.4.
2
1234567890 ‘’“” 042084 doi:10.1088/1742-6596/1087/4/042084
(a) Wind speed (b) Active power
(c) Reactive power

(d) Dc bus voltage (e) Harmonic voltage
Figure 4. Change of electric energy index under wind speed fluctuation
From Fig.4 (a) to (e), it can be seen that when the wind speed is higher or lower than the nominal
value, the change trend of the active power output from the wind farm is similar to that of the wind
speed. But there is a certain lag. The reactive power, DC bus voltage and harmonics all fluctuates in a
small range. From the simulation images obtained, it can be seen that the change of wind speed will
affect the power quality of the system to some extent.
3. Reactive power compensation device for wind farm

The control of wind farm power quality is usually embodied in reactive power compensation and
harmonic suppression. In terms of reactive power compensation, it can be divided into static reactive
power compensation and dynamic reactive power compensation [3].
3.1. Wind farm management based on static var compensator (SVC)

SVC can control itself to absorb or emit reactive power according to the changes in control parameters of
power system. It has fast response, reliable operation, wide application range and low price, which is
widely used in power systems [4].
In order to absorb and adjust the inductive or capacitive reactive power at the same time in common
power systems, TCR and TSR are usually used to form reactive power compensation devices. The single-
phase schematic diagram of TCR and TSC is shown in Fig.5.
Figure 5. Thyristor controlled reactor and thyristor switching capacitor single-phase schematic
diagram
The wind farm simulation model with SVC installed for power quality control is shown in Fig.6.
3
1234567890 ‘’“” 042084 doi:10.1088/1742-6596/1087/4/042084
Figure 6. Wind farm simulation model with SVC

In the simulation model, the parameters of the SVC module are set to be consistent with the overall
parameters of the wind farm. The transformer capacity is set to 5.5 MVA, and the primary side rated
voltage is 154 KV and the secondary side is 33 KV. The fault type is designed as a three-phase short
circuit to ground. The fault occurs at 2s and lasts for 0.1s. The waveform diagram of the system
simulation is shown in Fig.7 to Fig.11.
Figure 7. Active power of wind farm grid connection before and after SVC installation
Figure 8. Reactive power of wind farm grid connection before and after SVC installation
Figure 9. Outlet voltage wind farm grid connection before and after SVC installation
Figure 10. DC bus voltage of wind farm grid connection before and after SVC installation
Figure 11. Harmonic voltage of wind farm grid connection before and after SVC installation
4
1234567890 ‘’“” 042084 doi:10.1088/1742-6596/1087/4/042084
According to the simulation waveform, after adding SVC, the power quality is significantly
improved. It can be seen that the active power fluctuation caused by the failure of the wind farm is
obviously smoother after installing SVC. At the same time, the reverse spike of reactive power is not
as steep as it was. Meanwhile, the export voltage of the wind farm also returns to normal quickly, and
the fluctuation of DC bus voltage decreases obviously. Therefore, after adding SVC, when the voltage
of the wind farm drops, SVC sends out reactive power, which makes the voltage of the wind farm
outlet rise and keep constant. When the wind farm voltage rises, it absorbs reactive power and lowers
the voltage value. SVC smoothes the voltage fluctuation and has a good control effect on the power
quality.
3.2. Wind farm management based on static synchronous compensator (STATCOM)

STATCOM is a reactive power dynamic compensation device composed of bridge power electronic
converter and large capacity capacitor [5]. The dynamic reactive power compensation can be realized
by changing the magnitude of the voltage applied to the reactor and controlling whether STATCOM
draws ahead or lags behind the current from the power grid. Compared with SVC, STATCOM has
faster response speed, wider operating range and smaller harmonic current content. Since the current
transmission lines are mostly three-phase four-wire systems, the most commonly used structure is the
three-bridge arm circuit.
Figure 12. Circuit of STATCOM with three-arm voltage type

Fig.12 shows the main circuit diagram of the voltage-type three-bridge arm STATCOM. Six power
electronic switches complete the converter function and the DC capacitor stores energy. Under normal
working conditions, the voltage of the capacitor connected to the DC side keeps constant, so it can be
regarded as a voltage source.
The single-phase equivalent circuit and vector diagram of STATCOM is shown in Fig.13
Figure 13. Equivalent circuit and vector diagram of STATCOM

Set the grid voltage to U S and STATCOM output voltage to U ST . The voltage U L on reactor X
between the grid and STATCOM is the vector difference between the two. Ignoring the active power
loss, only changing the magnitude of U ST can control whether STATCOM 's current IS drawn from
the power grid is ahead or behind. STATCOM absorbs capacitive reactive power when U ST is greater
than U S ; When U ST is less than U S , the voltage leads the current by 90 degrees and absorbs the
reactive power.
The wind farm simulation model with STATCOM is shown in Fig.14.
5
1234567890 ‘’“” 042084 doi:10.1088/1742-6596/1087/4/042084
Figure 14. Wind farm simulation model with STATCOM
In the simulation model, the parameters of STATCOM module are set to be consistent with the
overall parameters of the wind farm. The transformer capacity is set to 550 MVA, and the rated
voltage on the primary side and the rated voltage on the secondary side are 154 KV and 33 KV
respectively. The fault type is designed as a three-phase short circuit to ground. The fault occurs at 2s
and lasts for 0.1s [6]. The waveform diagram of system simulation is shown in Fig.15 to Fig.19.
Figure 15.Active power of wind farm grid connection before and after STATCOM installation
Figure 16.Reactive power of wind farm grid connection before and after STATCOM installation
Figure 17. Outlet voltage wind farm grid connection before and after STATCOM installation
Figure 18. DC bus voltage of wind farm grid connection before and after STATCOM installation
6
1234567890 ‘’“” 042084 doi:10.1088/1742-6596/1087/4/042084
The simulation results show that the active power and reactive power of wind farm is more stable
and the overshoot of active power is obviously reduced after installing STATCOM. Especially
reactive power, output curve and its stability show that STATCOM plays a great role in the reactive
power compensation process. For DC bus voltage, the output voltage waveform overshoot is small,
and the output voltage is not overshoot. At the same time, STATCOM has faster response speed than
SVC, so it can adjust the output voltage of the wind farm faster. As for harmonic voltage, the
waveform after adding STATCOM is smoother and softer, which shows that it has good control effect
on harmonic voltage.
4. Conclusions
The wind farm composed of doubly-fed asynchronous wind turbine is simulated and analyzed, and its
active power and reactive power output under different wind speeds is analyzed. Through simulation
analysis, it can be seen that when the wind speed changes, the active power output from the wind farm
will be greatly influenced, and there will be some lag, while the reactive power, DC bus voltage and
harmonic wave fluctuation will be relatively small.
SVC and STATCOM reactive power compensation devices are added to the established wind farm
model respectively, and the compensation functions of the two reactive power compensation devices
to the power quality are analyzed respectively. Through simulation research and analysis, both of the
two power compensation devices can have a positive impact on power quality control. The specific
performance is that they can reduce voltage fluctuation and voltage deviation and compensate the
reactive power output. Compared with SVC, STATCOM has faster response speed and better output
characteristics, thus it is more advanced and efficient.
Acknowledgments
The authors wish to thank the corporate researchers in the common laboratory for the generous help of
simulation program and experiment.
References
[1] Bai Hongbin,Wang Ruihong, “Influence of the Grid-connected Wind Farm on Power
Quality,”Proceedings of the CSU-EPSA,pp.120-124,Feb.2012.
[2] Feijoo, A.E., Cidras J, “Modeling of Wind Farms in the Load Flow Analysis,” IEEE Transactions
on Power Systems,pp.110-115, Vol.15,2000.
[3] Jiabing Hu, Yikang He, Lie Xu, et al, “Improved Control of DFIG Systems During Network
Unbalance Using PI-R Current Regulators,” IEEE Transactions on Industrial Electronics,pp.
439-451,Vol.56,2009.
[4] Cai Zhi, Liu Jianzheng, Wang Jian, Liu Shu, “Simulation Research on Control Modes of
Doubly-Fed Wind Power Generator Based on PSCAD,” Electrical Technology,pp.61-
64,May.2008.
[5] Bostjan Blazic; Igor Papic, “STATCOM Control for Operation with Unbalanced Voltages,” 2006
12th International Power Electronics and Motion Control Conference,pp.1454-1459,2006.
[6] Liu Wei, “Evalution and Control Measures on Power Quality of Wind Power Grid,” Beijing
Jiaotong University.2014.
7
IOP Conference Series: Earth and Environmental Science
PAPER • OPEN ACCESS Related content

- Simulation and optimization of wind and
Optimization of Wind Fan Selection and Operation diesel power supply systems
Maintenance in the Wind Power Stations - Discussion of Several Key Technologies

about Offshore Wind Power
- Maintenance Optimization Model for One

To cite this article: Qinghua Liu 2018 IOP Conf. Ser.: Earth Environ. Sci. 170 042017 kind of Three-Dimensional Radar Antenna
Array

2nd International Symposium on Resource Exploration and Environmental Science IOP Publishing
IOP Conf. Series: Earth and Environmental Science 170 (2018)
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
Optimization of Wind Fan Selection and Operation

Maintenance in the Wind Power Stations
Qinghua Liu*
Beijing Information Technology College, General Competencies Development
Department, Beijing, China
* Corresponding author e-mail: liuqinghua@bitc.edu.cn
Abstract. In order to assess the wind energy resources and their utilization of wind
power stations, it analyze wind speed probability distribution, the average wind speed,
average wind power density, effective utilization time of wind energy, the wind power
output radio, and concludes that the wind energy level of this wind farm is between
"available area" and "suboptimal area", where the effect of applying to the grid wind
power is not good, and the power generation efficiency is only 19.34%. Use capacity
coefficient to discuss the matching of wind energy resources and the fan, determines
the new fan type Ⅲ is the highest effective utilization of wind energy resources. From
the two goals of balanced assignment and good economic benefit, programming model
is established, which gives the schedule of maintenance personnel and the
maintenance plan of the fan.
1. Introduction
Among the many new energy sources, wind power has a unique natural advantage. As a clean and
non-pollution renewable energy, wind energy is one of the most widely used energy sources for
development and utilization. Under the background of world energy shortage, wind power as the most
mature new energy has become an important energy supplement, and it is worth developing.
Problem D in 2016 CUMCM [1] is to analyze and optimize the operation of a wind farm in China,
and mainly discuss three problems: 1. assess the wind energy resources of the wind farm and its
utilization; Compare the new and old types of fan, and judge which is more suitable; 3. Set up a
reasonable schedule of maintenance personnel and fan maintenance plan, in order to ensure that the
wind farm has better economic benefits.
2. Assessment of wind energy resources and its utilization

According to the national standard "wind energy resource assessment method of wind farm", the wind
energy resources and their utilization need to be evaluated by wind energy resource parameters such as
the average wind speed, the average wind power density, the effective utilization time of wind energy
and the ratio of wind power output to the wind farms, and then the wind energy resources can be
classified [2].
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
2.1. Approximate calculation of wind speed distribution

From the annex data, the wind speed distribution is known as a positive partial distribution, and the
common Waybill distribution [3] is used to fit. The Waybill probability distribution model and its
probability distribution density function can be expressed as:
k
v
  
F (v )   f (v)dv  1  e c (1)
0
k
k 1 v
k v  
f (v )    e c
(2)
cc
Where v is the wind speed, c is the scale parameters of the distribution, and k is the shape
parameters of the distribution. Here we use statistics (mean value  , standard variance  ) to estimate
the shape parameter K and the scale parameter C of the Waybill distribution [3]:
1.086
 
k   (3)


c
 1 (4)
 1  
 k
Put the annual wind speed data in it and obtain k  2.21 , c  6.40 ,and the corresponding Waybill
distribution curve is shown in Figure 1.
Figure 1. Waybill distribution curve of wind speed.
2.2. Assessment of wind energy resources
2.2.1 Average wind speed. The average wind speed is the parameter that can best reflect the local wind
energy resources. Due to the randomness of wind speed, the long-term observation data are closer to
the actual wind energy, so the annual average wind speed can better show the advantages and
disadvantages of wind energy resources. After calculation, the average annual wind speed is v  5.67 .
2.2.2 Wind power density. To measure the size of a local wind energy and to evaluate the potential of
the wind, the wind power density is the most suitable, most convenient and most valuable. The higher
the wind power density is, the better the wind energy resources is, and the higher the utilization of
wind energy is. The wind power density is directly proportional to the air density and wind speed cube,
so the change of air density is negligible when the accuracy requirement is not high. Therefore, the
probability density distribution of wind power density W is only determined by the probability
distribution characteristics of wind speed. The wind power density can be calculated by the next
formula [4] :
2
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
1 3
w pc  (3 / k  1) (5)
2
So, the annual average wind power density is 204.1685 W/m2.
According to China's wind power density scale and wind energy partition table issued by the state
authority [5], the wind energy resources of the wind farm can be classified. The measured height of
the wind speed was not given in the annex, we can estimate the wind energy level of this wind farm is
between the "available area" and “suboptimal area", but for its application in grid connected wind
power generation, wind energy utilization is smaller than rich area. Therefore, we conclude that the
wind energy resources of the wind farm do not reach a rich level, but it can still be used.
2.3. Assessment of the utilization of wind energy resources
2.3.1 The utilization rate of wind energy resources.The utilization ratio of wind energy resources, or
the effective utilization of wind energy resources, is the ratio of the effective utilization hours T of
annual wind energy to the annual total hours,
T
A (6)
8760
The effective utilization time of wind energy refers to the cumulative hours of wind speed in a
period of time which the wind speed is in the effective wind speed range. Because the output power is
0 when the wind speed is less than or equal to a specified value, in other words, the generator fan is
not rotate, which shows that wind energy in this wind speed cannot be used by the wind fan. The
effective wind speed is the wind speed range from the cut-in wind speed to cut-out wind speed.
This paper selected the effective wind speed range as 3m/s-25m/s to evaluate the wind energy
resources.By calculating the effective utilization hours of annual wind energy, the utilization rate of
wind energy resources is 85.40%.
Similarly, the utilization rate of monthly wind energy resources can be calculated by using the ratio
of the effective utilization hours of month wind energy to the total hours of each month to, see figure 2:
Figure 2. Variation tendency of the monthly effective utilization time of wind energy.
2.3.2 The actual utilization rate of wind energy resources. The actual utilization of wind energy
resources  is to measure the power of wind fan, or is named the output ratio of wind power. It can be
obtained by the ratio of the average annual output power P to the installed capacity of the fan P0 . The
data of Annex 1 can be used to calculate the average annual output power of 38.39MW. Annex 3 gives
the wind farm produced 25 fans in first-stage project and 99 fans in second-stage project, which the
rated power is 2000kW and 1500kW, respectively. So get
3
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
P 38.39
   19.34% (7)
P0 2000  25  1500  99
In the same way, the actual utilization rate of monthly wind energy resources is calculated as
Figure 3:
Figure 3. Monthly actual utilization of wind energy resources
The annual utilization rate of wind energy in the wind farm is 85.40%, while the actual utilization
rate of the whole year is only 19.34%. This indicates that the wind power utilization in the wind farm
is not very good, which is consistent with the previous assessment of wind energy resources.
At the same time can be seen, the highest rate of the actual utilization of wind energy resources is
28.63% in January, and the lowest is 10.98% in July. In winter (December to February) has the highest
rate, followed by the spring (March to May), and again in autumn (September to November again), in
summer (June to August) has the lowest rate. Therefore, the wind power utilization of the wind farm
in winter and spring have more potential, which is more suitable for wind power generation, while the
autumn and summer wind energy potential is relatively small, wind power can be used as an effective
supplement.
From the comparison of the utilization rate of wind energy resources and the actual utilization rate,
there were significant differences in December, April, July and September. In December and April, the
effective utilization time is less, but the actual electricity output ratio is higher. In July and September,
the effective utilization time is much, but the actual electricity output ratio is low. It may be the
monthly wind speed is different, causing the wind energy to be centralized and unstable.
3. The Model of fan selection

In the construction of wind farms, the selection of wind fan with reliable operation, high efficiency,
good control and good power supply is very important, which is directly related to the utilization of
wind power in wind farms. Corresponding to the same wind farm, we should select wind fan by the
characteristics of each observation point at different observation points, rather than select the same
type.
3.1. The establishment of the model

In order to respond more accurately to the economy of the wind farm we choose the capacity
coefficient as a measure of wind power to select the appropriate type of fan [6]. The greater the annual
average capacity coefficient of the wind farm, will result in the better economy, and the annual
average utilization rate of the fan will be higher.
Capacity coefficient of wind farm is the ratio of the actual generating capacity to the rated power of
the wind farm within a certain period of time, it is also the ratio of the annual average output power
Pw to the rated power Pr of the fan, that is,
4
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
Pw
Cf  (8)
Pr
vc
Pw   P (v ) Pw (v )dv (9)
vi
According to the relationship between the power and the wind speed in Annex 3, the next formula
is obtained.
 0, 0  v  vi
 (v) P , vi  v  vr

P (v )   (10)
r
 Pr , vr  v  vc
 0, v  vc
where, vi is the cut-in wind speed of the wind fan, vr is the rated wind speed, vc is cut-out wind
speed ,  (v) is the ratio of the output power and the rated power when the wind speed is between vi and
vc , which is a complex function related to the wind speed and it can reflect the output characteristic of
the fan.
There are three functions that approximate the output characteristics of the fan: linear function,
quadratic function and cubic function, which can be expressed respectively.
v  vi (v  vi ) 2 (v  vi )3
1 (v)  , 2 (v)   (v ) 
2 , 3
vr  vi (vr  vi ) (vr  vi )3
According to the curve shape of the relationship between the output power of the fan and the wind
speed and the result of the fitting, it is the cubic function relation, so we choose 3 (v) to calculate, put
the output characteristic of the fan 3 (v) and wind speed probability density function Pw (v ) into the
annual average output power formula, then get:
Pr Vr Vc
3 V
Pw  (v  Vi )3 Pw (v)dv  Pr  P（w v) dv (11)
(Vr  Vi ) i Vr
3.2. The solution of the model

According to the wind speed of six fans provided in Annex 2, we can get the mean and standard
deviation of annual mean wind speed, and the characteristic parameters of wind speed obeying
Waybill distribution, as shown in Table 1:
Table 1. Wind speed distribution parameters at six observational points.

4# 16# 24# 33# 49# 57#
 6.32 6.07 5.74 5.65 6.08 6.20
 3.27 3.32 3.45 2.98 3.39 3.62
k 3.10 2.86 2.79 2.99 2.77 2.55
c 7.07 6.81 6.45 6.33 6.83 6.98
Through previous discussions, it has been known that the design parameters of wind fan and the
data of wind energy at the selected observation points directly determine the size of the capacity
coefficient. The MATLAB program is written to calculate the capacity coefficient of each observation
point for each fan type, such as table 2:
5
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
Table 2. Capacity coefficient of each observation point matching each fan type.
4# 16# 24# 33# 49# 57#
Fan type Ⅰ 0.2205 0.2036 0.1856
Fan type Ⅱ 0.1310 0.1756 0.1919
Fan type Ⅲ 0.2465 0.2274 0.2065 0.1816 0.2304 0.2461
Fan type Ⅳ 0.2205 0.2036 0.1856 0.1599 0.2068 0.2227
Fan type Ⅴ 0.1970 0.1820 0.1666 0.1406 0.1854 0.2014
The six observation points, the capacity coefficient calculated differences by using different types
of fans, where the capacity coefficient of type III is greater than the original type (type I and type II) in
first-stage project and second-stage project; the capacity coefficient of type IV is equal or or not quite
different to type I in first-stage project, but it is greater than type II in second-stage project; the
capacity coefficient of type V models is less than type I in first-stage project, and it is greater than type
II in second-stage project. It can be seen that considering the wind energy resources and fan's
matching angle, the capacity coefficient of each fan type is ranked as follows:
Type III > type I = type IV > type V > type II
Therefore, without considering the cost and maintenance cost of each fan type, the original type
(type I, type II) can be replaced with type III. If the cost of type III is higher, the original model can be
replaced by the type IV or V in second-stage project, accordingly increasing the effective utilization of
wind energy resources. To sum up, the new type fan is more suitable than the existing wind fan under
the existing wind energy resources.
4. planning model for maintenance scheduling

For safety production needs, fans need two times of shutdown maintenance every year, and the
continuous working time between two maintenance is not more than 270 days. Each maintenance
needs a group of maintenance personnel to work 2 days continuously. At the same time, the wind farm
needs a group of maintenance personnel to be on duty every day to deal with emergencies. The four
groups of maintenance personnel in the wind farm can be engaged in duty or maintenance work, and
the continuous working hours (on duty or maintenance) for each group of maintenance personnel are
not more than 6 days. In order to ensure that the wind farm has better economic benefits and keep the
work tasks of the maintenance personnel relatively balanced, a reasonable schedule and fan
maintenance plan will be made to maximize the benefit.
According to the requirements, a programming model can be established for the schedule of
maintenance personnel and the maintenance plan of the fan. 0-1 variables are introduced:
1, group i work (maintenance or duty) on the k - th day
xik  
0, group i is not working on the k - th day
 1, the j - th fan work on the k - th day
y jk  
0, the j - th fan is not working(in maintenance)on the k - th day
Where, i  1,2,3,4 is the group of the maintenance personnel, j  1,...,124 is the number of the fans,
k  1,...,365 is the number of days per year.
The objective function to measure the relatively balanced task of the maintenance personnel of
each group can be expressed as:
minmaxmi  mi '  maxmi  (12)
6
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
365
mi   xik Express the one year working time (including maintenance and duty) of maintenance
k 1
personnel of group i , maxmi  mi '  express the difference between the maximum working time and the
minimum working time in each group.
The objective function to measure the good economic benefit of the wind farm can be expressed as:
 124 365 
min  pk (1  y jk ) (13)
 j 1 k 1 
Pk
pk 
124
Express the output power of each fan on the day k (assuming that the output power of each
The following constraints can be obtained from the requirements in the question:
1) One group of maintenance personnel must be on duty every day to respond to emergencies:
2) The continuous working hours (on duty or maintenance) for each group of maintenance
personnel are not more than 6 days:
3) The time interval between the two maintenance of the fan is not more than 270 days.
4) One group of maintenance personnel should be operated for 2 days for each maintenance of the
fan:
To sum up, the programming model can be obtained.
 124 365 
min maxmi  mi '  maxmi   pk (1  y jk )
 j 1 k 1 
4
 xik  1, k  1,2,...,365
 ki 16
 x  7, i  1,2,3,4,k  1,2,...,359
 k il
s.t. kl 270 (14)
  y jl  270, j  1,2,...,124, k  1,2,...,95
 l k
k 2
 (1  y jl ) xil  2, i  1,2,3,4, j  1,2,...,124, k  1,2,...,364
 l k
By using LINGO software, we can get the scheduling plan for each group of maintenance
personnel in one year and the maintenance plan for each fan,hich table 3 give some results of
maintenance personnel’s work, rest and maintenance in July and Figure 4 show the number of working
days and maintenance days per month for the maintenance personnel of each group.
Table 3. Schedule of maintenance personnel in July.

182 183 184 185 186 187 188 189 190 191 ... 210 211 212
July 1 2 3 4 5 6 7 8 9 10 ... 29 30 31
Group 1 1 0 1 1 1 1 1 1 1 0 ... 1 0 1
Group 2 0 0 1 0 1 1 1 1 1 1 ... 1 0 1
Group 3 0 1 1 1 1 1 1 0 1 1 ... 1 0 0
Group 4 0 1 1 1 1 1 0 1 0 1 ... 1 1 0
Note: 1 means the maintenance personnel work (maintenance or duty), and 0 means the
maintenance personnel do not work.
7
1234567890 ‘’“” 042017 doi:10.1088/1755-1315/170/4/042017
Figure 4. The number of working days and maintenance days per month for the maintenance
personnel of each group.
5. Conclusion
It can provide an important basis for wind farm site selection using wind energy resource assessment,
the assessment is accurate or not related to the wind farm economic benefits. We choose parameters
such as the average wind speed, average wind power density, effective utilization time of wind energy
and wind power output ratio to assess, the calculated results can represent the actual wind energy
resources in a certain extent.
It is relatively simple and accurate to use capacity coefficient to select wind fan, but different
functions of the output characteristic functions has a great influence on the calculation results of the
capacity coefficient. Based on the parameters of wind energy resources and fan, we use cubic function
to fit and compare new and old types of fan. A multi-objective 0-1 programming model is established,
which give the better schedule of maintenance personnel and the maintenance plan of the fan, with
wide and good applicability.
References
[1] Information on http://www.mcm.edu.cn/html_cn/node/6d026d84bd785435f92e3079b4a87a2b.
html
[2] GB/T 18710-2002, Assessment of wind power resources in wind farms, China Standard Press,
Beijing 2002
[3] Ye Fang, Chen Jichuan, Feng Hengchang. Calculation method of wind frequency Weibull
distribution parameters in wind power field and its application, Electric Manufacturing, 2010,
5:48-50
[4] Zhang Xule. Economic Research on the Optimization of Wind Machine Selection in Wind Farm
Construction, Hunan University, 2011
[5] Wang Xiaolin. Wind Power Density Classification and Revision of Wind Energy Zoning, China
Electric Power, 2010,19(8):78-82
[6] Lv Pengyuan, Deng Zhiyong. Type selection of wind turbines in the construction of wind farms,
Water Conservancy and Hydropower Technology, 2009,40(9):57-59
[7] Chen Hui, Wu Jie. Analysis and optimization of the operating conditions of the wind farm,
Journal of Sichuan University of Science and Engineering (NATURAL SCIENCE
EDITION), 2016,29 (6):33-38
[8] Wang Jijian. Analysis and optimization of wind farm operating conditions, Mathematical
Modeling and Application,2016,5 (4):50-62
[9] Qi Fang, Zhang Lizi, Fu Rong. Research and example analysis of assessment of wind resources,
Chinese Electrical Engineering Society annual meeting proceedings, Guangdong, 2007
8
Environmental Research Letters
LETTER • OPEN ACCESS Related content

- Atmospheric stability affects wind turbine
In situ observations of the influence of a large power collection
onshore wind farm on near-surface temperature, - Experimental study of the impact of large-
scale wind farms on land–atmosphere
exchanges
turbulence intensity and wind speed profiles
- IEA-Task 31 WAKEBENCH: Towards a
protocol for wind farm flow model
To cite this article: Craig M Smith et al 2013 Environ. Res. Lett. 8 034006 evaluation. Part 2: Wind farm wake models
Recent citations
View the article online for updates and enhancements. - Review on optimisation methods of wind
farm array under three classical wind
condition problems
F. Azlan et al
- Simulated wind farm wake sensitivity to

configuration choices in the Weather
Research and Forecasting model version
3.8.1
Jessica M. Tomaszewski and Julie K.
Lundquist
- Sensitivity of Wind Turbine Array

Downstream Effects to the
Parameterization Used in WRF
T. J. Shepherd et al

IOP PUBLISHING ENVIRONMENTAL RESEARCH LETTERS
Environ. Res. Lett. 8 (2013) 034006 (9pp) doi:10.1088/1748-9326/8/3/034006
In situ observations of the influence of a

large onshore wind farm on near-surface
temperature, turbulence intensity and
wind speed profiles
Craig M Smith1 , R J Barthelmie and S C Pryor
Atmospheric Science Program, Department of Geological Sciences, Indiana University, Bloomington,
IN, USA
E-mail: craig.smith@dri.edu
Received 19 March 2013

Accepted for publication 2 July 2013
Published 16 July 2013
Online at stacks.iop.org/ERL/8/034006
Abstract
Observations of wakes from individual wind turbines and a multi-megawatt wind energy
installation in the Midwestern US indicate that directly downstream of a turbine (at a distance
of 190 m, or 2.4 rotor diameters (D)), there is a clear impact on wind speed and turbulence
intensity (TI) throughout the rotor swept area. However, at a downwind distance of 2.1 km (26
D downstream of the closest wind turbine) the wake of the whole wind farm is not evident.
There is no significant reduction of hub-height wind speed or increase in TI especially during
daytime. Thus, in high turbulence regimes even very large wind installations may have only a
modest impact on downstream flow fields. No impact is observable in daytime vertical
potential temperature gradients at downwind distances of >2 km, but at night the presence of
the wind farm does significantly decrease the vertical gradients of potential temperature
(though the profile remains stably stratified), largely by increasing the temperature at 2 m.
Keywords: lidar, wind turbine wakes, surface temperature profiles, turbulence intensity, wind
shear
S Online supplementary data available from stacks.iop.org/ERL/8/034006/mmedia
1. Introduction layer and surface-atmosphere fluxes of momentum, sensible

heat and latent heat. These changes may have implications
Consistent with historical and continuing use of wind for crop yields if wind farms are located in agricultural
propellers to prevent freezing in vineyards by enhancing regions (e.g. by changing water availability), and may alter
vertical exchange of warmer air from aloft (e.g. Crawford downstream atmospheric properties (see detailed literature
1965), reduction of wind speed and enhancement of review provided in the supplementary materials available at
turbulence in the lee of wind turbines (i.e. wind turbine wakes) stacks.iop.org/ERL/8/034006/mmedia).
(e.g. Barthelmie et al 2010), modifies the planetary boundary Idealized mesoscale models, in which wind turbines
are parameterized as sinks of momentum and sources of
Content from this work may be used under the terms of turbulent kinetic energy (Baidya Roy 2011), and large eddy
the Creative Commons Attribution 3.0 licence. Any further
distribution of this work must maintain attribution to the author(s) and the
resolving simulations (LES) of an idealized infinite wind
title of the work, journal citation and DOI. farms (Lu and Porté-Agel 2011, Calaf et al 2011), have been
1 Address for correspondence: Department of Geological Sciences, 702 used in efforts to quantify the impact of wind turbines on
North Walnut Grove, Indiana University, Bloomington, IN 47405-7100, USA. atmospheric properties within and downstream of multi-MW
1748-9326/13/034006+09$33.00 1 c 2013 IOP Publishing Ltd Printed in the UK

Environ. Res. Lett. 8 (2013) 034006 C M Smith et al
Figure 1. Overview of the wind farm, including close-up views of the NE and SW locations, and schematic of the unwaked (#1, black),
SW farm-waked (#2, red), SW direct-waked (#3, blue), NE farm-waked (#4, green) and NE direct-waked (#5, magenta) wind direction bins.
The frequency with which flow greater than 4 ms−1 was observed in the five directional sectors during 04/04/2012–05/20/2012 is 6.9, 16.6,
7.9, 14.2 and 1.8%, respectively.
wind turbine deployments. Results from these studies are located in the southwest (SW) and northeast (NE) corners
to some degree inconsistent; while some modeling groups of the wind farm which is collocated with soy and corn
have found ‘statistically significant impacts on near-surface crops (which were between 5 and 25 cm tall during the field
air temperature and humidity as well as surface sensible and campaign). Widely scattered houses and poplar trees exist
latent heat fluxes’ (Baidya Roy 2011), the simulations of an throughout the farm as well, but none were located nearby
infinite wind farm for tall (120 m) turbines presented in Lu our measurement sites, and soil and crop conditions close to
and Porté-Agel (2011) show no significant effect on surface both meteorological masts were relatively similar throughout
temperatures. As discussed in the supplementary materials, the campaign. Duplicate anemometers deployed on different
there are relatively few in situ observational data sets with sides each meteorological mast allowed for identification
which to evaluate these and other model simulations. In this and reduction of mast shadow effects (in which the tower
paper we present data from a large operational onshore wind superstructure upwind of a sensor perturbs the flow field)
farm located in flat homogeneous terrain with uniform land by conditionally sampling the data to select measurements
use and analysis to quantify the impact of this development from the sensor upwind of the tower. Additional data from
on near-surface profiles of wind speed, turbulence intensity two vertically pointing continuous wave ZephIR lidars and a
and temperature. Galion scanning pulsed lidar are presented for 7–20 May. As
discussed further below, wind speeds estimated using these
2. Data and methods lidars exhibit very good agreement with data from the cup
anemometers (Peña et al 2009) and sonic anemometers (Sathe
2.1. Observational data et al 2011). Further, scanning lidars have been shown to be
capable of capturing the −5/3 Kolmogorov spectrum (Iungo
Data analyzed herein were collected during 4 April–20 May et al 2013). However, estimates of turbulence intensity (TI)
2012, at a large wind farm in the Midwest (table S1 and (and momentum flux, e. g. Mann et al 2010) measured by the
supplementary materials available at stacks.iop.org/ERL/8/ ZephIR lidars over a scanning volume is attenuated relative
034006/mmedia). The array comprises ∼300 1.5+ MW wind to point measurements, and the degree of attenuation varies
turbines with a hub height of 80 m. Confidentiality agreements with mean wind speed (Peña et al 2009), height (scanning
prohibit discussion of some details, but the overall shape of volume) and stability (Sathe et al 2011). For the purposes of
the installation is roughly rectangular, and the layout is such this paper all wind speed and direction measurements taken
that in the west–east direction the wind turbine spacing is between 75–80 m are referred to as representing a nominal
∼5 rotor diameters (D) (where D = 80 m, and the range height of 80 m (i.e. turbine hub height). Unless otherwise
of spacing is 3.8–5.5D), and in the south–north direction stated, all times quoted are in local standard time (LST), the
the turbine spacing is ∼20D (range; 18–22D) (figure 1). date format is month/day/year, the data integration period is
Meteorological data were collected on or near two 80 m masts 10 min, and all data have been conditionally sampled to select
2
periods when the wind farm was operating and producing and direct-waked bins, the SW location is fully unwaked.
power (i.e. free-stream hub-height wind speeds >4 ms−1 ). Thus, in each case (2–5) one meteorological mast location can
In operational wind farms, including the one studied herein, be used to provide ‘free-stream’ (i.e. unwaked) conditions,
it typical that a small percentage of wind turbines will not while the other mast can be used to characterize conditions
be operational due, for example to maintenance issues or in the wake of an individual turbine, or the entire wind farm.
curtailment events by the grid operator, however, in the data To avoid situations in which the downstream edge of wind
period shown here the turbines closest to the meteorological turbines was not operating when free-stream wind speed was
masts were fully operational. near cut-in, we present data only from periods when the
Characteristics of wind turbine wakes, such as the hub-height wind speed at both the NW and SE meteorological
incremental increase in TI, are non-linear functions of incident masts exceeded 4 ms−1 . These criteria ensured that the wind
wind speed due to the dependence on the wind turbine thrust turbines proximal to the SW meteorological mast (#1 and
coefficient. Wake induced TI is generally maximized for #2, see figure 1) were operating and producing power >99%
wind speeds between the cut-in wind speed (when the wind of the time for the waked cases studied herein, which we
turbine starts turning) and the rated wind speed (when power verified through inspection of power production from each
output ceases to be a function of wind speed) (Barthelmie of these two turbines (not shown). The frequency with which
et al 2012). The median free-stream wind speed for data flow greater than 4 ms−1 was observed in the five directional
presented herein is 7.5 ms−1 , which is associated with sectors (1: both masts unwaked, 2: wind farm wake at SW,
the highest thrust coefficients for turbines on this farm, 3: direct wake at SW, 4: wind farm wake at NE, 5: direct wake
and hence the measurement period is characterized by a at NE) during 04/04/2012–05/20/2012 is 6.9, 16.6, 7.9, 14.2
high frequency of conditions during which wakes will be and 1.8%, respectively.
pronounced and the largest relative velocity deficits will be
observed. Thus the results presented herein may slightly 3. Results
over-estimate downstream effects relative to wind farms in
higher wind speed regimes. 3.1. Average conditions in the near wake, wind farm wake
and unwaked sectors
2.2. Differentiating unwaked, near wake and wind farm wake
conditions Low level wind shear (α) is characterized using the power law:
α
u2 z2
Wind farm geometry was used to identify wind direction = , (1)
sectors associated with outflow from the entire wind farm u1 z1
(wind farm wake), with a direct wake from the nearest wind where u is the wind speed, z is the height, and subscripts 1
turbine (near wake) and with flow that represents free-stream and 2 indicate 40 m and 80 m measurements, respectively.
conditions (unwaked). The nearest turbines to the SW mast In the unwaked wind directions the average shear exponent
are located 2.4D (193 m) to the northwest (299◦ ), 2.4D at SW = 0.05 during the day (11:00–17:00 LST, hereafter
(189 m) to the northeast (62◦ ), and 6.8D to the east (81◦ ) referred to as daytime) and 0.36 at night (00:00–06:00 LST,
(figure 1). Thus for the SW location a direct wake (sector hereafter referred to as nighttime), while at the NE mast
3 shown in figure 1) is characterized by flow centered on α = 0.05 during the day and 0.33 at night (table 1). Both are
turbines 1 and 2 (where each sector is assumed to have a ∼15◦ thus consistent with low shear (high turbulence regimes) in
half width, see figure S1 (available at stacks.iop.org/ERL/8/ the day and higher shear at night. The average shear exponent
034006/mmedia) for further evaluation of the wind direction from all unwaked data is 0.13 and 0.13 at the SW and NE
sectors). For wind directions between 352◦ and 035◦ the SW locations, respectively, which is consistent with the value of
location is waked by flow through a large portion of the wind 0.14 used in the wind turbine design standards (IEC standard
farm with the closest turbines displaced 26D (2.1 km) away 61400-3), but lower than the shear exponent of 0.23 computed
(with the next row located 44–48D (∼3.8 km) away). Thus, from measurements at 50 and 90 m from an Indiana tall tower
flow from these directions is used to describe the entire wind in a low roughness regime (Elliott et al 2008).
farm wake (SW farm waked, sector 2). Wind directions from Negative shear (decreasing wind speeds between 40
318◦ to 350◦ are fully unwaked at both the SW and NE and 80 m) is clearly evident in near wake wind directions
meteorological masts, and thus are used to define ‘unwaked’ (i.e. sector 3 at the SW mast, and sector 5 at the NE mast)
(i.e. sector 1) conditions. For the NE location, there is a (table 1 and figure S1 available at stacks.iop.org/ERL/8/
line of turbines aligned at ∼255◦ (with the closest turbine 034006/mmedia) consistent with the hub-height maximum
at 2.4D). This direction represents the centerline of the NE velocity deficit (Politis et al 2012). Also consistent with
direct-waked sector (#5). The nearest row south of the NE detailed wind turbine wake studies which have indicated
mast does not have turbines extending all the way to the wind relatively rapid recovery of wind turbine wakes in high
farm eastern boundary, and the nearest wind turbine is located turbulence regimes (Barthelmie et al 2012), the shear
33D (2.7 km) away at 218◦ . Thus the sector from 190◦ to exponent in the wind farm-waked sectors is only modestly
240◦ is used to define the wind farm wake for the NE location lower than that for the unwaked sectors (median daytime
(sector 4). For the SW farm-waked and direct-waked bins, the values for unwaked and wind farm-waked sectors are; 0.05
NE location is fully unwaked, and for the NE farm-waked versus 0.07 for SW, and 0.05 versus 0.07 for NE, see table 1).
3
Table 1. Median values of daytime (11:00–17:00 LST) and nighttime (00:00–06:00 LST) turbulence intensity (TI) (%), shear exponent (α)
(unitless), and vertical potential temperature gradient (1θ/1z) (◦ C m−1 ) for each wind sector (see figure 1).
Mast & wind direction sector Day TI Night TI Day α Night α Day 1θ/1z Night 1θ/1z
SW: Sector #1 unwaked 14.2 5.7 0.05 0.36 −0.018 0.048
SW: Sector #2 farm wake 14.8 7.7 0.07 0.20 −0.020 0.022
SW: Sector #3 direct wake 25.7 24.7 −0.05 −0.24 −0.021 0.026
NE: Sector #1 unwaked 12.5 5.4 0.05 0.33 — —
NE: Sector #4 farm wake 12.5 6.6 0.07 0.23 — —
NE: Sector #5 direct wake 25.3 24.3 0.01 0.02 — —
gradient is present during either the day or night according

to the Wilcoxon rank sum test applied with 95% confidence
level. This is consistent with the flat terrain and uniform
land use, and validates use of this location for an assessment
of the magnitude of whole wind farm wake effects on the
near-surface layer. For wind directions associated with direct
wakes at SW but no wake at NE (sector 3), wind speeds at the
SW mast are significantly lower (according to the Wilcoxon
rank sum test) for all hours of the day, especially during the
nighttime. The difference is smallest in the mid-afternoon
when wake recovery due to vertical momentum transfer is
maximized, presumably due to the unstable conditions and
efficient vertical mixing. Similar behavior (NE mast measured
wind speeds are lower than SW) is also noted when the NE
mast is direct waked and the SW location is unwaked (sector
#5), for bi-hourly bins containing more than 8 data points.
For sectors associated with whole wind farm wake conditions
(sectors 2 and 4), horizontal differences of hub-height wind
speed during the daytime are fairly small, but at least for
sector 4 mean wind speed difference is 1.0 ms−1 during
nighttime, when stable conditions and low ambient TI tend
to limit wake recovery (Schepers et al 2012, Barthelmie
Figure 2. Mean (a) hub-height difference in wind speed (SW–NE),
et al 2012). The horizontal wind speed differences for the
(b) difference in hub-height TI (SW–NE) and (c) mean potential direct-waked sector are in agreement with previous wind
temperature gradient for the SW location only, binned as an average tunnel studies (Zhang et al 2013) and field data (Magnusson
of each two hours as a function of local hour of day (LST) for the and Smedman 1994, Iungo et al 2013). The horizontal
unwaked (#1, black) SW farm-waked (#2, red), SW direct-waked wind speed differences at wind turbine hub height shown in
(#3, blue), NE farm-waked (#4, green) and NE direct-waked (#5,
magenta) wind direction bins. The error bars denote one standard
figure 2(a) are also qualitatively in agreement with predictions
deviation from the mean. Also plotted in (c) are events for which from the simplified Park model (Katic et al 1986), which
u < 3 ms−1 for the SW farm-waked (#2, yellow) and SW describes wake deficits, 1uwd (i.e. the difference between the
direct-waked (#3, cyan) direction sectors. free-stream velocity at hub height, uhub , and the velocity in the
wake at a given downstream distance) as;
√
For a wind speed at 40 m of 6.0 ms−1 , this change in shear 1 − 1 − CT
4 uwd (x) = uhub ∗ 2 , (2)
exponent due to farm waking, from 0.05 to 0.07, equates to 1+ D2kx
an absolute change in wind speed at hub height of <0.1 ms−1
which is near measurement uncertainty for wind speeds from where CT is the thrust coefficient and k is an empirically
cup anemometers, and is thus close to the limit of detection. determined linear expansion coefficient (here k = 0.07), and
Thus, the inference is the whole wind farm wake is virtually x/D is the rotor normalized distance downstream from the
undetectable in wind speed shear between 40 and 80 m. turbine, equal to 26 and 2.4 for sectors 2 and 3, respectively.
Data from cup anemometers mounted on the meteorolog- For uhub = 6 ms−1 and CT = 0.75, the wake deficit for a
ical masts (i.e. SW location minus NE for all sectors), indicate single turbine at hub height calculated from (2) is 0.14 and
that for the unwaked directions (sector 1) mean wind speeds 1.7 ms−1 for downwind distances equal to those for sectors 2
are almost identical (figure 2(a)). No horizontal velocity and 3, respectively (cf values shown in figure 2). This implies
4
that the mean hub-height velocity deficit derived from these

measurements in sector 3 (where wind speeds from the NE
mast is used as the free-stream and the SW mast is used to
characterize the wake depth) is dominated by the influence of
the closest wind turbine, and shows little additional impact of
the entire wind farm. This is consistent with analyses from
offshore wind farms which have shown that the wake velocity
deficit at hub height is dominated by the impact of the nearest
turbine (Barthelmie et al 2009).
TI is defined as the ratio of the standard deviation (σ )
of horizontal wind speed to the mean value (u) during the
(10 min) sampling interval
σ
TI =. (3)
u
Previous investigations (Yahaya and Frangi 2004) have shown
that TI derived from cup anemometers can exhibit bias
Figure 3. Probability distribution of TI at the (a) SW and (b) NE
at low wind speeds relative to sonic anemometers due to locations during the day (solid) and night (dashed) for the unwaked
inertia, and at high wind speeds due to over-speeding. TI (#1, black) SW farm-waked (#2, red), SW direct-waked (#3, blue),
estimates from a cup or sonic anemometer and those derived NE farm-waked (#4, green) and NE direct-waked (#5, magenta)
using the ZephIR are not directly comparable because the wind direction sectors.
ZephIR scans one height for 3 s over an 18 s repeat cycle.
During each 3 s cycle, 150 line of site measurements are
taken over the scanning volume and averaged together to TI of 12–15%, similar to those shown in Schepers et al 2012),
find one mean wind speed and horizontal velocity variance. and lowest values overnight (TI ∼ 5–7%) (table 1). Consistent
Thus 10 min mean estimates of the mean and variance of with models of rotor induced turbulence and previous work
horizontal wind speed derive from sampling during only (Barthelmie et al 2007), TI is significantly higher in the direct
1 min and 40 s. Further, the scanning volume increases wake (sectors 3 and 5) in all hours of the day. For example,
with height. For an 80 m measurement height, the horizontal median TI for the direct wake sector at the SW location
diameter of the scanning volume is 80 m (the scanning
is much greater than that at the NE location during both
angle of the ZephIR is roughly 30◦ from vertical), and
nighttime hours (median TI at night = 24.7% SW versus
the Doppler returns are considered from an along-beam
5.4% NE) and during the day (25.7% SW versus 12.5%
Lorentzian-weighted range gate. The large scanning volume
NE) (table 1). Conversely, conditions in the wind farm-waked
means there may be substantial variations of streamwise
sectors (#2 and #4) show little enhancement of TI relative
velocity within the sampled volume. Thus care should be
to the unwaked (ambient) TI (sector 1). The enhancement of
taken in interpreting TI and mean velocity measurements
nighttime TI in farm-waked conditions relative to unwaked
from the ZephIR particularly for the direct-waked bins since
conditions is relatively small (5.7 versus 7.7% at the SW mast,
the 30◦ wake widths shown in figure S1 imply a full wake
width of 100 m (190 m downstream of the proximal wind and 5.4 versus 6.6% at the NE mast). Comparable values for
turbines). TI at 80 m from the ZephIR and cup anemometer daytime are 14.2–14.8% and 12.5–12.5%, respectively. Thus,
on the SW mast exhibits a relatively high degree of accord wind farm waking generates at most only a modest increase
(for u > 4 ms−1 , R2 = 0.77) (figure S2 available at stacks. in TI (<2.0% difference between waked and unwaked values,
iop.org/ERL/8/034006/mmedia), as in Peña et al (2009 figure see figure 2(b) and table 1). This implies the cumulative
9), but again re-emphasizes that there are differences in TI turbine-added turbulence is small relative to the ambient TI at
in a scanned volume versus a point measurement. TI from a downwind distance from the wind farm edge of ∼20–30D.
the ZephIR generally exceed those from cup anemometers Probability distributions of TI show a clear signal of
during the daytime (figure S2, red circles) when turbulence higher TI for the direct wake sectors but, consistent with
length scales can be larger than the scanning volume, but the discussion above, the impact of the entire wind farm
ZephIR derived TI are generally less than those from the cup wake (where all turbines are displaced at least ten-times as
anemometer during night (figure S2, blue circles), when stable far away from the measurement point, i.e. 26D downwind,
stratification reduces turbulence length scales. The values of versus for the direct wake when the downstream distance
TI shown here are larger than typically seen in offshore wind is 2.4D) is not detectable (figure 3 and table 1). Direct
farms, but are consistent with previous reports for onshore waking is characterized by higher median TI, and a broader
wind farms (cf figure 7, Wharton and Lundquist 2012). Thus, distribution of TI, both day and night (figure 3). The majority
in the following we only compare wind speed and TI derived of nighttime direct-waking events were characterized by high
from identical instruments (i.e. cup anemometers or ZephIR TI and negative shear (figure S3 available at stacks.iop.
lidars). org/ERL/8/034006/mmedia). Both at night and during the
TI from cup anemometers in the unwaked sector exhibits day the modal value of TI is shifted to higher values in
a typical diurnal pattern with an afternoon maxima (median sectors impacted by the wind farm wake than in unwaked
5
conditions. However, consistent with the median TI values

shown in table 1, the overall change in the modal value
is small. Taken in combination these analyses thus indicate
that stochastic variability in TI (i.e. difference in median
TI between the two mast locations for unwaked sectors of
1–2%, see table 1) is of similar magnitude to the wind farm
effect at a downstream distance of >26D, or 2.1 km, from
the closest turbine (i.e. median farm-waked TI versus median
unwaked TI at both the NE and SW mast is also 1–2%).
These results indicate slightly more rapid dissipation of the
wake than is evident in analysis of data from wind tunnels
which suggest the added streamwise TI behind the turbine
rotor is ∼8% at a downstream distance of 10D under unstable
and neutral stratification (Zhang et al 2013) and is ∼8%
at 20D downstream under stable conditions (Chamorro and
Porté-Agel 2010). A partial explanation for this difference
is that the ambient TI in the wind tunnel experiments was
considerably lower (<10%) than the values experienced in the
field observations.
The vertical potential temperature gradient (1θ/1z) is
quantified using: Figure 4. Hub-height velocity (a) for the SW (solid) and NE
(dashed) location, and wind direction (b), low level shear (c) and TI
4θ θ2 − θ1 (d) for the SW location versus time of day (LST) on 04/20/2012.
= , (4)
4z z2 − z1 The dashed lines in panel (b) indicates the extent of wind sector #3
where the SW mast is impinged upon by the adjacent wind turbine.
where θ1 and θ2 denote the air potential temperature computed The color vertical lines in frame (a) show the times for which
from temperature measurements at z1 = 2 m, and z2 = vertical profiles are shown in figure 5.
75.7 m at the SW mast. 1θ/1z for the unwaked direction
sector indicates unstable conditions during the daytime
(−0.018 ◦ C m−1 ) consistent with positive sensible heat flux,
effects on free-stream vertical potential temperature structure.
and stable (+0.048 ◦ C m−1 ) at night consistent with negative
Contrary to the observed signals of TI and shear, which are
sensible heat flux (figure 2(c) and table 1). Consistent with
highest for the individual wake at 2.4D and appear to decay
remote sensing data (Zhou et al 2012), the largest changes
rather rapidly downstream from the wind farm, the influence
in 1θ/1z due to wind turbine wakes are observed late
of the entire wind farm (as manifest in wind direction sector
at night (figure 2(c)), and during the daytime the change
2) on nighttime vertical potential temperature gradients is
in median 1θ/1z between the farm-waked, direct-waked
larger than for the individual direct wake case (sector 3) (see
and unwaked sector is too small to be able to detect any
figure 2(c)).
impact of adjacent wind turbines or the wind farm. The
1θ/1z overnight in the direct wake and wind farm wake
sectors (+0.026 ◦ C m−1 and +0.022 ◦ C m−1 ) differ at the 3.2. Case studies
95% confidence level from those in the unwaked sectors
(+0.036 to +0.047 ◦ C m−1 , see table 1). However, it must To further illustrate the differences in flow characteristics
be noted that; (i) the individual temperature measurements associated with impacts from direct wind turbine wakes versus
have an accuracy of ±0.1 ◦ C (∼0.002 ◦ C m−1 in our vertical whole wind farm effects, a number of case studies were
potential temperature gradient calculations) and (ii) these analyzed. Two are discussed in detail here.
potential temperature gradients are not compared to the Flow during 04/20/2012 illustrates an example of direct
free-stream and therefore might be related to the larger waking on TI and shear. Initially the wind direction was
scale synoptic meteorology instead of, or in addition to, southerly, veering to become westerly and then northerly by
the wind turbine/wind farm. Thus this relatively weak noon (figure 4(b)). Between 11:00 and 12:30 the SW mast
signal in direct wakes and farm wakes on daytime vertical was waked by turbine #1, located 2.4D away. During this
potential temperature gradients (of ∼0.002 ◦ C m−1 ) is barely time power production at the turbines nearest to the SW and
above the limit of detection. Nevertheless, vertical potential NE locations was relatively constant, and hub-height wind
temperature gradients (figure 2(c)) for SW farm-waked and velocity at the NE location was between 8 and 10 ms−1 while
SW direct-waked direction sectors during calm conditions hub-height velocity at the SW location dropped to 4–5 ms−1
(u < 3 ms−1 ) are considerably larger than those during wind (figure 4(a)). At the same time, TI at the SW location
farm operation (figure 2(c)). Hence this is consistent with the increased from 10% to 30% (figure 4(d)) and low level shear
reduction of 1θ/1z for farm-waked and direct-waked bins became negative (figure 4(c)). Before and after this direct
being due to the effect of wind turbines when the farm is waking there was relatively little difference between velocity
operating (u > 4 ms−1 ), and not being due to synoptic scale at hub height at the SW and NE locations. Profiles from the
6
Figure 6. Hub-height velocity (a), wind direction (b), shear (c), and
TI for the SW (solid) and NE (dashed) locations, and (e) hub height
(80 m, solid) and near-surface (02 m, dashed) potential temperatures
for the SW (red) and NE (blue) locations versus time of day (LST)
on 05/16–05/17/2012. The dashed line in panel (b) indicates the
extent of wind sector #3 where the SW mast is impinged upon by
the adjacent wind turbine. The color vertical lines in frame (a) show
the times for which vertical profiles are shown in figure S5
(available at stacks.iop.org/ERL/8/034006/mmedia).
Figure 5. Vertical profiles of (a) wind speed (u) (ms−1 ),

(b) turbulence intensity (TI) (%) from the ZephIR lidar at SW mast NE location. Between 01:00 and 05:00 (LST), there was
on 04/20/2012 at 09:05 (red), 10:45 (blue) and 14:25 (green). an increase in near-surface (i.e. 2 m) temperature at the
Profile times are indicated in figure 4(a). SW mast by 1.0–1.5 ◦ C which was not seen at the NE
location or at hub height at either location (figure 6(e)).
The magnitude of increase in surface temperature during the
ZephIR lidar show a reduction of wind speed (figure 5(a)) and event is consistent with the mixing of warmer air aloft down
increase in TIZephIR (figure 5(b)) at hub height and further to the surface and partially eroded the vertical gradient of
aloft at approximately 10:45 AM, but not before or after 2.5–3.0 ◦ C (and thus stable stratification) seen before and
the direct-waked event. The time at which the direct-waked after the event at the SW and NE masts to 1.5–2.0 ◦ C
profile is taken is consistent with the change in wind direction during the event. Hourly mean vertical profiles of wind
during this event and small offset between the SW mast and speed (figure S4(a) available at stacks.iop.org/ERL/8/034006/
the ZephIR lidar. The change in wind direction during this mmedia) and TIZephIR (figure S4(b)) from the ZephIR just
period was due to the passage of a cold front, which also before (05/16/2012 23:00), during (05/17/2012 04:00) and
strongly influenced air temperatures, and thus no temperature after the event (05/17/2012 08:00) reveal the extent to which
data are presented for this case study, since synoptic scale direct waking by the adjacent wind turbine 2.4D away
thermal advection rather than local influences dominated air changed the structure of the boundary layer (N.B.: further
temperature variability. details of the wind turbine wake during this time is shown
The period 05/16/2012 22:00–05/17/2012 10:00 is by the plan position indication scan from the Galion lidar
characteristic of periods during which unwaked, wind farm collected at 01:06 on 05/17/2012 (figure S5)). Direct waking
waked and direct wake conditions were observed. During this at the SW mast led to a large decrease in hub-height velocity,
time the wind direction veered from northerly to easterly creating negative shear between 40 and 80 m (figure 6(c)),
(figure 6(b)). Thus before midnight the SW location was and increasing shear in the 80–120 m level. The increase
subject to wind directions associated with impact from the in shear in both layers at 04:00 at the SW location led to
whole wind farm wake, and from 01:00 to 05:00 (LST) enhanced TI at the SW location (i.e. hub-height TI was 14%
the SW location was impacted by a direct wake. Prior to greater at the SW location than at NE, see figures 6(d) and
the direct waking, shear and TI conditions at the SW mast S4(b)), throughout the rotor swept area (40–120 m). Before
were similar to those at the NE mast, and there was little and after the event, there was relatively little difference for TI
difference in hub-height velocity between the two locations. and velocity between the SW and NE locations.
As the wind continued to veer, the SW mast experienced
a direct wake from the proximal turbine, and a large 4. Concluding remarks
drop in hub-height wind speed was observed (figure 6(a))
simultaneously with a decrease in low level shear (figure 6(c)) This study presents observations of flow in the near-surface
and increase in hub-height TI (figure 6(d)) relative to the layer close to a large Midwestern wind farm co-located with
7
corn and soybean crops during spring. Data are conditionally agency thereof, nor any of their employees, makes any
sampled to examine gradients of wind speed, turbulence warranty, express or implied, or assumes any legal liability
intensity, shear and potential temperature under conditions or responsibility for the accuracy, completeness, or usefulness
of free-stream flow (unwaked conditions), downstream of the of any information, apparatus, product, or process disclosed,
entire wind farm, and for direct impingement of wakes from or represents that its use would not infringe privately owned
proximal wind turbines. In accord with a priori expectations rights. Reference herein to any specific commercial product,
and a wealth of prior research, for direct wake cases process, or service by trade name, trademark, manufacturer,
(i.e. where measurements are taken 2.4D (192 m) downstream or otherwise does not necessarily constitute or imply its
of a turbine), there is a clear signal of a decreased wind endorsement, recommendation, or favoring by the United
speeds and increased TI at hub height and throughout the rotor States Government or any agency thereof. The views and
swept area. However, the impact of the whole wind farm wake opinions of authors expressed herein do not necessarily state
on vertical shear and TI is virtually indistinguishable from or reflect those of the United States Government or any
background stochastic variability at distances of 2.1 km or agency thereof’. This manuscript was greatly improved by the
greater. constructive comments of three reviewers who are gratefully
Vertical potential temperature gradients were not strongly acknowledged.
affected by direct wake impingement or whole wind
farm waking during the daytime, but nighttime potential References
temperature gradients were reduced relative to unwaked
conditions. Mean 1θ/1z was 0.026 ◦ C m−1 for the direct Baidya Roy S 2011 Simulating impacts of wind farms on local
wake, 0.022 ◦ C m−1 in the wind farm wake and 0.048 ◦ C m−1 hydrometeorology J. Wind Eng. Ind. Aerodyn. 99 491–8
in unwaked wind directions. Hub-height temperatures varied Barthelmie R J, Frandsen S T, Nielsen M, Pryor S, Rethore P E and
little with wind direction, thus the change in nighttime vertical Jørgensen H 2007 Modelling and measurements of power
losses and turbulence intensity in wind turbine wakes at
potential temperature gradients was due almost solely to an
Middelgrunden offshore wind farm Wind Energy 10 517–28
increase in 2.5 m temperature, which was +1.6 ◦ C greater in Barthelmie R J, Hansen K S and Pryor S C 2012 Meteorological
the direct-waked case than in the unwaked case, and +1.9 ◦ C controls on wind turbine wakes Proc. IEEE 99 1–10
greater in the farm-waked case than in the unwaked cases. The Barthelmie R J et al 2009 Modelling and measuring flow and wind
increases in air temperature are consistent with case studies turbine wakes in large wind farms offshore Wind Energy
presented in Rajewski et al (2013) who reported warming at 12 431–44
9 m height of 0–1.5 ◦ C in direct wakes during the nighttime, Barthelmie R J, Pryor S C, Frandsen S T, Hansen K S,
Schepers J G, Rados K, Schlez W, Neubert A, Jensen L E and
and the analysis of MODIS data in Zhou et al (2012), who Neckelmann S 2010 Quantifying the impact of wind turbine
reported winter nighttime warming due to wind farms of wakes on power output at offshore wind farms J. Atmos.
∼0.45 ◦ C. Ocean. Technol. 27 1302–17
Our observations suggest that wind speed deficits and Calaf M, Parlange M and Meneveau C 2011 Large eddy simulation
increases in turbulence intensity created by very large onshore study of scalar transport in fully developed wind-turbine array
wind farms co-located with agricultural crops during the boundary layers Phys. Fluids 23 126603
Chamorro L P and Porté-Agel F 2010 Effects of thermal stability
spring recovery rapidly, likely within 2 km of the downstream
and incoming boundary-layer flow characteristics on
edge and have little impact on the potential temperature wind-turbine wakes: a wind-tunnel study Bound.-Layer
structure from the surface to hub height or by association Meteorol. 136 515–33
surface sensible heat fluxes (for a 80 m hub height) during the Crawford T T 1965 Frost protection with wind machines and
day. However, operating wind turbines may reduce nighttime heaters Meteorol. Monogr. 6 81–7
vertical potential temperature gradient within and downstream Elliott D, Schwartz M and Scott G 2008 Wind shear and resources
at elevated heights: Indiana and Iowa case studies Presented at
(<2 km) of large wind farms potentially altering near-surface
WINDPOWER 2008 (Houston, TX)
scalar fluxes and reducing risk from early season freezing Iungo G V, Wu Y and Porté-Agel F 2013 Field measurements of
events. Thus, in this study the influence of the downstream wind turbine wakes with lidars J. Atmos. Ocean. Technol.
influence of the wind farm on atmospheric properties and 30 274–87
profiles was relatively modest, but further investigations Katic I, Højstrup J and Jensen N O 1986 A simple model for cluster
during other seasons are warranted. efficiency Proc. EWEC’86 (Rome) vol 1, pp 407–10
Lu H and Porté-Agel F 2011 Large-eddy simulation of a very large
wind farm in a stable atmospheric boundary layer Phys. Fluids
Acknowledgments 23 065101
Magnusson M and Smedman A S 1994 Influence of atmospheric
This research was supported in part by the National Science stability on wind turbine wakes Wind Energy 18 139–52
Foundation (grant no. 1067007), by the Department of Energy Mann J, Peña A, Bingöl F, Wagner R and Courtney M S 2010 Lidar
scanning of momentum flux in and above the atmospheric
under Award Number #DE-EE0005379, and by the Nell J surface layer. J. Atmos. Ocean. Technol. 27 959–76
Redfield Foundation. Peña A, Hasager C B, Gryning S, Courtney M, Antoniou I and
Mikkelsen T 2009 Offshore wind profiling using light
Disclaimer: ‘This report was prepared as an account detection and ranging instruments Wind Energy 12 105–24
of work sponsored by an agency of the United States Politis E S, Prospathopoulos J, Cabezon D, Hansen K S,
Government. Neither the United States Government nor any Chaviaropoulos P K and Barthelmie R J 2012 Modeling wake
8
effects in large wind farms in complex terrain: the problem, the Wharton S and Lundquist J K 2012 Assessing atmospheric stability
methods and the issues Wind Energy 15 161–82 and its impacts on rotor-disk wind characteristics at an onshore
Rajewski D A et al 2013 Crop Wind Energy Experiment (CWEX): wind farm Wind Energy 15 525–46
observations of surface-layer, boundary-layer, and mesoscale Yahaya S and Frangi J P 2004 Cup anemometer response to the
interactions with a wind farm Bull. Am. Meteorol. Soc. wind turbulence–measurement of the horizontal wind variance
94 655–72 Ann. Geophys. 22 3363–74
Sathe A, Mann J, Gottschall J and Courtney M S 2011 Can wind Zhang W, Markfort C D and Porté-Agel F 2013 Wind-turbine
lidars measure turbulence? J. Atmos. Ocean. Technol. wakes in a convective boundary layer: a wind-tunnel study
28 853–68 Bound.-Layer Meteorol. 146 161–79
Schepers J G, Obdam T S and Prospathopoulos J 2012 Analysis of Zhou L, Tian Y, Roy S B, Thorncroft C, Bosart L F and Hu Y 2012
wake measurements from the ECN Wind Turbine Test Site Impacts of wind farms on land surface temperature Nature
Wieringermeer, EWTW Wind Energy 15 575–91 Clim. Change 2 539–43
9
Journal of Physics: Conference Series
PAPER • OPEN ACCESS
Cluster Wind Power Uncertainty Model and Operation Simulation

Method
To cite this article: Xuxia Li et al 2020 J. Phys.: Conf. Ser. 1449 012077

ISPECE 2019 IOP Publishing
Journal of Physics: Conference Series 1449 (2020) 012077 doi:10.1088/1742-6596/1449/1/012077
Cluster Wind Power Uncertainty Model and Operation

Simulation Method
Xuxia Li1, Yingying Hu2, Yao Wang3, Jiaojiao Deng4

1
Planning and Review Center, Economic and Electrical Research Institute of Shanxi
Electrical Power Company of SGCC, TaiYuan, ShanXi, 030002, China
2
3
4
*
Corresponding author’s e-mail: 779656332@qq.com
Abstract. The rational modelling of wind power uncertainty is the premise of relevant analysis
and decision-making. The uncertainty, intermittent nature of cluster wind power and the spatial
correlation between different wind farms are the key factors to determine the impact of cluster
wind power on the power system. The rationality of the wind power uncertainty model will
determine the validity of the wind power uncertainty analysis and decision-making model and
the credibility of the analysis conclusions. The development of wind power in China is very
rapid, but the accumulation of wind power and meteorological statistics is insufficient. The
research on the uncertainty of cluster wind power output is not deep-going. For this reason, the
wind power uncertainty model is taken as the entry point. This paper will establish the
uncertain model of wind power and its simulation method. This method can generate time
series of wind farm output in accordance with wind power uncertainty characteristics, and
effectively analyze the overall output characteristics of large-scale cluster wind power to be
planned in the future, which provides a new idea for studying the impact of wind power on
future power systems.
1. Overview
The rational modelling of wind power uncertainty is the premise of relevant analysis and decision-
making. The uncertainty, intermittent nature of cluster wind power and the spatial correlation between
different wind farms are the key factors to determine the impact of cluster wind power on the power
system. The rationality of the wind power uncertainty model will determine the validity of the wind
power uncertainty analysis and decision-making model and the credibility of the analysis conclusions.
The development of wind power in China is very rapid, but the accumulation of wind power and
meteorological statistics is insufficient. The research on the uncertainty of cluster wind power output
is not deep-going.
At present, there are three main problems in the analysis of cluster wind power uncertainty:
(1) The analysis stays at the level of actual data and does not deep into the height of the model and
method. At present, the existing research on wind power uncertainty is based on the statistical analysis
of the collected wind power historical data, but does not propose a widely applicable model. When the
object of analysis is changed, the reference and significance of it are limited.
(2) The detailed modelling of the spatial correlation of wind farms is less. For cluster wind power,
the spatial correlation between wind farms is an important factor in determining its overall output
characteristics and its impact on the power system. At present, the research on the spatial correlation
of wind farms only stays in the qualitative analysis and the calculation of several simple indicators,
and there is no mature theoretical support and sufficient empirical analysis.
(3) The perfect wind power operation simulation model is insufficiency. Many wind power
research needs to use time series data of wind speed or wind power output, but these data are often
difficult to obtain in practice: the wind farms studied may not have been fully put into operation or the
wind tower is not built, resulting in insufficient accumulation of wind data; The dispatching operation
of wind power farm may be affected by various factors, and its output cannot reflect its true random
characteristics. At present, there is a lack of a time series output simulation method that
comprehensively considers various operational factors of wind farms.
For this reason, the wind power uncertainty model is taken as the entry point. This paper will
establish the uncertain model of wind power and its simulation method. A probabilistic sequence
model of load and wind power dependence is established. At the same time, based on the study of
wind power intermittent, a set of cluster wind power operation simulation methods [1] that can
consider the random characteristics of wind speed, intermittent and spatial correlation are proposed.
The relationship between the research content of this paper can be summarized in figure 1.
This paper uses the way of the combination of empirical analysis and mathematical modeling.
Because of the wind speed and wind power prediction error time series data used in the analysis are
not enough in China, the Wind Integration Datasets[2] of US National Renewable Energy Laboratory
(NERL) is selected, which contains wind data from the US and offshore 30,000 wind towers from
2004 to 2006, meanwhile, the database also use the IEC wind power standard power characteristic
curve to indicate the "predicted" output and "predicted" forecast output of wind farm, providing
important data support for analyzing the uncertainty analysis and modeling of large-scale wind power
output. This paper will select the corresponding time series data from the 10 wind towers on the East
Coast of the United States in 2004 for an empirical analysis. The geographic locations and their
coordinate IDs of the 10 wind towers are shown in figure 2. It should be noted that although the US
wind power data is used in this paper, the proposed analytical methods and analytical conclusions are
equally applicable to wind farms in China.
2. Probability sequence modelling of power system generation and load dependence

The core of power system uncertainty analysis is considering the model and calculation method of
uncertainty. In the uncertainty analysis of power system considering wind power, the uncertainty of
power generation and load is usually considered at the same time. To this end, this section will
establish a probabilistic sequence model of power system generation and load dependence including
wind power, load and generator, which will lay the foundation for the uncertainty analysis and
decision-making method proposed in the following paper. Copula theory expresses the joint
distribution of multiple random variables as the "join" of their respective edge distributions. Therefore,
using this theory, the related multiple random variables can be modelled separately according to their
edge distribution and dependent structure, which provides great convenience for studying multiple
related random variables.
2.1. Univariate Edge Distribution Sequence Modeling

Taking the uncertainty of the long-term output of a single wind farm as an example, taking the
capacity discretization common factor as 𝐶̄ , the wind power output discretization sequence length is:
𝑁𝑤 =< 𝐶𝑤 /𝐶̄ > (1)
𝐶𝑤 in formula (1) is the installed capacity of the wind power farm. According to the theory of
sequence operations, the wind power farm has a total of states of 𝑁𝑤 + 1, wherein the available
2
capacity of the i state is (𝑖 − 1)𝐶̄ . In the process of serialization of wind power farm output,
according to historical statistics, the frequencies of the output of each discretized interval wind power
farm are separately counted as an estimate of the probability that the output of the wind farm falls into
the discretization interval. Taking the statistical one-hour hourly output in one year as an example, the
𝑃𝑤 (𝑗) is the output of wind power farm in the j hour, the wind power output sequence 𝑆𝑤 (𝑖) can be
calculated by the following formula:
∑8760 ̄
𝑗=1 𝐼𝑖 [<𝑃𝑤 (𝑗)/𝐶 >]
𝑆𝑤 (𝑖) = 8760
, 𝑖 = 0,1, . . . , 𝑁𝑤 (2)
In equation (2), < 𝑃𝑤 (𝑗)/𝐶̄ > indicates discretization of wind power output data, 𝐼𝑖 (𝑥) is an
explanatory function.
The modelling principle of load sequence model is the same as the wind power output sequence
model. The system load sequence can be generated by using the historical value of the statistical
system load and formula (2). This method is also applicable to the serialization modelling of the node
load.
In addition to the use of historical statistical values to generate a load sequence model, many
studies tend to consider the load as a normal distribution [4-19]. Therefore, a sequence model of load can
be established by load mean and variance information. Assume that the load mean value is 𝜇 and the
variance is 𝜎 2 . Since the normal distribution is unbounded distribution, it needs to be truncated, and
the probability distribution within the range of 𝜇 ± 5𝜎 is serialized. That is 𝑖𝑚𝑖𝑛 =< 𝜇 − 5𝜎 > ,
𝑖𝑚𝑎𝑥 =< 𝜇 + 5𝜎 >, then the sequence model of the load 𝑆𝑑 (𝑖) is:
𝑖𝐶
(𝑥−𝜇)2
1
∫ 2
𝑒 2𝜎2 𝑑𝑥 𝑖 = 𝑖𝑚𝑖𝑛
√2𝜋𝜎
−∞
𝜋
(𝑥−𝜇)2
1 −
𝑆𝑑 (𝑖 ) = ∫ √2𝜋𝜎 2
𝑒 2𝜎2 𝑑𝑥 𝑖 = 𝑖𝑚𝑖𝑛 + 1, 𝑖𝑚𝑖𝑛 + 2, … , 𝑖𝑚𝑎𝑥 − 1 (3)
(𝑖−1)𝐶
+∞
(𝑥−𝜇)2
1 −
∫ 𝑒 2𝜎2 𝑑𝑥 𝑖 = 𝑖𝑚𝑎𝑥
√2𝜋𝜎2
{ (𝑖−1)𝐶
In the uncertainty analysis of power system, the conventional generator set is generally modelled as
a two-state model considering the forced outage rate (FOR). Assume that the installed capacity of the
unit is 𝐶𝑠 , the forced outage rate is 𝐹𝑠 . The available capacity sequence length of the unit is 𝑁𝑠 =<
𝐶𝑠 /𝐶̄ >. The available capacity sequence 𝑆𝑠 (𝑖) of the unit can be expressed as:
𝐹𝑠 𝑖 = 0
(0)
𝑆𝑠 (𝑖) = {0, 𝑖 = 1,2, . . . , 𝑁𝑠 − 1 (4)
1 − 𝐹𝑠 , 𝑖 = 𝑁𝑠
For hydroelectric units or thermal power units that consider the operating state of the falling force,
they are often modelled as multi-state units[16], and their serialization methods are consistent with the
above-described modelling principles.
It should be noted that in the process of serializing the above-mentioned conventional generator set,
load, turbine generator system and virtual unit, a unified capacity discretization common factor 𝐶̄ is
required, which will affect the accuracy of the uncertainty analysis.
2.2. Dependent structure modelling

Power system power generation and load-dependent structure refers to the complex relationship
between the available capacity of different power generation components and the load of different
nodes. It can be seen from the modelling process of the dependent probability sequence which the
complex joint distribution function between random variables with different edge distributions can be
transformed into the composite function form of their respective edge distributions and their
3
dependent structures by Copula theory. Wherein, the dependent structure can be easily represented by
the Copula function. Thus, the modelling process of the dependent structure is transformed into the
fitting and parameter identification process of the Copula function, which is then transformed into the
process of determining the coefficient of rank correlation between each random variable. Assume that
the coefficient of rank correlation matrix between wind power and load is 𝐶𝐷 . In this section, the
modeling of the dependent structure is divided into two cases: the dependent structure modeling
considering only active and the dependent structure modeling considering both active and reactive.
Only the active power of power generation and load is considered in the unit combination model
and the DC power flow model. At this time, the rank correlation coefficient matrix 𝐶𝐷 of wind power
and load can be expressed as:
𝐶𝑊𝑃 𝐶𝑊𝑃−𝐿𝑃
𝐶𝐷 = [ 𝑇 ] (5)
𝐶𝑊𝑃−𝐿𝑃 𝐶𝐿𝑃
In the formula (5), the sub-blocks 𝑪𝑾𝑷 represent the rank correlation coefficients of the active
outputs of different wind farms, and the magnitudes thereof can be estimated by the distance between
the wind farms. 𝑪𝑳𝑷 indicates the correlation between load and active power of different nodes, and it
is mainly determined by the type of load (industrial load, commercial load, etc.) supplied by the node
and the power supply scale of the node. 𝑪𝑾𝑷−𝑳𝑷represents the correlation coefficient between wind
power and load and is determined by the daily characteristics of the load and wind power. The
identification of the above correlation coefficients and the selection of the Copula function can be
implemented by the method.
2.3. The dependent structure modelling of active and reactive power at the same time
In the calculation of the AC tidal current model, it is necessary to consider both the active power and
the reactive power of the power generation and the load. Therefore, the modelling of the dependent
structure is more complicated than considering only the active power. In the dependent structures
modelling considering active and reactive at the same time, the structure of the rank correlation
coefficient matrix 𝑪𝐷 between load and wind power output is:
𝑪𝑊𝑃 𝑪𝑊𝑃−𝑊𝑄 𝑪𝑊𝑃−𝐿𝑃 𝑪𝑊𝑃−𝐿𝑄
𝑇 𝑪𝑊𝑄−𝐿𝑃 𝑪𝑊𝑄−𝐿𝑄
𝑪𝑊𝑃−𝑊𝑄 𝑪𝑊𝑄
𝑪𝐷 = 𝑇 𝑇 (6)
𝑪𝑊𝑃−𝐿𝑃 𝑪𝑊𝑄−𝐿𝑃 𝑪𝐿𝑃 𝑪𝐿𝑃−𝐿𝑄
[ 𝑪𝑊𝑃−𝐿𝑄 𝑪𝑊𝑄−𝐿𝑄
𝑇 𝑪𝑇𝐿𝑃−𝐿𝑄 𝑪𝐿𝑄 ]
The meaning and calculation method of 𝑪𝑊𝑃 , 𝑪𝑊𝑃−𝐿𝑃 and 𝑪𝐿𝑃 in equation (6) are the same as it in
the formula (5). The other sub-blocks in the matrix (6) are related to reactive power. In theory, these
matrices can be obtained by analyzing the synchronized time series with the actual data. However, in
practice, the data of wind power reactive power and reactive load are not easy to obtain. Therefore, the
rank correlation coefficient sub-matrix related to reactive power can be estimated by.
𝑪𝑊𝑄 and 𝑪𝐿𝑄 respectively indicate the rank correlation coefficient of reactive power between the
different node loads and the different wind power farms, it can generally be considered to be the same
as the corresponding active rank correlation coefficient 𝑪𝑊𝑃 and 𝑪𝐿𝑄 .
𝑪𝑊𝑃−𝑊𝑄 indicates the rank correlation coefficient between the active output and the reactive output
of different wind farms. The diagonal element indicates the co-correlation between the active output
and the reactive output of a certain wind farm, which is related to the wind farm control strategy.
Generally speaking, wind farms are all powered by a constant power factor, so the diagonal elements
of 𝑪𝑊𝑃−𝑄 are often close to 1, and the non-diagonal elements of 𝑪𝑊𝑃−𝑄 indicate the correlation
between the active output of one wind farm and the reactive power of another wind farm. Its’ amount
can be estimated indirectly by the active power output correlation of two wind power farms and the
rank correlation coefficient between active and reactive power output of the wind farm. Assume that
the element in 𝑪𝑊𝑃 is 𝑐 𝑊𝑃 , the element in 𝑪𝑊𝑃−𝑊𝑄 is 𝑐 𝑊𝑃−𝑊𝑄 , then the non-diagonal element in
𝑪𝑊𝑃−𝑊𝑄 can be estimated by the following formula:
4
𝑊𝑃−𝑊𝑄 𝑊𝑃 𝑊𝑃−𝑊𝑄
𝑐𝑖𝑗 = 𝑐𝑖𝑗 𝑐𝑗𝑗 (7)
That is, the rank correlation coefficient of the active output of the i wind farm and the reactive
power of the 𝑗 wind farm can be approximated as the product of corresponding element in 𝑪𝑊𝑃 and
the corresponding diagonal element to the j wind farm in 𝑪𝑊𝑃−𝑊𝑄 .
𝑪𝐿𝑃−𝐿𝑄 indicates the rank correlation coefficient of the active load and the reactive load between
different nodes, and its’ diagonal element indicates the rank correlation coefficient between the active
load and the reactive load of the node. Generally, the number of users carried by the node is related to
the load characteristic, and The fewer users carried by the node, the more single for the load type, then
the stronger the correlation between the active load and the reactive load. The non-diagonal elements
of 𝑪𝐿𝑃−𝐿𝑄 represent the correlation between the active load of one node and the reactive load of
another node, where the size of each element can be estimated indirectly by 𝑪𝐿𝑃 and𝑪𝐿𝑃−𝐿𝑄 :
𝐿𝑃−𝐿𝑄 𝐿𝑃 𝐿𝑃−𝐿𝑄
𝑪𝑊𝑄−𝐿𝑄 indicates the co-correlation between a wind power reactive output and a node reactive
load. Generally, it can be considered that the corresponding active rank correlation coefficients are the
same.
𝑪𝑊𝑃−𝐿𝑄 represents the correlation between the active output of a wind farm and the reactive load of
a node. The size of each element can be estimated indirectly by 𝑪𝑊𝑃−𝐿𝑃 and 𝑪𝐿𝑃−𝐿𝑄 :
𝑊𝑃−𝐿𝑄 𝑊𝑃−𝐿𝑃 𝐿𝑃−𝐿𝑄
𝑪𝑊𝑄−𝐿𝑃 indicates the correlation between the reactive output of a certain wind farm and the active
output of a certain node. The principle of is similar to determination of𝑪𝑊𝑃−𝐿𝑄 , as follows:
𝑊𝑄−𝐿𝑃
𝑐𝑖𝑗 = 𝑐𝑖𝑖𝑊𝑃−𝑊𝑄 𝑐𝑖𝑗 𝑊𝑃−𝐿𝑃
(10)
3. Cluster wind power operation simulation model and method research

Effective cluster wind power operation simulation is of great significance for studying and planning
the output characteristics of wind farms and their impact on the power system. The key to the
simulation of cluster wind power operation is to consider the randomness of wind speed, the
correlation of multiple wind farms and the volatility of wind speed at the same time. In addition,
factors such as wind turbine reliability and wind farm wake effect should be considered.
Based on the study of wind power intermittent, this section uses the stochastic differential equation
model introduced in [20] to establish a cluster wind power operation simulation model and method.
3.1. Intermittent analysis of wind power

The intermittent nature of wind power comes from the volatility of wind speed. The effective way to
characterize the volatility of wind speed is to use the autocorrelation function, which is the linear
correlation coefficient of the sequence of time series and its own time shift. Set 𝑣𝑡 is a time series, 𝑣̄
and 𝜎 2 is its’ mean and variance, then the autocorrelation function 𝜌(𝑘) of 𝑣𝑡 is:
1 ∑𝑇 (𝑣 −𝑣̄ )(𝑣 −𝑣̄ )
𝜌(𝑘) = 𝑇 𝑡=1 𝑡 𝜎2 𝑡−𝑘 , 𝑘 = 1,2,3. .. (11)
The autocorrelation function is a measure of the temporal correlation of time series, reflecting the
magnitude of the volatility of the time series. In general, the value of the autocorrelation function is
attenuated by increasing the time difference, and the more intense the time series fluctuation, the faster
the autocorrelation function decays. Studies have shown that the autocorrelation function of wind
speed can be numerically represented by a negative exponential function:
𝜌(𝑘) = 𝑒 −𝜃𝑘 , 𝜃>0,𝑘 = 1,2,3. .. (12)
In equation (12), the size of 𝜃 is determined by the speed of the autocorrelation function decay,
which in turn can reflect the severity of wind speed fluctuations.
The wind speed time series of a wind tower is selected, and the time interval of the data is 1 hour.
The autocorrelation function of the wind speed time series is fitted by the negative exponential
5
function. The result is shown in figure 1. The fitting value 𝜃 of the attenuation coefficient of the wind
speed autocorrelation function =0.0904
Figure 1. Autocorrelation function and its fitting result of the wind farm (attenuation coefficient
=0.0904)
3.2. Single wind farm operation simulation model

The cluster wind power operation simulation method proposed in this section uses the stochastic
differential equation model introduced in [20]. The model is capable of simulating a stochastic process
that yields a given edge distribution and the autocorrelation function obeys exponential decay. Among
them, the single wind farm operation simulation model can be briefly described as follows:
Theorem 1 If the probability density function 𝑓(𝑥) is greater than zero, continuous and the
variance is limited in its’ domain of definition(𝑙, 𝑢), its mathematical expectation 𝐸(𝑥) = 𝜇 , the
stochastic process 𝑋𝑡 satisfies the following stochastic differential equation:
𝑑𝑋𝑡 = −𝜃(𝑋𝑡 − 𝜇)𝑑𝑡 + √𝜐(𝑋𝑡 )𝑑𝑊𝑡 , 𝑡 ≥ 0 (13)
Wherein 𝜃 ≥ 0, Wt is the standard Brownian motion, 𝜐(𝑋𝑡 ) is a non-negative function defined on
(𝑙, 𝑢):
2𝜃 𝑥
𝜐(𝑥) = ∫𝑙 (𝜇 − 𝑦)𝑓(𝑦) 𝑑𝑦, 𝑥 ∈ (𝑙, 𝑢) (14)
𝑓(𝑥)
Then there are the following conclusions:
(1) The stochastic process 𝑋 is the ergodic of each state and the corresponding edge probability
density function 𝑓(𝑥) of its state at each moment.
(2) The stochastic process 𝑋 is mean-reverting and its autocorrelation function is consistent with:
𝑐𝑜𝑟𝑟(𝑋𝑠+𝑡 , 𝑋𝑠 ) = 𝑒 −𝜃𝑡 , 𝑠, 𝑡 ≥ 0 (15)
A stochastic process model can be established by using the above theorem. The corresponding
probability distribution of the state of the stochastic process at each moment (hereinafter referred to as
the probability distribution of the stochastic process) obeys the Weibull distribution, and its
autocorrelation function obeys the exponential decay. The sample with the fixed time interval of the
random process is generated as the wind speed time series, and the running simulation of the single
wind farm can be realized by combining the wind turbine standard power characteristic curve, the
reliability model and the wind farm wake effect.
Set the wind speed meet the Weibull distribution with the scale parameter c and the shape
parameter 𝑘, then the average wind speed is 𝜇:
1
𝜇 = 𝐸(𝑋) = 𝑐𝛤 (𝑘 + 1) (16)
The expression of the Weibull distribution and the definition of the function 𝜐(𝑋𝑡 ) according to
equation (1):
6
2𝜃 𝑥 2𝜃 𝑥
𝜐(𝑥) = ∫ (𝜇 − 𝑦)𝑓(𝑦) 𝑑𝑦 = (𝜇𝐹(𝑥) − ∫𝑙 𝑦𝑓(𝑦) 𝑑𝑦)
𝑓(𝑥) 𝑙 𝑓(𝑥)
2𝜃 1 𝑥 𝑘 𝑥 𝑘 1
= 𝑓(𝑥) (𝑐𝛤 (𝑘 + 1) (1 − 𝑒𝑥𝑝 [− ( 𝑐 ) ]) − 𝑐𝛤 (( 𝑐 ) , 𝑘 + 1)) (17)
Wherein, 𝐹(𝑥) is the corresponding distribution function for𝑓(𝑥), 𝛤(𝑎 )is a gamma function
+∞
𝛤(𝑎 ) = ∫0 𝑦 𝑎−1 𝑒 −𝑦 𝑑𝑦 (18)
𝛤(𝑥, 𝑎), 𝑥 ≥ 0 is incomplete gamma function
𝑥
𝛤(𝑥, 𝑎) = ∫0 𝑦 𝑎−1 𝑒 −𝑦 𝑑𝑦 (19)
In summary, a single wind farm output time series 𝑣̂𝑡 ,𝑡 = 1,2, . . . , 𝑇can be calculated iteratively by
the following formula:
𝑣̂𝑡 = 𝑣̂𝑡−1 + 𝑑𝑋𝑡 (20)
The wind speed sequence of the wind farm is not a completely random process. For the weather
reasons, the wind speed of the wind farm in different seasons is different and has certain regularity
(such as small in winter and large in summer). In one day, the average wind speed at different times in
the day is different (such as large at night and small during the day), due to the different surface
temperatures of the wind farm. In order to consider the seasonality of the wind speed and the
regularity in one day of the wind farm, the generated wind speed sequence 𝑣𝑡 is corrected as follows:
𝑣𝑡 = 𝑘ℎ 𝑘𝑚 𝑣̂𝑡 (21)
Among them, 𝑘𝑚 (𝑚 = 1,2, . . . ,12) is the seasonal factor for wind speed, 𝑘ℎ (ℎ = 1,2, . . . ,24) is the
daily characteristic factor for wind speed.
Using the corrected wind speed sequence𝑣𝑡 , considering the output characteristic curve of wind
turbine generator, the wind farm wake effect and the wind farm wind turbine output reliability, the
wind farm timing output curve can be generated by the following formula:
𝑃𝑡 = 𝑛𝑡 (1 − 𝜂)𝐶 (𝑣𝑡 𝑘ℎ 𝑘𝑚 ) (22)
Among them, 𝐶(𝑥) is the output characteristic curve of wind turbine generator,𝜂 is the wind farm
wake effect coefficient, which indicates that the wind farm loses its output due to the wake effect,
which is usually 5%~10%, 𝑛𝑡 is the number of units available for the wind farm, which is a random
variable and is the number of wind turbines available in the wind farm. If it is assumed that the unit
fault in the wind farm is subject to an independent exponential distribution, then for any time t , the
available wind turbines in the wind farm is subject to Bernoulli distribution, which can be obtained by
random sampling.
3.3. Cluster wind farm operation simulation model

The key to the simulation of cluster wind farm operation is to generate wind speed sequences
containing spatial correlation. According to the theorem and its properties of using multiple stochastic
differential equations to generate multiple stochastic processes in [20], it is easy to prove that
when 𝑓(𝑥) is a normal distribution and𝑊𝑡 is a non-independent multiple Brownian motion. Each
dimension edge probability distribution corresponding to the time state in its’ established
multidimensional normal stochastic processes still satisfies the two conditions in the conclusion of
Theorem 1, and at the same time, the linear correlation coefficient matrix of the joint probability
distribution corresponding to the state of the multi-dimensional normal stochastic process at each
moment is the same with the𝑑𝑊𝑡 , and the specific proof process for the above conclusions is given in
Appendix C. Using the above conclusions, it is possible to design a cluster wind farm operation
simulation model similar to the principle of section 2. The cluster wind farm wind speed sequence
(1) (2) (𝑁)
𝑋𝑡 = [𝑋𝑡 , 𝑋𝑡 , . . . , 𝑋𝑡 ] can be generated by a number of stochastic differential equations as
follows:
(1) (1) (1) (1)
𝑋𝑡 = −𝜃 (1) (𝑋𝑡 − 𝜇 (1) )𝑑𝑡 + √𝜐 (1) (𝑋𝑡 )𝑑𝑊𝑡
(2) (2) (2) (2)
𝑋𝑡 = −𝜃 (2) (𝑋𝑡 − 𝜇 (2) )𝑑𝑡 + √𝜐 (2) (𝑋𝑡 )𝑑𝑊𝑡
7
...
(𝑁) (𝑁) (𝑁) (𝑁)
𝑋𝑡 = −𝜃 (𝑁) (𝑋𝑡 − 𝜇 (𝑁) )𝑑𝑡 + √𝜐 (𝑁) (𝑋𝑡 )𝑑𝑊𝑡
(23)
(1) (2) (𝑁)
𝜃 ,𝜃 ,...,𝜃 indicates the autocorrelation function attenuation coefficient of wind speed of
each wind farm, 𝜇 (1) , 𝜇 (2) , . . . , 𝜇 (𝑁) indicates the average wind speed of each wind farm. The
(1) (2) (𝑁)
function𝜐 (1) (𝑋𝑡 ),, 𝜐 (2) (𝑋𝑡 ),…, , 𝜐 (𝑁) (𝑋𝑡 ) is determined by the wind speed parameters of each
(1) (2) (𝑁)
wind farm by formula (24). 𝑊𝑡 , 𝑊𝑡 , . . . , 𝑊𝑡 are 1-dimensional Brownian motions, whose linear
(1) (2) (𝑁)
correlation coefficient in increments 𝑑𝑊𝑡 , 𝑑𝑊𝑡 , . . . , 𝑑𝑊𝑡 per unit time is 𝛴.
It can be seen from Theorem 1 that the edge probability distributions of each dimension of
(1) (2) (𝑁)
𝑋𝑡 , 𝑋𝑡 , . . . , 𝑋𝑡 satisfy the two conditions in the conclusion of Theorem 1 respectively. Set the
(1) (2) (𝑁)
linear correlation coefficient of 𝑋𝑡 , 𝑋𝑡 , . . . , 𝑋𝑡 in the joint probability distribution corresponding to
each time state is 𝛴′ . The conclusion in Appendix C is for𝑓(𝑥) in normal distribution and therefore
𝛴 = 𝛴′ cannot be derived. In fact, for the Weibull distribution, this conclusion is not strictly
established. However, since the shape of the Weibull distribution is similar to the normal distribution,
it can be inferred that 𝛴′ is approximately equal to 𝛴 in the numerical value. The actual example in this
section shows that the linear correlation coefficient matrix 𝛴′ corresponding to the wind speed
generated by equation (23) is very similar to 𝛴.
It should be noted that the matrix obtained when evaluating the spatial correlation of wind speed is
the Kendall rank correlation coefficient matrix. When the simulation is performed by using equation
(23), the input matrix is a linear correlation coefficient matrix. After obtaining the wind speed curves
of the wind farms, the output curves of multiple wind farms can be generated by the same method as
the single wind farm operation simulation.
4. Summary
When studying the problem of large-scale wind power consumption in the power system in the future,
the actual wind power data is difficult to obtain, which makes it difficult to conduct research on the
impact of wind power uncertainty on the power system. Therefore, establishing a reasonable and
effective cluster wind farm operation simulation model is a necessary tool for subsequent research. To
this end, this paper uses the autocorrelation function model to study the intermittence of wind power,
on this basis, using a stochastic differential equation theory, a set of cluster wind power simulation
methods that can consider the random characteristics of wind speed, intermittent and spatial
correlation are proposed. This method can generate time series of wind farm output in accordance with
wind power uncertainty characteristics, and effectively analyze the overall output characteristics of
large-scale cluster wind power to be planned in the future, which provides a new idea for studying the
impact of wind power on future power systems. This paper not only verifies the effectiveness of
modelling of wind power using the theory of dependent probability sequence operation, but also
reflects the necessity of considering wind power spatial correlation in uncertain analysis and decision-
making of cluster wind power.
Acknowledgement
Science and technology support project of State Grid Corporation of China, Contract number:
SGSXJY00PSJS1800016
References
[1] Zhang N, Kang C, Duan C, et al. (2010) Simulation methodology of multiple wind farms
operation considering wind speed correlation. International Journal of Power and Energy
Systems, vol.30(4), pp.264-173.
8
[2] National Renewable Energy Laboratory (NERL)．(2011) Wind Integration Datasets [2011-04-
20], http://www.nrel.gov/wind/integrationdatasets/, webpage.
[3] Hannele H. (2005) Hourly wind power variations in the Nordic countries. Wind Energy,vol. 8(2),
pp.173-195.
[4] Borkowska B. (1974) Probabilistic Load Flow. IEEE Transactions on Power Apparatus and
Systems, PAS-93(3), pp.752-759 .
[5] Ding M, Li H, Huang K, (2001) Probabilistic Load Flow Analysis Based On Monte-Carlo
Simulation, Power System Technology, vol.25(11),pp.10-14,22.
[6] Leite da Silva A M, Arienti V L. (1990) Probabilistic load flow by a multilinear simulation
algorithm. IEE Generation, Transmission and Distribution, vol.137(4), pp. 276-282.
[7] Breipohl A, Lee F N, Huang J, et al. (1990) Sample size reduction in stochastic production
simulation. IEEE Transactions on Power Systems, vol.5(3), pp. 984-992.
[8] Yu H, Zhong Z, Huang J, et al. (2009)_A Probabilistic Load Flow Calculation Method with Latin
Hypercube Sampling, Automation Of Electric Power Systems, vol.33(21), pp.32-35,81.
[9] Li J, Zhang B, (2001) Probabilistic Load Flow Based on Improved Latin Hypercube Sampling
With Evolutionary Algorithm, Proceedings of the CSEE, vol.31(25), pp.90-96.
[10] Rosenblueth E. (1981) Two-point estimates in probability. Applied Mathematical Modelling,
vol.5, pp.329-335.
[11] Chun-Lien S. (2005) Probabilistic load-flow computation using point estimate method. IEEE
Transactions on Power Systems, vol.20(4), pp.1843-1851.
[12] Harr M E. (1989) Probabilistic estimates for multivariate analysis. Applied Mathematical
Modelling, vol.5(5),pp.313-318.
[13] Morales J M, Perez-Ruiz J. (2007) Point estimate schemes to solve the probabilistic power flow.
IEEE Transactions on Power Systems, vol.22(4), pp.1594-1601.
[14] Hong H P. (1998) An efficient point estimate method for probabilistic analysis. Reliability
Engineering & System Safety, vol.59,pp.261-267.
[15] Allan R N, Leite da Silva A M, Burchett R C. (1981) Evaluation Methods and Accuracy in
Probabilistic Load Flow Solutions. IEEE Transactions on Power Apparatus and Systems,
PAS-100(5),pp.2539-2546.
[16] Wang X, (1988) Optimal Planning of Power System, Hydropower and Water Conservancy Press.
1990.
[17] Wang X. (2013) Stochastic power flow analysis of power system, Journal of Xi'an Jiaotong
University, vol.22(3), pp.87-97.
[18] Allan R N, Al-Shakarchi M R G. (1977)Probabilistic techniques in a.c. load-flow analysis.
Proceedings of the Institution of Electrical Engineers, vol.124(2), pp.154-160.
[19] Tian W D, Sutanto D, Lee Y B, et al. (1989)Cumulant based probabilistic power system
simulation using Laguerre polynomials. IEEE Transactions on Energy Conversion, vol.4(4),
pp.567-574.
[20] Bibby B M, Skovgaard M, Sørensen M. (2005) Diffusion-type models with given marginal
distribution and autocorrelation function. Bernoulli Journal, vol. 11(2), pp.191-220
9
news & views
WIND ENERGY
Taking advantage of modern turbines

Wind facilities have generally deployed turbines of the same power and height in regular uniform arrays. Now, the
modern generation of turbines, with customer-selectable tower heights and larger rotors, can significantly increase
wind energy’s economic potential using less land to generate cheaper electricity.
Robert Thresher
W
ind energy is a rapidly growing the turbine swept area. The engineering
global source of electricity challenges of increases in turbine scale
generation, because it has the had to be overcome while finding ways
environmental advantage of producing low to minimize the cyclic loading on the
carbon electricity at a very low price. The turbine rotor to prevent fatigue damage and
Global Wind Energy Council estimates increased structural failures. The overall
that the total global installed capacity is impact of these design innovations made
about 540 GW, and installed capacity will the wind turbines heavier and costlier on
continue to grow at about 8% per year for a per-turbine basis. However, at the same
the foreseeable future1. Offshore installation time, these innovations have made available
of wind facilities is growing, driven by a series of turbines with customer-selectable
the limited suitable locations onshore and rotor diameter and tower heights that can
by the higher, more persistent offshore be tailored to the specific attributes of an
winds. However, the development of a new installation site, as shown in Fig. 1, greatly
generation of turbines, both higher and reducing the overall cost of energy.
larger than previous generations, makes it Erkka Rinne and colleagues have
feasible to install turbines in lower wind performed a comprehensive wind
areas that were previously considered potential modelling study for Finland that
uneconomic. Now, writing in Nature Energy, demonstrates the significant increase in
Erkka Rinne and colleagues from the VTT generation potential of this new generation
Technical Research Centre in Finland have of larger and taller turbines. For the study,
quantified the increased potential of this the land area of interest was subdivided
modern generation of turbines, based on a into a calculation grid of small cells, and the
resource potential study for Finland2. wind resource for each cell was
The power output of a wind turbine is derived from the Finnish Wind Atlas
proportional to the product of its power (http://www.tuuliatlas.fi/en/). Then, a Fig. 1 | A complex wind farm layout in Lowville,
coefficient, the air density, the rotor swept turbine model with custom tower height New York. The complex layout leads to low
area and the cube of the free-stream wind and rotor diameter was selected to disruption of other land use activities. Credit:
velocity3. This simple relationship provides estimate the energy capture. The study Michael J. Okoniewski, National Renewable
insight into the evolution of the technology. included restrictions and buffer zones, Energy Laboratory (NREL) / US Department
Modern wind turbines have been perfected where turbines were excluded, that of Energy
to the point that power coefficients of depended on the land use. The energy
nearly 0.5 are typical, which is about 85% generation potential was calculated both
of the theoretical maximum. With this high for an older set of turbines from about 2002
conversion efficiency, turbine engineers and for modern turbines from 2015. Two generation of turbines is a significant
and wind farm developers have focused land-use scenarios were considered: the increase in the energy generated per unit
on making the swept area larger and first one was optimistic in that very lenient of land area, even though the larger rotors
increasing the wind velocity that the turbine land use restrictions were applied; and the required a greater separation distance
experiences. The velocity cubed term has the second one used restrictive assumptions between turbines. However, at the high
biggest impact on power production: a 10% regarding land usage. The study also energy wind locations, the older generation
change in wind velocity gives a 30% change accounted for turbine energy losses due to of turbines with shorter towers and smaller
in power. This means that placing the down times (for example from maintenance rotors were still competitive based on the
turbine in even a slightly better wind flow or or blade icing) or to other energy losses, cost of energy but used more land per
making the tower taller to get up higher in such as wake blockage (upstream turbines MWh produced. Considering the diversity
the boundary layer, away from the drag of block the wind to downstream turbines, of wind turbine models now available, the
the ground that slows the wind, can have a reducing energy capture). The cost study shows that it is critical to tailor the
big payoff. In accord with taller towers, there sensitivity to roads, transmission, utility turbine size and height to the site wind
has been a series of design innovations to interconnection, turbine height and other characteristics and terrain features to
make turbine rotor diameters larger, driven variables was also assessed. The results maximize the wind energy potential for
by the diameter-squared relationship for show that the major advantage of the newer a region.
Nature Energy | www.nature.com/natureenergy
© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
news & views
Now, it is possible to update and the landscape with less disruption of other Robert Thresher
improve past wind potential estimates land use activities, as illustrated in Fig. 1. National Renewable Energy Laboratory, Denver, CO,
for land-based wind energy following With additional research, the modelling USA.
the modelling approach used by Rinne approach of Rinne could be enhanced e-mail: Robert.Thresher@nrel.gov
and colleagues. Countries with regions of to include additional features that
relatively modest wind speeds will find would make wind energy facilities Published: xx xx xxxx
that they have significantly more cost even more productive. For example, https://doi.org/10.1038/s41560-018-0168-2
competitive wind potential than they ongoing research shows promise for References
had previously estimated. In addition, mitigating blockage effects of upwind 1. Global Wind Statistics 2017 (The Global Wind Energy Council,
they will be able to generate more energy turbines employing turbine yaw control 2018); http://gwec.net/wp-content/uploads/vip/GWEC_
PRstats2017_EN-003_FINAL.pdf
from a given area of land while reducing to steer upwind wakes away from 2. Rinne, E., Holttinen, H., Kiviluoma, J. & Rissanen, S. Nat. Energy
the number of turbines on the landscape, downwind turbines4, which would https://doi.org/10.1038//s41560-018-0137-9 (2018).
reducing their visibility. The ability to tailor provide an additional increase in 3. Manwell, J., McGowan, J. & Rogers, A. Wind Energy Explained –
Theory Design and Application 2nd edn, Ch. 3 (John Wiley and
the turbine height and rotor size means that energy capture and the wind potential for Sons, Chichester, 2009).
turbines can be more easily introduced into a given area of land. ❐ 4. Fleming, P. et al. Wind Energy 18, 2135–2143 (2015).
© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Analysis
https://doi.org/10.1038/s41560-020-00685-6
Effects of technology complexity on the

emergence and evolution of wind industry
manufacturing locations along global value chains
Kavita Surana 1 ✉, Claudia Doblinger 2, Laura Diaz Anadon3 ✉ and Nathan Hultman 1
Wind energy can contribute to national climate, energy and economic goals by expanding clean energy and supporting econo-
mies through new manufacturing industries. However, the mechanisms for achieving these interlinked goals are not well under-
stood. Here we analyse the wind energy manufacturing global value chain, using a dataset on 389 component supplier firms
(2006–2016) that work with 13 original equipment manufacturers. We assess how technology complexity, that is, the knowl-
edge intensity and difficulty of manufacturing components, shapes the location of suppliers. For countries without existing
wind industries, we find evidence of the emergence of suppliers for only low-complexity components (for example, towers and
generators). For countries with existing wind industries, we find that suppliers’ evolution, that is, changes in their international
supply relationships, is less likely for high-complexity components (for example, blades and gearboxes). Our findings show the
importance of understanding technologies along with firms and countries within global value chains for achieving policy goals.
T
he global market for wind power technology is large and GVC. This gap is present not only for wind energy but also more
growing. Installed wind capacity has grown worldwide from broadly for clean energy technologies and other manufacturing
73 GW in 2006 to 623 GW in 20191. Mid-century projections industries where GVCs are increasingly the subject of policy dis-
expect continued expansion2 for two reasons: first, the increasing cussions on globalization and manufacturing19–21. With limited evi-
ambition for clean energy deployment by the governments of many dence on the firm level, tensions have been prevalent as countries
countries; and second, the cost decreases driven by, among other try to promote or protect domestic manufacturing, especially in
factors, technological advances and manufacturing improvements clean energy industries9.
at the component level3–5—such as in blades, towers, gearboxes Here we examine in depth the manufacturing GVC of the wind
and bearings. energy industry to understand the technological drivers behind
This expanding market for wind power has created co-benefit location of manufacturing. We operationalize this enquiry by
opportunities for policymakers interested in coupling energy and focusing on what we call the technology complexity—that is, the
economic development goals. The ability to develop a domestic combination of design, processes, skills, resources and institutions
manufacturing component supply chain and generate employ- required to manufacture, transport and integrate individual com-
ment is particularly attractive. Examples of governments explicitly ponents (such as towers, blades, gearboxes, control systems and
trying to advance energy, climate and industrial goals in the wind more) into a wind turbine3,22. We analyse the link between technol-
sector include the Offshore Wind Sector Deal (United Kingdom)6 ogy complexity and two key factors: where and why new manufac-
and local content requirement for onshore wind deployment (in turing companies emerge over time; and how existing companies
Brazil, and previously in China)7. Articulating these co-benefits evolve in response to the international changes in the GVC.
for improving domestic energy technology industries has been The emphasis on components is critical for a global analy-
instrumental in the political dialogue on, and public support for, sis of the wind industry because wind turbines are customized
energy policy8–10. engineering-intensive goods where technology innovation and cost
Despite growing research and policy interest in clean energy reductions occur mainly at the individual component level rather
manufacturing and global value chains (GVCs)11–14, there is a lack of than the final product level23,24.
understanding of the global manufacturing patterns of wind energy
technologies (and other clean energy technologies). In the past two The full manufacturing GVC for wind energy
decades, changes in the manufacturing (and deployment) location The focus on components compels an assessment of the full manu-
of a few, large original equipment manufacturers (OEMs)—that is, facturing GVC, comprising around 13 large OEMS and the hun-
lead companies that assemble, and occasionally manufacture, com- dreds of supplier firms that manufacture components for the OEMs.
ponents for wind turbines—have reshaped the global industry, with Thus far, research has focused on public policies and technol-
countries such as China and India catching up to first movers in ogy strategies using the final wind turbine or the OEM as the unit
Europe and the United States15–18. However, there is an absence of of analysis, often examining how OEMs emerged in new countries
comprehensive datasets or analyses to understand these changes at or evolved with changing global markets (for example, refs. 15–17,25).
the industry-specific firm level, that is, comprising both component Hundreds of supplier firms manufacture components for the large
manufacturers and the OEMs that constitute the manufacturing OEMs and play a pivotal but often neglected role in shaping the
Center for Global Sustainability, School of Public Policy, University of Maryland, College Park, MD, USA. 2Campus Straubing for Biotechnology and
1
Sustainability and School of Management, Technical University of Munich, Straubing, Germany. 3Centre for Environment, Energy and Natural Resource
Governance, Department of Land Economy, University of Cambridge, Cambridge, UK. ✉e-mail: ksurana@umd.edu; lda24@cam.ac.uk

Analysis NaTure Energy
Forgings
States (General Electric), Japan (Mitsubishi), and later in China
(for example, Goldwind) and India (Suzlon). We then combined
Tower
this dataset with the technology complexity of components to
Control system Country of assess the emergence and evolution of manufacturing locations of
Type of component
supplier
Power converter Germany
component suppliers.
Denmark
Generator Spain
United States Technology complexity variation in wind turbine
Blade Japan
China
components
Nacelle India Our analysis takes into account the inherent technological differ-
Other
Gearbox ences across the different turbine components, instead of treating
the end product, that is, the wind turbine as a technology black box
Bearing
as in most previous research focused on countries or OEMs (excep-
0 25 50 75 100 125 150 tions include, for example, ref. 24).
Number of suppliers To capture differences in the components, we quantify the vari-
ability in the technology complexity of each of the nine components
Fig. 1 | Diversity in number and geographic spread of suppliers by wind in our dataset (Fig. 2). Complexity can be measured in different
turbine component. The figure shows the total number and country of ways but there is no universal consensus metric or terminology (for
suppliers for each component that were active at least once in the period example, refs. 28–34) We use the Product Complexity Index (PCI)
between 2006 and 2016 in our dataset, including OEMs’ in-house suppliers. developed by Hausmann et al.28 as a measure of technology com-
There are 389 suppliers in our dataset, but because some suppliers plexity. We found that compared with most other approaches to
manufacture multiple components (see Methods), they are listed under measure complexity that focus on knowledge competences, the PCI
each component in this figure. better reflects the broad set of real-world perspectives—including
country economy and contexts, knowledge requirements, manufac-
turing skills, resources and costs—for manufacturing, transporting
industry and the GVC11,13,14. Suppliers are often small and medium and integrating wind components35 (see Methods for comparison
enterprises—the main employers that often constitute the backbone of different complexity metrics, how they compare with insights
of many economies26—who must develop competences or strate- specific to wind turbine technologies and why we chose the PCI).
gies to stay competitive in rapidly changing local and international We continue to refer to technology complexity rather than product
markets. Yet, there is limited evidence on where suppliers emerge complexity in the rest of this paper, because wind turbine technol-
or how they respond to the broader changes in global wind indus- ogy is the end product and includes multiple components or prod-
try markets, with some case-study-based exceptions pointing to the ucts, which in turn comprise other subcomponents or products.
importance of technology characteristics in determining supplier The PCI is based on the hypothesis that more complex tech-
activity11. Given the importance of suppliers in the clean energy nologies with greater knowledge intensity are manufactured and
industry, understanding their behaviour is key to coupling energy, exported by countries that have higher knowledge intensity, and
climate and industry policy goals. that these countries are also able to manufacture and export other
We developed a database of the component suppliers in the high-complexity technologies (that is, with a higher PCI28). We cal-
wind energy technology GVC (see Methods for details). Our data- culate the PCI metric indicating technology complexity by assign-
set builds on industry reports27 and captures data on 389 suppli- ing to each wind turbine component a relevant Harmonized System
ers involved in over 2,000 supplier–OEM market relationships (HS) code(s). We then calculate the average PCI of that component
with 13 major OEMs occurring between 2006 and 2016 for 9 key based on PCI estimates derived from the approach of Hausmann
components identified in industry reports (Fig. 1)27. The OEMs et al. using global trade data on the component level36 (Methods,
are located in Europe (for example, Siemens, Vestas), the United Supplementary Tables 1–3 and Supplementary Fig. 1).
2.5
2.0
1.5
Technology complexity
1.0
0.5
−0.5
−1.0
−1.5
Gearbox Bearing Blade Nacelle Control system Forgings Generator Power converter Tower
Wind turbine component
Fig. 2 | Technology complexity estimates of wind turbine components. Wind turbine components have differences in technology complexity, as estimated
using the PCI method based on Hausmann et al.28. For each component, in the box plot, the thick horizontal line indicates the median and the red dots
indicate the mean from 2006 to 2016 (also see Supplementary Table 2). The bottom line in the box indicates the 25th percentile and the top indicates the
75th percentile. The whiskers indicate the observations that lie within 1.5 times the interquartile range and the black dots indicate outliers.

NaTure Energy Analysis
Table 1 | Regression results on the relationship between technology complexity and evolution
International evolution Controls All European US suppliers Chinese Suppliers to Suppliers to Suppliers
(change in the fraction of suppliers suppliers suppliers European US OEMs to Chinese
supplier relationships with OEMs OEMs
international OEMs)
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8
Technology complexity −0.118** −0.088 −0.281 −0.131* −0.099 −0.266*** −0.014
(0.054) (0.066) (0.239) (0.075) (0.078) (0.089) (0.062)
t = −2.191 t = −1.337 t = −1.175 t = −1.752 t = −1.263 t = −2.993 t = −0.222
P = 0.029 P = 0.182 P = 0.240 P = 0.080 P = 0.207 P = 0.003 P = 0.825
Wind specialization −1.141*** −1.105*** 0.307 0.011 −1.943*** 0.291* −12.733*** −0.381
(0.390) (0.390) (0.368) (0.299) (0.649) (0.173) (4.602) (1.550)
t = −2.927 t = −2.836 t = 0.834 t = 0.036 t = −2.995 t = 1.682 t = −2.767 t = −0.246
P = 0.004 P = 0.005 P = 0.405 P = 0.972 P = 0.003 P = 0.093 P = 0.006 P = 0.806
Component diversification 0.136** 0.107 −0.055 0.449*** 0.309*** −0.105 0.04 0.088
(0.069) (0.068) (0.081) (0.148) (0.119) (0.097) (0.120) (0.093)
t = 1.984 t = 1.572 t = −0.676 t = 3.027 t = 2.588 t = −1.079 t = 0.337 t = 0.946
P = 0.048 P = 0.116 P = 0.500 P = 0.003 P = 0.010 P = 0.281 P = 0.737 P = 0.345
Patenting international 0.003 0.003 −0.0004 0.082** 0.627** 0.001 0.007** 0.007***
(0.003) (0.003) (0.002) (0.041) (0.280) (0.004) (0.003) (0.002)
t = 1.101 t = 1.132 t = −0.148 t = 2.011 t = 2.241 t = 0.267 t = 2.200 t = 2.813
P = 0.271 P = 0.258 P = 0.883 P = 0.045 P = 0.026 P = 0.790 P = 0.028 P = 0.005
Patenting home −0.001 0.003 −0.001 −0.132 0.045*** 0.009 −0.0002 0.001
(0.005) (0.005) (0.005) (0.244) (0.014) (0.011) (0.016) (0.009)
t = −0.116 t = 0.505 t = −0.166 t = −0.539 t = 3.091 t = 0.874 t = −0.015 t = 0.142
P = 0.908 P = 0.614 P = 0.868 P = 0.590 P = 0.002 P = 0.383 P = 0.989 P = 0.888
Size −0.547** -0.516** 0.156** −0.396 −1.977*** 0.064 10.710*** −0.343
(0.216) (0.214) (0.078) (0.281) (0.635) (0.110) (3.303) (1.342)
t = −2.534 t = −2.408 t = 2.016 t = −1.409 t = −3.111 t = 0.578 t = 3.243 t = −0.255
P = 0.012 P = 0.017 P = 0.044 P = 0.159 P = 0.002 P = 0.564 P = 0.002 P = 0.799
Age 0.008*** 0.008*** 0.002 −0.057*** −0.031 0.001 −0.680*** −0.01
(0.002) (0.002) (0.001) (0.011) (0.023) (0.001) (0.228) (0.033)
t = 3.736 t = 3.697 t = 1.079 t = −4.958 t = −1.367 t = 1.053 t = −2.989 t = −0.293
P = 0.0002 P = 0.0003 P = 0.281 P = 0.00000 P = 0.172 P = 0.293 P = 0.003 P = 0.770
Supplier dependence on OEM −0.018 −0.016 0.127 −0.092 −0.075 −0.09 −0.108
(0.067) (0.067) (0.077) (0.130) (0.072) (0.083) (0.083)
t = −0.266 t = −0.240 t = 1.640 t = −0.708 t = −1.034 t = −1.076 t = −1.300
P = 0.791 P = 0.810 P = 0.102 P = 0.479 P = 0.302 P = 0.283 P = 0.194
Constant 4.703*** 4.551*** −0.940** 4.457*** 14.883*** −0.046 −50.519*** 3.054
(1.604) (1.595) (0.447) (1.691) (4.896) (0.635) (15.602) (10.390)
t = 2.931 t = 2.853 t = −2.103 t = 2.637 t = 3.039 t = −0.072 t = −3.238 t = 0.294
P = 0.004 P = 0.005 P = 0.036 P = 0.009 P = 0.003 P = 0.943 P = 0.002 P = 0.769
Country FE Yes Yes No No No Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1,227 1,227 479 169 455 637 421 485
R2 0.426 0.429 0.456 0.548 0.487 0.457 0.449 0.509
Adjusted R2 0.288 0.291 0.319 0.397 0.338 0.307 0.297 0.362
Suppliers of high-complexity components are likely to have low evolution, that is, change in fraction of relationships with international OEMs. The model results are from OLS regressions. Numbers in
parentheses indicate robust standard errors. *P < 0.1; **P < 0.05; ***P < 0.01. FE, fixed effects.
Under our assessment, blades and gearboxes are among the most solar photovoltaic cells have a relatively high PCI of 0.89, while
complex technologies (PCI > 1), while towers are among the least biofuels have low complexity with a PCI of −1.1 (refs. 37,38). Our
complex (PCI < 0). For reference, using a similar methodology, findings on the relatively high complexity of blades are consistent

Complexity
Year: 2006
Tower (n = 17) Generator (n = 18) Control system (n = 14) Blade (n = 25) Gearbox (n = 26)
Germany
Germany Germany
Germany Germany
Germany Germany
Germany
Germany
Germany
Denmark Denmark
Denmark
Denmark Denmark Denmark
Denmark Spain
Denmark Spain
Denmark Spain United
Spain Spain
States
Spain Japan
Spain United Japan
Spain States
Spain
Austria Spain United
Japan States
United States United States Belgium United States
United States Japan
France
Japan India
Japan Japan
Japan Japan India Finland Japan
Switzerland Netherlands
India India
India India Chile India India
Chile Brazil Switzerland
Year: 2016
Tower (n = 275) Generator (n = 94) Control system (n = 104) Blade (n = 78) Gearbox (n = 59)
Germany
Germany
Germany Germany Germany
Denmark Germany
Spain Denmark Germany
United States Spain
Denmark Denmark Denmark
United States Germany Denmark
China Japan Germany Spain Spain
Spain Spain
Denmark Spain
India China Denmark United States United
Finland Germany United States
Japan Spain United States
France
Spain States Japan United States
Italy India
Denmark United
Netherlands Austria China Japan
United States States
Portugal
Japan
United Kingdom Spain Finland Japan China
India China
Canada Japan
France Austria
Brazil United States China China
Norway Belgium
South Korea Italy China
Japan China
Indonesia Switzerland India
Australia China Finland
Sweden
Turkey United Kingdom India
India India India Brazil India
Egypt Brazil Ireland Czech rep.
Supplier country OEM country International supplier–OEM relationship Domestic supplier–OEM relationship
Fig. 3 | Change in international supplier–OEM relationships between 2006 and 2016 with increasing technology complexity. The figure shows the
country (in the solid black rectangle) of the component suppliers that sold components to OEMs from specific countries (denoted by the white rectangles).
The green bands denote an international relationship between component suppliers and OEMs (that is, supplier and OEM from different countries) and
the grey bands denote a domestic relationship between suppliers and OEMs (that is, supplier and OEM from the same country). n represents the number
of relationships in the dataset. The number of countries involved in the manufacturing of low-complexity components increased substantially between
2006 and 2016. This was not the case for high-complexity components. Low-complexity components such as towers and generators experienced a greater
diversification (or number) of supplier locations and more international relationships compared with high-complexity components such as blades and
gearboxes.
with the intensive requirements of blade manufacturing that countries that also had the largest wind deployment markets in the
require high-technology equipment, more time and advanced skills. study period (that is, Germany, Denmark, Spain, the United States,
Similarly, our findings on the low complexity of towers are consis- China, India and Japan).
tent with research that indicates that tower manufacturing involves Second, the emergence of suppliers in new locations, especially in
more standard industrial processes24,27 (Supplementary Table 1). countries without an OEM, relates to the technology complexity of
the components. Although new countries became part of the wind
The emergence of new component suppliers GVC over time, the extent of this emergence and the consequent
Our analysis quantifies the relationships between wind component global diversification of the GVC was inversely linked to the com-
suppliers and their large OEM partners in the GVC between 2006 plexity of the component (Figs. 3 and 4). For low-complexity compo-
and 2016. These interactions highlight three findings on the charac- nents (that is, towers and generators), suppliers from new locations
teristics of the GVC and the emergence of wind component suppli- in developing economies emerged (including countries in Africa,
ers within our study period. Latin America and Asia–Pacific regions). For high-complexity
First, OEMs and suppliers were dispersed globally in 34 coun- components (that is, blades and gearboxes), the emergence of new
tries, but their relationships remained largely domestic, albeit with supplier countries was much lower, potentially because more com-
some exceptions discussed below. In our study period, 78% of sup- plex products required suppliers with skilled manufacturing, higher
pliers (305 out of 389) were in countries that had a large OEM and absorptive capacity and tacit knowledge that may be more difficult
58% of relationships between OEMs and suppliers (1,239 out of for suppliers originating in developing and emerging economies39.
2,121) were domestic, that is, involving suppliers and OEMs from Towers are large and their shipping costs are high, incentivizing
the same country (see, for example, Fig. 3). Our analysis, which manufacturing closer to demand, but such incentives may also be
starts in 2006, suggests that a domestic manufacturing supply chain present for labour-intensive components such as blades40. Our find-
initially developed in countries with large OEMs, which were the ing on the greater emergence of low-complexity towers rather than

Low complexity,
of blades in most countries indicate the importance of transport as high number of supplier countries Number of
one of many factors, along with knowledge and skills28, that shape 20 supplier firms
Number of supplier countries

Tower
costs and decisions in the location of manufacturing. Generator 25
Third, a larger fraction of high-complexity component suppliers
15 Forgings
interacted exclusively with OEMs from their own country or with 50
OEMs from other industrialized countries. For example, we found Power converter Control system
75
that German OEMs primarily sourced blades from other German 10 Blade
Gearbox
suppliers (or had subsidiaries or in-house production in Germany) Nacelle 100
Bearing
or from suppliers in other industrialized countries (for example, 5 High complexity, 125
Denmark, United States). This implies that higher-complexity low number of supplier countries
components that require more skills and expertise were probably
manufactured by only a few specialized suppliers in industrialized 0
−0.50 −0.25 0 0.25 0.50 0.75 1.00 1.25 1.50 1.75
countries (Supplementary Fig. 2). The emergence of a diverse and
Average complexity
large number of countries with component suppliers for towers
(a low-complexity component) contrasts with the fewer specialized
countries with suppliers working on gearboxes (a high-complexity Fig. 4 | Relationship between the number of supplier countries of each
component) (Fig. 3). component and the average complexity of the component. The dot
Overall, our analysis implies that, for most countries (and in size indicates the number of firms for each component in our dataset.
particular developing countries that face institutional, financial and Low-complexity components experienced emergence of suppliers
operational risks and uncertainties17), the emergence of suppliers (including OEMs’ in-house suppliers).
manufacturing high-complexity components with higher value add
may be a challenging endeavour without active policy interventions, size of the effect in our sample of components, a one-unit increase
which we discuss later in this paper. in technology complexity is the difference in complexity measured
using the PCI separating a low-complexity component such as tow-
International evolution of suppliers ers (−0.24) from a higher-complexity component such as control
The globalization of the wind energy industry was evident in the systems (0.69) or blades (1.00) (Supplementary Table 2).
shift of the initial leadership of Europe and the United States in In addition, the international evolution of supplier firms may be
deployment and OEMs (in 2006) to increasing deployment and new associated with their own country or with the OEMs that they work
OEMs in China and India (by 2016) with large and growing demand with (for example, differences in countries’ incentives for manufac-
in those countries15–18,41. turing or the OEMs’ strategy)11,49,50. We developed two separate sets
The changes in the broader industry affected the traditional or of models distinguishing our results based on the origin country
existing suppliers in countries with OEMs, as these suppliers faced of suppliers (Fig. 5a, models 3–5) and the countries of their target
increasing competition from new markets (and new suppliers). OEMs (in Europe, the United States and China; Fig. 5b, models
These existing suppliers had opportunities to work with both OEMs 6–8). We note that while in many cases target OEMs are associated
from the suppliers’ countries (domestic or local OEMs) and those with the deployment markets in the countries of those OEMs, such
from other countries (international OEMs). However, the most assumptions may not always be true as, for example, several interna-
strategic and competitive suppliers probably delivered components tional OEMs were also present in India and China17.
to international OEMs and increased such international relation- We found that the additional and statistically significant
ships over time42–44, in what we refer to as evolution. We estimate firm-level predictors of international evolution, that is, low spe-
this evolution by calculating the change (over a two-year time lag) cialization in wind and smaller size (in model 1, Table 1), seem
in the fraction of each supplier’s market or contractual relationships to be primarily driven by Chinese suppliers (see model 5). This is
with international OEMs (Methods). potentially because larger state-owned firms may have established
We assessed the relationship between technology complexity relationships with Chinese OEMs17 whereas smaller firms in China
and evolution with a detailed statistical analyses using ordinary may work more with international OEMs. In addition, with the
least squares (OLS) regressions (model 1 and model 2, see Methods, large manufacturing base in China, many firms may not special-
Table 1 and Supplementary Table 4), where we controlled for vari- ize in wind energy but rather manufacture components that have
ous factors that may affect evolution such as firm characteristics and applications in multiple industries (for example, generators, power
firm strategic decisions43,45–48. These characteristics include wind converters) and supply these to international OEMs.
specialization (activities only in wind and not in any other sectors), Even after accounting for the OEM- or country-related factors
component diversification (supply of multiple wind components), that could affect suppliers’ behaviour (as seen in the automotive
age (number of years since company founding), size (number of sector, for example50), we find continued evidence of the negative
employees) and knowledge stock (measured through international relationship between technology complexity and international evo-
and home-country patents). We also controlled for the governance lution. Distinguishing the origin countries of suppliers, we find that
of the GVC49—that is, whether suppliers supply to individual OEMs high technology complexity decreases the likelihood of interna-
or have been acquired by them (for example, ‘captive’ suppliers or tional evolution by 13% for suppliers from China, 9% for suppliers
those that are part of vertically integrated OEMs) or whether they from Europe (that is, Germany, Denmark and Spain) and 28% for
supply to multiple OEMs in a more competitive market by estimat- suppliers from the United States (the latter two are not significant)
ing the supplier dependence on OEMs through in-house or out- (Fig. 5a and Table 1, models 3–5). Distinguishing the OEM coun-
sources relationships. In addition, we use fixed effects to account tries that suppliers have market relationships with, we find that high
for any firm-, country- and time-specific features (see Methods for technology complexity decreases the likelihood of international
details on the variables). evolution by up to 27% when suppliers work with OEMs from dif-
The OLS regression analysis demonstrates that, as technology ferent countries, most notably the United States (Fig. 5b and Table 1,
complexity increases by one unit, the likelihood of international models 6–8). Our results are robust across different model specifi-
evolution (that is, increase in fraction of relationships with inter- cations and additional robustness checks, such as time lags of one
national OEMs) decreases by 12%, even after controlling for other and three years and using different complexity metrics (Methods
important characteristics (model 1 in Table 1). To give a sense of the and Supplementary Tables 5 and 6).

a b
Technology Technology
complexity complexity
Component Component
diversification diversification
Independent and control variables

Independent and control variables
International International
patents patents
Domestic Domestic
patents patents
Firm size Firm size
Firm age Firm age
Supplier Supplier
dependence dependence
on OEM on OEM
−0.4 −0.2 0 0.2 0.4 −0.4 −0.2 0 0.2 0.4

Model coefficients Model coefficients
Location of suppliers Location of OEMs that work with suppliers (target market)
All (model 2, n = 1,221) Europe (model 3, n = 479) All (model 2, n = 1,227) Europe (model 6, n = 637)
United States (model 4, n = 169) China (model 5, n = 455) United States (model 7, n = 421) China (model 8, n = 485)
Fig. 5 | Coefficient plots showing the relationship between international evolution, technology complexity and other control variables. The figure shows
results from OLS regressions, where the size of the regression coefficients is represented as dots and standard errors as bars. a, Models with relationships
grouped by location of the suppliers. b, Models with relationships grouped by location of the OEMs that suppliers work with (that is, the primary target
market). Some of the larger coefficients of the OEM or country-related factors are not depicted due to their large values but are shown in Table 1.
Together, our quantitative findings on the evolution of suppli- enhanced by policy efforts that target both emergence of new sup-
ers show that international competitiveness (proxied with increase pliers and, eventually, the evolution of existing ones (Fig. 6).
in suppliers’ relationships with international OEMs) increased For example, in China, a local content requirement policy that
for low-complexity components for suppliers from all countries. started in 2003 mandated domestic manufacturing of some com-
Overall, this is consistent with the findings on the emergence of ponents until 2009 to make them eligible for deployment incen-
manufacturing. tives17. Partly to meet this requirement, high-complexity blade
manufacturing began with the Danish OEM Vestas establishing a
Implications for wind component technology new manufacturing location in China. A large number of domes-
manufacturing tic suppliers emerged following the Renewable Energy Law of 2006
Our analysis provides a comprehensive view of the wind energy that supported rapid, large-scale wind power deployment while
manufacturing GVC, with central emphasis on the technological parallel policies supported the domestic development of larger tur-
characteristics of components and suppliers—as opposed to just the bines17. With growing demand and because of the presence of other
turbines and OEMs covered by previous research. industries with relevant transferable knowledge and skills, our data-
As countries expand wind turbine manufacturing and domestic set shows that the manufacturing of high-complexity components
supply chains for both onshore and offshore wind, our findings sug- such as gearboxes quickly emerged, led by the China High Speed
gest that governments and private firms would benefit from devel- Transmission Equipment Group Company that supplied to both
oping targeted, technology-specific approaches to participate in the Chinese and international OEMs from 2008.
wind energy manufacturing GVC. This requires designing policies India provides a second example. A sizeable domestic market
that consider the technology complexity of individual components was already in place in 2006, along with some incentives for manu-
and the domestic capabilities of the country rather than simply the facturing, leading to the emergence of several domestic component
end product (that is, the turbine). In turn, it means tailoring local suppliers for low-complexity components17. Higher-complexity
industry support, skills development and national policies to the components such as blades were manufactured in 2006 through
specific characteristics of component technologies18,51. Suzlon, a large Indian OEM, rather than through international
To support expanded wind manufacturing in develop- suppliers. Although overall only a few high-complexity domes-
ing countries and newly industrialized economies, we find that tic component manufacturers emerged in India, the existing
low-complexity towers are a promising entry point. Even in larger low-complexity base coupled with policies attracted the emergence
market countries such as China and India, the majority of domes- of manufacturing for higher-complexity gearboxes in 2010 through
tic suppliers that initially emerged manufactured low-complexity subsidiaries of European suppliers.
components (Fig. 3). Of course, many countries would like to sup- A third example is from Brazil, which developed higher-
port industries that upgrade beyond low complexity. To this end, complexity component manufacturing even without a large domes-
over the decade we studied, we also find evidence that a base of tic OEM. Brazil’s domestic manufacturing expanded potentially
lower-complexity technologies may provide a gateway to upgrade to because of its large market size, a local content requirement policy,
more complex technologies. This programme of ‘catching up’ can be and existing industry strengths as exemplified by companies such

1.75
Gearbox
Country of
in the long term given that many other countries are already able
to successfully produce low-complexity towers at competitive costs.
Maximum complexity in a country
1.50 supplier
1.25 Bearing Brazil The high labour costs in the United States mean that high tariffs
1.00
China may help US producers of low-complexity components only for the
Blade Egypt domestic market but are unlikely to be helpful in expanding the
0.75 India reach of US manufacturing to sell such components internation-
0.50 Indonesia ally. Instead, a more effective, long-term strategy may be to support
0.25 Forgings Mexico domestic innovation and industry in more complex components as
Turkey
0 the lead time for other countries to enter the competition can be
Vietnam
−0.25
longer and may require more systematic efforts on their part as well.
Tower
−0.50
2006 2008 2010 2012 2014
Implications for GVCs
The evidence provided in this paper on how technology complex-
Year
ity shapes the emergence and evolution of the full manufacturing
Fig. 6 | The highest complexity of wind turbine components in a country
value chain (that is, both suppliers and OEMs) is valuable for under-
in a given year. The figure is based on data on a country’s suppliers and
standing the interactions of domestic energy and industrial policies.
the available data on any international subsidiaries. The data points are
It specifically underscores the importance of supporting an initial
staggered around the year they represent to allow visualization of multiple
base of manufacturing, usually through a low-complexity manufac-
countries for each component in a year. Developing countries and emerging
turing entry point in latemover countries, to provide a gateway for
economies have been able to manufacture more complex components over
upgrading to higher-complexity manufacturing in conjunction with
time. We cannot rule out the possibility that China, India and Brazil—all
carefully scoped policies.
appearing in the top part of the graph as manufacturing higher-complexity
As countries try to develop clean energy industries and meet
components—started manufacturing lower-complexity components well
climate and energy goals, it has become increasingly evident
before 2006, as shown by the current state of manufacturing complexity in
that effective and lasting policies will depend on simultaneously
wind in Indonesia, Vietnam and Egypt.
addressing economic development goals, including manufactur-
ing8,10. By including technology and GVC perspectives in clean
energy policy design, countries can take the opportunity to develop
as Tecsis, a blade manufacturer that emerged as a spin-off from the clean energy industries that will probably expand both manufac-
existing aviation industry52. turing and deployment over time. From this perspective, our work
Within countries, such opportunities for emergence and upgrad- on wind turbines can be extended to other similar clean energy
ing in the GVC have been possible over time through an overlapping industries that require high design capabilities for innovation but
system of domestic and international clean energy policies that spur relatively low manufacturing capabilities51 and involve ‘lumpy’
market demand, incentivize domestic manufacturing, and catalyse investments55. Such technologies include geothermal, concentrated
existing industrial and knowledge bases or support new skills. In solar, large hydropower stations, offshore wind, grid infrastruc-
addition, from the examples of China and India, we also note that tures, electric vehicles and large buildings (as consumers of energy
the emergence of manufacturing in more complex technologies may technologies)51,55.
also be enabled through subsidiaries of suppliers from other coun- Our findings also underscore the central role of component
tries who come in and exploit potential business opportunities in a technology characteristics at the supplier level—in addition to firms
large market16,17. Although we had limited data on suppliers’ foreign and countries—in understanding GVCs. We found that technology
subsidiaries to be included in our quantitative, statistical analysis, complexity shapes both the emergence and evolution of suppliers
we found multiple examples where such approaches were used and the location of manufacturing, even as industries develop glob-
(these are included in Fig. 6). ally over time. To incentivize the development of new manufacturing
Our emphasis on ‘domestically owned companies’ is neverthe- opportunities in the clean energy industry or upgrading along the
less valuable. Foreign firms with local manufacturing facilities may GVC, our findings imply that policies should have a targeted focus
provide employment and tax revenues but not necessarily the same on manufacturing that considers existing local industrial strengths
level of know-how, intellectual property or support for local technol- and suppliers, GVC dynamics and the technology complexity of
ogy transfer40. In contrast, as domestic firms develop know-how and components. Without such an integrated approach, countries may
can meet standard manufacturing requirements (Supplementary need to temper expectations for moving from lower-complexity to
Table 1), even with the help of international firms, they can eventu- higher-complexity components.
ally get access to international markets, for example, observed in We note three needs for future research that also address some of
blade manufacturing in China53. the limitations of our work. First, future work needs to remedy the
For wind manufacturing in more industrialized countries, we absence of detailed industry datasets. Such datasets should capture
observed a larger emergence of high-complexity component manu- granularity on the full location of the GVC, over an extended set of
facturing and lower evolution in high-complexity component sup- components and a longer period of time. This includes a global net-
ply, even as emerging economies suppliers emerged and evolved. work of multinational companies and their subsidiaries, small busi-
Our finding that suppliers are more likely to work with international nesses and downstream firms and the quantity of supply between
OEMs for low-complexity components suggests that the continued different firms and of different components. Our own approach was
domestically owned manufacturing of such components would limited in using the location of component suppliers rather than
need additional policy incentives to be competitive in international the location of manufacturing (for example, supplier subsidiaries in
markets. This means that for countries trying to retain existing other countries) and lacked details on supply quantities because of
manufacturing in low-complexity components (for example, some limited data availability. Second, more mixed-methods research is
countries in Europe or the United States) through the evolution of needed to understand the relationships between technology com-
existing firms, policies would need to be targeted towards specific plexity, governance of GVCs and upgrading of supplier firms in dif-
technologies or components. While protectionist policies such as the ferent country contexts, especially for developing countries. Third,
United States considering imposing trade tariffs on tower imports given that location of manufacturing may be influenced by technol-
are one such near-term approach54, they may not be effective ogy complexity, but can also affect technology innovation, future

research needs to analyse the direction of research and development The patents were then classified by the component they most closely relate to
and technology transfer in the GVC and its implications for devel- using a machine learning approach in R (version 3.6.2), as described in the following.
The patent information was prepared for text-based analysis using the text mining
oping countries (see, for example, ref. 53). package tm57 for pre-processing of the text corpus in the title and description text
Finally, GVC research and policy need to be specifically developed (for example, by removing redundant words in patent language such as ‘section’
for knowledge-intensive clean energy industries. Evidence-based or ‘description’, which are likely to be present in most patents, but do not add any
insights that capture technology, along with supplier firm and coun- notable meaning to the technical content of the invention). We then used probabilistic
topic modelling with latent Dirichlet allocation in the R topicmodels package58. The
try characteristics within, are needed to inform policy design that
topicmodels package allowed us to differentiate the technological focus of innovation
couples energy, climate and economic development goals. in the patent as we generated 26 topics or categories of patents by clubbing together
those with similar word occurrences59,60. We first estimated a probability for patents to
Methods be related to each topic. We then mapped each of the 26 topics to the 9 components
Wind supplier database development. We developed an original global database (and an additional category ‘other’) to identify which component a patent most
of component suppliers to major OEMs for wind turbines. The database was closely links to. Our results were robust to changes in the number of topics.
manually developed by analysing, in detail, text-based industry reports on the
wind GVC and tabulating relevant information at the firm level27. We obtained Technology complexity of components. Researchers have developed multiple
time series data using biennial reports from Navigant Consulting (2006, 2008, approaches to quantify technology complexity (examples in refs. 28,31,33,34). Many of
2010, 2012 and 2014), with each relationship reported for a three-year horizon— these approaches are based on the concepts of knowledge diversity and technology
for example, the 2014 industry report identified supplier–OEM relationships interfaces and few approaches take into account the skills, capabilities or costs
from 2014 to 2016. In this step, we tabulated information on all major component associated with manufacturing (for example, comparing the production of bulky
suppliers (active between 2006 and 2014), the OEMs they supply to (and are and heavy components such as blades or towers with gearboxes).
expected to supply to until 2016), the outsourcing strategies of the OEM firms Wind energy industry reports27 suggest that gearboxes and blades are likely to
(either in-house development of components or outsourced to external supplier) have high complexity while towers are the least complex (Supplementary Table 1).
and the geographical location of the supplier firms. These industry perspectives have also been reflected in empirical literature on
Our dataset captures nearly a decade of rapid advances and international wind turbine components’ design hierarchy35.
changes in wind energy manufacturing and deployment (for example, refs. 16,17,42)— As there is no single or consensus metric in the literature that uniquely captures
however, it does not capture the emergence of suppliers before 2006 in the technology complexity, we tested three approaches to identify a quantitative
formative stages of the wind energy industry in countries worldwide (for example, metric that would most closely match the real-world challenges of designing,
ref. 25). It also does not capture more recent advances—such as the merger between manufacturing, integrating and transporting each of the nine wind turbine
two large OEMs, Siemens and Gamesa in 2016—or new technological challenges components analysed in this paper and their complexity over time (2006 to 2016).
related to grid integration and storage that suppliers and OEMs now work on56. First, we used the PCI developed by Hausmann et al.28,36. The PCI quantifies
Nonetheless, our dataset also includes part of the period before onshore wind the knowledge intensity of a technology by considering the knowledge intensity of
was highly commoditized and is relevant for many other clean energy industries its exporting countries (thus also capturing countries’ economic and institutional
that are still at a formative stage, trying to establish domestic suppliers and to contexts). We estimate the products or technologies associated with wind
participate in GVCs. components by mapping each wind component with the HS code that they are
After an initial cleaning of this dataset and excluding missing or incomplete globally exported under and averaging the reported PCIs in the database across all
data points, we had information on 389 suppliers and 9 components (that is, our mapped codes for each year. As components may be exported under different
towers, blades, nacelle, gearboxes, generators, control systems, power converters codes, we compiled these codes from literature and from a deeper review of
including transformers, bearings and forgings) including information on which of code descriptions that were verified by two technical experts (see Supplementary
the 13 OEMs the suppliers worked with for in-house or outsourced manufacturing. Table 2 for the mapping of component codes)38. Technologies with higher
All analyses in this study were conducted on this dataset. complexity are manufactured in (and exported by) fewer countries with diversified
The OEMs were firms with the greatest global market shares between 2006 and manufacturing and reflect higher levels of skills and knowledge. Conversely,
2016 and were based in Germany (Siemens, Nordex, Enercon, REPower/Senvion), technologies with lower complexity are manufactured in and exported by a larger
Denmark (Vestas), Spain (Gamesa), the United States (General Electric), Japan number of countries that may not necessarily be diversified in their manufacturing
(Mitsubishi), China (Goldwind, Mingyang, Dongfang, United Power) and India capabilities. This metric also captures the fact that while some technologies with
(Suzlon). In addition, in some cases, suppliers also had multiple subsidiaries with higher complexity may be bulkier and have higher transportation costs resulting in
manufacturing locations outside of their home country—for example, ABB from more countries tempted to manufacture them locally, they would still require the
Switzerland manufactured in the United States and Rothe Erde from Germany domestic skills for manufacturing28. We estimated the PCI from the 2002 HS trade
manufactured in India, France, China, the United Kingdom and others—but classification (HS02) as well as the 2007 HS trade classification (HS07). HS07 values
a complete dataset on such additional subsidiaries or locations is not publicly were available in and after 2008 (we assumed HS02 numbers for 2006 and 2007).
available or verifiable and was not used for this assessment. Overall, the suppliers Second, we use an approach developed by Fleming and Sorenson34 that
represent a global distribution of firms from major countries home to OEMs as quantifies technology as a complex adaptive system. This metric is based on
well as others that are trying to develop domestic wind manufacturing capabilities the interdependence of technologies and modularity of interfaces as assessed
in components and/or OEMs (for example, France, UK). by international patent classification codes. We use the simple interpretation
applied by Broekel31, which evaluates the ratio of patent subclass co-occurrences
Database expansion. We obtained additional data on each supplier firm from (ten-digit international patent classification codes) of patents in a given year (with
additional datasets and company website searches (Bloomberg, Orbis, Amadeus) a three-year moving average) to the cumulative patent subclass co-occurrences in
on firm size, founding year and specialization—that is, whether the firm supplies all previous years (starting from 1994). To find the complexity of each of the nine
to industries beyond the wind industry or whether the firm supplies multiple components, we averaged these ratios over all patents of each component.
components. Wind companies experienced multiple mergers and acquisitions in the Third, structural diversity is a metric developed by Broekel31, inspired by the
timeframe of our study (for example, Suzlon, REPower and Senvion) and following notion that technologies are combinatorial networks of technology and knowledge.
previous research we considered them as individually operating companies if they This complexity metric intends to capture the diversity of a technology’s
were not integrated and continued to operate under a different brand. subnetworks, captured through patents. We apply a simplified interpretation
We also estimated the knowledge stock, that is, previous research and of Broekel’s structural diversity approach. We use the probability of patents
development activity of each firm for each component domestically (that is, in its association with each component (as explained earlier) and assume that this
home country) and internationally using patent information. We first searched for probability reflects technology design and knowledge, in that it captures when
wind technology patents for each supplier firm (that is, where the supplier was an components are closely related to other components in a patent description by
assignee on the patent) based on a detailed and previously tested keyword search assigning a probability to each component. We estimated the subnetworks of each
of the patent text and its Cooperative Patent Classification24,35 from the Derwent component in a year by extracting all the patents for a given component in that
World Patent Index database. We extracted patent information (for example, year (with a three-year moving average). In this component subnetwork, we use
title, abstract including translated abstracts, technology classification, priority social network analysis (weighted degree centrality) to estimate the co-occurrence
country where patent was first filed and date of application) on each of the firms. of each component pair (where components are nodes and their co-occurrences
Our search methodology limits patent results to wind energy technologies and are edges) weighted by the intensity of association between the component pair61.
components and minimizes influences from those suppliers and OEMs that are The edge-weight is the product of the probability of each component in a patent,
involved in multiple industries (for example, large conglomerates such as Siemens relative to the maximum probability of any component in that patent. To find
and General Electric). Although our approach may not yield patenting activity the complexity of each of the nine components across the patent dataset, we
in components that are not unique to wind energy, we expect our approach to be averaged the degree of all components and divided it by the total patents for each
thorough as our analysis emphasizes on the content of the patent in its linkages to component to account for the differences in the number of patents. We used the
wind-specific research and development. igraph62 package in R for the social network analysis.

In comparing these metrics, we found that the PCI approach of Hausmann et al. component, 23 (7.5%) offered two and 5 (1.6%) firms offered three compo-
best captures actual challenges of manufacturing and integrating wind components nents. Those 5 firms offered 8 of the 9 distinct components, so there is no bias
as reflected by technology roadmaps and the broader literature on wind power in a certain direction.
technologies27,35 (Supplementary Fig. 1 and Supplementary Data 1). A correlation • Patenting international (x4), which captures the cumulative number of
analysis of these metrics across our study period (Supplementary Table 3) reveals that international patents per component by each supplier, depreciated by 15%
all of them are positively correlated; however, the PCI-based metric has the strongest annually41. We used the patent data and classification as described above and
correlation with international evolution (which we describe in the following). mapped whether the country where each patent was first registered matches
Although our interpretation of Fleming and Sorenson’s approach demonstrates the country of origin of the supplier.
similar trends to the PCI approach, it assigns a slightly higher complexity to towers • Patenting domestic (x5) similarly captures the cumulative number of
that contrasts with the insights from the literature on wind turbine manufacturing. home-country patents applied for by each supplier.
Our interpretation of Broekel’s structural diversity index differed from the • Size (x6), which estimates the number of employees (logged). This informa-
understanding of complexity reflected by the specific academic literature on tion was obtained during the database expansion from Orbis, Amadeus,
knowledge transfer and manufacturing in wind. This could be because of differences Bloomberg and the suppliers’ webpages. We used the last available number of
specific to the wind sector and/or because of limitations in our simplified approach full-time employees (or equivalents) given that many of the covered suppliers
for estimating the index. For these reasons, we used the PCI-based approach as the are private firms where time-varying data are not available.
main measure of technology complexity in our analysis. Our primary results report • Age (x7), which represents the time interval since the founding year of the sup-
HS02 values as these were reported for each year from 2006 to 2016. plier. This information was also obtained during the database expansion.
We note that the PCI has two main caveats. One, the PCI relies on international • Supplier dependence on OEM (x8), which captures the different outsourcing or
trade (and export) data and may not fully capture what is produced for local use— insourcing strategies applied by the OEMs and indicates how dependent each
but it is likely that countries only export what they are good at producing, for both supplier is on the OEM. The importance of including this variable as a control
domestic and international use28. Two, resulting from the dependence on trade stems from the fact that OEMs have different approaches for procuring com-
flows, the PCI values for individual components may see variations over time. ponents from suppliers, that is, the governance of the value chain: some sup-
However, we found that data on the different complexity metrics was correlated pliers are in-house or through acquired companies, and some are outsourced
and the PCI was still the best suited for our study. For the purposes of our research, to international suppliers who, despite being part of an OEM, continued to
the PCI provides a suitable estimate of manufacturing wind turbine components, brand their products differently. This is a continuous variable ranging from 0
and of how technology characteristics that capture more than technology or (only in-house relationships) to 1 (only outsourced relationships).
knowledge competences determine the location of manufacturing. In the regression results, the change in international evolution (Yi) for supplier
i is estimated using the following OLS model:
Mixed-methods analysis of emergence of suppliers. We used network
analysis techniques to visualize the relationships between OEM and component ðY i Þ ¼ β0 þ β1 x1i þ β2 x2i þ β3 x3i þ β4 x4i þ β5 x5i
supplier firms over different reported time periods (that is, 2006 and 2014). The ð1Þ
þβ6 x6i þ β7 x7i þ β8 x8i þ φi þ γ i þ t
networks-based approach is increasingly used to visualize and quantify GVCs as
scholars recognize that GVCs are better represented by multidimensional networks where β1 is the coefficient of interest on technology complexity and β2–8 the
rather than linear chains63. We use the term ‘relationships’ to describe interfirm coefficients of the control variables. φi, γi and t are fixed effects for the supplier (φi),
linkages (for example, Vestas (OEM) with Titan Wind (supplier) for towers in the country of the supplier (γi) and year (t).
2014) and intrafirm linkages (for example, Vestas (OEM) with Vestas (in-house The same set of explanatory and control variables in equation (1) are used
manufacturing) for nacelles)). We use a Sankey (alluvial) diagram to visualize in all cases. Table 1 reports values of (β) and Supplementary Table 4 shows the
the proportional flow between nodes of the network (that is, the location of the descriptive statistics and correlations. In total, we have an unbalanced panel of 1,227
supplier and the location of the OEM) using R (version 3.6.2) package ggforce64. observations of 318 suppliers from Denmark, Germany, Spain, the United States,
Japan, China and India (out of the 389 global suppliers, from 2006 to 2016). Model 1
Statistical analysis for evolution of suppliers. To estimate the links between measures the impact of all control variables on international evolution, and model 2
technology complexity and suppliers’ ability to be strategic and competitive in adds the effect of our main independent variable technology complexity, suggesting
international markets, we conduct a set of OLS regression analyses from 2006 to a significant negative impact on international evolution (β = −0.118, P = 0.029).
2016 using statistical modelling in R (version 3.6.2) and output using the stargazer Model 3 includes the same variables but limits the firms to the 114 suppliers from
package65. Germany, Denmark and Spain (excluding the United States and Chinese suppliers).
The dependent variable is the evolution, estimated as the difference over two Model 4 captures the same for 37 US suppliers, and model 5 for 138 Chinese
years (that is, a two-year time lag) in the fraction of supplier’s market relationships suppliers. Given the low number of suppliers from India (21) and Japan (8), we did
with OEMs from a different country (international OEMs), as a proxy for suppliers’ not calculate separate models for these countries. Model 6 limits the dataset to only
ability to compete in international markets. We used the network analysis technique capture relationships of suppliers with OEMs from the European Union, model 7 to
(as described in the previous section) to first quantify the market relationships OEMs from the United States and model 8 from China.
between suppliers with OEMs. In a given year t, a value of 0 reflects that suppliers In addition, we conducted several robustness checks for our model
work only with OEMs from the same country while 1 reflects that suppliers only specifications. These include different complexity metrics, time lags and
work with OEMs from a different country (international OEM). Then, to estimate interaction effects (Supplementary Tables 5 and 6). Our results are robust to all
the change over time, where an increase in international relationships indicates model specifications.
an increase in competitiveness, we calculated the difference with year t + 2. The Finally, we undertook a three-step approach to address endogeneity concerns
final variable for evolution ranges from −1 to 1, where a negative value indicates a that the complexity will shape how countries export and internationalize, while the
decrease in the fraction of international relationships, 0 indicates no change and 1 PCI-based complexity measure is also calculated based on countries that are able to
indicates an increase in the fraction of international relationships. manufacture and export a technology. First, complexity (the independent variable)
We combined the data on supplier–OEM relationships with home-country is measured at the component level based on broader mapping of HS codes from
information of the suppliers and OEMs. We manually collected the addresses the PCI approach of Hausmann et al. (where other components unrelated to wind
of each supplier and OEM by searching databases such as Orbis, Amadeus or may also be traded under a particular component code). International evolution
Bloomberg and verified and extended this information with a manual search (our dependent variable) is estimated on the supplier-component level of the wind
on the suppliers’ webpages. We used headquarter addresses in the case of larger energy industry. This eliminates the use of the same data and unit of analysis for
companies with multiple facilities. We calculated changes in the fraction of the two variables. Second, we use other complexity metrics that are based on patent
international relationships on a two-year basis, and also on a yearly basis as a data and do not rely on country information. Our results are again robust to these
robustness check. While the results using two-year and one-year changes revealed other complexity metrics (Supplementary Table 5). Third, we use time lags of two
robust estimates, we decided to focus on two-year changes that are likely to capture years in our main model (model 1 and model 2) and multiple other time lags for
actual strategic changes of suppliers’ evolution to a greater extent. robustness checks (Supplementary Table 6). This reduces the relationship in a
The main independent variable is technology complexity (x1), measured using particular year between the dependent variable and the complexity-independent
the product complexity index as described above for each component and year. variable. Our results are robust under different specifications.
In addition, we used the following supplier-specific control variables:
• Wind specialization (x2), which is a binary variable that measures whether the Reporting Summary. Further information on research design is available in the
supplier specialized in wind energy (1) or was active in other sectors outside Nature Research Reporting Summary linked to this article.
of the wind industry (0). We obtained this information during our efforts of
expanding the original dataset by manually coding all suppliers based on an Data availability
analysis of their webpages and databases such as Bloomberg. The database on the global manufacturing value chain developed for this
• Component diversification (x3) is a variable that measures the number of study was built on third-party reports published by Navigant Consulting, with
wind components supplied by a firm to wind OEMs, which we derived from additional details obtained from Orbis, Amadeus, Bloomberg and Derwent
our original dataset. In our database, 279 (90.9%) suppliers only offered one World Patents Index. Restrictions apply to the availability of these third-party

data and so the dataset is not publicly available. Data are however available 25. Garud, R. & Karnoe, P. Bricolage versus breakthrough: distributed
upon reasonable request from the corresponding author. Supplier data (without and embedded agency in technology entrepreneurship. Res. Policy 32,
the supplier company name) are available at https://github.com/kavsurana/ 277–300 (2003).
tech-complexity-project/. The source data underlying Figs. 1–6 are provided as 26. OECD SME and Entrepreneurship Outlook 2019 (OECD, 2019).
source data. The source data underlying Supplementary Figs. 1 and 2 are provided 27. Supply Chain Assessment —Wind Energy 2006–2014
as Supplementary Data 1 and 2. (Navigant Research, 2014).
28. Hausmann, R. et al. The Atlas of Economic Complexity (MIT Press, 2013).
Code availability 29. McNerney, J., Farmer, J. D., Redner, S. & Trancik, J. E. Role of design
The source and code to replicate the analysis are available at https://github.com/ complexity in technology improvement. Proc. Natl Acad. Sci. USA 108,
kavsurana/tech-complexity-project/. 9008–9013 (2011).
30. Novak, S. & Eppinger, S. D. Sourcing by design: product complexity and the
supply chain. Manag. Sci. 47, 189–204 (2001).
Received: 24 February 2020; Accepted: 30 July 2020; 31. Broekel, T. Using structural diversity to measure the complexity of
Published: xx xx xxxx technologies. PLoS ONE 14, e0216856 (2019).
32. Mealy, P., Farmer, J. D. & Teytelboym, A. Interpreting economic complexity.
References Sci. Adv. 5, eaau1705 (2019).
1. Renewable Capacity Statistics 2020 (IRENA, 2020); https://www.irena.org/ 33. Balland, P.-A. & Rigby, D. The geography of complex knowledge. Econ. Geogr.
publications/2020/Mar/Renewable-Capacity-Statistics-2020 93, 1–23 (2017).
2. World Energy Outlook 2019 (IEA, 2019); https://www.iea.org/reports/ 34. Fleming, L. & Sorenson, O. Technology as a complex adaptive system:
world-energy-outlook-2019 evidence from patent data. Res. Policy 30, 1019–1039 (2001).
3. U.S. Wind Energy Manufacturing and Supply Chain: A Competitiveness 35. Huenteler, J., Ossenbrink, J., Schmidt, T. S. & Hoffmann, V. H. How a
Analysis (Global Wind Network, 2014); https://energy.gov/sites/prod/ product’s design hierarchy shapes the evolution of technological knowledge—
files/2014/09/f18/U.S.%20Wind%20Energy%20Manufacturing%20and evidence from patent-citation networks in wind power. Res. Policy 45,
%20Supply%20Chain%20Competitiveness%20Analysis_0.pdf 1195–1217 (2016).
4. Wiser, R. et al. Expert elicitation survey on future wind energy costs. 36. Observatory of Economic Complexity—Product Complexity Rankings (MIT
Nat. Energy 1, 16135 (2016). Media Lab, 2011); https://atlas.media.mit.edu/en/rankings/product/
5. Wiser, R. et al. 2018 Wind Technologies Market Report 103 (US Department hs07/?year_range=2011-2016
of Energy, Office of Energy Efficiency and Renewable, 2018). 37. Paisemi, F. EU Energy Technology Trade: Import and Export (Publications
6. Industrial Strategy: Offshore Wind Sector Deal (HM Government, 2019); Office of the European Union, 2017); https://setis.ec.europa.eu/sites/default/
https://assets.publishing.service.gov.uk/government/uploads/system/uploads/ files/reports/eu_energy_technology_trade.pdf
attachment_data/file/790950/BEIS_Offshore_Wind_Single_Pages_web_ 38. Wind, I. HS Codes and the Renewable Energy Sector. Research and Analysis
optimised.pdf (International Centre for Trade and Sustainable Development, 2008).
7. Hansen, U. E., Nygaard, I., Morris, M. & Robbins, G. The effects of local 39. Asheim, B. T. & Coenen, L. Knowledge bases and regional
content requirements in auction schemes for renewable energy in developing innovation systems: comparing Nordic clusters. Res. Policy 34,
countries: aA literature review. Renew. Sustain. Energy Rev. 127, 109843 (2020). 1173–1190 (2005).
8. Schmidt, T. S., Schmid, N. & Sewerin, S. Policy goals, partisanship and 40. Lewis, J. I. & Wiser, R. H. Fostering a renewable energy technology industry:
paradigmatic change in energy policy — analyzing parliamentary discourse in an international comparison of wind industry policy support mechanisms.
Germany over 30 years. Clim. Policy 19, 771–786 (2019). Energy Policy 35, 1844–1857 (2007).
9. Lewis, J. I. The rise of renewable energy protectionism: emerging trade 41. Qiu, Y. & Anadon, L. D. The price of wind power in China during its
conflicts and implications for low carbon development. Glob. Environ. Polit. expansion: technology adoption, learning-by-doing, economies of scale, and
14, 10–35 (2014). manufacturing localization. Energy Econ. 34, 772–785 (2012).
10. Stokes, L. C. & Warshaw, C. Renewable energy policy design and framing 42. Awate, S., Larsen, M. M. & Mudambi, R. Accessing vs sourcing knowledge: a
influence public support in the United States. Nat. Energy 2, 17107 (2017). comparative study of R&D internationalization between emerging and
11. Haakonsson, S. J. & Kirkegaard, J. K. Configuration of technology networks advanced economy firms. J. Int. Bus. Stud. 46, 63–86 (2015).
in the wind turbine industry. A comparative study of technology management 43. Nieto, M. J. & Rodríguez, A. Offshoring of R&D: looking abroad to improve
models in European and Chinese lead firms. Int. J. Technol. Manag. 70, innovation performance. J. Int. Bus. Stud. 42, 345–361 (2011).
281–299 (2016). 44. Dunning, J. H. & Lundan, S. M. The internationalization of corporate R&D: a
12. Binz, C. & Truffer, B. Global innovation systems—a conceptual framework review of the evidence and some policy implications for home countries. Rev.
for innovation dynamics in transnational contexts. Res. Policy 46, Policy Res. 26, 13–33 (2009).
1284–1298 (2017). 45. Doblinger, C., Dowling, M. & Helm, R. An institutional perspective of public
13. Pietrobelli, C. & Rabellotti, R. Global value chains meet innovation systems: policy and network effects in the renewable energy industry: enablers or
are there learning opportunities for developing countries? World Dev. 39, disablers of entrepreneurial behaviour and innovation? Entrep. Reg. Dev. 28,
1261–1269 (2011). 126–156 (2016).
14. Jurowetzki, R., Lema, R. & Lundvall, B.-Å. Combining innovation systems 46. Cox Pahnke, E., McDonald, R., Wang, D. & Hallen, B. Exposed: venture
and global value chains for development: towards a research agenda. Eur. J. capital, competitor ties, and entrepreneurial innovation. Acad. Manag. J. 58,
Dev. Res. 30, 364–388 (2018). 1334–1360 (2014).
15. Lema, R. & Lema, A. Technology transfer? The rise of China and India in 47. Doblinger, C., Surana, K. & Anadon, L. D. Governments as partners: the
green technology sectors. Innov. Dev. 2, 23–44 (2012). role of alliances in U.S. cleantech startup innovation. Res. Policy 48,
16. Lewis, J. I. Building a national wind turbine industry: experiences from 1458–1475 (2019).
China, India and South Korea. Int. J. Technol. Global. 5, 281–305 (2011). 48. Schilling, M. A. & Phelps, C. C. Interfirm collaboration networks: the impact
17. Surana, K. & Anadon, L. D. Public policy and financial resource mobilization of large-scale network structure on firm innovation. Manag. Sci. 53,
for wind energy in developing countries: a comparison of approaches and 1113–1126 (2007).
outcomes in China and India. Glob. Environ. Change 35, 340–359 (2015). 49. Gereffi, G., Humphrey, J. & Sturgeon, T. The governance of global value
18. Binz, C., Gosens, J., Hansen, T. & Hansen, U. E. Toward technology-sensitive chains. Rev. Int. Polit. Econ. 12, 78–104 (2005).
catching-up policies: insights from renewable energy in China. World Dev. 50. Sturgeon, T., Van Biesebroeck, J. & Gereffi, G. Value chains, networks and
96, 418–437 (2017). clusters: reframing the global automotive industry. J. Econ. Geogr. 8,
19. World Development Report 2020: Trading for Development in the Age of Global 297–321 (2008).
Value Chains (World Bank, 2020). 51. Schmidt, T. S. & Huenteler, J. Anticipating industry localization effects of
20. Multinational Enterprises in the Global Economy—Heavily Debated but Hardly clean technology deployment policies in developing countries. Glob. Environ.
Measured (OECD, 2018); https://www.oecd.org/industry/ind/ Change 38, 8–20 (2016).
MNEs-in-the-global-economy-policy-note.pdf 52. Brauch, H. G., Spring, Ú. O., Grin, J. & Scheffran, J. Handbook on
21. Antràs, P. Conceptual Aspects of Global Value Chains NBER Working Paper Sustainability Transition and Sustainable Peace Vol. 10 (Springer, 2016).
No. 26539 (NBER, 2019); https://doi.org/10.3386/w26539 53. Haakonsson, S., Kirkegaard, J. K. & Lema, R. The decomposition of
22. Wind Energy Industry Manufacturing Supplier Handbook (AWEA, 2011). innovation in Europe and China’s catch-up in wind power technology: the
23. Hobday, M. Product complexity, innovation and industrial organisation. role of KIBS. Eur. Plan. Stud. https://doi.org/10.1080/09654313.2020.1712329
Res. Policy 26, 689–710 (1998). (2020).
24. Huenteler, J., Schmidt, T. S., Ossenbrink, J. & Hoffmann, V. H. Technology 54. Utility Scale Wind Towers from Canada, Indonesia, Korea, and Vietnam
life-cycles in the energy sector—technological characteristics and the role of Investigation Nos. 701-TA-627-629 and 731-TA-1458-1461 (Preliminary)
deployment for innovation. Technol. Forecast. Soc. Change 104, 102–121 (2016). (USITC, 2019); https://www.usitc.gov/publications/701_731/pub4952.pdf

55. Wilson, C. et al. Granular technologies to accelerate decarbonization. Science Acknowledgements
368, 36–39 (2020). Funding for this research was provided by the US National Science Foundation under
56. Future of Wind: Deployment, Investment, Technology, Grid Integration and grant number 1829252; and the UK Economic and Social Research Council under grant
Socio-economic Aspects (IRENA, 2019); https://www.irena.org/-/media/Files/ number ES/S010688/1. M. George, L. He, A. Hammerstingl and F. Traimer helped with
IRENA/Agency/Publication/2019/Oct/IRENA_Future_of_wind_2019.pdf cleaning and verifying the dataset. D. Li and M. Vigil provided valuable feedback on the
57. Feinerer, I., Hornik, K. & Meyer, D. Text mining infrastructure in R. J. Stat. concepts behind this paper.
Softw. 25, 1–54 (2008).
58. Grün, B. & Hornik, K. topicmodels: an R package for fitting topic models. Author contributions
J. Stat. Softw. 40, 1–30 (2011). K.S, C.D. and L.D.A developed the research idea and concept. K.S. and C.D collected
59. Chan, G. The Commercialization of Publicly Funded Science: How Licensing and analysed the data. K.S., C.D., L.D.A. and N.H. interpreted the results and conducted
Federal Laboratory Inventions Affects Knowledge Spillovers. Doctoral thesis, policy analysis. K.S. and C.D. wrote the manuscript. L.D.A. and N.H. edited the
Harvard Univ. (2015); http://isites.harvard.edu/fs/docs/icb.topic1459278.files/ manuscript. K.S., L.D.A. and N.H. secured project funding.
CHAN-Gabriel_11-21-14_JMP%20-%20National%20Lab%20Patent%20
Licensing.pdf Competing interests
60. Kim, G., Park, S. & Jang, D. in Soft Computing in Big Data Processing (eds. The authors declare no competing interests.
Lee, K. M. et al.) 71–80 (Springer International, 2014).
61. Barrat, A., Barthélemy, M., Pastor-Satorras, R. & Vespignani, A. The
architecture of complex weighted networks. Proc. Natl Acad. Sci. USA 101, Additional information
3747–3752 (2004). Supplementary information is available for this paper at https://doi.org/10.1038/
62. Csardi, G. & Nepusz, T. The igraph software package for complex network s41560-020-00685-6.
research. InterJ. Complex Syst. 1695, 1–9 (2006). Correspondence and requests for materials should be addressed to K.S. or L.D.A.
63. Taglioni, D. & Winkler, D. Making Global Value Chains Work for Development Reprints and permissions information is available at www.nature.com/reprints.
(The World Bank, 2016); https://doi.org/10.1596/978-1-4648-0157-0
64. Pedersen, T. L. ggforce: accelerating ‘ggplot2’. R package version 0.1 2 (2019). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
65. Hlavac, M. Stargazer: well-formatted regression and summary statistics tables. published maps and institutional affiliations.
R package version 5 (2015). © The Author(s), under exclusive licence to Springer Nature Limited 2020

nature research | reporting summary
Corresponding author(s): Kavita Surana
Last updated by author(s): Jul 20, 2020
Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.
Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.
Software and code

Policy information about availability of computer code
Data collection Data collection was conducted from .pdf reports and tabulated into Microsoft Excel. Data with anonymised company name is available on
GitHub at https://github.com/kavsurana/tech-complexity-project/
Data analysis Data analysis was conducted using R (version 3.6.2). The code and data to replicate statistical analysis and create figures is available on Github
at https://github.com/kavsurana/tech-complexity-project/
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
April 2020
- A list of figures that have associated raw data

- A description of any restrictions on data availability
The database on the global manufacturing value chain developed for this study was built on third-party reports published by Navigant Consulting, with additional
details obtained from Orbis, Amadeus, Bloomberg, and Derwent World Patents Index. Restrictions apply to the availability of these third-party data and so the
dataset is not publicly available. Data are however available upon reasonable request from the corresponding author. Supplier data (without the supplier company
name) are available at [https://github.com/kavsurana/tech-complexity-project/] along with the source and code to replicate the analysis. The source data
underlying Figs. 1–6 and Supplementary Figs. 1–2 are provided as a Source Data file.
1
nature research | reporting summary
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf
Behavioural & social sciences study design

All studies must disclose on these points even when the disclosure is negative.
Study description The study is based on a dataset of the manufacturing global value chain of wind energy technologies, with data on component
suppliers and the large original equipment manufacturers (OEMs) they supply to. The study uses mixed-methods with time series
data on supplier-OEM relationships, analyzed using qualitative approaches and statistics.
Research sample We obtained time series data using biennial reports from Navigant Consulting (2006, 2008, 2010, 2012, and 2014), with each
relationship reported for a 3-year horizon—for example, the 2014 industry report identified supplier-OEM relationships from 2014
through 2016. In this step, we tabulated information on all major component suppliers (active between 2006 and 2014), the OEMs
they supply to (and are expected to supply to until 2016), the outsourcing strategies of the OEM firms (either in-house development
of components or outsourced to external supplier), and the geographical location of the supplier firms.
Sampling strategy Our dataset captures nearly a decade of rapid advancements and international changes in wind energy manufacturing and
deployment. Because the source of the data is detailed industry reports from Navigant Consulting that tracked this information over
time in extensive detail, we believe that the most important supplier-OEM relationships have been captured by the dataset.
Data collection The data was coded from text-based industry reports obtained as pdfs, and tabulated in excel. The initial data was independently
verified by two authors KS and CD, with additional information added from multiple third-party sources.
Timing The time period is 2006 to 2016.
Data exclusions No data were excluded in the qualitative analysis. Data with missing variables / information were excluded in the statistical analysis.
Non-participation This study was not an experiment and participation is not applicable here.
Randomization This study was not an experiment and randomization is not applicable here.
Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.
Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology and archaeology MRI-based neuroimaging
Animals and other organisms
Human research participants
Clinical data
Dual use research of concern
April 2020
2
Ain Shams Engineering Journal 11 (2020) 377–387
Contents lists available at ScienceDirect
Ain Shams Engineering Journal

journal homepage: www.sciencedirect.com
Electrical Engineering
Dynamic reluctance air gap modeling and experimental evaluation of

electromagnetic characteristics of five-phase permanent magnet
synchronous generator for wind power application
Raja Ram Kumar, S.K. Singh, R.K. Srivastava, R.K. Saket ⇑
Indian Institute of Technology (Banaras Hindu University), Varanasi, UP, India
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents the design and analysis of electromagnetic characteristics of a five phase permanent
Received 2 June 2018 magnet synchronous generator for direct drive wind energy conversion system (WECS). In this study,
Revised 16 September 2019 simple and accurate Dynamic Reluctance Network Modeling is used for design and optimization of gen-
Accepted 18 September 2019
erator. The anisotropic structure of stator and rotor for the accurate prediction of flux distribution in the
Available online 14 November 2019
air gap and electromagnetic performance is accounted by the dynamic variation of air gap reluctance in
an electrical period. This model considers the leakage flux paths for machine design optimization to
Keywords:
achieve better performance. In this context, three permanent magnet (PM) materials namely NdFeB,
Wind power
Five phase
SmCo and ferrite are considered to evaluate the generated voltage. The dimensions of these permanent
Permanent magnet synchronous generator magnets are varied and performance under normal and saturated core condition is evaluated. A prototype
Dynamic reluctance network modeling is developed in the machine laboratory of the IIT(BHU) Varanasi (India) and results obtained in accor-
Finite element method dance with Finite Element Method.
Ó 2019 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Ain Shams University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-
nd/4.0/).
1. Introduction high efficiency [5], high power density, light weight and ease of use
as compared to conventional systems [6,7]. The power density can
Due to environmental issues, renewable energy has become the be further improved by using multi-phase systems. Multi-phase
current trend of energy generation [1]. Energy sources such as generators possess fault tolerance capability since they can operate
solar, wind, biomass, geothermal, etc are quite popular since they stably even when more than one phase is under fault, with accept-
are locally available but are indefinite source of energy. Wind able output [8]. It has inherent features of minimizing the space
power is a sustainable energy source that yields cheaper energy harmonics, torque ripples and suppressing the vibrations, due to
production [2,3]. To harness maximum amount of energy from which it is fit for direct drive wind power applications [9,10]. With
wind for variable speed, a suitable generating system is required. optimized design the torque ripple and harmonics in the generated
Permanent magnet synchronous generator (PMSG) is most suited voltage could be further reduced.
for this purpose since it is a direct drive system and generates volt- To design and analyse a generator, mainly two techniques are
age for wide range of speeds, whereas conventional generators reported, analytical and finite element analysis. In Finite Element
operate at a single fixed speed and have issues of frequent gear Method (FEM), for the calculation of electromagnetic field distribu-
box and slip ring maintenance [4]. PMSG possesses advantages like tion, the geometry has to be discretized and thus the solution
become more complex, time consuming but at the same time accu-
⇑ Corresponding author. racy improves [11]. Analytical models are important tools for
E-mail addresses: rajaram.rs.eee@iitbhu.ac.in (R.R. Kumar), sksingh.eee@iitbhu. design, analysis and optimization of electrical machines because;
ac.in (S.K. Singh), rksrivastava.eee@iitbhu.ac.in (R.K. Srivastava), rksaket.eee@iitb- these are very quick in computing the electromagnetic field solu-
hu.ac.in (R.K. Saket). tions. In [12,13] authors have reported an analytical model and
Peer review under responsibility of Ain Shams University. computed the performance of PM machines based on sub domain
modeling technique but it requires enough information about
boundary conditions and mathematical calculations are rigorous.
Reluctance network modeling technique is very popular owing to
Production and hosting by Elsevier its simplicity, fast and accurate prediction of performance for all
https://doi.org/10.1016/j.asej.2019.09.004
2090-4479/Ó 2019 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Ain Shams University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
378 R.R. Kumar et al. / Ain Shams Engineering Journal 11 (2020) 377–387
Nomenclature
Nc no. of turns per coil

List of symbols Ia,b,c,d,e phase current
lm length of magnet g airgap length
L core length of generator geff effective airgap length
Li iron core length Nph number of turns per phase
lfe relative permeability of iron f frequency
lo 4p 107 Her rotor rotation electrical angle
Hpm height of magnets b angular length of magnet
Hc coercivity of PM
kind of machines. In this context, machine is represented by lumped 2. Proposed topology and machine description
parameter reluctance network and reluctance of air gap is modelled
considering the effective air gap length using the carter coefficient. The proposed topology of the five phases PMSG, consisting of
This accounts for the slotting effect of machine for the performance stator and rotor with 8 poles is shown in Fig. 1 and Fig. 2.
evaluation [14]. The leakage flux affects the performance and the There are 8 permanent magnets mounted on the surface of
analysis is effectively evaluated [15] in terms of flux density distri- rotor. Magnetic poles are having 36° arc-span and 2.5 mm thick-
bution. Today, researchers are working on variable air gap reluc- ness. Stator is having 60 slots consisting of double layer fractional
tance network which implies reluctance of air gap varies with slot winding of five phases and 8 poles. Each phase of the winding
variation in the position of rotor which in turn provides actual vari- is mutually 72° phase apart. It consists of 240 turns placed in 12
ation of presence of the air gap. This accounts for actual distribution coils. Which are short-pitched by 12°. The winding connection
of flux in the air gap. For computation of interaction between stator details for phase A is presented in Table 1. The voltage developed
teeth and PMs, many techniques have been proposed [16]. in each of the 12 coils (ca, cb, cc, cd, ce, cf, cg, ch, ci, cj, ck and cl) with
Although, many literatures incorporated the effects of satura- rotor position is shown in Fig. 3. The phasor sum of all these coil
tion, slotting and armature reaction with some assumptions to voltages is shown in Fig. 4. The specification and dimensional
avoid the complexity in the machine reluctance network model details of the generator are given in Table 2 and Table 3,
due to which they compromise with the accuracy of electromag- respectively.
netic performance results. The improvement of model accuracy,
evaluation of power density of the optimized design and perfor-
mance studies under saturation condition are most motivating fac- 3. Dynamic reluctance network model
tors for the authors. This paper focuses on design and the
evaluation of electromagnetic performance of the five phase per- To predict the flux distribution and electromagnetic perfor-
manent magnet synchronous generator (FP-PMSG) for direct- mance of the generator, a dynamic reluctance network model is
drive wind power applications. Basically, it is highlighted to the adopted. DRNM comprises linear, non-linear and variable reluc-
dynamic reluctance network modelling for model accuracy, tances together with Magneto-Motive Force (MMF) sources which
adverse effect of core saturation on the performance, power den- are shown in Fig. 5.
sity and electromagnetic performance of FP-PMSG. The prediction In this model, permanent magnet and five-phase armature
of performances is evaluated using dynamic reluctance network windings are the main sources of flux. The armature winding flux
modeling (DRNM). This network comprises linear, non-linear and
dynamic reluctances together with flux sources. Armature winding
and permanent magnets are two flux sources whose Flux ampli-
tude varies with current density, material properties and dimen-
sional parameters. In this paper, authors have taken three
different PM materials namely ferrite, SmCo and NdFeB to evaluate
their effect on the performance of generator, where the focus
would be on air gap reluctance. The air gap reluctance varies with
rotor position while core reluctance varies with saturation of mag-
netic core material. The saturation of core leads to adverse effect
on generated voltage. To get the optimal performance, the leakage
flux reluctance is also considered in the dynamic reluctance net-
work. To validate the calculated results using DRNM, Finite ele-
ment analysis is performed and results are found in good
agreement. Furthermore, a prototype is fabricated for the experi-
mental validation of the predicted and simulated results.
This Paper is organized in the following manners. The Section 2
of this manuscript introduces the proposed topology, winding
details, voltage developed and design details. The Section 3 elabo-
rates the dynamic reluctance network modeling for the proposed
generator and integrates the effect of rotor rotation within the
model. The Section 4 of the paper describes the experimental
set-up along with results. The comparisons of predicted, FEM and
experimental results are discussed in Section 5. Finally, Section 6
summarises the concluding remarks. Fig. 1. Stator.
R.R. Kumar et al. / Ain Shams Engineering Journal 11 (2020) 377–387 379
Linear reluctance (Rl)

ll
Rl ¼ ð1Þ
l0 lr Al
where ll is the length of segment in which flux flows, Al is the cor-
responding area of the flux tube, lr is the relative permeability of
the material.
Non-linear reluctance (Rnl)
lnl
Rnl ¼ ð2Þ
l0 lfeBnl Anl
where lnl is the length of segment in which flux flows, Anl is the cor-
responding area of the flux tube, lfe Bnl is permeability function
dependent upon the flux density in the material.
MMF sources connected in the stator tooth due to wind-
ing
0 1
2 0 0 0 0
B 0 0 0 2 0 C
B C
B C
B 0 1 0 1 0 C
B C
B 0 C
B 2 0 0 0 C
B C
Fig. 2. Rotor. B 0 0 0 0 2 C
B C0 1
B 0 0 1 C
B 0 1 C Ia
depends upon the current density, whereas the magnet flux is B CB C
B 0 0 C
B 0 2 0 CB Ib C
decided by the magnetic material and its dimensional parameters. B CB C
0 C:B Ic C
T
MMFw ¼ ½f 1 f 2 f 3 :::f 15 ¼ Nc :B 2 0 0 0
The variable reluctance in the model is corresponds to air gap B CBB C
C
reluctance. The reluctance of air gap varies with rotor position B 1 0 0 1 C
0 C @ Id A
B
and interaction of PM with stator teeth, as demonstrated in DRNM. B C
B 0 0 0 2 0 C Ie
B C
Non-linear reluctances are stator and rotor core reluctances origi- B 0 2 0 0 0 C
B C
nating due to saturation of ferromagnetic material. Stator teeth are B C
B 0 1 0 0 1 C
the most sensitive part with respect to saturation. Due to satura- B C
B 0 0 0 0 2 C
tion, the permeability of core starts to decrease which derates B C
B C
the machine performance. Linear reluctances are stator tooth tip @ 0 0 2 0 0 A
to tip, permanent magnet reluctance and their leakage flux path 1 0 1 0 0
reluctances in model. Fig. 6 represent the one-fourth linear model
ð3Þ
of generator in which three fluxes are shown, one for the main and
other two for the leakage fluxes. Main flux traverses through PM- where f1, f2, f3, . . .f15 are the MMFw sources connected correspond-
air-stator core-rotor core whereas leakage flux travels from ing to stator tooth branch.
magnet-air-rotor iron, and magnet-air-magnet. Reluctance of air gap (Rgi) due to interaction of PM with ith sta-
DRNM only shows one fourth model because of its quarter sym- tor tooth
metry. It consists of 34 nodes, 79 branches and 46 loops corre- g
sponding to t = 0 sec. The linear reluctance is calculated using Eq. Rgi ¼ ð4Þ
l0 :width:L
(1) and non-linear reluctances of different sections are calculated
using Eq. (2). There are 15 stator slots having an upper and lower where width = rg.ht-pm and ht-pm is the interaction angle between
layer of winding in the model. This winding shows mmf sources tooth and PM and rg is the radius of air gap.
which are present in the model with tooth reluctance. These MMF due to PM
mmf sources are computed from the phase currents using Eq. (3), MMFpm ¼ Hc :Hpm ð5Þ
where Nc is the number of turns per coil. The air gap reluctance
is not easy to calculate because it changes dynamically as rotor where Hc is the coercivity and Hpm is the height of PM in the direc-
rotates. It is based on the interaction of PMs and stator tooth. There tion of magnetization.
are 15 reluctances with one PM interaction in the reluctance net- The RN method results in a set of linear equations which should
work and all reluctances can be computed using Eq. (4). The PMs be solved to obtain magnetic fluxes.
are the main MMF sources in this network and can be computed -1
½/ ¼ ½R ½MMF ð6Þ
with Eq. (5). The leakage reluctances of magnet to magnet and
magnet to rotor iron can be computed using the circular arc- 0 1
R1;1 ... R1;46
straight technique. A detailed description of this method is B .
½ R ¼ B .. .. C C
explained in Sections 3.1 and 3.2 respectively. The flux in each part @ .. . . A ð7Þ
of the model can be found using Eq. (6). Eq. (7) represents the R46;1 R46;46
reluctance matrix corresponding to the model.
Table 1
Winding connection for phase A.
Coil 1 8 9 16 23 24 31 38 39 46 53 54
In 1 15 16 16 30 31 31 45 46 46 60 1
Out 8 8 9 23 23 24 38 38 39 53 53 54
Table 3
Design parameters.
Design parameter Value

Shaft radius 20 mm
Rotor outer radius 95 mm
Height of magnet 2.5 mm
Air gap length 2.3 mm
width of sleeve 0.2 mm
Inner radius of stator 100 mm
Outer radius of stator 175 mm
Area of slot 123.27 mm2
where R is (46 46), MMF is (46 1) and the flux is (46 1) matrix
corresponding to model.
3.1. Reluctance calculation for magnet to magnet leakage path
The leakage reluctance of magnet to magnet flux path can be

computed using circular arc-straight technique for considering
the flux accurately [17]. The leakage paths reluctance is calculated
on the basis of their dimensional details and material properties
using Eq. (9). Fig. 7 shows the path of leakage flux between mag-
nets to magnets.
Permeance of magnet to magnet leakage flux path (Pmm)
Z g eff
l0 L 1
Pmm ¼ df ¼ ð8Þ
0 d þ pf Rmm
Fig. 3. Coil voltage. Reluctance of magnet to magnet leakage flux path (Rmm)

p
Rmm ¼ ð9Þ
l0 Lln 1 þ pgdeff
where geff is the effective airgap length and d is the gap between
the the magnets
3.2. Reluctance calculation for magnet to rotor iron leakage path
The leakage reluctance of flux paths magnet to rotor iron is

computed on the basis of dimensional details and their material
properties using Eq. (11). Similarly, using circular arc-straight tech-
nique Fig. 8 shows the path of leakage flux between magnets to
rotor iron.
Permeance of magnet to rotor iron leakage flux path (Pmi)
Z g eff
l0 L 1
Pmi ¼ df ¼ ð10Þ
0 Hpm þ pf Rmi
Reluctance of magnet to rotor iron leakage flux path (Rmi)

p
Rmi ¼ ð11Þ
l0 Lln 1 þ pHgpmeff
3.3. Magnetic field distribution and electromagnetic performances
The magnetic flux distribution in air gap is flat-topped due to

Fig. 4. Phase voltage. the arc shaped magnets mounted on the rotor. This results in
odd space harmonic components of the flux density. These har-
monics can be reduced by proper selection of arc span of the
Table 2
PMs. The pole arc width is selected as 144°E which eliminates mul-
Rating of machine.
tiples of 5th order space harmonics. This short pitching of magnets
Parameter Value reduces the weight of magnets and makes the machine cost
Power 3.42 kVA effective.
stator per phase voltage 171.0 V Magnetic ux density due to PM’s in the air gap
phase current 4A
Number of phases 5 X
1
Speed 400 rpm Bg ðher Þ ¼ Bgk sinðkher Þ ð12Þ

k¼1;3;5
Fig. 5. DRNM of FP-PMSG.
Fig. 8. Magnet to rotor iron flux path.

Fig. 6. Linear model.
with reduction in ripple torque and thereby enhances the overload

capability and performance of the machine. No-load generated
EMF of stator and average electromagnetic torque required the
generator are represented by Eq. (14) and Eq. (17) respectively.
Generated voltage in the stator winding (E(t))
X
1
EðtÞ ¼ Ek sinðkwe tÞ ð14Þ
k ¼ 1;3;7
where peak value of voltage (Ek) is
16 p
Ek ¼ 2:9563 Nc Li r s we Bgk cos k ð15Þ
P 30
Fig. 7. Magnet to magnet flux path. The total output power (P(t)) is
PðtÞ ¼ Ea ðtÞIa ðtÞ þ Eb ðtÞIb ðtÞ þ Ec ðtÞIc ðtÞ þ Ed ðtÞId ðtÞ þ Ee ðtÞIe ðtÞ
where peak value of flux density (Bgk ) is ð16Þ
p The average electromagnetic torque (T(t)) is

4 B
Bgk ¼ p 2g cosðkbÞsin k ð13Þ
p b k 2 PðtÞ
2 TðtÞ ¼ Tcogg þ ð17Þ
we
Moreover, elimination of space harmonics in the air gap flux
density decreases the harmonics in the generated voltage along where we is the electrical speed of rotor.
Cogging torque (Tcogg)is voltage is the fundamental building block for grid connected or
stand alone generators, so loading of generator with the rectifier
1 2 dðRðhÞÞ
Tcogg ¼ / ð18Þ provides its actual behaviour. The no-load phase voltage and corre-
2 g dðhÞ sponding rectified output DC voltage at 26.81 Hz are 195 V and
The periodicity of cogging torque is estimated to be the least 390 V, respectively as shown in Fig. 10.
common multiple of the number of stator slots and number of
rotor poles. 5. Results and validation
4. Experimental setup The analytical results are predicted using the dynamic reluc-
tance network method. The network consists of MMF sources
To test the FP-PMSG, the experimental setup was developed as along with linear, non-linear and variable reluctances. Dynamic
shown in Fig. 9. Generator is driven by a prime-mover which fulfils reluctance modeling of air gap reluctance ensures the accuracy of
the power requirement for cogging torque and the electric load. prediction about magnetic flux distribution due to anisotropic
The cogging torque is an inherent behaviour of PM machine which structure of stator and rotor. Since distribution of the fluxes
is present throughout its operation. Machine is loaded using a five directly associates itself with the electromagnetic performance,
phase rectifier (FP-rectifier) circuit. The rectified direct current link so its accuracy is dependent upon its accurate prediction. The
Fig. 9. Experimental set-up.
Fig. 10. DC voltage and phase voltage at 26.81 Hz.

Fig. 11. Flux density.
generator performance is largely dependent upon the core satura- 1.28 Tesla at the stator teeth and is highest at the corner of
tion and its proper design goes a long way in saving the material magnet-rotor interface due to leakage flux. The flux density in
cost and enhancing its performance. The performance analysis the air gap is evaluated using analytical and FEM analysis and is
under loading condition of FP-PMSG is carried out using FP- shown in Fig. 12. The predicted air gap flux density, found to be
rectifier. The regulation of terminal voltage of generator can be car- 0.5621 Tesla, is 1.282% lesser than that found using FEM. The close-
ried out by direct electrical loading of rectified DC voltage at its ter- ness of these results reveals the effectiveness of dynamic magnetic
minals. The analytical results are validated using FEM analysis. The circuit modeling.
Ansoft Maxwell software is used for the FEM analysis. However,
the modeling is done for one fourth portion of the machine because
of its quarter symmetry. The reduced model decreases the struc- 5.2. Variation of EMF
tural complexity as well as time consumption of performance anal-
ysis. In addition the no load analysis in FEM is carried out by using The No-load generated EMF using analytical, FEM and experi-
magnetostatic solver whereas the loading analysis is carried out by mental results at 400 rpm are shown in Fig. 13. The predicted
using time step transient solver. For the magnetostatic analysis, result is found to be 187.56 V, which, is 1.284% and 3.28% lesser
the boundary condition at all the surfaces is taken as vector poten- than FEM and experimental results, respectively at 400 rpm.
tial whose value is zero. The total number of mesh elements is There are two line voltages: adjacent and non-adjacent in FP-
1374 in triangular shape with different sizes which are generated PMSG. These voltages are shown in Fig. 14 and Fig. 15. The peak
automatically by setting of edge length for different sections in value of both are same and is twice the peak value of phase voltage
the model. The solution is converged after two passes. The profile whereas the fundamental value of adjacent and non-adjacent line
data shows that the total real time is taken 5 s and the CPU time
is taken 3 s for the convergence of the simulation. On the contrary,
the transient analysis is made based on the boundary conditions
which are vector potential, master and slave. The total number of
mesh elements is 3142 of triangular shape with different sizes in
the model. From the excitation, the generator is externally loaded
together with eddy current effect is also provided by this. The time
steps are set to 0.0002 s for the simulation. The profile data show
that the total real time taken is 56.40 s whereas the time taken
by CPU is 41.20 s for the simulation. FEM results are accurate but
more time consuming compared to analytical methods. Further,
these results are validated with experimental results.
5.1. Magnetic field distribution
Due to anisotropic structure of stator and rotor, reluctance of air

gap is not fixed therefore the magnitude of flux density varies with
rotor position. For the analytical prediction, 3600 rotor positions
have been considered for flux density computation. Fig. 11 repre-
sents the flux density plot in the machine and is approximately Fig. 12. FEM and analytical result of air gap flux density.
Fig. 13. Comparison of phase voltage of analytical, FEM and Hardware result.
Fig. 16. Generated voltage per phase with speed.
Fig. 14. Adjacent line voltage.
Fig. 17. Generated voltage with different grade of magnets with height of magnet.
and FEM results are quite close to the experimental results. The
electromagnetic performance is calculated for different PM materi-
als like NdFeB, SmCo and ferrite. The generated voltage due to fer-
rite is lowest and that with NdFeB is highest because the value of
BHmax and remanence flux density (Br) is least for ferrite and high-
est for NdFeB.
Fig. 17 represents the variation of generated voltage with the
height of magnet. It is observed that the voltage due to ferrite mag-
net increases linearly as the height of magnet increases from 1 to
3 mm and at a slightly higher rate in between 3 and 4.5 mm
because of the dominance of ferrite reluctance change over the
air reluctance change in the main flux path. On the other hand, it
is observed that the linear relation holds only upto 2.5 mm for
NdFeB and SmCo magnets. Further, non-linearity creeps in as a
consequence of saturation of stator teeth, as the height of magnet
is increased from 2.5 to 3.5 mm in case of NdFeB and the maximum
Fig. 15. Non-adjacent line voltage. voltage is found to be 215.13 V at 3.5 mm. The similar non-linear
behaviour is exhibited in SmCo in the range of height from 2.5 to
4 mm and maximum voltage is found to be 187.6 V at 4 mm. Fur-
voltages are 1.176 and 1.902 times the fundamental value of phase ther, voltage starts to decrease in range of 4–4.5 mm height of
voltage, respectively. magnet and attains 196.7 and 185.9 V for NdFeB and SmCo, respec-
Fig. 16, represents the generated voltage-speed characteristic tively. The analytical and FEM results for NdFeB are verified and
and the relation is found to be linear. It is found that predicted found to be in close agreement and is shown in Fig. 18.
Fig. 18. Generated voltage with height of NdFeB magnet.

Fig. 20. Torque Vs current density.
5.3. Variation of torque and rectified power
The torque requirement of FP-PMSG is fulfilled by prime-mover.

The torque under loaded condition comprises, average electromag-
netic torque and cogging torque. Cogging torque has a negative
impact during operation of machine and increases torque ripple.
Fig. 19 represents an analytical and FEM variation of torque. The
average torque predicted by analytical method is 78.885 N m and
is 79.89 N m in case of FEM, at rated current. The percentage ripple
in the torque is 9.05% and 7.4% for predicted and FEM respectively.
Variation of magnitude of average torque with current density is
also carried out and a linear relationship is established between
the two quantities as shown in Fig. 20.
Average torque which has an initial value at zero current density,
owes its existence to the cogging torque. The cogging torque is the
inherent torque of PM machine and is present throughout the oper-
ation of generator and its variation is shown in Fig. 21. Its predicted
result is 4.615 N m, and is 2.55% lesser than FEM. It is further vali-
dated and found 6.46% lesser than the experimental result. Genera-
tor is loaded through rectifier and variation of output DC electric Fig. 21. Comparison of cogging Torque of analytical, FEM and Hardware result.
power with resistive loading is carried out which is shown in Fig. 22.
5.4. Comparative studies between three and five phase PMSG
A comparative study between 60 slots, 8 poles, 3 phase and 60

slot, 8 pole, 5 phase PMSG is done and their performance parameters
are enlisted in Table 4. For the comparison, all the dimensions,
Fig. 22. Output DC Electric power with resistive load.
material properties and current rating of the generator are kept con-
stant. It is found that the winding phase spread of five phase PMSG is
lesser by 24° than three phase generator which limits the leakage
flux and minimises the leakage inductance. The number of turns
Fig. 19. Average torque. per phase is lesser which reduces the phase resistance that limits
Table 4 predicted and FEM respectively. Hence, cogging torque was evalu-
Comparative performance parameters. ated as 4.615 N m under no-load condition which was found that
Parameter 5-Phase PMSG 3-Phase PMSG predicted results were 2.55% and 6.46% lesser as compared to FEM
Winding phase spread angle 24 48 and experimental studies respectively. Generator was loaded
No. Of turn per phase 456 760 through rectifier and the variation in electric power with resistive
Torque 78.8850 77.3909 loading was in good agreement with experimental results. In addi-
%Ripple torque 9.05% 39.6556% tion, the power density and efficiency for the FP-PMSG were found
%Efficiency 92.0632 91.9002
Power to weight ratio 83.4993 81.6746
to be 83.499 W/kg and 92.06%, respectively which are higher than
the conventional three phase generator.
References
[1] Goh HH, Lee SW, Chua QS, Goh KC, Teo KTK. Wind energy assessment
considering wind speed correlation in Malaysia. Renew Sustain Energy Rev
2016;54:1389–400.
[2] Mahdizadeh Amin, Schmid Robert, Oetomo Denny. Enhanced energy capture
of wind turbines by exact output regulation. In: IEEE 56th annual conference
on decision and control (CDC), Melbourne, Australia. p. 1446–51.
[3] Sindhuja B. A proposal for implementation of wind energy harvesting system
in trains. In: International conference on control, instrumentation, energy and
communication. p. 696–702.
[4] Zhou Jiawei, Li Shengquan, Li Juan, Zhang Jin. A combined control strategy of
wind energy conversion system with direct-driven PMSG. 31st Youth
academic annual conference of Chinese association of automation, Wuhan,
China; 2016. p. 369–74.
[5] Aubree Rene, Auger Franois, Mace Michel, Loron Luc. Design of an efficient
small wind-energy conversion system with an adaptive sensorless MPPT
strategy. Renew Energy 2016;86:280–91.
[6] Li H, Chen Z. Design optimization and site matching of direct-drive permanent
magnet wind power generator systems. Renew Energy 2009;34:1175–84.
[7] Blow Fredrik, Eriksson Sandra, Bernhoff Hans. No-load core loss prediction of
PM generator at low electrical frequency. Renew Energy 2012;43:389–92.
[8] Kumar Raja Ram, Singh Santosh K, Srivastava RK. Thermal modelling of dual-
Fig. 23. Comparison of Torque of three phase and Five Phase PMSG. stator five-phase permanent magnet synchronous generator. In: IEEE
transportation electrification conference(ITEC-India). p. 1–6.
[9] Trabelsi Mohamed, Nguyen Ngac Ky, Semail Eric. Real-time switches fault
the losses compared to three phase generator. The efficiency and diagnosis based on typical operating characteristics of five-phase permanent-
power density for the five phase PMSG are found to be 92.06% and magnetic synchronous machines. IEEE Trans Ind Electron 2016;63
83.499 W/kg, respectively which is higher than the conventional (8):4683–94.
[10] Youssef Abdel-Raheem, Sayed Mahmoud A, Abdelwhab MN, Gaber Shabib.
three phase generator. Fig. 23 presents the electromagnetic torque Generator side converter for five-phase PMSG based wind turbine generators
required for five and three phase generator and it is found that for using perturbation and observation technique. Int J Power Eng Energy (IJPEE)
similar torque, the percentage ripple factor is 9.05% and 39.65% 2016;7(1):610–8.
[11] Barhoumi EM, Wurtz F, Chillet C. Efficient reluctance network formulation for
respectively. Higher ripple factor shows adverse effect in terms of modeling design and optimization of linear hybrid motor. IEEE Trans Magn
higher vibration and acoustic noise in the system. 2016;52(3). Article. 7002404.
[12] Dalal Ankit, Nekkalapu Sameer, Kumar Praveen. 2-D analytical sub-domain
model for hybrid dual-rotor motor. IEEE Trans Magn 2016;52(6). Aricle no.
6. Conclusion 8103609.
[13] Hannon Bert, Sergeant Peter, Dupr Luc. 2-D analytical subdomain model of a
slotted PMSM with shielding cylinder. IEEE Trans Magn 2014;50(7). Article.
This paper studies the design and analysis of electromagnetic
8101410.
behaviour of a FP-PMSG for wind power application. DRNM has [14] Mirahki Hooshang, Moallem Mehdi. Torque calculation in interior permanent
been used for design and optimization of generator. The reluctance magnet synchronous machine using improved lumped parameter models.
model predicted the flux distribution in the air gap and electromag- Prog Electromag Res M 2014;39:133–9.
[15] Kumar Raja Ram, Singh SK, Srivastava RK. Effect of magnetic trajectories in a
netic performance in a precise and accurate manner. In comparison magnetically coupled dual stator five phase PMSG. International conference on
with FEM, for any saturation level, predicted results are quite close industrial technology (IEEE ICIT2015), Seville, Spain; 2015. p. 720–25.
to the value of generated voltage. The predicted flux density at mid- [16] Dogan H, Garbuio L, Nguyen-Xuan H, Delinchant B, Foggia A, Wurtz F.
Multistatic reluctance network modeling for the design of permanent-magnet
dle of air gap and phase voltage is 0.5621 T and 187.56 V at the rated synchronous machines. IEEE Trans Magn 2013;49(5):2347–50.
speed respectively which are 1.282% and 1.284% lesser than the [17] Hanselman DC. Brushless, permanent-magnet motor design. New
obtained FEM results. Furthermore, experiment was conducted for York: McGraw-Hill; 1994.
generated phase voltage using the fabricated generator and was
found to be 3.39% higher than the predicted result. Accuracy of
Dr. Raja Ram Kumar received the B. Tech. Degree in
DRNM was high as it not only considered dynamic reluctance of
Electrical Engineering from the West Bengal University
air gap but also considered reluctance of leakage flux paths which of Technology, Kolkata, India in 2010 and M. Tech. as
led to the improved design. The permeability of core material varies well as Ph.D. Degree in Electrical Machines & Drives
abruptly with saturation, thus for realisation of its adverse affect on from the Indian Institute of Technology (Banaras Hindu
University), Varanasi, India in 2012 and 2018, respec-
the generated voltage, different PM materials were taken like
tively. He is currently working as an Assistant Professor
NdFeB, SmCo and ferrite and their dimensions were varied. The gen- (Under TEQIP-III) in the Department of Electrical Engi-
erated voltage at 4.5 mm of PM thickness was found to be decreased neering, Jorhat Engineering College, Assam. He received
by 8.5% and 0.9% from its peak value for NdFeB and SmCo respec- institute research award by IIT (BHU) Varanasi in 2016
tively. The predicted voltage with NdFeB was in close agreement and second prize paper award in IACC Annual Meeting
2018 at Portland during IAS Annual Meeting. He was the
with the FEM results. In addition, the predicted electromagnetic tor-
chair of student branch of IEEE-IAS at IIT (BHU) Varanasi. He is the author of several
que was calculated under loaded conditions was 78.88 N m which research papers which are published in international/national journals and con-
was 1.25% lesser than that evaluated FEM. The ripple percentage ference proceedings. His current research interests include renewable energy,
in electromagnetic torque came out to be 9.05% and 7.4% for electrical machines, electric drives, power electronic converter design and control.
Dr. Santosh Kumar Singh received Ph.D. degree from electromagnetic propulsion, linear induction machines, special purpose electrical
the Department of Engineering, University of Cambridge machines, permanent magnets motors and generators and Eddy current brake. He
(UK). He is currently an Associate Professor with the had travelled to USA, Japan and Indonesia. He is Senior Member of IEEE (USA) and
Department of Electrical Engineering, Indian Institute of an Adviser and Founder of the IEEE Industry Applications Society Student Chapter at
Technology (BHU), Varanasi (UP) India. He is a senior IIT (BHU), Varanasi (India).
member of IEEE (USA), member IEI (India) and Life
member ISTE (India). He has published several research
papers in international/national journals and confer- Professor (Dr.) R.K. Saket is working in the Department
ence proceedings. His research interests include silicon of Electrical Engineering at Indian Institute of Technol-
carbide converters, power electronic topologies, electric ogy (Banaras Hindu University), Varanasi (UP) India.
drives, and permanent magnet generators. Previously, he was an Assistant Lecturer in Electrical
and Electronics Engineering group at Birla Institute of
Technology and Science, Pilani (Rajasthan), India; Lec-
turer in University Institute of Technology, Rajiv Gandhi
University of Technology, Bhopal (MP), India and
Professor (Dr.) R.K. Srivastava was born on December Assistant Professor in the Faculty of Engineering and
06, 1960 in Mirzapur (UP) India. He received the B. Tech. Technology, Sam Higginbottom University of Agricul-
Degree in Electrical Engineering, the M. Tech. Degree in ture, Technology and Sciences, Allahabad (UP), India. He
Electrical Machines and Drives, and the Ph. D. Degree in has more than twenty years of teaching and research
Electrical Machines from the erstwhile Institute of experience. He is the author/co-author of approximately 80 scientific articles, book
Technology, Banaras Hindu University, Varanasi (UP) chapters and research papers in prestigious international journals and conference
India in 1983, 1985, and 2000 respectively. He has proceedings. Dr. Saket is a fellow of the Intuition of Engineers (India), SMIEEE (USA),
served as teaching faculty in Pant College of Technology, MIET (UK), LMISTE (India) and an editorial board member of the Engineering,
Pantnagar and Institute of Engineering & Technology Technology and Applied Science Research (Greece). He is awarded by GYTI Award -
(IET), Lucknow (UP) India. He is currently working as a 2018 by Hon’ble President of India at Rashtrapati Bhavan, New Delhi (India) on
Professor in Electrical Machines and Drives in Indian March 19, 2018, and Design Impact Award by Padma Vibhushan Ratan Tata at
Institute of Technology (Banaras Hindu University), Mumbai (India) on July 24, 2018. His research interests include power system
Varanasi (UP) India. He has published several research papers in international/na- reliability, electrical machines and drives, reliability enhancement of energy sys-
tional journals and conference proceedings. He has authored two books. His spe- tems, and renewable sources of electrical energy.
cializations are the electromagnetic field applied to electrical machines,
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/326340286
Wind Energy: A Review Paper
Article · July 2018

DOI: 10.21058/gjet.2018.42004
CITATIONS READS
0 5,101
3 authors, including:
Bishwajeet Pandey
Gyancity Research Lab
231 PUBLICATIONS 1,268 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Energy Efficient Sensor Design View project
Transformation of Model into HDL Code View project
All content following this page was uploaded by Bishwajeet Pandey on 16 January 2019.
The user has requested enhancement of the downloaded file.

Gyancity Journal of Engineering and Technology,
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
Wind Energy: A Review Paper

1 2 3
Atul Kumar Muhammad Zafar Ullah Khan Bishwajeet Pandey
M S College, University of Management & Gyancity Research Lab,
1
Motihari, India Technology, Sialkot Campus, Pakistan Motihari, India
1
atulmth1@gmail.com 2
acd.dir@skt.umt.edu.pk gyancity@gyancity.com
Abstract- This review paper examined the outline of wind innovation, where the approach depends on standards and down to
earth executions. Wind vitality is the second biggest wellspring of sustainable power source after hydropower. It is incredibly
reasonable, yet it is discontinuous. Even though the abuse of twist goes back a few centuries, the cutting edge wind vitality
industry started amid the oil emergency of the seventies. Most these days wind turbines are onshore; however others are
fabricated seaward, more often than not in wind ranches. Since wind vitality is discontinuous, it must be upheld by different
wellsprings of power. Wind vitality can be productive as a rule. However, it has not yet accomplished full matrix equality with
fossil vitality sources.
Keywords- Literature Survey; Wind Forecasting Categories; Wind Speed and Power Forecasting Methods.
1. Wind Energy- Introduction:

Rising oil costs feature the abuse of sustainable power source applications. Wind vitality is a standout amongst
the most appealing sustainable power source advancements on account of its high proficiency and low contamination
[1]. Be that as it may, since the vitality created by wind vitality transformation frameworks (WECS) changes with
environmental meteorology and wind speed [2-3], surprising varieties in WECS vitality generation may expand the
working expenses of the electrical structure in light of the fact that the stores will be developed and the potential dangers
will be put for the unwavering quality of the power supply [4]. Power lattice administrators need to anticipate changes
in wind control age to program turning save limit and oversee arrange tasks [4]. To lessen hold limit and increment wind
infiltration, precise gauging of wind speed is required [5]. What's more, the forecast of wind vitality assumes a vital part
in the portion of balance control. What's more, the breeze vitality conjecture is utilised for the day by day programming
of conventional power plants and the commercialisation of power in the spot advertise [6]. Even though the conjecture
precision of the breeze vitality figure is lower than the expectation exactness of the heap gauge. Wind vitality gauges
still assume a crucial part in tackling the issues of misusing power supply. As of late, a few techniques have been utilised
for the forecast of wind vitality. Various written works have been dedicated to enhancing wind vitality anticipating
approaches by analysts with broad involvement in field preliminaries. A few techniques for estimating wind vitality
have been created and propelled on wind ranches.
We can characterise into six gatherings to anticipate wind vitality: tirelessness strategy, physical technique,
factual strategy, spatial relationship strategy, human-made reasoning strategy and cross breed approach. [7-15] As
indicated by the latest World Wind Energy Association (WWEA) information, even in 2009, the time of the worldwide
money related emergency, the world's aggregate introduced limit will achieve 152,000 MW before the finish of 2009
[3] This implies in 2009, there will be 30,300 MW of newly introduced limit, which compares to a development of 25%
contrasted with the earlier year (Fig.1).
Fig. 1. Worldwide growing (red) and worldwide yearly (green) Fig. 2.Recently mounted power capacity in EU, 2008
mounted wind capacity. (2009 predicted value).
29
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
As per the report of The European Wind Energy Association (EWEA), the year 2008 in the EU was the chief year in
which more breeze control was introduced than some other power creating innovation, (Fig.2).
2. Wind Energy:
Wind vitality is a changed over type of sunlight based vitality that is created by the atomic combination of
hydrogen (H) into helium (He) in its core. The H → He dissolving process makes streams of warmth and
electromagnetic radiation from the sun to space every which way. Albeit the Earth catches just a little division of sun-
powered radiation, it gives the more significant part of the Earth's vitality needs. Wind vitality is a noteworthy
wellspring of cutting-edge dynamism, and a notable player in the worldwide vibrancy advertise. As a best in class
vitality innovation, the specialised development and quick organisation of wind vitality are perceived, just like the
absence of a down to earth furthest point of confinement for the level of wind that can be coordinated into the
electrical framework [1]. It has been assessed that the aggregate sun oriented vitality got by the Earth is around 1.8 ×
10 11 MW. Of this sun based info, just 2% (3.6 × 10 9 MW) is changed over to wind vitality, and around 35% of wind
vitality scatters inside 1,000 m of the Earth's surface [2]. Subsequently, accessible breeze vitality that can be changed
over into different types of energy is roughly 1.26 × 10 9 MW. Since this esteem speaks to 20 times the present rate of
worldwide vitality utilisation, wind vitality could on a fundamental level fulfil the everyday vitality needs of the
world. Contrasted with familiar vitality sources, wind vitality has numerous favourable circumstances and advantages.
Not at all like petroleum derivatives that radiate explosive gases and atomic dynamism that creates radioactive waste,
wind vitality is a clean and naturally amicable wellspring of energy. As an endless and free wellspring of life, it is
accessible and plentiful in many parts of the world. Furthermore, more widespread utilisation of wind vitality would
help diminish the interest for non-renewable energy sources, which could be exhausted sooner or later in this century,
contingent upon their present usage. Moreover, the cost per kWh of wind vitality is much lower than that of sun
oriented energy [3]. In this way, as the most encouraging vitality source, it is trusted that breeze vitality assumes an
essential part in global vitality supply in the 21st century.
3. Wind Energy Equation:
a) Wind speed-
The measure of vitality in the breeze shifts with the solid shape of the breeze speed, at the end of the
day, if the breeze speed copies, there is eight times more vitality in the breeze ( ). Little
varieties in wind speed mostly affect the measure of energy accessible in the breeze.
b) The density of the air-
The denser the sky, the more vitality the turbine gets. The thickness of air shifts with height and
temperature. The atmosphere is less thick at high altitude than adrift level, and warm air is less compressed
than chilly air. Every single other thing being equivalent, turbines will deliver more power at bringing down
heights and in places where average temperatures are colder.
c) The swept area of the turbine-
The bigger the region cleared (the measure of the rotational part of the rotor), the more noteworthy the
power that the turbine can get from the breeze. Since the cleared territory is, the place span of the rotor, a little
increment in edge length brings about a more significant increase in the accessible power for the turbine.
4. Wind Turbines:
Wind turbines fabricate power to abuse the moderate intensity of the breeze to drive a generator. The breeze
could be a clean and property supply of fuel, it doesn't create discharges and it'll ne'er run out since it's interminably
restored with the vitality of the sun. In some ways, wind turbines zone unit the character advancement of old windmills,
anyway presently they here and there have three cutting edges, that spin around a flat shape at the most elevated of a
steel tower. One in all the chief usual and vogue turbine styles could be a metal pinnacle with a three-sharp edge rotor
appeared in Figure 3.
30
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
Fig.3. NEG Micon 1.5MW Wind Turbine, 68m diameter Rotor, Tubular Steel Tower. Blades are made up
of fibreglass. The Rotor blades can be Turned to optimise power & the entire tower is rotated to face the wind.
Most breeze turbines begin creating power at twist speeds in the request of 3-4 meters for each second (m/s) (8 miles
for every hour); produce an "ostensible" greatest power at around 15 m/s (30 mph); and close to stop the harm of the
tempest at twenty-five m/s or more (50 mph).
a) Almost all the power on Earth is made with any rotating engine,
b) Turbine: change of the rectilinear stream development to the turn of the pole through pivoting streamlined
surfaces
Underneath Table 1 demonstrates the unique kinds of age and it is rotating the motor and Figure 4 indicates diverse
sorts of turbines.[18]
Table 1. Different types of generation and its Turbine type

5. Wind Turbine Technology:
Regardless of whether wind ranches can fit the new matrix codes relies upon the innovation of the breeze turbines.
There territory unit 3 basic styles of revolving engine|turbine}s utilised these days: the mounted speed twist turbine
with confine Induction Generator, the variable speed turbine with Doubly Fed Induction Generator and in this
manner the variable speed turbine with Synchronous Generator. The mounted speed confines Induction Generator
expends receptive power and can't add to voltage administration. Hence, however static capacitance administration
may allow twist ranches with this sort of generators to deliver responsive power, this sort of generators territory unit
bound to vanish from wind turbines. The variable speed turbine with Doubly Fed Induction Generator are regularly
management led to deliver recurrence and voltage control with a succeeding converter inside the rotor. Control code
redesign and equipment adjustments region unit essential, a considerable measure of precision, the convertorratings
may be expanded for recurrence reaction [39]. This kind of generator shows a few troubles in going through voltage
bounces, since the consequences of the voltage drop create high voltages and streams in the rotor circuit and,
therefore, the power converter can break. This can be the principal broadened variable speed turbine innovation and
creators as of now furnish this sort of twist turbines with blame ride-through abilities. The variable speed turbine
with Synchronous Generator is associated through a succeeding converter to the lattice. This gives most adaptability,
facultative full genuine and receptive power administration and blames ride-through capacity all through voltage
plunges. Once more, control code update and minor equipment change zone unit essential to add to the framework
soundness. Different elements like site particular load coordinating (when the yearly breeze profile correlates the
heap) and a high scope of twist turbines inside the power station encourage wash the task of the network.
31
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
Fig.4. Wind Turbine Types [18]

6. Power Curve:
Capacity Factor
a) The part of the year the turbine generator is working at evaluated (top) control
Limit Factor = Average Output/Peak Output ≈ 30%
b) CF depends on both the attributes of the turbine and the site qualities (ordinarily 0.3 or above for a
decent site)
Figure 5 demonstrates a power bend for a 1500 kW turbine.
Fig.5. (a) Power curve for a 1500 kW turbine, (b) Wind Frequency Distribution
7. Fundamental Equation of Wind Power:

Wind Power depends on the amount of air (volume), the speed of air (velocity), the mass of air (density),
flowing through the area of interest (flux)-
Kinetic Energy definition: Ke= 𝑚𝑣
Power is KE per unit time. p = 𝑚𝑣
Fluid mechanics gives mass flow rate (density * volume flux):
dm/dt = ρ* A * v
Thus: Power ~ cube of velocity
P = ρ𝐴𝑣 Power ~ air density
 Power ~ rotor swept area , [A= π𝑟2 ]
7.1 Wind speed-

Wind speed is one in everything about chief basic qualities in elective energy generation. Wind speed
changes in each time and house, controlled by a few components equal geographic and climatic conditions.
because of wind speed could be a variable parameter; estimated wind speed information regularly takes care of
exploitation connected science techniques. Wrongdoing waves commonly outline the diurnal varieties of normal
breeze speeds. As partner illustration, diurnal varieties of hourly breeze speed esteem, which are the run of the
mill figured qualities that help data in the vicinity of 1970 and 1984, in Dhahran, Asian country demonstrated
the curved design. The wind speeds progressed in the daytime and in this way the most velocity happens at
concerning threep.m., demonstrating that the daytime wind speed is relating to the nature of light. George et al.
reportable that breeze speed at urban concentration, Lone-Star State is close relentless all through dull hours,
32
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
and takes after a twisted illustration all through sunlight hours. A short time later, George et al. have
incontestable that diurnal breeze plans at five regions inside the prairie take after an illustration the same as that
educated in. Maintained the breeze speed was learning for the total 1970– 2003 from up to sixty-six inland areas
around the United Kingdom, Sinden has over that month to month average breeze speed is correspondingly
propositional to the month to ordinary month temperature, i.e. it's higher inside the winter and minor inside the
pre-summer. The most extreme breeze speed happens in the Gregorian date-book month and like this the base
in August. Hassan and Hill have reportable that the month-to-month assortment of low breeze speed regards
over the measure of 1970– 1984 at Dhahran, the Asian nation has exhibited the wavy illustration. Regardless,
as a result of the assortment in temperature at Dhahran is negligible over the whole year, there's no a clear
connection between's breeze speed and temperatures. The year-to-year assortment of yearly mean breeze speeds
depends extraordinarily on picked zones as there's no first association with anticipating it. Perhaps, alongside
various years, the yearly mean breeze speeds decrease all the technique from 1970 to 1983 at Dhahran, Saudi
Arabia. In the UK, this theatrical presentation in an outstandingly a lot of American state actuated matter for
the total 1970– 2003.
Mostly, A critical variable in the average yearly breeze speed over a 20-year time span (1978-1998) is
to be noted, and the more significant part of the base qualities begin from under 7.8 to very nearly 9.2 m/s. The
semipermanent learning of the breeze (1978-2007) got from the concise perception framework controlled by
the mechanical meteorological observatories was investigated and announced by KO et al. The outcomes
demonstrate that the change of the mean yearly breeze speed occurs at the exact destinations; it tends to diminish
somewhat on Jeju Island, while the 2 contradicting locales have irregular trends.[23-28]
7.2 Wind Direction:
Wind heading Wind course is one in everything about breeze attributes. Connected science learning of
twist headings over an expanded measure of your chance is amazingly vital inside the site decision of intensity
plant and subsequently the design of twist turbines inside the power plant. The climate graph chart might be a
formidable apparatus of dissecting wind learning that square measure concerning twist headings at a specific
location over a chose principal amount (year, season, month, week, and so on.). This round outline shows the
recurrence of twist headings in eight or sixteen foremost bearings. As partner degree case appeared in Fig. 6,
there square measure sixteen outspread lines inside the climate outline graph, with 22.5° except for each other.
The length of each line is corresponding to the recurrence of wind bearing. The repetition of quiet or near breeze
is given as assortment inside the focal circle. Some climate graph outlines may moreover contain the information
of wind speeds.[36-42]
Fig. 5. Windrose diagram for wind directions.

7.3 Wind turbine controls:
Wind turbine administration frameworks still assume fundamental parts for ensuring turbine stable
and safe task and to upgrade wind vitality catch. The most administration frameworks in an exceedingly
popular turbine body pitch administration, slow down administration (uninvolved and dynamic), yaw
administration, and others. Beneath wind speed conditions, the capacity yield from a turbine may surpass its
appraised worth. Along these lines, control administration is expected to manage the capacity yield among
reasonable American state uctuations for maintaining a strategic distance from rotating motor mischief and
settling the capacity yield. There are two essential administration routes inside the power administration: pitch
administration and slow down control. The turbine control framework is utilised to manage the capacity yield
among reasonable variances.
7.4 Wind turbine configuration:
33
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
A significant portion of the modern monstrous breeze turbines ar flat pivot turbines with for the most part
3 cutting edges. As appeared in Fig. 6, a turbine is included a walled in area, that is situated on the most
astounding of a breeze tower, lodging the principal rotating motor components inside. Three cutting edges
(not appeared) mounted on the rotor centre point, that is associated using the most shaft to the rigging case.
The rotor of the generator is related to the yield shaft of the apparatus case. In this way, the moderate turning
pace of the rotor centre point is misrepresented to a coveted high pivoting rate of the generator rotor.
Exploitation the pitch framework, each sharp edge is pitched independently to improve the approach of the
cutting edge for allowing the following vitality catch in the shared task and for shielding the revolving motor
components (edge, tower, and so forth.) from harming in crisis things. With the input info corresponding to
estimated moment wind course and speed from the weathervane, the yaw system provides the yaw
introduction administration for ensuring the rotational motor unendingly against the wind.[30-35]
Fig. 6. A horizontal-axis wind turbine configuration (horizontal-axis wind)
7.5 Wind turbine lifetime:

Present day wind turbines are intended for an administration lifetime of 20-30 years. Rotating motor
manufacturers and elective vitality plants confront the critical test of knowing the best approach to achieve benefit
life objectives while limiting support and repair costs. Notwithstanding, up operational trustworthiness and
expanding the lifetime of wind turbines are unpleasantly troublesome assignments for some reasons:
a) Wind turbines ought to be presented to changed antagonistic conditions worship extreme temperatures,
wind speed variances, moistness, dust, radiation, lightning, saltiness, and successive storms of rain, hail,
snow and sand.
b) The contemporary turbine includes a sizable measure of components and frameworks; everything about has
his own life. By the Cannikin law, disappointment ought to happen first inside the part or structure with the
briefest life.
c) A turbine is liable to a massive kind of great masses given twist vacillations in rapidity & course and
different begins and rests of the structure. Specific real mechanisms should face critical weakness masses.
d) Advanced high malleable and exhaustion safe materials are fundamental for a couple of critical components
of substantial fashionable breeze turbines due to the constant increment in cutting edge length, container
stature and turning motor weight.
e) As a rich building framework, a {wind turning engine|turbine} ought to be outlined at the framework level
rather than at the part/part level as a standard application for a couple of turbine makers.[29]
8. Wind Energy- Challenges:
8.1 Technical:
According to statics, the added substance wind age ranches ability in the Asian nation were found
around one,380 MW before 2002. Right now wind age represents eight.7 p.c of the put in control ability in the
Asian country anyway it exclusively contributes one. Six p.c of the office generated12. Indian breeze age as of
now includes a lower Plant proportion (PLF) contrasted with fuel, atomic and hydropower plants and it's
furthermore low if we tend to compare it and universal principles. The centre reason for this issue is because of
the more significant part of windage cultivates in the Asian nation are come to up to its charged sum and needs
repowering. Repowering won't exclusively encourages them to remain beneficial anyway also may create a shot
34
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
of intensity age capacity change to their best playing locales. According to ponders, repowering of late breeze
homesteads may build the breeze vitality PLFproportion significantly from fifteen p.c to thirty p.c. It's been
found that inside the nonappearance of right government arrangements structure and appropriations, a few
breeze age partnerships don't appear to will to repower their plants that are imperative to beat this impediment.
MNRE ought to got to inspire such late breeze ranches for repowering their capacity by supporting with brilliant
and log terms approaches.
8.2 Infrastructural:
Vacillations in network recurrence and voltage deliver challenges in powerhouse activities and cut
back the potential outcomes for fruitful breeze vitality lattice entrance, that is shown in reports13. Since of the
constraint of a network framework, it's been discovered that the quantity of vitality made from wind ranches
couldn't be viably transmitted all through to clients cause wastage of life. MNRE has detected this issue and
aggregated in its provide details regarding "Environmentally friendly power Energy Corridors", that
distinguishes the foundation interest for departure and transmission of the sustainable power source, together
with wind14 and additionally wanted Germany's participation to convey propelled matrix joining innovation
to the Asian nation.
8.3 Economic:
High borrowing prices in Asian country creates the obstacle for wind energy sector growth. The project
funding methodology applied for a majority of wind generation comes area unit planned with 70: thirty debt
equity magnitude relation, that additionally with high-interest rates that create a rich debt under
troublesome political economy conditions of the Asian country.
Other Issues in Wind Power Development in India:
a) Forest arrive freedom, e.g. in Karnataka,
b) Creation of a framework for control clearing and transmission offices,
c) Availability of land for wind ranches,
d) Potential arrive generally obtained by producers/engineers,
e) Implementation of updated levy according to CERC rules,
f) Development of planning and anticipating framework (industry and LDCs not yet wholly arranged),
g) Withdrawal of Accelerated Depreciation.
9. Trends in wind turbine Developments and wind power:
Wind revolving motor innovation has been produced by continually upgrading turbine style, rising turning
motor execution, and improving general rotating motor efficiency. There are numerous ages of advancement and change
in turbine innovation, focusing on cutting edges, generators, coordinate drive methods, pitch and yaw administration
frameworks, et cetera. to deliver extra power from twist innovation inside the following numerous decades, it needs:
a) Developing inventive procedures
b) Decreasing wind turbine costs through innovation progression
c) Optimizing fabricating forms and upgrading producing activities
d) Improving wind turbine execution and efficiency
e) Reducing working and support costs
f) Expanding wind turbine generation limits. The present significant patterns in the advancement of
wind turbines are towards a higher power, higher productivity and consistent quality, and lower
cost per kilowatt machines.
10. Conclusion:
A sustainable solution is evident that the utilisation of wind energy as a permanent resolution to this
world energy considerations may well be property. Even so, conditions for the property are evaluated. As a
result, albeit the resource in its current state of technology is useful enough to be able
to support numerous developments within the business, achievements of vast technological opportunities might find
yourself creating the resource unlimited. At the financial level, wind energy has proved to be not solely
environmentally however additionally socially profitable to financially reinforce wind business whereasceasing to price
competition. Many governments square measure of the read that the wind businesses ready to require up to the opened
business, with a new certificate market taking over all the favour. Even so, about the little market, there ought to be the
maintenance of a set value system. Socially, the reality that the wind business is taking part in native development
35
Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
encourages for its property. Besides, its checked authentic influence on the native inhabitants
might facilitate in crippling the general public temperament. Finally, it's necessary to push
for more analysis regarding potential environmental analysis. It is, therefore, judicious to 1st rethink results of
studies associated ecological impact analysis once thinking of golf shot up a replacement power plant or reconsidering
a previous one.
Acknowledgements
The authors would like to dedicate this paper to His Parents for his valuable guidance and continuous
encouragement to successfully completion of this work.
Reference
[1] Wang, X.C., Guo, P. also, Huang, X.B. (2011) A Review of Wind Power Forecasting Models. Vitality Procedia, 12, 770-
778. http://dx.doi.org/10.1016/j.egypro.2011.10.103
[2] Zhao, D.M., Zhu, Y.C. also, Zhang, X. (2011) Research on Wind Power Forecasting in Wind Farms. Procedures of the
2011 IEEE Power Engineering and Automation Conference, Wuhan, 8-9 September 2011, 175-178.
http://dx.doi.org/10.1109/PEAM.2011.6134829
[3] Sideratos, G. also, Hatziargyriou, N.D. (2007) An Advanced Statistical Method for Wind Power Forecasting. IEEE
Transactions on Power Systems, 22, 258-265. http://dx.doi.org/10.1109/TPWRS.2006.889078
[4] Ma, L., Luan, S.Y., Jiang, C.W., Liu, H L. also, Zhang, Y. (2009) A Review on the Forecasting of Wind Speed and
Generated Power. Inexhaustible and Sustainable Energy Reviews, 13, 915-920.
http://dx.doi.org/10.1016/j.rser.2008.02.002
[5] Lange, M. what's more, Focken, U. (2008) New Developments in Wind Energy Forecasting. Procedures of the 2008
IEEE Power and Energy Society General Meeting, Pittsburgh, 20-24 July 2008, 1-8.
[6] Zhao, X., Wang, S.X. What's more, Li, T. (2011) Review of Evaluation Criteria and Main Methods of Wind Power
Forecasting? Vitality Procedia, 12, 761-769. http://dx.doi.org/10.1016/j.egypro.2011.10.102
[7] Chong Han, Alex Q. Huang, Wayne Litzenberger, Loren Anderson, Abdel-Aty Edris "STATCOM Impact Study on the
Integration of a Large Wind Farm into a Weak Loop Power System", (1266-1272), 2006.
[8] Katsuhisa Yoshimoto, Toshiya. Nanahara, Gentaro Koshimizu, Yoshihsa Uchida "New Control Method for Regulating
State-of-Charge of a Battery in Hybrid Wind Power/Battery Energy Storage System", (1244-1251), 2006.
[9] http://www.wwindea.org
[10] EWEA yearly report, "Winning with European Wind Creating power, helping the earth", 2008.
[11] B.C. Ummels, E. Pelgrum, W.L. Kling "Joining of huge scale wind power and utilisation of vitality stockpiling in the
Netherlands' power supply", 2007.
[12] Dany G. "Power save in interconnected frameworks with high breeze control generation", 2001.
[13] H. Holttinen et al. "Outline and Operation of Power Systems with Large Amounts of Wind Power, first aftereffects of
IEA joint effort", 2007.
[14] Mary Black, Goran Strbac "Estimation of capacity in giving adjusting administrations to power age frameworks with
high wind penetration", 2005.
[15] WWEA, "World breeze vitality report 2008", South Korea, 2009.
[16] http://www.pikeresearch.com/inquire about/vitality stockpiling technology markets
[17] Magnus Korpås "Appropriated Energy Systems with Wind Power and EnergyStorage",2004
[18] Wind Power Wind Power Fundamentals Presented by Alex Kalmikov and Katherine Dykes With commitments from
Araujo PhD Candidates, MIT Mechanical Engineering,
[19] MNRE got to at http://bit.ly/14woYJX (2013)
[20] http://bit.ly/15WqKDm (2013)
[21] India wind vitality standpoint 2012, got to at http://www.gwec.net/wp-content/transfers/2012/11/IndiaWind-Energy-
Outlook-2012.pdf
[22] A Review of Wind Energy Scenario in India, International Research Journal of Environment Sciences, ISSN 2319– 1414
Vol. 3(4), 87-92, April (2014) Int. Res. J. Condition Sci.
[23] [Siddiqi, A.H., Khan, S. and Rehman, S., Wind speed reenactment utilising wavelets. American Journal of Applied
Sciences, 2(2), pp. 557– 564, 2005.
[24] George, J.M., Peterson, R.E., Lee, J.A. and Wilson, G.R., Modeling wind and relative dampness impacts on air quality.
Int. Claim to fame Conf. on Aerosols and Atmospheric Optics: Radiative Balance and Visual Air Quality, Snowbird,
Utah,
36
View publication stats

Vol.4, No.2, pp.29-37, July 2018
ISSN:2456-0065 DOI: 10.21058/gjet.2018.42004
[25] George, J.M., Wilson, G.R. and Vining, R.C., Modeling hourly and day by day wind and relative dampness. Int. Conf. on
Air Pollution from Agricultural Operations, pp. 183– 190, Ames, Iowa, 1996.
[26] Sinden, G., Characteristics of the UK wind asset: long-haul examples and relationship to power request. Vitality Policy,
35(1), pp. 112– 127, 2007.
[27] European Wind Energy Association. Wind vitality – the certainties, part I: Technology, The yearly fluctuation of wind
speed 2009
[28] [Hassanm, Y.H. and Hill, D.R., Islamic Technology: An Illustrated History, Cambridge University Press, 1986.
[29] Fundamentals of wind vitality Wei Tong Kollmorgen Corporation, Virginia, USA. Mind Transactions on State of the Art
in Science and Engineering, Vol 44, © 2010 WIT Press
[30] [Adam, D., Floating breeze turbines ready to outfit sea winds,
http://www.guardian.co.uk/condition/2008/jul/16/windpower. renewable energy
[31] Sway, Principles of the idea, http://sway.no/index.php?id=16
[32] [Vanderover, J.S. and Visser, K.D., Analysis of a contra-pivoting propeller driven transport airship,
ftp://ftp.clarkon.edu/.depts/mae/public_html/papers/vanderover.pdf
[33] Schönball, W., Electrical generator course of action, US Patent 3,974,396, 1976.
[34] McCombs, J.C., Machine for changing over breeze vitality to electrical vitality, US Patent 5,506,453, 1996.
[35] Wachinski, A., Drive gadget for a windmill gave two counter-rotative propellers, US Patent 7,384,239, 2008.
[36] Shin, C. and Hur, M.C., Over-drive equip gadget, US Patent 5,222,924, 1993.
[37] Shin, C., Multi-unit rotor cutting-edge framework coordinated breeze turbine, US Patent 5,876,181, 1999.
[38] Kowintec Corp., http://www.kowintec.com/english/items/main.html
[39] Korane, K.J., Flexible apparatuses support wind-turbine dependability. Machine Design, 79(15), pp. 24– 28, 2007.
[40] Fox, G., Epicyclic apparatus framework, US Patent 7,056,259, 2006.
[41] Musial, W., Butterfi eld, S. and McNiff, B., Improve wind turbine gearbox dependability. 2007 European Wind Energy
Conf. , Milan, Italy, 2007.
[42] Gburˇ sick, P., Gburˇ sick, V., Gavrilov, M., Srdanoviˇ c, V. and Mastilovic', S., Complementary administrations of sun
oriented and twist vitality in Serbia. Geographica Pannonica, 10, pp. 22– 25, 2006.
37

Trabajo de Investigación Formativa

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Trabajo de Investigación Formativa

Încărcat de

Drepturi de autor:

Formate disponibile

TERMODINAMICA I

UNIVERIDAD NACIONAL DE TRUJILLO

UNIVERSIDAD NACIONAL DE TRUJILLO

Kumar, Raja; Singh, S; Srivastava, R; Saket, R. (14 de 09 de 2019). Dynamic

Lui, Q. (17 de 07 de 2018). Optimization of Wind Fan Selection and Operation.

Smith, Craig; Barthelmie, R; Pryor, S. (16 de 07 de 2013). In situ observations of the

Thresher, R. (14 de 04 de 2018). Taking advantage of modern turbines. Obtenido de

PAPER • OPEN ACCESS

Influence of wind power grid connection on power quality and reactive

View the article online for updates and enhancements.

This content was downloaded from IP address 181.64.57.90 on 13/09/2020 at 16:27

Influence of wind power grid connection on power quality and

2. Wind speed model and its influence on grid connection

2.1. Wind speed model

Figure 1. Gust Figure 2. Gradient wind Figure 3. Random wind model

2.2. Influence of wind speed change on grid connection

(a) Wind speed (b) Active power

(c) Reactive power

3. Reactive power compensation device for wind farm

3.1. Wind farm management based on static var compensator (SVC)

Figure 6. Wind farm simulation model with SVC

3.2. Wind farm management based on static synchronous compensator (STATCOM)

Figure 12. Circuit of STATCOM with three-arm voltage type

Figure 13. Equivalent circuit and vector diagram of STATCOM

Figure 14. Wind farm simulation model with STATCOM

PAPER • OPEN ACCESS Related content

Maintenance in the Wind Power Stations - Discussion of Several Key Technologies

- Maintenance Optimization Model for One

View the article online for updates and enhancements.

This content was downloaded from IP address 181.64.57.90 on 13/09/2020 at 16:28

Optimization of Wind Fan Selection and Operation

* Corresponding author e-mail: liuqinghua@bitc.edu.cn

2. Assessment of wind energy resources and its utilization

2.1. Approximate calculation of wind speed distribution

Figure 1. Waybill distribution curve of wind speed.

2.2. Assessment of wind energy resources

2.3. Assessment of the utilization of wind energy resources

Figure 3. Monthly actual utilization of wind energy resources

3. The Model of fan selection

3.1. The establishment of the model

3.2. The solution of the model

Table 1. Wind speed distribution parameters at six observational points.

4. planning model for maintenance scheduling

Table 3. Schedule of maintenance personnel in July.

LETTER • OPEN ACCESS Related content

- Simulated wind farm wake sensitivity to

- Sensitivity of Wind Turbine Array

This content was downloaded from IP address 181.65.155.130 on 12/09/2020 at 20:31

In situ observations of the influence of a

Received 19 March 2013

1. Introduction layer and surface-atmosphere fluxes of momentum, sensible

1748-9326/13/034006+09$33.00 1 c 2013 IOP Publishing Ltd Printed in the UK

gradient is present during either the day or night according

that the mean hub-height velocity deficit derived from these

conditions. However, consistent with the median TI values

Figure 5. Vertical profiles of (a) wind speed (u) (ms−1 ),

PAPER • OPEN ACCESS

Cluster Wind Power Uncertainty Model and Operation Simulation

View the article online for updates and enhancements.

This content was downloaded from IP address 181.64.57.90 on 13/09/2020 at 15:53

Cluster Wind Power Uncertainty Model and Operation

Xuxia Li1, Yingying Hu2, Yao Wang3, Jiaojiao Deng4

2. Probability sequence modelling of power system generation and load dependence

p The average electromagnetic torque (T(t)) is