710 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 26, NO.

2, MAY 2011

Enhancing Frequency Response Control by DFIGs

in the High Wind Penetrated Power Systems
Le-Ren Chang-Chien, Member, IEEE, Wei-Ting Lin, and Yao-Ching Yin

Abstract—As wind power retains a significant proportion of gen- their grid codes for this operating requirement. For example,
eration mix in the electric system, we foresee that less fossil fuel ESB national grid has updated the grid code to set out rules for
units will be on service. Under this situation, it may impact system’s wind power to provide balancing services which were originally
frequency security due to the lack of frequency support from units.
To make up for such a system change, wind turbines should ac- supported from conventional plants, such as on-load governor
tively provide frequency response upon request. For wind turbines response, set-point control, frequency response, and ramp rate
to achieve this goal, active power output of the wind turbine should [2], [3]. Regarding the operating reserve, the grid operation may
be controllable so that the generating margin could be preserved require that wind power should provide reserve of Delta produc-
for frequency response. Therefore, the presentation of this paper tion to account for the situation when spinning reserve is inad-
will begin with the demonstration of the control scheme for the
doubly fed induction generator (DFIG). Simulation results vali- equate from conventional plants [4], [5].
date the effectiveness of the methodology in regulating wind power To realize frequency response from wind turbines, control of
upon operator’s request, while stability is assured under variant active power output in response to system frequency is a neces-
wind speeds. Following that, this paper will explore several oper- sity. Recently, the active power control of the variable speed
ating strategies for DFIGs to support system frequency with less wind turbines is becoming more mature with the advanced
sacrifice in the wind energy production. Simulation results show
that the proposed strategies can enhance regulation performance power electronic technology. Topics about frequency control in
in the high wind penetrated power system. wind turbines could be found in some literatures. References
[6]–[10] introduced the inertial and speed-droop controls of
Index Terms—Doubly fed induction generator (DFIG), fre-
quency response, regulation, wind power. wind turbines for frequency response. The developed control
loop within the wind turbine is like the speed governor which
automatically adjusts power output in response to a frequency
I. INTRODUCTION drop. However, due to the wind variability, the speed-droop
control has its own limitation. The effectiveness of the pro-

W IND energy is one of the prominent renewable supplies

for the electrical system around the world. For areas
with abundant wind energy potentials, wind generation is grad-
posed control scheme is dependent on the operating stability
and available generating margin.
Because wind power is an intermittent motive source which
ually replacing the capacity of the fossil-fuel generating plants differs from the source applying to the traditional plants, the
[1]. Whenever wind turbines gradually replace the traditional control of wind turbine output is much more challenging. For
plants under this trend, system operators will encounter the fre- example, when an operator’s request is beyond the limit of the
quency control challenge. Currently, most of the wind farms available wind power, stability problem may arise when rotor’s
are operating in the maximum power production for pursuing speed is not running within the stable operating region [11].
the maximum economic benefit. However, this type of opera- Therefore, an appropriate coordination between stability and
tion does not help to preserve the generating margin that is re- controllability of active power in wind turbines should be fur-
sponding to frequency. The situation may be worse in an iso- ther defined.
lated island operation. Especially when wind power production Coming from the similar idea of the traditional units, the
is relatively high during the off-peak hours, fossil fuel units speed-droop response control requires the wind turbine to pre-
may need to be forced offline due to the excessive wind power. serve generating margin. Such a strategy inevitably sacrifices
This situation results in less frequency response reserve in the the maximum power production if the storage system is not
system. For security concern, wind turbines are expected to pro- considered. However, the high installation cost of the storage
vide frequency response. Therefore, some utilities have updated system in a large scale wind farm may not have its economical
interest. This situation inspires us to contemplate a new strategy
for wind turbines to provide regulation service with less sacri-
Manuscript received October 25, 2009; revised February 20, 2010; accepted
April 30, 2010. Date of publication July 19, 2010; date of current version April
fice in wind energy production.
22, 2011. This work was supported by the National Science Council of Taiwan, Discussions of the frequency response control on wind tur-
R.O.C. under grants: NSC 96-2628-E-006-251-MY3 and NSC 98-3114-E-006- bines will start from the active power control of a doubly fed
011-. Paper no. TPWRS-00840-2009.
L.-R. Chang-Chien and W.-T. Lin are with the Department of Electrical En-
induction generator (DFIG). As operators expect wind turbines
gineering, National Cheng Kung University, Tainan, Taiwan. to provide frequency response for the system, the control of the
Y.-C. Yin is with the Power Division of Taiwan Semiconductor Manufac- DFIG should be able to trace operator’s command while sta-
turing Company, Tainan, Taiwan. bility is also maintained in all wind conditions [12]. After going
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. through a comprehensive introduction of the control scheme,
Digital Object Identifier 10.1109/TPWRS.2010.2052402 this paper will lead to the next discussion on the strategies for
0885-8950/$26.00 © 2010 IEEE
CHANG-CHIEN et al.: ENHANCING FREQUENCY RESPONSE CONTROL BY DFIGS 711

Fig. 1. Active power control scheme for a DFIG.

DFIGs to provide frequency regulation, which is the main theme

of this paper. Testing results from simulation validate the pro-
posed operating strategies for enhancing the frequency quality. Fig. 2. Performance of the command tracing when the DFIG is operating in FR
Investigation on unit output reversals also confirms the contri- mode at medium wind condition. (a) Wind velocity. (b) Output power tracking.
(c) Rotor speed.
bution of wind frequency response in alleviating regulation bur-
dens on traditional plants.

II. ACTIVE POWER CONTROL OF A DFIG speed [Fig. 2(c)] running up and down. So long as the rotor is
running within the allowable operating range, the output power
Currently, the general approach of active power control on
(solid line in Fig. 2(b)) of the DFIG was successfully tracing the
a DFIG is to use the back-to-back converter for regulating the
power command (dash line in Fig. 2(b)).
d-axis rotor current, . With proper configuration, regu-
lates the electromagnetic torque that impedes the turbine torque,
B. High Wind Condition—Pitch Angle Control Mode
which in turn determines the rotor speed. The regulated electro- (PAC Mode)
magnetic torque with the controlled rotor speed determines the
active power output of a DFIG [13]–[16]. It is not easy to make the constant power control and keep a
DFIG running at stable condition. If is not properly set,
A. Medium Wind Condition—Free Running Mode (FR Mode) rotor speed may run out of the upper or lower speed limit and
Fig. 1 demonstrates the control scheme for a DFIG to trace cause stability problem [11]. Therefore, some actions should be
operator’s command. A set point command, , is first sent made to avoid this situation. For example, if wind speed keeps
to the DFIG. With the rotor speed feedback, a torque command, on rising when the DFIG is tracing the constant power command
, is calculated for generating the direct axis rotor current . with a fixed pitch angle, relation of will make
With proper reference frame arrangement, will come up with increase. When is about to reach the upper speed limit,
an electromagnetic torque, , against turbine’s mechanical should be reduced by decreasing . An important way to
torque, , at some rotor speed. Due to the fixed pitch angle decrease is to change wind acquisition by adjusting the
of the turbine blades, the mechanical power may not be the same blade pitch angle, . When is approaching the upper speed
as the commanded power. Consequently, the torque difference limit, an upper speed reference, , is taken to compare
between the and will make the rotor accelerate or with . If the error signal is positive, increases to reduce
decelerate according to the following relation: . Consequently, will be kept below the upper speed
limit, . It is noted that the speed regulator is designed
(1) using the proportional control (P controller). As wind power
decreases, will gradually reduce to 0 degree and operation
where is the rotor speed. Change of rotor speed could affect is switched back to FR mode. Since this kind of operation is
turbine’s tip speed ratio and then change the power coefficient, to adjust the pitch angle of turbine blades, it is named as pitch
which further results in a change of power acquisition from the angle control mode. This idea is also used to limit the maximum
DFIG [14], [17]. If equilibrium between and could wind power acquisition when wind speed may boost up power
be reached, rotor speed will rest at a new state. output above the nameplate rating. The schematic diagram of
Equation (1) infers that the moment of inertia in the turbine- pitch angle control is presented in Fig. 3. It is noted that with
generator system acts to absorb or release the energy difference P controller, a proper speed margin between and
caused by the wind fluctuation. However, rotor speed may be should be carefully chosen to ensure that will not exceed
drifting in a wide range. Fig. 2 presents the performance of the . In our controller model, and are set 1.2
constant power output control with a medium wind velocity. p.u and 1.3 p.u., respectively. The gain of P controller is set
The fluctuating behavior of wind speed [Fig. 2(a)] makes rotor between 3 and 10 to suit for a proper pitch response. The other
712 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 26, NO. 2, MAY 2011

Fig. 3. Schematic diagram of the proposed pitch angle control.

parameters of pitch angle controller are: equals to 0.11 and

equals to 10.

C. Low Wind Condition—Modified Linear Slope Control

Mode (MLSC Mode)
If the available wind power is lower than DFIG’s command,
maintaining the DFIG in stable condition is the first priority. As
wind power is lower than operator’s set point , should
be reduced with the decreasing . Therefore, a linear slope
Fig. 4. Modified linear slope control (MLSC) strategy.
strategy is applied to prevent further reduction of rotor speed.
Fig. 4 shows how the modified linear slope control (MLSC) is
used for the case of low wind condition. As wind power might
be lower than due to the low wind condition, MLSC ensures
the following relation:

(2)

where is the rotor speed corresponding to the set point .

is defined as

(3)

where and are the cut-in rotor speed and cut-in

power of the DFIG, respectively. Equation (2) shows that
is adjusted by the change of rotor speed . When is lower
than , the second term in the right-hand side of the equal sign
in (2) becomes negative, which will reduce . Consequently,
the reduced will reduce . It is noted that line Fig. 5. Schematic diagram of the coordination for the proposed control strate-
gies.
should intersect the right half side of the low-wind power curves
to ensure the sufficient torque reduction. The idea of the linear
slope strategy is to automatically adjust counter-torque for re- The FR and PAC modes ensure the steady power tracing for
straining the drift of rotor speed so that stability is well main- the medium and high wind speeds. When wind drops to the low
tained. This idea is inspired from [11], whose idea is meant to speed range, MLSC mode is activated for stabilizing the DFIG.
restrict rotor speed running within a narrower range. However, To coordinate these three control modes, setting should
the linear slope strategy in [11] is primarily used for smoothing comply with the following rule:
wind power fluctuation around the set point . The MLSC
mode introduced in this paper is to further define the linear slope
term in (2)—see (3)—so as to ensure stable control of rotor
(4)
speed when the set point is higher than the available wind
power. It is noted that the linear slope calculated from (3) is When is approaching , MLSC mode is activated to re-
definitely larger than that in [11] for the stability concern. duce for stabilizing the DFIG.
The pitch angle of the DFIG is kept at 0 degrees in both
III. COORDINATION OF THE CONTROL SCHEMES MLSC and FR modes. One reason is that fixing pitch angle
FOR ALL WIND-VELOCITY REGIONS could avoid excessive pitch angle adjustments, and avoid more
The proposed control modes are designed to serve both pur- wear and tear. The other reason is that keeping pitch angle at 0
poses in command tracing and stable operation under various degrees not only optimizes power coefficient at some tip speed
wind speeds. The schematic diagram for controls in all wind-ve- ratios but also makes the constant power control easier. Table I
locity regions is shown in Fig. 5. Fig. 5 depicts that the execution concludes the operating conditions for DFIGs to perform the
of control modes is determined by the measured rotor speed. proposed strategies. The performance of the DFIG operating at
CHANG-CHIEN et al.: ENHANCING FREQUENCY RESPONSE CONTROL BY DFIGS 713

TABLE I
CONDITIONS FOR DFIGS TO OPERATE IN THE INTRODUCED THREE CONTROL MODES

Fig. 7. Schematic diagram of the frequency response control on a DFIG.

from the wind source. Therefore, a detailed investigation on the

tradeoff between regulation and energy production will be dis-
Fig. 6. Performance of the DFIG operating in all wind conditions. (a) Wind cussed in the following sections. Before we move further, a fre-
velocity. (b) Power tracking curves. (c) Rotor speed. (d) Pitch angle.
quency response control scheme is first introduced.

A. Mechanism of Frequency Response Control in the DFIG

all wind conditions is demonstrated in Fig. 6. The DFIG was
tested under medium, low, and high wind speeds subsequently Fig. 7 illustrates the frequency response control on a DFIG.
as shown in Fig. 6(a). Fig. 6(b) shows that the DFIG output As the DFIG receives a command , this command is ad-
could keep on tracing the power command except the MLSC justed by the speed-droop response to come up with a set
mode. During this operation mode, because setting was too point . This reference signal is then turning into a refer-
high, the DFIG’s command was then modified to match the low ence torque , which is further adjusted by the inertial re-
wind power for maintaining stability. That is why the power sponse loop. The adjusted signal then drives the DFIG to feed
fluctuations were obvious during the operation at MLSC mode. the grid with output . For the given wind speed, the gener-
Fig. 6(c) shows that as rotor speed was running close to the ating margin of the DFIG is calculated as follows:
upper speed limit , blade pitch controller started to in-
(5)
crease the pitch angle as shown in Fig. 6(d) for restraining the
increase of rotor speed. Fig. 6(c) also indicates that the large The maximum power output of the DFIG is esti-
rotor speed range may put extra burden on the converter design. mated by the following equation:
Such a wide speed range will require higher rotor voltages and
consequently higher dc voltages. (6)

IV. CONTINUOUS FREQUENCY REGULATION where is wind speed, is air density, is the cross section
PROVIDED FROM WIND POWER of rotor for the DFIG, and is the power coefficient which is
Units participating the regulation service, no matter what is the function of pitch angle and tip speed ratio .
meant for continuous frequency regulation or for contingency, It is noted that should be lower than to pre-
should reserve a certain amount of generating margin for gov- serve . Assume the amount of will be used
ernor actions. The similar idea also applies to the wind turbines for supporting the off nominal frequency band by , say
which are selected to provide frequency regulation. In order to , the following equation expresses how the equivalent
provide generating margin, the deloaded operation is necessary. speed-droop is set according to the preset frequency
Generally, the larger the deloaded margin, the better the regu- band and reserve:
lation ability. However, such a deloaded operation may contra-
dict wind power’s operating principle for acquiring more power (7)
714 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 26, NO. 2, MAY 2011

TABLE II
CONDITIONS FOR A DFIG TO OPERATE IN EITHER THE FIXED OR VARIABLE SPEED-DROOP CONTROL

step is to investigate how the preserved generating margin

affects the frequency regulation capability. In order to keep the
wind power procurement as high as possible, we propose to set
operator’s command that follows the wind trend by using the
moving average estimation of turbine output with multiples of
deviation factor. The measure of moving average with multiple
deviation is commonly adopted to predict the moving trend of
a short-term load following or to allocate spinning reserve in
the utilities [18], [19].
The wind farm’s command signal is set as follows:

(10)

Fig. 8. Load frequency control loop for the frequency regulation test, system where is the moving average of the maximum wind tur-
parameters: R = 0 05 : ,T =02s: ,F =03 : ,T = 7s ,T = bine output in the averaging interval , is the
03s
: ,H = 5s = 1
,D .
weighting number. is the deviation according to the av-
erage value of
As a result, as system frequency deviates at , speed-droop
will give out the power adjustment as follows: (11)
(8) where is calculated as

Consequently, the set point of the DFIG is calculated as

follows: (12)

(9)
According to a normal possibility distribution, the multiple
The proposed variable speed-droop mechanism is similar to deviation metric could determine the possibility of wind reserve
those equipped within the traditional unit and those in [9] and to be realized during the interval. A higher setting ensures
[10]. The advance of our proposed method is that is higher possibility of wind reserve to be allocated. However, the
changing with . According to (7), variable speed-droop reserve quantity is uncertain.
insures to contribute the overall to the sup-
port of frequency band within . A fixed droop setting C. Simulation Results of Wind Frequency Regulation
does not make the same way. For a better speed-droop response, A simulation study was conducted for examining the influ-
a better fixed-droop setting may come from tedious testings by ence of K weighting in supporting the frequency response. For
trial and error. Table II lists the comparison between the vari- this purpose, a simulated power system with load frequency
able speed-droop and fixed speed-droop from simulation. In this control regulated by a representative mix of a conventional plant
case, the fixed speed-droop was set 5% for a comparison with a and a wind farm was created as Fig. 8. The conventional steam
variable speed-droop setting. It is noted that using the variable turbine was built based on [20]. The scenario was to simulate an
speed-droop mechanism, the number of unit output reversals as isolated island system, which has abundant wind resource but
well as RMS of frequency deviation are less than those with limited frequency support from fossil fuel units. The wind
using the fixed speed-droop mechanism. It is because penetration to the generation mix in the system was about 30%.
may not be totally utilized to support the frequency band The system inertia and speed-droop response were proportion-
when the fixed speed-droop is applied. ally reduced by 30% compared to the case with 100% traditional
units. The moving average interval was set to 1 min.
B. Investigation on Setting of Wind Reserve for Continuous Because the simulation was conducted to observe the effect
Frequency Regulation of wind regulation on system frequency, the variation of wind
In most cases, DFIGs are operating in the maximum wind power was designed a little larger than the variation of system
power production unless some security constrains apply. If load. Using (10), different K’s were assigned. Simulation re-
some wind turbines are selected to provide frequency response sults with K’s equal to 0 and 3 are shown in Figs. 9 and 10, re-
for system, these wind turbines need to be deloaded instead spectively. Figs. 9(a) and 10(a) show the total generation in the
of running at the maximum wind power production. The next system, wind farm output, and traditional unit output. The base
CHANG-CHIEN et al.: ENHANCING FREQUENCY RESPONSE CONTROL BY DFIGS 715

TABLE III
SIMULATION RESULTS OF DIFFERENT K WEIGHTING

10(c) show the retained wind reserve for the system. It indicates
that the more K was assigned, the higher possibility of more
wind reserve was kept for the system. Figs. 9(d) and 10(d) show
the speed-droop response ( in 1000-MW
base) of DFIGs. The response curve shows that the more wind
reserve is retained, the more frequency response is contributed
from the DFIGs. Summaries of the simulation results are listed
in Table III.
Table III lists the influence of different K settings on system’s
frequency, units’ output reversals, and wind energy production.
The data indicate that a higher K setting gives the wind turbine
more confidence to provide regulation for reducing unit’s output
reversals as well as frequency deviation. However, more sacri-
fice in wind energy production is inevitable. Therefore, there is
tradeoff between wind regulation and energy production. Oper-
ators are advised to adaptively tune the K setting for the required
frequency response subject to a satisfactory wind production.
Another benefit of K setting is relating to the alleviation of the
Fig. 9. Simulation results with K = 0. impact from wind’s abrupt ramp rate. The observation was made
from Fig. 10(b). A larger K deviation alleviates wind ramp im-
pact on the system. According to the wind operating experience,
many wind penetrated systems suffer from the steep wind ramp
over 10-min time frame [21]. Our tests in the larger K devia-
tion cases prove its effectiveness in smoothing the ramping wind
power. It is foreseen that the smoothing on the down ramping
wind power is not as effective as that on the up ramping wind
power due to DFIGs’ stability concern.

V. OCCASION FOR ACTIVATING WIND

FREQUENCY REGULATION

A. Impact of Wind Power Variation on System Frequency

Another observation we found in wind operation is that wind
power may adversely impact system frequency if wind power
variation is larger than that of system load. This phenomenon
happens more often in the island [22], [23]. Another simulation
was conducted to gauge such an impact. Table IV summarizes
the results of the simulation scenario when the variation of wind
Fig. 10. Simulation results with K = 3. power is larger than that of system load. A counterpart situation
is shown in Table V. From Tables IV and V, the RMS value
of (0.114) in Table IV is larger than that of (0.073)
of the generating power is 1000 MW per one p.u. The star sym- in Table V. A relatively smaller RMS value of indicates a
bols marked in the solid line indicate the occurrence of units’ better frequency performance in the system. It confirms the im-
output reversals (turnovers). Generally, the higher number of pact of wind gust on system frequency if wind power fluctuation
unit output reversals indicates that more unit output trends are is relatively larger than load variation. Another inference we ob-
reversed and thus cause more wear and tear. Figs. 9(b) and 10(b) tained from Tables IV and V is that when wind regulation is ac-
show the maximum wind farm output estimation and the real tivated, the RMS values of are smaller than those without
wind farm output. Since we used to control the DFIG wind regulation. However, the wind regulation effect does not
output, the real output was lower than maximum wind farm show its significance in Table V ( reduction from
output estimation. It also indicates that the higher the K as- 0.073 to 0.072) compared to the result in Table IV (
signed, the lower the wind farm output obtained. Figs. 9(c) and reduction from 0.114 to 0.053). The result makes us conclude
716 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 26, NO. 2, MAY 2011

TABLE IV
WIND FREQUENCY REGULATION TEST AS WIND POWER VARIATION IS LARGER THAN LOAD VARIATION

TABLE V
WIND FREQUENCY REGULATION TEST AS WIND POWER VARIATION IS SMALLER THAN LOAD VARIATION

that the moment for wind power to provide frequency regula-

tion is selective. As variation of wind power is larger than that
of system load, there is a good opportunity for wind to provide
regulation. Regarding to the reduction of output reversals made
from traditional units, both results of Tables IV and V indicate
that wind regulation can make significant reduction. It is be-
cause the responses of wind turbines are faster than those of
fossil fuel units. However, the reduction of wind energy pro-
duction is inevitable as wind regulation is activated even when
K was set to 0.
B. Variance Decision Control
Tables IV and V imply that variation indices of both systems’
load and wind power can be used to determine another right
moment for wind farms to activate speed-droop control. To im-
plement this concept, both variance indices of wind power and
system load should be calculated and then compared with each
other. The variance index is defined in the following equation:

(13)
Fig. 11. Decision flow chart for activating wind frequency regulation.

where is the variance index, is the th sample, is the

average of samples, and is the total number of samples. is
the variable that comes from the maximum wind power estimate the solid line, the operation was switched to speed-droop reg-
(MWPE) or system load. ulation mode. The sequence of the switching signal is shown
If the variance of MWPE is larger than that of system load, in Fig. 12(b). Fig. 12(c) shows the real-time variance track of
DFIGs should follow the command setting derived from (10) wind farm output. The wind farm output variance (solid line)
and activate speed-droop for regulation. Otherwise, the speed- was kept below the load variance (dash line) most of the time. It
droop should remain deactivated for DFIGs to pursue maximum indirectly confirms the effectiveness of variance decision con-
wind power procurement. The decision flow chart is illustrated trol in alleviating possible impacts on system frequency due to
in Fig. 11. This control can be coordinated between the system the gusty winds.
operator and the DFIG power commanding system. Generally,
system load data reside with the operator’s side. The system op- VI. GET READY FOR SYSTEM CONTINGENCY
erator is likely in the best position to send control mode requests
to the individual plants. The individual plants then execute either Equation (10) not only is useful in the continuous frequency
maximum output mode or frequency regulation mode according regulation but also is useful in supporting the contingent event.
to (6) or (10), respectively. For instance, when wind power production is relatively high
compared with system load at the off-peak hours, in order to
C. Simulation Results keep minimum fossil fuel units online, extraneous wind power
A simulation test was conducted to verify the effectiveness may be curtailed for security concern. Therefore, operators may
of the proposed variance decision control. Fig. 12(a) shows the expect that the excessive wind power could serve as reserve for
real-time variance estimate of system load (dash line) and the contingent event [4].
real-time variance estimate of MWPE (solid line). As the dash The only difference between the wind contingency reserve
line was relatively higher, DFIGs were operating at the max- and wind frequency regulation reserve is the setting of devia-
imum power output mode. As the dash line became lower than tion K in (10). To prepare for system contingency, more wind
CHANG-CHIEN et al.: ENHANCING FREQUENCY RESPONSE CONTROL BY DFIGS 717

power reserve, which is designed for providing the frequency

response to the system. Simulation results validate its contri-
bution to system’s frequency regulation at a small expense on
the wind power procurement. In addition, the variance decision
control introduced in this paper provides an operating example
to activate wind frequency regulation as gusty winds may ad-
versely impact system frequency. Related simulation confirms
its contributions to the frequency quality and to the relief on
unit’s regulation stress. Our next step is planning to implement
our regulation methodology in the high-wind penetrated island
system and gathering experience to enhance system’s frequency
response in the grid operation.

ACKNOWLEDGMENT
The authors would like to thank the reviewers for their valu-
Fig. 12. Operating conditions of the testing on the variance decision control. able comments on the manuscript that they rely on heavily in
(a) Variance estimate of load and MWPE. (b) Sequence of the switching signal. revising this paper.
(c) Variance calculation of load and wind power output (generating base is 1000
MW per p.u.).
REFERENCES
[1] R. Thresher, M. Robinson, and P. Veers, “To capture the wind,” IEEE
reserve is retained using a larger deviation K. Con- Power Energy Mag., vol. 5, no. 6, pp. 34–46, Nov./Dec. 2007.
sequently, a larger frequency band (say 0.5 ) in (7) [2] E. Fagan, S. Grimes, J. McArdle, P. Smith, and M. Stronge, “Grid code
should be set so that the overall is responding to a larger provisions for wind generators in Ireland,” in Proc. IEEE PES General
Meeting, Jun. 2005, vol. 2, pp. 1241–1247.
frequency event. Since wind turbine would be expected to pro- [3] “Controllable wind farm power station grid code provisions,” EirGrid
vide both frequency regulation and contingency at this moment, (Ireland) Grid Code, ver. 3.1, May 2008.
setting a larger is important for avoiding over-reaction of [4] T. Ackermann, J. R. Abbad, I. M. Dudurych, I. Erlich, H. Holttinen, J.
R. Kristofferson, and P. E. Sørensen, “European balancing act,” IEEE
frequency regulation in wind turbines. Otherwise, a larger fre- Power Energy Mag., vol. 5, no. 6, pp. 90–103, Nov./Dec. 2007.
quency deadband should be applied to the speed-droop loop in [5] I. M. de Alegria, J. Andreu, J. L. Martin, P. Ibanez, J. L. Villate, and
Fig. 7 to distinguish contingency service from frequency regu- H. Camblong, “Connection requirements for wind farms: A survey on
technical requirements and regulation,” Renew. Sustain. Energy Rev.,
lation. vol. 11, no. 8, pp. 1858–1872, Oct. 2007.
From system’s point of view, the wind farm reserve [6] A. Mullane and M. O’Malley, “The inertial response of induction-ma-
is estimated by the following relation: chine-based wind turbines,” IEEE Trans. Power Syst., vol. 20, no. 3,
pp. 1496–1503, Aug. 2005.
[7] G. Lalor, A. Mullane, and M. O’Malley, “Frequency control and wind
(14) turbine technologies,” IEEE Trans. Power Syst., vol. 20, no. 4, pp.
1905–1913, Nov. 2005.
[8] O. Anaya-Lara, F. M. Hughes, N. Jenkins, and G. Strbac, “Contribu-
tion of DFIG-based wind farms to power system short-term frequency
where is the number of DFIGs operating for reserve response regulation,” Proc. Inst. Elect. Eng., Elect. Power Appl., vol. 153, no. 2,
within the wind farm. pp. 164–170, Mar. 2006.
[9] R. G. de Almeida and J. A. Peças Lopes, “Participation of doubly
fed induction wind generators in system frequency regulation,” IEEE
VII. CONCLUSION Trans. Power Syst., vol. 22, no. 3, pp. 944–950, Aug. 2007.
[10] J. F. Conroy and R. Watson, “Frequency response capability of full
The activation of wind frequency regulation is dependant on converter wind turbine generators in comparison to conventional gen-
the system condition. As there is an opportunity to expect fre- eration,” IEEE Trans. Energy Syst., vol. 23, no. 2, pp. 649–656, May
quency response from wind, wind turbines should be ready to 2008.
[11] C. Luo, H. Banakar, S. Baike, and B.-T. Ooi, “Strategies to smooth
contribute their inherent potential. wind power fluctuations of wind turbine generator,” IEEE Trans. En-
This paper first presents a stable active power control of the ergy Convers., vol. 22, no. 2, pp. 341–349, Jun. 2007.
DFIG at all wind speed regions. Depending on the rotor speed [12] L.-R. Chang-Chien and Y.-C. Yin, “Strategies for operating wind
power in a similar manner of conventional power plant,” IEEE Trans.
condition, the DFIG can be controlled in three operating modes Energy Convers., vol. 24, no. 4, pp. 926–934, Dec. 2009.
to trace operator’s active power command. Simulation results [13] R. G. de Almeida, E. D. Castronuovo, and J. A. Peças Lopes, “Optimum
confirm the feasibility of the control on various wind speed con- generation control in wind parks when carrying out system operator
requests,” IEEE Trans. Power Syst., vol. 21, no. 2, pp. 718–725, May
ditions. Notion of the test shows that output power of the DFIG 2006.
was following the DFIG command so long as the DFIG com- [14] J. L. Rodriguez-Amenedo, S. Arnalte, and J. C. Burgos, “Automatic
mand was set equal or below the available wind power. In the generation control of a wind farm with variable speed wind turbines,”
IEEE Trans. Energy Convers., vol. 17, no. 2, pp. 279–284, Jun. 2002.
case of insufficient wind power compared to the DFIG com- [15] R. Pena, J. C. Clare, and G. M. Asher, “Doubly fed induction generator
mand, test result also shows that DFIG’s stability is still assured. using back-to-back PWM converters and its applications to variable-
In the second part, this paper presents strategies for wind speed wind-energy generation,” Proc. Inst. Elect. Eng., Elect. Power
Appl., vol. 143, no. 3, pp. 231–241, May 1996.
power frequency regulation. The moving-average with K devia- [16] J. Ekanayake, L. Holdsworth, and N. Jenkins, “Control of DFIG wind
tion method is introduced to preserve a certain amount of wind turbines,” Power Eng. J., vol. 17, no. 1, pp. 28–32, Feb. 2003.
718 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 26, NO. 2, MAY 2011

[17] E. N. Hinrichsen, “Controls for variable pitch wind turbine generators,” Wei-Ting Lin received the B.S. degree in electrical
IEEE Trans. Power App. Syst., vol. PAS-103, pp. 886–892, Apr. 1984. and computer science engineering from National
[18] E. Ela, M. Milligan, B. Parsons, D. Lew, and D. Corbus, “The evolution Sun-Yet-Sen University, Kaohsiung, Taiwan, in 2007
of wind power integration studies: Past, present, and future,” in Proc. and the M.S. degree from National Cheng Kung
IEEE PES General Meeting, Jul. 2009, pp. 1–8. University, Tainan, Taiwan, in 2009.
[19] H. Holttinen, M. Milligan, B. Kirby, T. Acker, V. Neimane, and T. His research interests are in system operation and
Molinski, “Using standard deviation as a measure of increased opera- wind applications to power systems.
tional reserve requirement for wind power,” Wind Eng., vol. 32, no. 4,
pp. 355–378, 2008.
[20] P. Kundar, Power System Stability and Control, 1st ed. New York:
McGraw-Hill, 1994.
[21] S.-H. Huang, D. Maggio, K. McIntyre, V. Betanabhatla, J. Dumas,
and J. Adams, “Impact of wind generation on system operations in
the deregulated environment: ERCOT experience,” in Proc. IEEE PES
General Meeting, Jul. 2009, pp. 1–8. Yao-Ching Yin received the B.S. degree in electrical
[22] T. Senjyu, R. Sakamoto, N. Urasaki, T. Funabashi, H. Fujita, and H. and computer science engineering from National
Sekine, “Output power leveling of wind turbine generator for all op- Taiwan University of Science and Technology,
erating regions by pitch angle control,” IEEE Trans. Energy Convers., Taipei, Taiwan, in 2006 and the M.S. degree from
vol. 21, no. 2, pp. 467–475, Jun. 2006. National Cheng Kung University, Tainan, Taiwan, in
[23] T. Kaneko, T. Senjyu, A. Yona, M. Datta, T. Funabashi, and C.-H. 2008.
Kim, “Output power coordination control for wind farm in small power His research interests are in power system opera-
system,” in Proc. 14th Int. Conf. Intelligent Systems Applied to Power tion and renewable applications.
Systems (ISAP’07), Nov. 2007, pp. 51–56.

Le-Ren Chang-Chien (M’03) received the B.S.

degree in engineering science from National Cheng
Kung University, Tainan, Taiwan, in 1993, the
M.S.E.E. degree from the University of Wis-
consin-Madison in 1998, and the Ph.D. degree from
Purdue University, West Lafayette, IN, in 2002.
He joined the Department of Electrical Engi-
neering at National Cheng Kung University as an
Assistant Professor in 2003 and became an Associate
Professor in 2008. His research interests include
electric machines, power system operation, control,
and reliability.