Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Research ArticleThe Virtual Resistance Control Strategy for HVRT ofDoubly Fed Induction Wind Generators Based on ParticleSwarm Optimization
Zhen Xie and Xue Li
School of Electric Engineering and Automation Hefei University of Technology Hefei 230009 China
Correspondence should be addressed to Zhen Xie ppsd2003xiesinacom
Received 17 January 2014 Revised 9 May 2014 Accepted 25 May 2014 Published 10 July 2014
Academic Editor Marcelo M Cavalcanti
Copyright copy 2014 Z Xie and X Li This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Grid voltage swell will cause transient DC flux component in the doubly fed induction generator (DFIG) stator windings creatingserious stator and rotor current and torque oscillation which is more serious than influence of the voltage dip It is found thatvirtual resistance manages effectively to suppress rotor current and torque oscillation avoid the operation of crowbar circuit andenhance its high voltage ride through technology capability In order to acquire the best virtual resistance value the excellentparticle library (EPL) of dynamic particle swarm optimization (PSO) algorithm is proposed It takes the rotor voltage and rotorcurrent as two objectives which has a fast convergence performance and high accuracy Simulation and experimental results verifythe effectiveness of the virtual resistance control strategy
1 Introduction
Many studies have focused on the impact of grid voltagesags [1ndash4] on wind turbine generators and low voltage ride-through (LVRT) technology but insufficient attention hasbeen given to the influence of grid voltage swell on thesame generators and high voltage ride-through (HVRT)technology In wind power systems grid voltage swell may becaused by single-phase line-to-ground faults sudden removalof large loads and the use of a large capacitor compensator [56] The commonly used DFIG [7] which is connected to thegrid through the converter follows the principle of grid volt-age sag faults When grid voltage swells its transient processalso causes the current or voltage to swell in the stators androtors of the DFIG To avoid these problems and protect theconverter wind turbine generators can be automatically takenoff-grid and perform grid-connected operation after failurerecovery However this off-grid strategy no longer meets thestandards of the grid connection of modern large-scale windpower generation [8]Mostwind power integration standardsrequire the LVRT ability of wind farms the expansion of windpower generation capacity and the improvement of integra-tion standards will soon make HVRT ability a prerequisite
for wind farms The wind power integration standards ofEON in Germany require the HVRT ability of wind tur-bine generators when grid voltage increases to 12 pu windturbine generators should maintain long-term fault ride-through and absorb a certain amount of reactive power inhigh-voltage conditionsMoreover the ratio between reactivecurrent and the change rate of grid voltage should be 2 1Thisstandard is the control requirement of reactive compensationin high-voltage conditions Australia first established HVRTstandards for grid-connected wind turbine generators [9]When grid voltage increases to 130 of the nominal voltagewind turbine generators should maintain fault ride-throughin 60ms andprovide a sufficiently high fault recovery currentThis standard requires the capability of resisting and runningthrough the high voltage of wind turbine generators
DFIG can achieve HVRT by increasing hardware circuitsand improving system control (Figure 1) A chopper circuiton the DC side of the converter was used by [5] to preventvoltage on the DC side from increasing because of the reverseflow of the grid side converter energy triggered by a grid volt-age swell Reference [10] proposed that the series resistanceof current transformers on the rotor side could prevent rotorovercurrent thus preventing the rotor side converter from
Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2014 Article ID 350367 11 pageshttpdxdoiorg1011552014350367
2 Mathematical Problems in Engineering
is
+
Rs L1s L1r
s r
minus
Lm120595s 120595r
j120596120595r
+
minus
Rrir
+ minus
Figure 1 DFIG equivalent circuit at static stator-oriented referenceframe
losing control over the generator because crowbar resistancewill be used if a grid fault occurs This series resistancecan also continuously provide reactive power to the gridduring the fault and reduce torque ripple References [611] examined and compared the static synchronous com-pensator (STATCOM) and dynamic voltage restorer (DVR)schemes during a grid voltage swell In the DVR schemegenerator voltage is maintained by compensating for thevoltage difference between normal and fault conditions Inthe STATCOM scheme grid voltage decreases by controllingthe reactive current injected into the power grid Especiallywhen grid impedance is relatively low both schemes signif-icantly increase the cost of using the whole set of hardwareThe STATCOM scheme is uneconomical because it requirescurrent far stronger than the rated current of wind turbinegenerators The enhanced controlling strategy of the DFIGunder a grid voltage swell is proposed in [12] decouplingcontrol in fault conditions is achieved by establishing anaccuratemodel of the dynamic process of theDFIG rotor fieldcurrent This model considers the compensation dosage ofthe rotor field current change on the basis of the conventionalcontroller Proportional-integral control should be used innormal operations and vector hysteresis control in grid faults[13]
In the HVRT of DFIG converter overcurrent can beavoided by using the crowbar on the rotor side thusrapidly increasing the generator torque and generating anuncontrollable output of active and reactive power whenthe grid voltage swells To solve these problems and ensurethe excellent HVRT of the DFIG this study analyzes theelectromagnetic transient course of the DFIG under the gridvoltage swell and examines the control strategy of activedamping in the rotor converter according to the passivedamping scheme of the series resistance of the generator rotor[14 15] Thus rotor overvoltage is managed when the gridvoltage swells the rotor overcurrent is maximally controlledand the HVRT of the DFIG is improved To acquire the bestvirtual resistance value this paper adopts the EPL of dynamicPSO algorithm for the multiobjective optimization design ofparameters It takes the rotor voltage and rotor current astwo objectives for the integrated optimization As the resultof the multiple iterations the algorithm can be caught inlocal optimumThus the algorithm is improved in this paperwhich is on the basis of the excellent library and dynamicimprovement This makes the algorithm more suitable forthe dynamic changes of the actual cases Therefore the localoptimum of PSO algorithm can not only be avoided but alsothe speed and accuracy of the optimum can be improved
PI
PI
SVPWM
DC+
+
minus
abc
ilowastrd
ilowastrq
rdc
rdc
rqc
rqc
lowastrd
lowastrq
lowastrD
lowastrQ
2r2s
3s2r
3s2r
+
120579r
120579r
120579r
120579s120579sr
ird
irq
isd
isq
DFIG
PLLC
AB
Feedforwarddecoupling
Statorflux
calculation
Ra
Ra
d
dt
d
dt
120596r
120596sr
120595sd
120595sq iAiB iC
us
iaib
ic
Figure 2 Block diagram of the DFIG control system based onvirtual resistance
Simulation and experimental results confirm the feasibility ofthis damping scheme
2 Dynamic Analysis of DFIG duringGrid Voltage Swell
DFIG equivalent circuit is shown in Figure 1 to analyze thedynamic behavior during grid voltage swell For simplicityall rotor variables are converted to stator side and linearmagnetic circuits are assumed [16] With the application ofmotor convention the DFIG model can be expressed as
V119904= 119877119904
119894119904+
119889
119889119905
119904 (1)
V119903= 119877119903
119894119903+
119889
119889119905
119903minus 119895120596119903119903 (2)
119904= 119871119904
119894119904+ 119871119898
119894119903 (3)
119903= 119871119898
119894119904+ 119871119903
119894119903 (4)
The symbols and their corresponding meanings are asfollows
V voltage vector119894 current vector120596119903 rotor electrical speed
flux vector119877 resistance119871 inductance
Subscripts 119903 and 119904 are the rotor and stator respectivelyThe rotor voltage can be calculated from (2) as follows
V119903=
119871119898
119871119904
(119901 minus 119895120596119903) 119904+ [119877119903+ 120590119871119903(119901 minus 119895120596
119903)]
119894119903 (5)
where 120590 = 1 minus 1198712119898119871119904119871119903
Mathematical Problems in Engineering 3
Under normal condition the rotor voltage term inducedby the stator flux is expressed as
V1199030= 119895120596119904119897
119871119898
119871119904
119904= 119904
119871119898
119871119904
119881119904119890119890
119895120596119904119905
(6)
where 120596119904119897= 120596119904minus120596119903and 119904 is the slip (119904 = 120596
119904119897120596119904) and119881
119904119890is the
stator voltage amplitudeThe rotor voltage at rotor terminal generated by the
converter is expressed as
V119903=
119871119898
119871119904
119904V119904+ [119877119903+ 120590119871119903(119901 minus 119895120596
119903)]
119894119903 (7)
At 119905 = 1199050 a fault appears in the grid The corresponding
stator side voltage is expressed as
V119904(119905 lt 1199050) = 119881119904119890119890
119895120596119904119905
V119904(119905 ge 1199050) = (1 + 119902)119881
119904119890119890
119895120596119904119905
(8)
where 119902 is the grid voltage swell ratio If the resistance 119877119904is
neglected the steady-state stator flux is proportional to thevoltage
119904(119905 lt 1199050) =
119881119904119890
119895120596119904
119890
119895120596119904119905
119904(119905 ge 1199050) = (1 + 119902)
119881119904119890
119895120596119904
119890
119895120596119904119905
(9)
The stator flux under grid voltage swell is
119904(119905 ge 1199050) = (1 + 119902)
119881119904119890
119895120596119904
119890
119895120596119904119905
minus 119902
119881119904119890
119895120596119904
119890
1198951205961199041199050
119890
minus(119905minus1199050)120591119904
(10)
where 120591119904= 120590119871119904119877119904
The stator flux of (10) can be divided into forced fluxand natural flux The forced flux is proportional to the gridvoltage and rotates at synchronous speed The natural fluxis a transient flux which guarantees that no discontinuityappears in the stator flux during grid voltage swell Thenatural flux is proportional to the grid voltage swell ratio anddoes not rotateThe time constant of stator winding decreasesexponentially the amplitude of the natural flux to zero
If 1199050
= 0 is assumed the induced voltage in rotorreference frame during grid voltage swell can be obtainedby substituting (10) in the first term of the right hand of (5)(neglecting the term 1120591
119904due to its low value)
V1199030=
119871119898
119871119904
119881119904119890[119904 (1 + 119902) 119890
119895120596119904119897119905
+ (1 minus 119904) 119902119890
minus119895120596119903119905
119890
minus(119905minus1199050)120591119904
] (11)
Each component of the stator flux induces a correspond-ing component A small component proportional to the slipis first induced by the forced flux The small componentrotates at slip frequency and achieves someHertzThe secondcomponent induced by natural flux is important because itis proportional to 1 minus 119904 In addition the second componentrotates at rotor electrical speed 120596
119903
3 HVRT Control Strategy of DFIG
Virtual impedance control strategy is used to enhance HVRTcapability and suppress the oscillation of DFIG during gridvoltage swell [17]
31 Stability Analysis of the DFIG System Using stator fluxas state variables rotor current and stator voltage as inputvariables stator flux state equation can be expressed as
119889
119889119905
120595119904119902= minus
119877119904
119871119904
120595119904119902minus 120596119904120595119904119889+
119871119898
119871119904
119877119904119894119903119902+ 119906119904
119889
119889119905
120595119904119889= minus
119877119904
119871119904
120595119904119889+ 120596119904120595119904119902+
119871119898
119871119904
119877119904119894119903119889
(12)
The characteristic equation is expressed as
120582
2
+ 2
119877119904
119871119904
120582 +
119877
2
119904
119871
2
119904
+ 120596
2
119904= 0 (13)
The damping factor 120585 and natural oscillation frequency120596119899are
120585 =
119877119904
120596119899119871119904
120596119899=radic
119877
2
119904
119871
2
119904
+ 120596
2
119904
(14)
Given the relatively low stator resistance of megawattgenerators DFIG have the feature of underdamping Thesystem is prone to oscillation when the grid voltage swellsbecause its natural oscillation frequency is close to thegrid frequency Therefore system damping control includ-ing both passive and active damping controls should beconsidered Thus the virtual resistance control strategy wasemployed to increase the system damping and improve thedynamic HVRT performance of DFIG
32 PassiveDampingControl Strategy In aDFIG electromag-netic transient process motivated by a grid voltage swell therotor inducts strong back electromotive force (EMF) once thegrid voltage swells because the rotor resistance and leakageimpedance of the MW DFIG are relatively low This strongEMF causes rotor overcurrent Therefore passive dampingcontrol is introduced that is dynamic series resistance isallocated to the side of the generator rotor When the gridvoltage swells the passive damping of the dynamic seriesresistance of this generator rotor efficiently increases thesystem damping reduces the peak oscillation of the rotorcurrent decreases the time constant of the rotor circuit andquickly attenuates the rotor current However introducing adynamic resistance increases system loss and reduces systemefficiency thus further complicating the resistance designTherefore a control algorithm that imitates the dynamicresistance (ie active damping control) should be considered
4 Mathematical Problems in Engineering
ilowastr+minus minus
+
minusE
irC(s) T(s) G(s)
Ra
r
Figure 3 Block diagram of the current control system based onvirtual resistance
33 Virtual Resistance Control Strategy To address the short-comings of the practical dynamic resistance active dampingcontrol based on virtual resistance can be adopted under agrid voltage swell to increase system resistance and improvethe HVRT of the DFIG A structural chart of the DFIGactive damping control based on virtual resistance is shownin Figure 2
For the analysis the rotor side voltage can be reexpressedas
V119904= (119877119903+ 119877119904
119871
2
119898
119871
2
119904
) 119894119903+ 120590119871119903
119889
119889119905
119894119903+ 119895120596119904119897120590119871119903119894119903+ 119864
119864 =
119871119898
119871119904
120595119904minus
119871119898
119871119904
(
119877119904
119871119904
+ 119895120596119903)120595119904
(15)
where 119864 is the back electromotive force which is a distur-bance component for the rotor current control loop Virtualresistance control strategy is used to reduce the time constantof the rotor current control loop improve its dynamicresponse and inhibit the effect of the disturbance on thedynamic performance of the rotor current The rotor currentcontrol loop with virtual resistance control strategy is shownin Figure 3
The rotor current closed loop controlled object transferfunction 119866(119904) can be expressed as
119866 (119904) =
119894119903(119904)
V119903(119904)
=
119870119892
119879119892119904 + 1
119879119892=
120590119871119903
(119877119903+ 119871
2
119898119877119904119871
2
119904)
119870119892=
1
(119877119903+ 119871
2
119898119877119904119871
2
119904)
(16)
The switching frequency of the rotor side converter isusually high thus the delay of inertia can be neglected tosimplify the analysis
119879 (119904) = 119870119905 (17)
Assume that the virtual resistance is equal to119877119886 the rotor
current closed loop controlled object transfer function canthen be obtained using (16) and (17) as shown in Figure 3
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119892119870119905
119879119892119904 + 1 + 119877
119886119870119892119870119905
(18)
After the introduction of virtual resistance when (16) and(18) are compared the time constant of the rotor currentclosed loop is reduced by 1 + 119877
119886119870119892119870119905times
The PI controller of the rotor current loop is designedby the internal model principle before the introduction ofvirtual resistance the transfer function is expressed as
119862 (119904) =
119870119888119879119892119904 + 119870119888
119870119905119870119892119904
(19)
where119870119888is the bandwidth of the rotor current closed loop
After the introduction of virtual resistance the PI con-troller transfer function is expressed as
119862
1015840
(119904) =
119870119888119879119892119904 + 119870119888+ 119877119886119870119892119870119905119870119888
119870119905119870119892119904
(20)
Before the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866119901(119904) =
119894119903(119904)
119894
lowast
119903(119904)
=
119904
119904 + 119870119888
(21)
After the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866
1015840
119901(119904) = 119866
119901(119904) (22)
Before the introduction of virtual resistance the distur-bance transfer function is expressed as
119866119864119868(119904) =
119894119903(119904)
119864 (119904)
=
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1)
(23)
After the introduction of virtual resistance the distur-bance transfer function is expressed as
119866
1015840
119864119868(119904) =
119894119903(119904)
119864 (119904)
= minus
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1 + 119877
119886119870119892119870119905)
(24)
from (16) the following expression is obtained
119866 (119904) =
119894119903(119904)
V119903(119904)
=
1
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904
(25)
from (18) the following expression is obtained
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119905
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904+ 119870119905119877119886
(26)
From (25) and (26) the introduction of the feedbackcoefficient 119877
119886is equivalent to series resistance 119870
119905119877119886in the
rotor circuit of the DFIG The equivalent circuit of the rotorbased on virtual resistance can be shown in Figure 4
It is known that regulating virtual resistance can changethe dynamic response of the rotor current control loopduring grid voltage swell Figure 5 is bode graph before andafter the introduction of virtual resistance Compared tononvirtual resistance strategy in the low frequency regionactive damping strategy has better inhibitory characteristicon the disturbance In the high frequency region bothhave the same disturbance inhibitory characteristic Figure 6is the step response of the disturbance transfer function
Mathematical Problems in Engineering 5
998400r
KtRa120590Lr
r E
Rr +L2mL2s
Rs
Figure 4 The rotor equivalent circuit based on virtual resistance
10minus1 100 101 102 103minus90
minus45
0
45
90
Phas
e (de
g)
0
10
20
30
40
50
Mag
nitu
de (d
B)
Bode diagram
Frequency (rads)
Without virtual resistanceWith virtual resistance
Figure 5 Bode graph on disturbance with and without activedamping strategy
with and without virtual resistance compared to nonvirtualresistance strategy the overshoot was obviously inhibited bythe virtual resistance strategy under step load disturbancewhich can improve the capability of disturbance suppressionunder grid voltage fluctuations Figure 7 is the root locusof the disturbance transfer function with and without vir-tual resistance after the introduction of virtual resistancethe system root is near the real axis which can increasethe system damping and suppress the oscillation of thesystem
4 The Selection of Virtual Resistance
41 The Establishment of Objective Function The rotor cur-rent objective function is as follows
1198911(119877119886) = 119894119889119902
119894119889119902= radic119894
2
119903119889119899+ 119894
2
119903119902119899(119899 = 1 2 3 4)
(27)
where
1198941199031198891
= minus
119871119898120596slip1205951199041199020
119871119904(120590119877119903+ 119877119886)
1198941199031199021
=
119871119898120596slip1205951199041198890
119871119904(120590119877119903+ 119877119886)
1198941199031198892
= minus
119871119898120596slip (1205951199041198890 minus 1205951199041198892)
119871119904(120590119877119903+ 119877119886)
0
10
20
30
40
50
60
70
80
90Step response
Time (s)
Am
plitu
de
Without virtual resistanceWith virtual resistance
0 05 1 15 2 25
Figure 6 Step response of the disturbance transfer function withand without virtual resistance
1198941199031199022
= minus
119871119898120596slip (1205951199041199020 minus 1205951199041199022)
119871119904(120590119877119903+ 119877119886)
1198941199031198893
=
119871119898120596119903[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199023
=
119871119898120596119903[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031198894
=
119871119898120596119903119886119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199024
=
119871119898120596119903119887119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
(28)
where119898 = 120590119871119903 119899 = (120590119877
119903+ 119877119886) 119889 = 119898119899 119886 = (119889 minus 120590)(120595
1199041198890minus
1205951199041198892)minus120596119904(1205951199041199020minus1205951199041199022) and 119887 = (dminus120590)(120595
1199041198890minus1205951199041198892) + 120596119904(1205951199041199020minus
1205951199041199022) 119894119903119900
is the initial state of current 120572119904and 120572
119889119902represent
the closed loop and the open loop bandwidth of the controlmodel 120590 denotes the damping time constant 119871
119898 119871119904 and
119871119903are the mutual inductance and the stator and rotor self-
inductances 120596119904is the synchronous speed 120595
1199041198890and 120595
1199041199020are
the steady state stator flux of 119889 and 119902 axes prior to the voltagedrop 120595
1199041198892and 120595
1199041199022are the steady state stator flux of 119889 and 119902
axes after the voltage dropThe rotor voltage objective function is as follows
1198912(119877119886) = V119889119902
V119889119902= radicV2119889119899+ V2119902119899
(119899 = 1 2 3)
(29)
where
V1199031198891
= minus
120596slip119877119886119871119898
119899119871119904
1205951199041199020 V
1199031199021= minus
120596slip119877119886119871119898
119899119871119904
1205951199041198890
V1199031198892
=
120596slip119877119886119871119898 (1205951199041199020 minus 1205951199041199022)
119899119871119904
V1199031199022
=
120596slip119877119886119871119898 (1205951199041198890 minus 1205951199041198892)
119899119871119904
6 Mathematical Problems in Engineering
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520
024046064078087093097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(a) Without virtual resistance
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520 024046064078087093
097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(b) With virtual resistance
Figure 7 The root locus of the disturbance transfer function with and without virtual resistance
V1199031198893
=
120596119903119877119886119871119898[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
V1199031199023
=
120596119903119877119886119871119898[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
(30)
where119898 119886 and 119887 are the same as the above description 120596slipis the slip frequency and 120596slip = 119904 sdot 120596119904 119904 denotes the slip
Obviously there are two objective functions the currentand voltage This paper adopts the method of importantobjective and satisfied constraints to simply the objectivefunctions Let the current objective function as the firstoptimization goal when the grid voltage swell is less than03 pu Let the voltage objective function as the first goalwhen the grid voltage swell is greater than 03 pu Meanwhilemake the satisfaction function as a constraint condition Theconstraint value is described as follows
max 1198911(119877119886) le 1205721
max 1198912(119877119886) le 1205722
119869 = 12057311198911(119877119886) + 12057321198912(119877119886)
(31)
where 1205721 1205722are the satisfaction function values and equal to
the maximum current allowed to flow through the converterWhen the swell amplitude reduces 120573
1decreases and 120573
2
increases Thus we can seek out the best solution
42 The EPL of Dynamic PSO Algorithm In particle swarmoptimization (PSO) algorithm the particles are optimizedby changing the speed and position of the particles As thealgorithm proceeds the effective distance between particleswill be diminished and the particle concentration increasedThen particles with small Euclidean distance will iteraterepeatedly prone to premature convergence and lead thePSO algorithm to the local best solution which affect thespeed of convergence and the precision of algorithm Henceit is necessary to adjust the particle concentration in PSOalgorithm The particle concentration is associated with the
degree of similarity between particles The degree of similar-ity between particles is decided by two aspects the Euclideandistance between two particles 119889(119909
119894 119909119895) and the difference
between their fitness function value |fit(119909119894) minus fit(119909
119895)| If 119909
119894=
(1199091198941 1199091198942 119909
119894119898) and 119909
119895= (1199091198951 1199091198952 119909
119895119898) are noted as two
particles of PSO algorithm the Euclidean distance betweenthem is
119889 (119909119894 119909119895) = radic
119898
sum
119896=1
(119909119894119896minus 119909119895119896)
2
(32)
When 119889(119909119894 119909119895) le 119863 and |fit(119909
119894) minus fit(119909
119895)| le 119891 the
two particles 119909119894and 119909
119895can be regarded as similarity 119863 is
the threshold value of distance and 119891 is the threshold valueof fitness The concentration of the particle 119894 is defined asfollows
119886119894=
119899
119873
(33)
where 119899 is the particle number which is similar to 119894119873 is thetotal particle number in the particle swarm
The probability of retained particles is
119875 (119909119894) =
fit (119909119894) 119886119894
sum
119873
119895=1[fit (119909
119895) 119886119894]
(34)
The probability of retained particles is determined by theconcentration and the fitness of particle 119894 The probabilityof retained particles is lower if the particles concentrationis larger and the probability is higher if the fitness valueis larger This method controls the evolution of the highparticle concentration and also prevents the occurrence ofldquoprematurerdquo phenomenon Concurrently the particles withlarger fitness value can evolve fully which can keep thepopulation diversity
The particles with larger fitness value inevitably containsome key information Hence this paper adopts probabilisticmethod to keep some excellent particles and then theseexcellent particles iterate into the next generation withlarge probability It not only avoids the blind search of thealgorithm but also can improve the convergence speed
Mathematical Problems in Engineering 7
The steps to establish the excellent particle library (EPL)are as follows after each iteration of particle swarm optimiza-tion (PSO) the PSO can produce two best solutions (denotedas fitness) which are 119875
119894best denoted as pbest and 119866best denotedas gbest 119875
119894best is the fitness which has been achieved tillcurrent iteration and 119866best is the global best solution Let thecapacity of the EPL be 119876 make the 119875
119894best sort in terms of thefitness of the particles from small to large and select119866best andthe first119876minus1 of119875
119894best to replace the original excellentmemoryparticles in the EPL Thus in processing a new generationof particles the particles of smaller fitness in PSO will bereplaced by more excellent particles Hence the fitness of thewhole particle groups is improved and the convergence speedof the algorithm is accelerated
In a static environment on the basis of local and globaloptimal solutions the particles can finally find the optimalvalue in the solution space through iterative formula Butbecause the rotor voltage and current of doubly fed inductiongenerator change at any time the dynamic improvement isintroduced to adapt to the changes of the current environ-ment
The achievement of dynamic PSO is as follows first ofall the solution space is divided into 119899
1uniform subspaces
In the subspaces 1198992particles sensitive to the environment
are initialized randomlyThen the fitness of sensitive particlesdenoted as 119891
119894is calculated separately in every iteration and
the difference of the adjacent two particles denoted as Δ119891119894is
calculated by (33) In the end all the absolute values ofΔ119891119894are
summed denoted as 119865 by (34) Consider
Δ119891119894= 119891119894(119896 + 1) minus 119891
119894(119896) (35)
119865 =
119899
sum
119894=1
1003816
1003816
1003816
1003816
Δ119891119894
1003816
1003816
1003816
1003816
(119899 = 1198991times 1198992) (36)
If 119865 is not equal to 0 it can be thought that the externalenvironment has been changed We can set a threshold valuedenoted as 119865
119900 When 119865 is greater than 119865
119900 the dynamic
response is triggered The response mechanism is used toreinitialize the position and velocity of particles proportion-ally which is described as follows
If 119865 gt 119865119900 then
V (119894) = rand (119872) times Vmax119909 (119894) = rand (119872) times 119909max
(37)
where rand(119872) are119872 dimensions random numbers in [0 1]119909119894= (119909
1198941 1199091198942 119909
119894119872) and V
119894= (V
1198941 V1198942 V
119894119872) are
denoted as the position and velocity of the 119894th particle selectedfrom the population of the new initialization119872 denotes thedimension 119883max denotes the maximum position and 119881maxdenotes the maximum velocity
43 Steps of the EPL of Dynamic PSO Algorithm Probabilityto be selected is related to the concentration of the particlesBoth the current and voltage in the objective function are tofind the minimum values so opposite to (34) it is replacedby (38) as follows
119901 (119909119894) =
1 (119886119894fit (119909119894))
sum
119873
119895=1[1 (119886
119894fit (119909119895))]
(38)
The detailed algorithm simulation steps are as follows
Step 1 Initialize the particle parameters which include thesearching space the particle numbers119873 initial position andvelocity All of them are random numbers meanwhile theposition and velocity are between 0 and 1 considering thatthe resistance cannot be negative
Step 2 Establish the EPL according to (31) to calculate thefitness of each particle and acquire the local optimal solution119875119894best and the global optimal solution 119866
119894best of the particle119909119894 Put 119866
119894best and the smaller fitness value into the EPL asexcellent particles Judge whether the particle completes theiterations and if it does the iteration will end
Step 3 Supposing the EPL is made up of 1198731particles and
random initialization 119872 sensitive particles calculate thefitness of sensitive particle 119894 by (31) and 119865 by (36) If 119865 gt 1198651then reinitialize according to (37) and to Step 2
Step 4 The velocity and position of each particle are updatedby the following equation
V119905+1119894119889
= 120596V119905119894119889+ 1198881times rand
1(sdot) times (119901
119894119889minus 119909
119905
119894119889)
+ 1198882times rand
2(sdot) times (119901
119892119889minus 119909
119905
119892119889)
119909
119905+1
119894119889= 119909
119905
119894119889+ V119905+1119894119889
(39)
where 120596 is used as an inertia weight and the initial value of120596 is set within the range [0 1] 120596 is updated by the followingequation
120596 = 120596max minus120596max minus 120596min
119879
sdot 119905 (40)
where120596max and120596min denote themaximum inertia weight andthe minimum 119879 is the maximum number of iterations and 119905is the current iteration number at the present number
Step 5 Regulate the particle concentration Select119873 particlesfrom the (119873 + 119872) particles generated in Steps 1 and 2 Theprobability of each particle is selected according to (38)
Step 6 Use the excellent particles in EPL to replace theparticles of larger fitness In this way a new generation ofparticle swarm is generated and then go to Step 2
Step 7 End the iteration
5 Results and Analysis
Supposing the initial voltage of the stator 119880119904= 690V the
initial voltage of the rotor 119880119903= 220V when the voltage swell
is 03 and the rotor speed is 1800 rmin the fitness curvesof the ordinary PSO algorithm the changing inertia weightPSO algorithm and the EPL of dynamic PSO algorithm areas Figure 8
From Figure 8 the ordinary PSO algorithm convergesfinally but the convergence speed is slow and the convergencetime is long Comparing with the ordinary PSO algorithmthe changing inertia weight PSO algorithm increases the
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Mathematical Problems in Engineering
is
+
Rs L1s L1r
s r
minus
Lm120595s 120595r
j120596120595r
+
minus
Rrir
+ minus
Figure 1 DFIG equivalent circuit at static stator-oriented referenceframe
losing control over the generator because crowbar resistancewill be used if a grid fault occurs This series resistancecan also continuously provide reactive power to the gridduring the fault and reduce torque ripple References [611] examined and compared the static synchronous com-pensator (STATCOM) and dynamic voltage restorer (DVR)schemes during a grid voltage swell In the DVR schemegenerator voltage is maintained by compensating for thevoltage difference between normal and fault conditions Inthe STATCOM scheme grid voltage decreases by controllingthe reactive current injected into the power grid Especiallywhen grid impedance is relatively low both schemes signif-icantly increase the cost of using the whole set of hardwareThe STATCOM scheme is uneconomical because it requirescurrent far stronger than the rated current of wind turbinegenerators The enhanced controlling strategy of the DFIGunder a grid voltage swell is proposed in [12] decouplingcontrol in fault conditions is achieved by establishing anaccuratemodel of the dynamic process of theDFIG rotor fieldcurrent This model considers the compensation dosage ofthe rotor field current change on the basis of the conventionalcontroller Proportional-integral control should be used innormal operations and vector hysteresis control in grid faults[13]
In the HVRT of DFIG converter overcurrent can beavoided by using the crowbar on the rotor side thusrapidly increasing the generator torque and generating anuncontrollable output of active and reactive power whenthe grid voltage swells To solve these problems and ensurethe excellent HVRT of the DFIG this study analyzes theelectromagnetic transient course of the DFIG under the gridvoltage swell and examines the control strategy of activedamping in the rotor converter according to the passivedamping scheme of the series resistance of the generator rotor[14 15] Thus rotor overvoltage is managed when the gridvoltage swells the rotor overcurrent is maximally controlledand the HVRT of the DFIG is improved To acquire the bestvirtual resistance value this paper adopts the EPL of dynamicPSO algorithm for the multiobjective optimization design ofparameters It takes the rotor voltage and rotor current astwo objectives for the integrated optimization As the resultof the multiple iterations the algorithm can be caught inlocal optimumThus the algorithm is improved in this paperwhich is on the basis of the excellent library and dynamicimprovement This makes the algorithm more suitable forthe dynamic changes of the actual cases Therefore the localoptimum of PSO algorithm can not only be avoided but alsothe speed and accuracy of the optimum can be improved
PI
PI
SVPWM
DC+
+
minus
abc
ilowastrd
ilowastrq
rdc
rdc
rqc
rqc
lowastrd
lowastrq
lowastrD
lowastrQ
2r2s
3s2r
3s2r
+
120579r
120579r
120579r
120579s120579sr
ird
irq
isd
isq
DFIG
PLLC
AB
Feedforwarddecoupling
Statorflux
calculation
Ra
Ra
d
dt
d
dt
120596r
120596sr
120595sd
120595sq iAiB iC
us
iaib
ic
Figure 2 Block diagram of the DFIG control system based onvirtual resistance
Simulation and experimental results confirm the feasibility ofthis damping scheme
2 Dynamic Analysis of DFIG duringGrid Voltage Swell
DFIG equivalent circuit is shown in Figure 1 to analyze thedynamic behavior during grid voltage swell For simplicityall rotor variables are converted to stator side and linearmagnetic circuits are assumed [16] With the application ofmotor convention the DFIG model can be expressed as
V119904= 119877119904
119894119904+
119889
119889119905
119904 (1)
V119903= 119877119903
119894119903+
119889
119889119905
119903minus 119895120596119903119903 (2)
119904= 119871119904
119894119904+ 119871119898
119894119903 (3)
119903= 119871119898
119894119904+ 119871119903
119894119903 (4)
The symbols and their corresponding meanings are asfollows
V voltage vector119894 current vector120596119903 rotor electrical speed
flux vector119877 resistance119871 inductance
Subscripts 119903 and 119904 are the rotor and stator respectivelyThe rotor voltage can be calculated from (2) as follows
V119903=
119871119898
119871119904
(119901 minus 119895120596119903) 119904+ [119877119903+ 120590119871119903(119901 minus 119895120596
119903)]
119894119903 (5)
where 120590 = 1 minus 1198712119898119871119904119871119903
Mathematical Problems in Engineering 3
Under normal condition the rotor voltage term inducedby the stator flux is expressed as
V1199030= 119895120596119904119897
119871119898
119871119904
119904= 119904
119871119898
119871119904
119881119904119890119890
119895120596119904119905
(6)
where 120596119904119897= 120596119904minus120596119903and 119904 is the slip (119904 = 120596
119904119897120596119904) and119881
119904119890is the
stator voltage amplitudeThe rotor voltage at rotor terminal generated by the
converter is expressed as
V119903=
119871119898
119871119904
119904V119904+ [119877119903+ 120590119871119903(119901 minus 119895120596
119903)]
119894119903 (7)
At 119905 = 1199050 a fault appears in the grid The corresponding
stator side voltage is expressed as
V119904(119905 lt 1199050) = 119881119904119890119890
119895120596119904119905
V119904(119905 ge 1199050) = (1 + 119902)119881
119904119890119890
119895120596119904119905
(8)
where 119902 is the grid voltage swell ratio If the resistance 119877119904is
neglected the steady-state stator flux is proportional to thevoltage
119904(119905 lt 1199050) =
119881119904119890
119895120596119904
119890
119895120596119904119905
119904(119905 ge 1199050) = (1 + 119902)
119881119904119890
119895120596119904
119890
119895120596119904119905
(9)
The stator flux under grid voltage swell is
119904(119905 ge 1199050) = (1 + 119902)
119881119904119890
119895120596119904
119890
119895120596119904119905
minus 119902
119881119904119890
119895120596119904
119890
1198951205961199041199050
119890
minus(119905minus1199050)120591119904
(10)
where 120591119904= 120590119871119904119877119904
The stator flux of (10) can be divided into forced fluxand natural flux The forced flux is proportional to the gridvoltage and rotates at synchronous speed The natural fluxis a transient flux which guarantees that no discontinuityappears in the stator flux during grid voltage swell Thenatural flux is proportional to the grid voltage swell ratio anddoes not rotateThe time constant of stator winding decreasesexponentially the amplitude of the natural flux to zero
If 1199050
= 0 is assumed the induced voltage in rotorreference frame during grid voltage swell can be obtainedby substituting (10) in the first term of the right hand of (5)(neglecting the term 1120591
119904due to its low value)
V1199030=
119871119898
119871119904
119881119904119890[119904 (1 + 119902) 119890
119895120596119904119897119905
+ (1 minus 119904) 119902119890
minus119895120596119903119905
119890
minus(119905minus1199050)120591119904
] (11)
Each component of the stator flux induces a correspond-ing component A small component proportional to the slipis first induced by the forced flux The small componentrotates at slip frequency and achieves someHertzThe secondcomponent induced by natural flux is important because itis proportional to 1 minus 119904 In addition the second componentrotates at rotor electrical speed 120596
119903
3 HVRT Control Strategy of DFIG
Virtual impedance control strategy is used to enhance HVRTcapability and suppress the oscillation of DFIG during gridvoltage swell [17]
31 Stability Analysis of the DFIG System Using stator fluxas state variables rotor current and stator voltage as inputvariables stator flux state equation can be expressed as
119889
119889119905
120595119904119902= minus
119877119904
119871119904
120595119904119902minus 120596119904120595119904119889+
119871119898
119871119904
119877119904119894119903119902+ 119906119904
119889
119889119905
120595119904119889= minus
119877119904
119871119904
120595119904119889+ 120596119904120595119904119902+
119871119898
119871119904
119877119904119894119903119889
(12)
The characteristic equation is expressed as
120582
2
+ 2
119877119904
119871119904
120582 +
119877
2
119904
119871
2
119904
+ 120596
2
119904= 0 (13)
The damping factor 120585 and natural oscillation frequency120596119899are
120585 =
119877119904
120596119899119871119904
120596119899=radic
119877
2
119904
119871
2
119904
+ 120596
2
119904
(14)
Given the relatively low stator resistance of megawattgenerators DFIG have the feature of underdamping Thesystem is prone to oscillation when the grid voltage swellsbecause its natural oscillation frequency is close to thegrid frequency Therefore system damping control includ-ing both passive and active damping controls should beconsidered Thus the virtual resistance control strategy wasemployed to increase the system damping and improve thedynamic HVRT performance of DFIG
32 PassiveDampingControl Strategy In aDFIG electromag-netic transient process motivated by a grid voltage swell therotor inducts strong back electromotive force (EMF) once thegrid voltage swells because the rotor resistance and leakageimpedance of the MW DFIG are relatively low This strongEMF causes rotor overcurrent Therefore passive dampingcontrol is introduced that is dynamic series resistance isallocated to the side of the generator rotor When the gridvoltage swells the passive damping of the dynamic seriesresistance of this generator rotor efficiently increases thesystem damping reduces the peak oscillation of the rotorcurrent decreases the time constant of the rotor circuit andquickly attenuates the rotor current However introducing adynamic resistance increases system loss and reduces systemefficiency thus further complicating the resistance designTherefore a control algorithm that imitates the dynamicresistance (ie active damping control) should be considered
4 Mathematical Problems in Engineering
ilowastr+minus minus
+
minusE
irC(s) T(s) G(s)
Ra
r
Figure 3 Block diagram of the current control system based onvirtual resistance
33 Virtual Resistance Control Strategy To address the short-comings of the practical dynamic resistance active dampingcontrol based on virtual resistance can be adopted under agrid voltage swell to increase system resistance and improvethe HVRT of the DFIG A structural chart of the DFIGactive damping control based on virtual resistance is shownin Figure 2
For the analysis the rotor side voltage can be reexpressedas
V119904= (119877119903+ 119877119904
119871
2
119898
119871
2
119904
) 119894119903+ 120590119871119903
119889
119889119905
119894119903+ 119895120596119904119897120590119871119903119894119903+ 119864
119864 =
119871119898
119871119904
120595119904minus
119871119898
119871119904
(
119877119904
119871119904
+ 119895120596119903)120595119904
(15)
where 119864 is the back electromotive force which is a distur-bance component for the rotor current control loop Virtualresistance control strategy is used to reduce the time constantof the rotor current control loop improve its dynamicresponse and inhibit the effect of the disturbance on thedynamic performance of the rotor current The rotor currentcontrol loop with virtual resistance control strategy is shownin Figure 3
The rotor current closed loop controlled object transferfunction 119866(119904) can be expressed as
119866 (119904) =
119894119903(119904)
V119903(119904)
=
119870119892
119879119892119904 + 1
119879119892=
120590119871119903
(119877119903+ 119871
2
119898119877119904119871
2
119904)
119870119892=
1
(119877119903+ 119871
2
119898119877119904119871
2
119904)
(16)
The switching frequency of the rotor side converter isusually high thus the delay of inertia can be neglected tosimplify the analysis
119879 (119904) = 119870119905 (17)
Assume that the virtual resistance is equal to119877119886 the rotor
current closed loop controlled object transfer function canthen be obtained using (16) and (17) as shown in Figure 3
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119892119870119905
119879119892119904 + 1 + 119877
119886119870119892119870119905
(18)
After the introduction of virtual resistance when (16) and(18) are compared the time constant of the rotor currentclosed loop is reduced by 1 + 119877
119886119870119892119870119905times
The PI controller of the rotor current loop is designedby the internal model principle before the introduction ofvirtual resistance the transfer function is expressed as
119862 (119904) =
119870119888119879119892119904 + 119870119888
119870119905119870119892119904
(19)
where119870119888is the bandwidth of the rotor current closed loop
After the introduction of virtual resistance the PI con-troller transfer function is expressed as
119862
1015840
(119904) =
119870119888119879119892119904 + 119870119888+ 119877119886119870119892119870119905119870119888
119870119905119870119892119904
(20)
Before the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866119901(119904) =
119894119903(119904)
119894
lowast
119903(119904)
=
119904
119904 + 119870119888
(21)
After the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866
1015840
119901(119904) = 119866
119901(119904) (22)
Before the introduction of virtual resistance the distur-bance transfer function is expressed as
119866119864119868(119904) =
119894119903(119904)
119864 (119904)
=
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1)
(23)
After the introduction of virtual resistance the distur-bance transfer function is expressed as
119866
1015840
119864119868(119904) =
119894119903(119904)
119864 (119904)
= minus
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1 + 119877
119886119870119892119870119905)
(24)
from (16) the following expression is obtained
119866 (119904) =
119894119903(119904)
V119903(119904)
=
1
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904
(25)
from (18) the following expression is obtained
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119905
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904+ 119870119905119877119886
(26)
From (25) and (26) the introduction of the feedbackcoefficient 119877
119886is equivalent to series resistance 119870
119905119877119886in the
rotor circuit of the DFIG The equivalent circuit of the rotorbased on virtual resistance can be shown in Figure 4
It is known that regulating virtual resistance can changethe dynamic response of the rotor current control loopduring grid voltage swell Figure 5 is bode graph before andafter the introduction of virtual resistance Compared tononvirtual resistance strategy in the low frequency regionactive damping strategy has better inhibitory characteristicon the disturbance In the high frequency region bothhave the same disturbance inhibitory characteristic Figure 6is the step response of the disturbance transfer function
Mathematical Problems in Engineering 5
998400r
KtRa120590Lr
r E
Rr +L2mL2s
Rs
Figure 4 The rotor equivalent circuit based on virtual resistance
10minus1 100 101 102 103minus90
minus45
0
45
90
Phas
e (de
g)
0
10
20
30
40
50
Mag
nitu
de (d
B)
Bode diagram
Frequency (rads)
Without virtual resistanceWith virtual resistance
Figure 5 Bode graph on disturbance with and without activedamping strategy
with and without virtual resistance compared to nonvirtualresistance strategy the overshoot was obviously inhibited bythe virtual resistance strategy under step load disturbancewhich can improve the capability of disturbance suppressionunder grid voltage fluctuations Figure 7 is the root locusof the disturbance transfer function with and without vir-tual resistance after the introduction of virtual resistancethe system root is near the real axis which can increasethe system damping and suppress the oscillation of thesystem
4 The Selection of Virtual Resistance
41 The Establishment of Objective Function The rotor cur-rent objective function is as follows
1198911(119877119886) = 119894119889119902
119894119889119902= radic119894
2
119903119889119899+ 119894
2
119903119902119899(119899 = 1 2 3 4)
(27)
where
1198941199031198891
= minus
119871119898120596slip1205951199041199020
119871119904(120590119877119903+ 119877119886)
1198941199031199021
=
119871119898120596slip1205951199041198890
119871119904(120590119877119903+ 119877119886)
1198941199031198892
= minus
119871119898120596slip (1205951199041198890 minus 1205951199041198892)
119871119904(120590119877119903+ 119877119886)
0
10
20
30
40
50
60
70
80
90Step response
Time (s)
Am
plitu
de
Without virtual resistanceWith virtual resistance
0 05 1 15 2 25
Figure 6 Step response of the disturbance transfer function withand without virtual resistance
1198941199031199022
= minus
119871119898120596slip (1205951199041199020 minus 1205951199041199022)
119871119904(120590119877119903+ 119877119886)
1198941199031198893
=
119871119898120596119903[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199023
=
119871119898120596119903[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031198894
=
119871119898120596119903119886119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199024
=
119871119898120596119903119887119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
(28)
where119898 = 120590119871119903 119899 = (120590119877
119903+ 119877119886) 119889 = 119898119899 119886 = (119889 minus 120590)(120595
1199041198890minus
1205951199041198892)minus120596119904(1205951199041199020minus1205951199041199022) and 119887 = (dminus120590)(120595
1199041198890minus1205951199041198892) + 120596119904(1205951199041199020minus
1205951199041199022) 119894119903119900
is the initial state of current 120572119904and 120572
119889119902represent
the closed loop and the open loop bandwidth of the controlmodel 120590 denotes the damping time constant 119871
119898 119871119904 and
119871119903are the mutual inductance and the stator and rotor self-
inductances 120596119904is the synchronous speed 120595
1199041198890and 120595
1199041199020are
the steady state stator flux of 119889 and 119902 axes prior to the voltagedrop 120595
1199041198892and 120595
1199041199022are the steady state stator flux of 119889 and 119902
axes after the voltage dropThe rotor voltage objective function is as follows
1198912(119877119886) = V119889119902
V119889119902= radicV2119889119899+ V2119902119899
(119899 = 1 2 3)
(29)
where
V1199031198891
= minus
120596slip119877119886119871119898
119899119871119904
1205951199041199020 V
1199031199021= minus
120596slip119877119886119871119898
119899119871119904
1205951199041198890
V1199031198892
=
120596slip119877119886119871119898 (1205951199041199020 minus 1205951199041199022)
119899119871119904
V1199031199022
=
120596slip119877119886119871119898 (1205951199041198890 minus 1205951199041198892)
119899119871119904
6 Mathematical Problems in Engineering
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520
024046064078087093097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(a) Without virtual resistance
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520 024046064078087093
097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(b) With virtual resistance
Figure 7 The root locus of the disturbance transfer function with and without virtual resistance
V1199031198893
=
120596119903119877119886119871119898[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
V1199031199023
=
120596119903119877119886119871119898[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
(30)
where119898 119886 and 119887 are the same as the above description 120596slipis the slip frequency and 120596slip = 119904 sdot 120596119904 119904 denotes the slip
Obviously there are two objective functions the currentand voltage This paper adopts the method of importantobjective and satisfied constraints to simply the objectivefunctions Let the current objective function as the firstoptimization goal when the grid voltage swell is less than03 pu Let the voltage objective function as the first goalwhen the grid voltage swell is greater than 03 pu Meanwhilemake the satisfaction function as a constraint condition Theconstraint value is described as follows
max 1198911(119877119886) le 1205721
max 1198912(119877119886) le 1205722
119869 = 12057311198911(119877119886) + 12057321198912(119877119886)
(31)
where 1205721 1205722are the satisfaction function values and equal to
the maximum current allowed to flow through the converterWhen the swell amplitude reduces 120573
1decreases and 120573
2
increases Thus we can seek out the best solution
42 The EPL of Dynamic PSO Algorithm In particle swarmoptimization (PSO) algorithm the particles are optimizedby changing the speed and position of the particles As thealgorithm proceeds the effective distance between particleswill be diminished and the particle concentration increasedThen particles with small Euclidean distance will iteraterepeatedly prone to premature convergence and lead thePSO algorithm to the local best solution which affect thespeed of convergence and the precision of algorithm Henceit is necessary to adjust the particle concentration in PSOalgorithm The particle concentration is associated with the
degree of similarity between particles The degree of similar-ity between particles is decided by two aspects the Euclideandistance between two particles 119889(119909
119894 119909119895) and the difference
between their fitness function value |fit(119909119894) minus fit(119909
119895)| If 119909
119894=
(1199091198941 1199091198942 119909
119894119898) and 119909
119895= (1199091198951 1199091198952 119909
119895119898) are noted as two
particles of PSO algorithm the Euclidean distance betweenthem is
119889 (119909119894 119909119895) = radic
119898
sum
119896=1
(119909119894119896minus 119909119895119896)
2
(32)
When 119889(119909119894 119909119895) le 119863 and |fit(119909
119894) minus fit(119909
119895)| le 119891 the
two particles 119909119894and 119909
119895can be regarded as similarity 119863 is
the threshold value of distance and 119891 is the threshold valueof fitness The concentration of the particle 119894 is defined asfollows
119886119894=
119899
119873
(33)
where 119899 is the particle number which is similar to 119894119873 is thetotal particle number in the particle swarm
The probability of retained particles is
119875 (119909119894) =
fit (119909119894) 119886119894
sum
119873
119895=1[fit (119909
119895) 119886119894]
(34)
The probability of retained particles is determined by theconcentration and the fitness of particle 119894 The probabilityof retained particles is lower if the particles concentrationis larger and the probability is higher if the fitness valueis larger This method controls the evolution of the highparticle concentration and also prevents the occurrence ofldquoprematurerdquo phenomenon Concurrently the particles withlarger fitness value can evolve fully which can keep thepopulation diversity
The particles with larger fitness value inevitably containsome key information Hence this paper adopts probabilisticmethod to keep some excellent particles and then theseexcellent particles iterate into the next generation withlarge probability It not only avoids the blind search of thealgorithm but also can improve the convergence speed
Mathematical Problems in Engineering 7
The steps to establish the excellent particle library (EPL)are as follows after each iteration of particle swarm optimiza-tion (PSO) the PSO can produce two best solutions (denotedas fitness) which are 119875
119894best denoted as pbest and 119866best denotedas gbest 119875
119894best is the fitness which has been achieved tillcurrent iteration and 119866best is the global best solution Let thecapacity of the EPL be 119876 make the 119875
119894best sort in terms of thefitness of the particles from small to large and select119866best andthe first119876minus1 of119875
119894best to replace the original excellentmemoryparticles in the EPL Thus in processing a new generationof particles the particles of smaller fitness in PSO will bereplaced by more excellent particles Hence the fitness of thewhole particle groups is improved and the convergence speedof the algorithm is accelerated
In a static environment on the basis of local and globaloptimal solutions the particles can finally find the optimalvalue in the solution space through iterative formula Butbecause the rotor voltage and current of doubly fed inductiongenerator change at any time the dynamic improvement isintroduced to adapt to the changes of the current environ-ment
The achievement of dynamic PSO is as follows first ofall the solution space is divided into 119899
1uniform subspaces
In the subspaces 1198992particles sensitive to the environment
are initialized randomlyThen the fitness of sensitive particlesdenoted as 119891
119894is calculated separately in every iteration and
the difference of the adjacent two particles denoted as Δ119891119894is
calculated by (33) In the end all the absolute values ofΔ119891119894are
summed denoted as 119865 by (34) Consider
Δ119891119894= 119891119894(119896 + 1) minus 119891
119894(119896) (35)
119865 =
119899
sum
119894=1
1003816
1003816
1003816
1003816
Δ119891119894
1003816
1003816
1003816
1003816
(119899 = 1198991times 1198992) (36)
If 119865 is not equal to 0 it can be thought that the externalenvironment has been changed We can set a threshold valuedenoted as 119865
119900 When 119865 is greater than 119865
119900 the dynamic
response is triggered The response mechanism is used toreinitialize the position and velocity of particles proportion-ally which is described as follows
If 119865 gt 119865119900 then
V (119894) = rand (119872) times Vmax119909 (119894) = rand (119872) times 119909max
(37)
where rand(119872) are119872 dimensions random numbers in [0 1]119909119894= (119909
1198941 1199091198942 119909
119894119872) and V
119894= (V
1198941 V1198942 V
119894119872) are
denoted as the position and velocity of the 119894th particle selectedfrom the population of the new initialization119872 denotes thedimension 119883max denotes the maximum position and 119881maxdenotes the maximum velocity
43 Steps of the EPL of Dynamic PSO Algorithm Probabilityto be selected is related to the concentration of the particlesBoth the current and voltage in the objective function are tofind the minimum values so opposite to (34) it is replacedby (38) as follows
119901 (119909119894) =
1 (119886119894fit (119909119894))
sum
119873
119895=1[1 (119886
119894fit (119909119895))]
(38)
The detailed algorithm simulation steps are as follows
Step 1 Initialize the particle parameters which include thesearching space the particle numbers119873 initial position andvelocity All of them are random numbers meanwhile theposition and velocity are between 0 and 1 considering thatthe resistance cannot be negative
Step 2 Establish the EPL according to (31) to calculate thefitness of each particle and acquire the local optimal solution119875119894best and the global optimal solution 119866
119894best of the particle119909119894 Put 119866
119894best and the smaller fitness value into the EPL asexcellent particles Judge whether the particle completes theiterations and if it does the iteration will end
Step 3 Supposing the EPL is made up of 1198731particles and
random initialization 119872 sensitive particles calculate thefitness of sensitive particle 119894 by (31) and 119865 by (36) If 119865 gt 1198651then reinitialize according to (37) and to Step 2
Step 4 The velocity and position of each particle are updatedby the following equation
V119905+1119894119889
= 120596V119905119894119889+ 1198881times rand
1(sdot) times (119901
119894119889minus 119909
119905
119894119889)
+ 1198882times rand
2(sdot) times (119901
119892119889minus 119909
119905
119892119889)
119909
119905+1
119894119889= 119909
119905
119894119889+ V119905+1119894119889
(39)
where 120596 is used as an inertia weight and the initial value of120596 is set within the range [0 1] 120596 is updated by the followingequation
120596 = 120596max minus120596max minus 120596min
119879
sdot 119905 (40)
where120596max and120596min denote themaximum inertia weight andthe minimum 119879 is the maximum number of iterations and 119905is the current iteration number at the present number
Step 5 Regulate the particle concentration Select119873 particlesfrom the (119873 + 119872) particles generated in Steps 1 and 2 Theprobability of each particle is selected according to (38)
Step 6 Use the excellent particles in EPL to replace theparticles of larger fitness In this way a new generation ofparticle swarm is generated and then go to Step 2
Step 7 End the iteration
5 Results and Analysis
Supposing the initial voltage of the stator 119880119904= 690V the
initial voltage of the rotor 119880119903= 220V when the voltage swell
is 03 and the rotor speed is 1800 rmin the fitness curvesof the ordinary PSO algorithm the changing inertia weightPSO algorithm and the EPL of dynamic PSO algorithm areas Figure 8
From Figure 8 the ordinary PSO algorithm convergesfinally but the convergence speed is slow and the convergencetime is long Comparing with the ordinary PSO algorithmthe changing inertia weight PSO algorithm increases the
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 3
Under normal condition the rotor voltage term inducedby the stator flux is expressed as
V1199030= 119895120596119904119897
119871119898
119871119904
119904= 119904
119871119898
119871119904
119881119904119890119890
119895120596119904119905
(6)
where 120596119904119897= 120596119904minus120596119903and 119904 is the slip (119904 = 120596
119904119897120596119904) and119881
119904119890is the
stator voltage amplitudeThe rotor voltage at rotor terminal generated by the
converter is expressed as
V119903=
119871119898
119871119904
119904V119904+ [119877119903+ 120590119871119903(119901 minus 119895120596
119903)]
119894119903 (7)
At 119905 = 1199050 a fault appears in the grid The corresponding
stator side voltage is expressed as
V119904(119905 lt 1199050) = 119881119904119890119890
119895120596119904119905
V119904(119905 ge 1199050) = (1 + 119902)119881
119904119890119890
119895120596119904119905
(8)
where 119902 is the grid voltage swell ratio If the resistance 119877119904is
neglected the steady-state stator flux is proportional to thevoltage
119904(119905 lt 1199050) =
119881119904119890
119895120596119904
119890
119895120596119904119905
119904(119905 ge 1199050) = (1 + 119902)
119881119904119890
119895120596119904
119890
119895120596119904119905
(9)
The stator flux under grid voltage swell is
119904(119905 ge 1199050) = (1 + 119902)
119881119904119890
119895120596119904
119890
119895120596119904119905
minus 119902
119881119904119890
119895120596119904
119890
1198951205961199041199050
119890
minus(119905minus1199050)120591119904
(10)
where 120591119904= 120590119871119904119877119904
The stator flux of (10) can be divided into forced fluxand natural flux The forced flux is proportional to the gridvoltage and rotates at synchronous speed The natural fluxis a transient flux which guarantees that no discontinuityappears in the stator flux during grid voltage swell Thenatural flux is proportional to the grid voltage swell ratio anddoes not rotateThe time constant of stator winding decreasesexponentially the amplitude of the natural flux to zero
If 1199050
= 0 is assumed the induced voltage in rotorreference frame during grid voltage swell can be obtainedby substituting (10) in the first term of the right hand of (5)(neglecting the term 1120591
119904due to its low value)
V1199030=
119871119898
119871119904
119881119904119890[119904 (1 + 119902) 119890
119895120596119904119897119905
+ (1 minus 119904) 119902119890
minus119895120596119903119905
119890
minus(119905minus1199050)120591119904
] (11)
Each component of the stator flux induces a correspond-ing component A small component proportional to the slipis first induced by the forced flux The small componentrotates at slip frequency and achieves someHertzThe secondcomponent induced by natural flux is important because itis proportional to 1 minus 119904 In addition the second componentrotates at rotor electrical speed 120596
119903
3 HVRT Control Strategy of DFIG
Virtual impedance control strategy is used to enhance HVRTcapability and suppress the oscillation of DFIG during gridvoltage swell [17]
31 Stability Analysis of the DFIG System Using stator fluxas state variables rotor current and stator voltage as inputvariables stator flux state equation can be expressed as
119889
119889119905
120595119904119902= minus
119877119904
119871119904
120595119904119902minus 120596119904120595119904119889+
119871119898
119871119904
119877119904119894119903119902+ 119906119904
119889
119889119905
120595119904119889= minus
119877119904
119871119904
120595119904119889+ 120596119904120595119904119902+
119871119898
119871119904
119877119904119894119903119889
(12)
The characteristic equation is expressed as
120582
2
+ 2
119877119904
119871119904
120582 +
119877
2
119904
119871
2
119904
+ 120596
2
119904= 0 (13)
The damping factor 120585 and natural oscillation frequency120596119899are
120585 =
119877119904
120596119899119871119904
120596119899=radic
119877
2
119904
119871
2
119904
+ 120596
2
119904
(14)
Given the relatively low stator resistance of megawattgenerators DFIG have the feature of underdamping Thesystem is prone to oscillation when the grid voltage swellsbecause its natural oscillation frequency is close to thegrid frequency Therefore system damping control includ-ing both passive and active damping controls should beconsidered Thus the virtual resistance control strategy wasemployed to increase the system damping and improve thedynamic HVRT performance of DFIG
32 PassiveDampingControl Strategy In aDFIG electromag-netic transient process motivated by a grid voltage swell therotor inducts strong back electromotive force (EMF) once thegrid voltage swells because the rotor resistance and leakageimpedance of the MW DFIG are relatively low This strongEMF causes rotor overcurrent Therefore passive dampingcontrol is introduced that is dynamic series resistance isallocated to the side of the generator rotor When the gridvoltage swells the passive damping of the dynamic seriesresistance of this generator rotor efficiently increases thesystem damping reduces the peak oscillation of the rotorcurrent decreases the time constant of the rotor circuit andquickly attenuates the rotor current However introducing adynamic resistance increases system loss and reduces systemefficiency thus further complicating the resistance designTherefore a control algorithm that imitates the dynamicresistance (ie active damping control) should be considered
4 Mathematical Problems in Engineering
ilowastr+minus minus
+
minusE
irC(s) T(s) G(s)
Ra
r
Figure 3 Block diagram of the current control system based onvirtual resistance
33 Virtual Resistance Control Strategy To address the short-comings of the practical dynamic resistance active dampingcontrol based on virtual resistance can be adopted under agrid voltage swell to increase system resistance and improvethe HVRT of the DFIG A structural chart of the DFIGactive damping control based on virtual resistance is shownin Figure 2
For the analysis the rotor side voltage can be reexpressedas
V119904= (119877119903+ 119877119904
119871
2
119898
119871
2
119904
) 119894119903+ 120590119871119903
119889
119889119905
119894119903+ 119895120596119904119897120590119871119903119894119903+ 119864
119864 =
119871119898
119871119904
120595119904minus
119871119898
119871119904
(
119877119904
119871119904
+ 119895120596119903)120595119904
(15)
where 119864 is the back electromotive force which is a distur-bance component for the rotor current control loop Virtualresistance control strategy is used to reduce the time constantof the rotor current control loop improve its dynamicresponse and inhibit the effect of the disturbance on thedynamic performance of the rotor current The rotor currentcontrol loop with virtual resistance control strategy is shownin Figure 3
The rotor current closed loop controlled object transferfunction 119866(119904) can be expressed as
119866 (119904) =
119894119903(119904)
V119903(119904)
=
119870119892
119879119892119904 + 1
119879119892=
120590119871119903
(119877119903+ 119871
2
119898119877119904119871
2
119904)
119870119892=
1
(119877119903+ 119871
2
119898119877119904119871
2
119904)
(16)
The switching frequency of the rotor side converter isusually high thus the delay of inertia can be neglected tosimplify the analysis
119879 (119904) = 119870119905 (17)
Assume that the virtual resistance is equal to119877119886 the rotor
current closed loop controlled object transfer function canthen be obtained using (16) and (17) as shown in Figure 3
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119892119870119905
119879119892119904 + 1 + 119877
119886119870119892119870119905
(18)
After the introduction of virtual resistance when (16) and(18) are compared the time constant of the rotor currentclosed loop is reduced by 1 + 119877
119886119870119892119870119905times
The PI controller of the rotor current loop is designedby the internal model principle before the introduction ofvirtual resistance the transfer function is expressed as
119862 (119904) =
119870119888119879119892119904 + 119870119888
119870119905119870119892119904
(19)
where119870119888is the bandwidth of the rotor current closed loop
After the introduction of virtual resistance the PI con-troller transfer function is expressed as
119862
1015840
(119904) =
119870119888119879119892119904 + 119870119888+ 119877119886119870119892119870119905119870119888
119870119905119870119892119904
(20)
Before the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866119901(119904) =
119894119903(119904)
119894
lowast
119903(119904)
=
119904
119904 + 119870119888
(21)
After the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866
1015840
119901(119904) = 119866
119901(119904) (22)
Before the introduction of virtual resistance the distur-bance transfer function is expressed as
119866119864119868(119904) =
119894119903(119904)
119864 (119904)
=
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1)
(23)
After the introduction of virtual resistance the distur-bance transfer function is expressed as
119866
1015840
119864119868(119904) =
119894119903(119904)
119864 (119904)
= minus
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1 + 119877
119886119870119892119870119905)
(24)
from (16) the following expression is obtained
119866 (119904) =
119894119903(119904)
V119903(119904)
=
1
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904
(25)
from (18) the following expression is obtained
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119905
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904+ 119870119905119877119886
(26)
From (25) and (26) the introduction of the feedbackcoefficient 119877
119886is equivalent to series resistance 119870
119905119877119886in the
rotor circuit of the DFIG The equivalent circuit of the rotorbased on virtual resistance can be shown in Figure 4
It is known that regulating virtual resistance can changethe dynamic response of the rotor current control loopduring grid voltage swell Figure 5 is bode graph before andafter the introduction of virtual resistance Compared tononvirtual resistance strategy in the low frequency regionactive damping strategy has better inhibitory characteristicon the disturbance In the high frequency region bothhave the same disturbance inhibitory characteristic Figure 6is the step response of the disturbance transfer function
Mathematical Problems in Engineering 5
998400r
KtRa120590Lr
r E
Rr +L2mL2s
Rs
Figure 4 The rotor equivalent circuit based on virtual resistance
10minus1 100 101 102 103minus90
minus45
0
45
90
Phas
e (de
g)
0
10
20
30
40
50
Mag
nitu
de (d
B)
Bode diagram
Frequency (rads)
Without virtual resistanceWith virtual resistance
Figure 5 Bode graph on disturbance with and without activedamping strategy
with and without virtual resistance compared to nonvirtualresistance strategy the overshoot was obviously inhibited bythe virtual resistance strategy under step load disturbancewhich can improve the capability of disturbance suppressionunder grid voltage fluctuations Figure 7 is the root locusof the disturbance transfer function with and without vir-tual resistance after the introduction of virtual resistancethe system root is near the real axis which can increasethe system damping and suppress the oscillation of thesystem
4 The Selection of Virtual Resistance
41 The Establishment of Objective Function The rotor cur-rent objective function is as follows
1198911(119877119886) = 119894119889119902
119894119889119902= radic119894
2
119903119889119899+ 119894
2
119903119902119899(119899 = 1 2 3 4)
(27)
where
1198941199031198891
= minus
119871119898120596slip1205951199041199020
119871119904(120590119877119903+ 119877119886)
1198941199031199021
=
119871119898120596slip1205951199041198890
119871119904(120590119877119903+ 119877119886)
1198941199031198892
= minus
119871119898120596slip (1205951199041198890 minus 1205951199041198892)
119871119904(120590119877119903+ 119877119886)
0
10
20
30
40
50
60
70
80
90Step response
Time (s)
Am
plitu
de
Without virtual resistanceWith virtual resistance
0 05 1 15 2 25
Figure 6 Step response of the disturbance transfer function withand without virtual resistance
1198941199031199022
= minus
119871119898120596slip (1205951199041199020 minus 1205951199041199022)
119871119904(120590119877119903+ 119877119886)
1198941199031198893
=
119871119898120596119903[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199023
=
119871119898120596119903[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031198894
=
119871119898120596119903119886119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199024
=
119871119898120596119903119887119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
(28)
where119898 = 120590119871119903 119899 = (120590119877
119903+ 119877119886) 119889 = 119898119899 119886 = (119889 minus 120590)(120595
1199041198890minus
1205951199041198892)minus120596119904(1205951199041199020minus1205951199041199022) and 119887 = (dminus120590)(120595
1199041198890minus1205951199041198892) + 120596119904(1205951199041199020minus
1205951199041199022) 119894119903119900
is the initial state of current 120572119904and 120572
119889119902represent
the closed loop and the open loop bandwidth of the controlmodel 120590 denotes the damping time constant 119871
119898 119871119904 and
119871119903are the mutual inductance and the stator and rotor self-
inductances 120596119904is the synchronous speed 120595
1199041198890and 120595
1199041199020are
the steady state stator flux of 119889 and 119902 axes prior to the voltagedrop 120595
1199041198892and 120595
1199041199022are the steady state stator flux of 119889 and 119902
axes after the voltage dropThe rotor voltage objective function is as follows
1198912(119877119886) = V119889119902
V119889119902= radicV2119889119899+ V2119902119899
(119899 = 1 2 3)
(29)
where
V1199031198891
= minus
120596slip119877119886119871119898
119899119871119904
1205951199041199020 V
1199031199021= minus
120596slip119877119886119871119898
119899119871119904
1205951199041198890
V1199031198892
=
120596slip119877119886119871119898 (1205951199041199020 minus 1205951199041199022)
119899119871119904
V1199031199022
=
120596slip119877119886119871119898 (1205951199041198890 minus 1205951199041198892)
119899119871119904
6 Mathematical Problems in Engineering
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520
024046064078087093097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(a) Without virtual resistance
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520 024046064078087093
097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(b) With virtual resistance
Figure 7 The root locus of the disturbance transfer function with and without virtual resistance
V1199031198893
=
120596119903119877119886119871119898[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
V1199031199023
=
120596119903119877119886119871119898[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
(30)
where119898 119886 and 119887 are the same as the above description 120596slipis the slip frequency and 120596slip = 119904 sdot 120596119904 119904 denotes the slip
Obviously there are two objective functions the currentand voltage This paper adopts the method of importantobjective and satisfied constraints to simply the objectivefunctions Let the current objective function as the firstoptimization goal when the grid voltage swell is less than03 pu Let the voltage objective function as the first goalwhen the grid voltage swell is greater than 03 pu Meanwhilemake the satisfaction function as a constraint condition Theconstraint value is described as follows
max 1198911(119877119886) le 1205721
max 1198912(119877119886) le 1205722
119869 = 12057311198911(119877119886) + 12057321198912(119877119886)
(31)
where 1205721 1205722are the satisfaction function values and equal to
the maximum current allowed to flow through the converterWhen the swell amplitude reduces 120573
1decreases and 120573
2
increases Thus we can seek out the best solution
42 The EPL of Dynamic PSO Algorithm In particle swarmoptimization (PSO) algorithm the particles are optimizedby changing the speed and position of the particles As thealgorithm proceeds the effective distance between particleswill be diminished and the particle concentration increasedThen particles with small Euclidean distance will iteraterepeatedly prone to premature convergence and lead thePSO algorithm to the local best solution which affect thespeed of convergence and the precision of algorithm Henceit is necessary to adjust the particle concentration in PSOalgorithm The particle concentration is associated with the
degree of similarity between particles The degree of similar-ity between particles is decided by two aspects the Euclideandistance between two particles 119889(119909
119894 119909119895) and the difference
between their fitness function value |fit(119909119894) minus fit(119909
119895)| If 119909
119894=
(1199091198941 1199091198942 119909
119894119898) and 119909
119895= (1199091198951 1199091198952 119909
119895119898) are noted as two
particles of PSO algorithm the Euclidean distance betweenthem is
119889 (119909119894 119909119895) = radic
119898
sum
119896=1
(119909119894119896minus 119909119895119896)
2
(32)
When 119889(119909119894 119909119895) le 119863 and |fit(119909
119894) minus fit(119909
119895)| le 119891 the
two particles 119909119894and 119909
119895can be regarded as similarity 119863 is
the threshold value of distance and 119891 is the threshold valueof fitness The concentration of the particle 119894 is defined asfollows
119886119894=
119899
119873
(33)
where 119899 is the particle number which is similar to 119894119873 is thetotal particle number in the particle swarm
The probability of retained particles is
119875 (119909119894) =
fit (119909119894) 119886119894
sum
119873
119895=1[fit (119909
119895) 119886119894]
(34)
The probability of retained particles is determined by theconcentration and the fitness of particle 119894 The probabilityof retained particles is lower if the particles concentrationis larger and the probability is higher if the fitness valueis larger This method controls the evolution of the highparticle concentration and also prevents the occurrence ofldquoprematurerdquo phenomenon Concurrently the particles withlarger fitness value can evolve fully which can keep thepopulation diversity
The particles with larger fitness value inevitably containsome key information Hence this paper adopts probabilisticmethod to keep some excellent particles and then theseexcellent particles iterate into the next generation withlarge probability It not only avoids the blind search of thealgorithm but also can improve the convergence speed
Mathematical Problems in Engineering 7
The steps to establish the excellent particle library (EPL)are as follows after each iteration of particle swarm optimiza-tion (PSO) the PSO can produce two best solutions (denotedas fitness) which are 119875
119894best denoted as pbest and 119866best denotedas gbest 119875
119894best is the fitness which has been achieved tillcurrent iteration and 119866best is the global best solution Let thecapacity of the EPL be 119876 make the 119875
119894best sort in terms of thefitness of the particles from small to large and select119866best andthe first119876minus1 of119875
119894best to replace the original excellentmemoryparticles in the EPL Thus in processing a new generationof particles the particles of smaller fitness in PSO will bereplaced by more excellent particles Hence the fitness of thewhole particle groups is improved and the convergence speedof the algorithm is accelerated
In a static environment on the basis of local and globaloptimal solutions the particles can finally find the optimalvalue in the solution space through iterative formula Butbecause the rotor voltage and current of doubly fed inductiongenerator change at any time the dynamic improvement isintroduced to adapt to the changes of the current environ-ment
The achievement of dynamic PSO is as follows first ofall the solution space is divided into 119899
1uniform subspaces
In the subspaces 1198992particles sensitive to the environment
are initialized randomlyThen the fitness of sensitive particlesdenoted as 119891
119894is calculated separately in every iteration and
the difference of the adjacent two particles denoted as Δ119891119894is
calculated by (33) In the end all the absolute values ofΔ119891119894are
summed denoted as 119865 by (34) Consider
Δ119891119894= 119891119894(119896 + 1) minus 119891
119894(119896) (35)
119865 =
119899
sum
119894=1
1003816
1003816
1003816
1003816
Δ119891119894
1003816
1003816
1003816
1003816
(119899 = 1198991times 1198992) (36)
If 119865 is not equal to 0 it can be thought that the externalenvironment has been changed We can set a threshold valuedenoted as 119865
119900 When 119865 is greater than 119865
119900 the dynamic
response is triggered The response mechanism is used toreinitialize the position and velocity of particles proportion-ally which is described as follows
If 119865 gt 119865119900 then
V (119894) = rand (119872) times Vmax119909 (119894) = rand (119872) times 119909max
(37)
where rand(119872) are119872 dimensions random numbers in [0 1]119909119894= (119909
1198941 1199091198942 119909
119894119872) and V
119894= (V
1198941 V1198942 V
119894119872) are
denoted as the position and velocity of the 119894th particle selectedfrom the population of the new initialization119872 denotes thedimension 119883max denotes the maximum position and 119881maxdenotes the maximum velocity
43 Steps of the EPL of Dynamic PSO Algorithm Probabilityto be selected is related to the concentration of the particlesBoth the current and voltage in the objective function are tofind the minimum values so opposite to (34) it is replacedby (38) as follows
119901 (119909119894) =
1 (119886119894fit (119909119894))
sum
119873
119895=1[1 (119886
119894fit (119909119895))]
(38)
The detailed algorithm simulation steps are as follows
Step 1 Initialize the particle parameters which include thesearching space the particle numbers119873 initial position andvelocity All of them are random numbers meanwhile theposition and velocity are between 0 and 1 considering thatthe resistance cannot be negative
Step 2 Establish the EPL according to (31) to calculate thefitness of each particle and acquire the local optimal solution119875119894best and the global optimal solution 119866
119894best of the particle119909119894 Put 119866
119894best and the smaller fitness value into the EPL asexcellent particles Judge whether the particle completes theiterations and if it does the iteration will end
Step 3 Supposing the EPL is made up of 1198731particles and
random initialization 119872 sensitive particles calculate thefitness of sensitive particle 119894 by (31) and 119865 by (36) If 119865 gt 1198651then reinitialize according to (37) and to Step 2
Step 4 The velocity and position of each particle are updatedby the following equation
V119905+1119894119889
= 120596V119905119894119889+ 1198881times rand
1(sdot) times (119901
119894119889minus 119909
119905
119894119889)
+ 1198882times rand
2(sdot) times (119901
119892119889minus 119909
119905
119892119889)
119909
119905+1
119894119889= 119909
119905
119894119889+ V119905+1119894119889
(39)
where 120596 is used as an inertia weight and the initial value of120596 is set within the range [0 1] 120596 is updated by the followingequation
120596 = 120596max minus120596max minus 120596min
119879
sdot 119905 (40)
where120596max and120596min denote themaximum inertia weight andthe minimum 119879 is the maximum number of iterations and 119905is the current iteration number at the present number
Step 5 Regulate the particle concentration Select119873 particlesfrom the (119873 + 119872) particles generated in Steps 1 and 2 Theprobability of each particle is selected according to (38)
Step 6 Use the excellent particles in EPL to replace theparticles of larger fitness In this way a new generation ofparticle swarm is generated and then go to Step 2
Step 7 End the iteration
5 Results and Analysis
Supposing the initial voltage of the stator 119880119904= 690V the
initial voltage of the rotor 119880119903= 220V when the voltage swell
is 03 and the rotor speed is 1800 rmin the fitness curvesof the ordinary PSO algorithm the changing inertia weightPSO algorithm and the EPL of dynamic PSO algorithm areas Figure 8
From Figure 8 the ordinary PSO algorithm convergesfinally but the convergence speed is slow and the convergencetime is long Comparing with the ordinary PSO algorithmthe changing inertia weight PSO algorithm increases the
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Mathematical Problems in Engineering
ilowastr+minus minus
+
minusE
irC(s) T(s) G(s)
Ra
r
Figure 3 Block diagram of the current control system based onvirtual resistance
33 Virtual Resistance Control Strategy To address the short-comings of the practical dynamic resistance active dampingcontrol based on virtual resistance can be adopted under agrid voltage swell to increase system resistance and improvethe HVRT of the DFIG A structural chart of the DFIGactive damping control based on virtual resistance is shownin Figure 2
For the analysis the rotor side voltage can be reexpressedas
V119904= (119877119903+ 119877119904
119871
2
119898
119871
2
119904
) 119894119903+ 120590119871119903
119889
119889119905
119894119903+ 119895120596119904119897120590119871119903119894119903+ 119864
119864 =
119871119898
119871119904
120595119904minus
119871119898
119871119904
(
119877119904
119871119904
+ 119895120596119903)120595119904
(15)
where 119864 is the back electromotive force which is a distur-bance component for the rotor current control loop Virtualresistance control strategy is used to reduce the time constantof the rotor current control loop improve its dynamicresponse and inhibit the effect of the disturbance on thedynamic performance of the rotor current The rotor currentcontrol loop with virtual resistance control strategy is shownin Figure 3
The rotor current closed loop controlled object transferfunction 119866(119904) can be expressed as
119866 (119904) =
119894119903(119904)
V119903(119904)
=
119870119892
119879119892119904 + 1
119879119892=
120590119871119903
(119877119903+ 119871
2
119898119877119904119871
2
119904)
119870119892=
1
(119877119903+ 119871
2
119898119877119904119871
2
119904)
(16)
The switching frequency of the rotor side converter isusually high thus the delay of inertia can be neglected tosimplify the analysis
119879 (119904) = 119870119905 (17)
Assume that the virtual resistance is equal to119877119886 the rotor
current closed loop controlled object transfer function canthen be obtained using (16) and (17) as shown in Figure 3
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119892119870119905
119879119892119904 + 1 + 119877
119886119870119892119870119905
(18)
After the introduction of virtual resistance when (16) and(18) are compared the time constant of the rotor currentclosed loop is reduced by 1 + 119877
119886119870119892119870119905times
The PI controller of the rotor current loop is designedby the internal model principle before the introduction ofvirtual resistance the transfer function is expressed as
119862 (119904) =
119870119888119879119892119904 + 119870119888
119870119905119870119892119904
(19)
where119870119888is the bandwidth of the rotor current closed loop
After the introduction of virtual resistance the PI con-troller transfer function is expressed as
119862
1015840
(119904) =
119870119888119879119892119904 + 119870119888+ 119877119886119870119892119870119905119870119888
119870119905119870119892119904
(20)
Before the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866119901(119904) =
119894119903(119904)
119894
lowast
119903(119904)
=
119904
119904 + 119870119888
(21)
After the introduction of virtual resistance the rotorcurrent closed loop transfer function is expressed as
119866
1015840
119901(119904) = 119866
119901(119904) (22)
Before the introduction of virtual resistance the distur-bance transfer function is expressed as
119866119864119868(119904) =
119894119903(119904)
119864 (119904)
=
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1)
(23)
After the introduction of virtual resistance the distur-bance transfer function is expressed as
119866
1015840
119864119868(119904) =
119894119903(119904)
119864 (119904)
= minus
119870119892119904
(119904 + 119870119888) (119879119892119904 + 1 + 119877
119886119870119892119870119905)
(24)
from (16) the following expression is obtained
119866 (119904) =
119894119903(119904)
V119903(119904)
=
1
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904
(25)
from (18) the following expression is obtained
119866
1015840
(119904) =
119894119903(119904)
V119903(119904)
=
119870119905
120590119871119903119904 + 119877119903+ (119871
2
119898119871
2
119904) 119877119904+ 119870119905119877119886
(26)
From (25) and (26) the introduction of the feedbackcoefficient 119877
119886is equivalent to series resistance 119870
119905119877119886in the
rotor circuit of the DFIG The equivalent circuit of the rotorbased on virtual resistance can be shown in Figure 4
It is known that regulating virtual resistance can changethe dynamic response of the rotor current control loopduring grid voltage swell Figure 5 is bode graph before andafter the introduction of virtual resistance Compared tononvirtual resistance strategy in the low frequency regionactive damping strategy has better inhibitory characteristicon the disturbance In the high frequency region bothhave the same disturbance inhibitory characteristic Figure 6is the step response of the disturbance transfer function
Mathematical Problems in Engineering 5
998400r
KtRa120590Lr
r E
Rr +L2mL2s
Rs
Figure 4 The rotor equivalent circuit based on virtual resistance
10minus1 100 101 102 103minus90
minus45
0
45
90
Phas
e (de
g)
0
10
20
30
40
50
Mag
nitu
de (d
B)
Bode diagram
Frequency (rads)
Without virtual resistanceWith virtual resistance
Figure 5 Bode graph on disturbance with and without activedamping strategy
with and without virtual resistance compared to nonvirtualresistance strategy the overshoot was obviously inhibited bythe virtual resistance strategy under step load disturbancewhich can improve the capability of disturbance suppressionunder grid voltage fluctuations Figure 7 is the root locusof the disturbance transfer function with and without vir-tual resistance after the introduction of virtual resistancethe system root is near the real axis which can increasethe system damping and suppress the oscillation of thesystem
4 The Selection of Virtual Resistance
41 The Establishment of Objective Function The rotor cur-rent objective function is as follows
1198911(119877119886) = 119894119889119902
119894119889119902= radic119894
2
119903119889119899+ 119894
2
119903119902119899(119899 = 1 2 3 4)
(27)
where
1198941199031198891
= minus
119871119898120596slip1205951199041199020
119871119904(120590119877119903+ 119877119886)
1198941199031199021
=
119871119898120596slip1205951199041198890
119871119904(120590119877119903+ 119877119886)
1198941199031198892
= minus
119871119898120596slip (1205951199041198890 minus 1205951199041198892)
119871119904(120590119877119903+ 119877119886)
0
10
20
30
40
50
60
70
80
90Step response
Time (s)
Am
plitu
de
Without virtual resistanceWith virtual resistance
0 05 1 15 2 25
Figure 6 Step response of the disturbance transfer function withand without virtual resistance
1198941199031199022
= minus
119871119898120596slip (1205951199041199020 minus 1205951199041199022)
119871119904(120590119877119903+ 119877119886)
1198941199031198893
=
119871119898120596119903[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199023
=
119871119898120596119903[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031198894
=
119871119898120596119903119886119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199024
=
119871119898120596119903119887119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
(28)
where119898 = 120590119871119903 119899 = (120590119877
119903+ 119877119886) 119889 = 119898119899 119886 = (119889 minus 120590)(120595
1199041198890minus
1205951199041198892)minus120596119904(1205951199041199020minus1205951199041199022) and 119887 = (dminus120590)(120595
1199041198890minus1205951199041198892) + 120596119904(1205951199041199020minus
1205951199041199022) 119894119903119900
is the initial state of current 120572119904and 120572
119889119902represent
the closed loop and the open loop bandwidth of the controlmodel 120590 denotes the damping time constant 119871
119898 119871119904 and
119871119903are the mutual inductance and the stator and rotor self-
inductances 120596119904is the synchronous speed 120595
1199041198890and 120595
1199041199020are
the steady state stator flux of 119889 and 119902 axes prior to the voltagedrop 120595
1199041198892and 120595
1199041199022are the steady state stator flux of 119889 and 119902
axes after the voltage dropThe rotor voltage objective function is as follows
1198912(119877119886) = V119889119902
V119889119902= radicV2119889119899+ V2119902119899
(119899 = 1 2 3)
(29)
where
V1199031198891
= minus
120596slip119877119886119871119898
119899119871119904
1205951199041199020 V
1199031199021= minus
120596slip119877119886119871119898
119899119871119904
1205951199041198890
V1199031198892
=
120596slip119877119886119871119898 (1205951199041199020 minus 1205951199041199022)
119899119871119904
V1199031199022
=
120596slip119877119886119871119898 (1205951199041198890 minus 1205951199041198892)
119899119871119904
6 Mathematical Problems in Engineering
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520
024046064078087093097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(a) Without virtual resistance
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520 024046064078087093
097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(b) With virtual resistance
Figure 7 The root locus of the disturbance transfer function with and without virtual resistance
V1199031198893
=
120596119903119877119886119871119898[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
V1199031199023
=
120596119903119877119886119871119898[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
(30)
where119898 119886 and 119887 are the same as the above description 120596slipis the slip frequency and 120596slip = 119904 sdot 120596119904 119904 denotes the slip
Obviously there are two objective functions the currentand voltage This paper adopts the method of importantobjective and satisfied constraints to simply the objectivefunctions Let the current objective function as the firstoptimization goal when the grid voltage swell is less than03 pu Let the voltage objective function as the first goalwhen the grid voltage swell is greater than 03 pu Meanwhilemake the satisfaction function as a constraint condition Theconstraint value is described as follows
max 1198911(119877119886) le 1205721
max 1198912(119877119886) le 1205722
119869 = 12057311198911(119877119886) + 12057321198912(119877119886)
(31)
where 1205721 1205722are the satisfaction function values and equal to
the maximum current allowed to flow through the converterWhen the swell amplitude reduces 120573
1decreases and 120573
2
increases Thus we can seek out the best solution
42 The EPL of Dynamic PSO Algorithm In particle swarmoptimization (PSO) algorithm the particles are optimizedby changing the speed and position of the particles As thealgorithm proceeds the effective distance between particleswill be diminished and the particle concentration increasedThen particles with small Euclidean distance will iteraterepeatedly prone to premature convergence and lead thePSO algorithm to the local best solution which affect thespeed of convergence and the precision of algorithm Henceit is necessary to adjust the particle concentration in PSOalgorithm The particle concentration is associated with the
degree of similarity between particles The degree of similar-ity between particles is decided by two aspects the Euclideandistance between two particles 119889(119909
119894 119909119895) and the difference
between their fitness function value |fit(119909119894) minus fit(119909
119895)| If 119909
119894=
(1199091198941 1199091198942 119909
119894119898) and 119909
119895= (1199091198951 1199091198952 119909
119895119898) are noted as two
particles of PSO algorithm the Euclidean distance betweenthem is
119889 (119909119894 119909119895) = radic
119898
sum
119896=1
(119909119894119896minus 119909119895119896)
2
(32)
When 119889(119909119894 119909119895) le 119863 and |fit(119909
119894) minus fit(119909
119895)| le 119891 the
two particles 119909119894and 119909
119895can be regarded as similarity 119863 is
the threshold value of distance and 119891 is the threshold valueof fitness The concentration of the particle 119894 is defined asfollows
119886119894=
119899
119873
(33)
where 119899 is the particle number which is similar to 119894119873 is thetotal particle number in the particle swarm
The probability of retained particles is
119875 (119909119894) =
fit (119909119894) 119886119894
sum
119873
119895=1[fit (119909
119895) 119886119894]
(34)
The probability of retained particles is determined by theconcentration and the fitness of particle 119894 The probabilityof retained particles is lower if the particles concentrationis larger and the probability is higher if the fitness valueis larger This method controls the evolution of the highparticle concentration and also prevents the occurrence ofldquoprematurerdquo phenomenon Concurrently the particles withlarger fitness value can evolve fully which can keep thepopulation diversity
The particles with larger fitness value inevitably containsome key information Hence this paper adopts probabilisticmethod to keep some excellent particles and then theseexcellent particles iterate into the next generation withlarge probability It not only avoids the blind search of thealgorithm but also can improve the convergence speed
Mathematical Problems in Engineering 7
The steps to establish the excellent particle library (EPL)are as follows after each iteration of particle swarm optimiza-tion (PSO) the PSO can produce two best solutions (denotedas fitness) which are 119875
119894best denoted as pbest and 119866best denotedas gbest 119875
119894best is the fitness which has been achieved tillcurrent iteration and 119866best is the global best solution Let thecapacity of the EPL be 119876 make the 119875
119894best sort in terms of thefitness of the particles from small to large and select119866best andthe first119876minus1 of119875
119894best to replace the original excellentmemoryparticles in the EPL Thus in processing a new generationof particles the particles of smaller fitness in PSO will bereplaced by more excellent particles Hence the fitness of thewhole particle groups is improved and the convergence speedof the algorithm is accelerated
In a static environment on the basis of local and globaloptimal solutions the particles can finally find the optimalvalue in the solution space through iterative formula Butbecause the rotor voltage and current of doubly fed inductiongenerator change at any time the dynamic improvement isintroduced to adapt to the changes of the current environ-ment
The achievement of dynamic PSO is as follows first ofall the solution space is divided into 119899
1uniform subspaces
In the subspaces 1198992particles sensitive to the environment
are initialized randomlyThen the fitness of sensitive particlesdenoted as 119891
119894is calculated separately in every iteration and
the difference of the adjacent two particles denoted as Δ119891119894is
calculated by (33) In the end all the absolute values ofΔ119891119894are
summed denoted as 119865 by (34) Consider
Δ119891119894= 119891119894(119896 + 1) minus 119891
119894(119896) (35)
119865 =
119899
sum
119894=1
1003816
1003816
1003816
1003816
Δ119891119894
1003816
1003816
1003816
1003816
(119899 = 1198991times 1198992) (36)
If 119865 is not equal to 0 it can be thought that the externalenvironment has been changed We can set a threshold valuedenoted as 119865
119900 When 119865 is greater than 119865
119900 the dynamic
response is triggered The response mechanism is used toreinitialize the position and velocity of particles proportion-ally which is described as follows
If 119865 gt 119865119900 then
V (119894) = rand (119872) times Vmax119909 (119894) = rand (119872) times 119909max
(37)
where rand(119872) are119872 dimensions random numbers in [0 1]119909119894= (119909
1198941 1199091198942 119909
119894119872) and V
119894= (V
1198941 V1198942 V
119894119872) are
denoted as the position and velocity of the 119894th particle selectedfrom the population of the new initialization119872 denotes thedimension 119883max denotes the maximum position and 119881maxdenotes the maximum velocity
43 Steps of the EPL of Dynamic PSO Algorithm Probabilityto be selected is related to the concentration of the particlesBoth the current and voltage in the objective function are tofind the minimum values so opposite to (34) it is replacedby (38) as follows
119901 (119909119894) =
1 (119886119894fit (119909119894))
sum
119873
119895=1[1 (119886
119894fit (119909119895))]
(38)
The detailed algorithm simulation steps are as follows
Step 1 Initialize the particle parameters which include thesearching space the particle numbers119873 initial position andvelocity All of them are random numbers meanwhile theposition and velocity are between 0 and 1 considering thatthe resistance cannot be negative
Step 2 Establish the EPL according to (31) to calculate thefitness of each particle and acquire the local optimal solution119875119894best and the global optimal solution 119866
119894best of the particle119909119894 Put 119866
119894best and the smaller fitness value into the EPL asexcellent particles Judge whether the particle completes theiterations and if it does the iteration will end
Step 3 Supposing the EPL is made up of 1198731particles and
random initialization 119872 sensitive particles calculate thefitness of sensitive particle 119894 by (31) and 119865 by (36) If 119865 gt 1198651then reinitialize according to (37) and to Step 2
Step 4 The velocity and position of each particle are updatedby the following equation
V119905+1119894119889
= 120596V119905119894119889+ 1198881times rand
1(sdot) times (119901
119894119889minus 119909
119905
119894119889)
+ 1198882times rand
2(sdot) times (119901
119892119889minus 119909
119905
119892119889)
119909
119905+1
119894119889= 119909
119905
119894119889+ V119905+1119894119889
(39)
where 120596 is used as an inertia weight and the initial value of120596 is set within the range [0 1] 120596 is updated by the followingequation
120596 = 120596max minus120596max minus 120596min
119879
sdot 119905 (40)
where120596max and120596min denote themaximum inertia weight andthe minimum 119879 is the maximum number of iterations and 119905is the current iteration number at the present number
Step 5 Regulate the particle concentration Select119873 particlesfrom the (119873 + 119872) particles generated in Steps 1 and 2 Theprobability of each particle is selected according to (38)
Step 6 Use the excellent particles in EPL to replace theparticles of larger fitness In this way a new generation ofparticle swarm is generated and then go to Step 2
Step 7 End the iteration
5 Results and Analysis
Supposing the initial voltage of the stator 119880119904= 690V the
initial voltage of the rotor 119880119903= 220V when the voltage swell
is 03 and the rotor speed is 1800 rmin the fitness curvesof the ordinary PSO algorithm the changing inertia weightPSO algorithm and the EPL of dynamic PSO algorithm areas Figure 8
From Figure 8 the ordinary PSO algorithm convergesfinally but the convergence speed is slow and the convergencetime is long Comparing with the ordinary PSO algorithmthe changing inertia weight PSO algorithm increases the
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 5
998400r
KtRa120590Lr
r E
Rr +L2mL2s
Rs
Figure 4 The rotor equivalent circuit based on virtual resistance
10minus1 100 101 102 103minus90
minus45
0
45
90
Phas
e (de
g)
0
10
20
30
40
50
Mag
nitu
de (d
B)
Bode diagram
Frequency (rads)
Without virtual resistanceWith virtual resistance
Figure 5 Bode graph on disturbance with and without activedamping strategy
with and without virtual resistance compared to nonvirtualresistance strategy the overshoot was obviously inhibited bythe virtual resistance strategy under step load disturbancewhich can improve the capability of disturbance suppressionunder grid voltage fluctuations Figure 7 is the root locusof the disturbance transfer function with and without vir-tual resistance after the introduction of virtual resistancethe system root is near the real axis which can increasethe system damping and suppress the oscillation of thesystem
4 The Selection of Virtual Resistance
41 The Establishment of Objective Function The rotor cur-rent objective function is as follows
1198911(119877119886) = 119894119889119902
119894119889119902= radic119894
2
119903119889119899+ 119894
2
119903119902119899(119899 = 1 2 3 4)
(27)
where
1198941199031198891
= minus
119871119898120596slip1205951199041199020
119871119904(120590119877119903+ 119877119886)
1198941199031199021
=
119871119898120596slip1205951199041198890
119871119904(120590119877119903+ 119877119886)
1198941199031198892
= minus
119871119898120596slip (1205951199041198890 minus 1205951199041198892)
119871119904(120590119877119903+ 119877119886)
0
10
20
30
40
50
60
70
80
90Step response
Time (s)
Am
plitu
de
Without virtual resistanceWith virtual resistance
0 05 1 15 2 25
Figure 6 Step response of the disturbance transfer function withand without virtual resistance
1198941199031199022
= minus
119871119898120596slip (1205951199041199020 minus 1205951199041199022)
119871119904(120590119877119903+ 119877119886)
1198941199031198893
=
119871119898120596119903[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199023
=
119871119898120596119903[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)] 119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031198894
=
119871119898120596119903119886119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
1198941199031199024
=
119871119898120596119903119887119890
minus120590119905
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
(28)
where119898 = 120590119871119903 119899 = (120590119877
119903+ 119877119886) 119889 = 119898119899 119886 = (119889 minus 120590)(120595
1199041198890minus
1205951199041198892)minus120596119904(1205951199041199020minus1205951199041199022) and 119887 = (dminus120590)(120595
1199041198890minus1205951199041198892) + 120596119904(1205951199041199020minus
1205951199041199022) 119894119903119900
is the initial state of current 120572119904and 120572
119889119902represent
the closed loop and the open loop bandwidth of the controlmodel 120590 denotes the damping time constant 119871
119898 119871119904 and
119871119903are the mutual inductance and the stator and rotor self-
inductances 120596119904is the synchronous speed 120595
1199041198890and 120595
1199041199020are
the steady state stator flux of 119889 and 119902 axes prior to the voltagedrop 120595
1199041198892and 120595
1199041199022are the steady state stator flux of 119889 and 119902
axes after the voltage dropThe rotor voltage objective function is as follows
1198912(119877119886) = V119889119902
V119889119902= radicV2119889119899+ V2119902119899
(119899 = 1 2 3)
(29)
where
V1199031198891
= minus
120596slip119877119886119871119898
119899119871119904
1205951199041199020 V
1199031199021= minus
120596slip119877119886119871119898
119899119871119904
1205951199041198890
V1199031198892
=
120596slip119877119886119871119898 (1205951199041199020 minus 1205951199041199022)
119899119871119904
V1199031199022
=
120596slip119877119886119871119898 (1205951199041198890 minus 1205951199041198892)
119899119871119904
6 Mathematical Problems in Engineering
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520
024046064078087093097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(a) Without virtual resistance
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520 024046064078087093
097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(b) With virtual resistance
Figure 7 The root locus of the disturbance transfer function with and without virtual resistance
V1199031198893
=
120596119903119877119886119871119898[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
V1199031199023
=
120596119903119877119886119871119898[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
(30)
where119898 119886 and 119887 are the same as the above description 120596slipis the slip frequency and 120596slip = 119904 sdot 120596119904 119904 denotes the slip
Obviously there are two objective functions the currentand voltage This paper adopts the method of importantobjective and satisfied constraints to simply the objectivefunctions Let the current objective function as the firstoptimization goal when the grid voltage swell is less than03 pu Let the voltage objective function as the first goalwhen the grid voltage swell is greater than 03 pu Meanwhilemake the satisfaction function as a constraint condition Theconstraint value is described as follows
max 1198911(119877119886) le 1205721
max 1198912(119877119886) le 1205722
119869 = 12057311198911(119877119886) + 12057321198912(119877119886)
(31)
where 1205721 1205722are the satisfaction function values and equal to
the maximum current allowed to flow through the converterWhen the swell amplitude reduces 120573
1decreases and 120573
2
increases Thus we can seek out the best solution
42 The EPL of Dynamic PSO Algorithm In particle swarmoptimization (PSO) algorithm the particles are optimizedby changing the speed and position of the particles As thealgorithm proceeds the effective distance between particleswill be diminished and the particle concentration increasedThen particles with small Euclidean distance will iteraterepeatedly prone to premature convergence and lead thePSO algorithm to the local best solution which affect thespeed of convergence and the precision of algorithm Henceit is necessary to adjust the particle concentration in PSOalgorithm The particle concentration is associated with the
degree of similarity between particles The degree of similar-ity between particles is decided by two aspects the Euclideandistance between two particles 119889(119909
119894 119909119895) and the difference
between their fitness function value |fit(119909119894) minus fit(119909
119895)| If 119909
119894=
(1199091198941 1199091198942 119909
119894119898) and 119909
119895= (1199091198951 1199091198952 119909
119895119898) are noted as two
particles of PSO algorithm the Euclidean distance betweenthem is
119889 (119909119894 119909119895) = radic
119898
sum
119896=1
(119909119894119896minus 119909119895119896)
2
(32)
When 119889(119909119894 119909119895) le 119863 and |fit(119909
119894) minus fit(119909
119895)| le 119891 the
two particles 119909119894and 119909
119895can be regarded as similarity 119863 is
the threshold value of distance and 119891 is the threshold valueof fitness The concentration of the particle 119894 is defined asfollows
119886119894=
119899
119873
(33)
where 119899 is the particle number which is similar to 119894119873 is thetotal particle number in the particle swarm
The probability of retained particles is
119875 (119909119894) =
fit (119909119894) 119886119894
sum
119873
119895=1[fit (119909
119895) 119886119894]
(34)
The probability of retained particles is determined by theconcentration and the fitness of particle 119894 The probabilityof retained particles is lower if the particles concentrationis larger and the probability is higher if the fitness valueis larger This method controls the evolution of the highparticle concentration and also prevents the occurrence ofldquoprematurerdquo phenomenon Concurrently the particles withlarger fitness value can evolve fully which can keep thepopulation diversity
The particles with larger fitness value inevitably containsome key information Hence this paper adopts probabilisticmethod to keep some excellent particles and then theseexcellent particles iterate into the next generation withlarge probability It not only avoids the blind search of thealgorithm but also can improve the convergence speed
Mathematical Problems in Engineering 7
The steps to establish the excellent particle library (EPL)are as follows after each iteration of particle swarm optimiza-tion (PSO) the PSO can produce two best solutions (denotedas fitness) which are 119875
119894best denoted as pbest and 119866best denotedas gbest 119875
119894best is the fitness which has been achieved tillcurrent iteration and 119866best is the global best solution Let thecapacity of the EPL be 119876 make the 119875
119894best sort in terms of thefitness of the particles from small to large and select119866best andthe first119876minus1 of119875
119894best to replace the original excellentmemoryparticles in the EPL Thus in processing a new generationof particles the particles of smaller fitness in PSO will bereplaced by more excellent particles Hence the fitness of thewhole particle groups is improved and the convergence speedof the algorithm is accelerated
In a static environment on the basis of local and globaloptimal solutions the particles can finally find the optimalvalue in the solution space through iterative formula Butbecause the rotor voltage and current of doubly fed inductiongenerator change at any time the dynamic improvement isintroduced to adapt to the changes of the current environ-ment
The achievement of dynamic PSO is as follows first ofall the solution space is divided into 119899
1uniform subspaces
In the subspaces 1198992particles sensitive to the environment
are initialized randomlyThen the fitness of sensitive particlesdenoted as 119891
119894is calculated separately in every iteration and
the difference of the adjacent two particles denoted as Δ119891119894is
calculated by (33) In the end all the absolute values ofΔ119891119894are
summed denoted as 119865 by (34) Consider
Δ119891119894= 119891119894(119896 + 1) minus 119891
119894(119896) (35)
119865 =
119899
sum
119894=1
1003816
1003816
1003816
1003816
Δ119891119894
1003816
1003816
1003816
1003816
(119899 = 1198991times 1198992) (36)
If 119865 is not equal to 0 it can be thought that the externalenvironment has been changed We can set a threshold valuedenoted as 119865
119900 When 119865 is greater than 119865
119900 the dynamic
response is triggered The response mechanism is used toreinitialize the position and velocity of particles proportion-ally which is described as follows
If 119865 gt 119865119900 then
V (119894) = rand (119872) times Vmax119909 (119894) = rand (119872) times 119909max
(37)
where rand(119872) are119872 dimensions random numbers in [0 1]119909119894= (119909
1198941 1199091198942 119909
119894119872) and V
119894= (V
1198941 V1198942 V
119894119872) are
denoted as the position and velocity of the 119894th particle selectedfrom the population of the new initialization119872 denotes thedimension 119883max denotes the maximum position and 119881maxdenotes the maximum velocity
43 Steps of the EPL of Dynamic PSO Algorithm Probabilityto be selected is related to the concentration of the particlesBoth the current and voltage in the objective function are tofind the minimum values so opposite to (34) it is replacedby (38) as follows
119901 (119909119894) =
1 (119886119894fit (119909119894))
sum
119873
119895=1[1 (119886
119894fit (119909119895))]
(38)
The detailed algorithm simulation steps are as follows
Step 1 Initialize the particle parameters which include thesearching space the particle numbers119873 initial position andvelocity All of them are random numbers meanwhile theposition and velocity are between 0 and 1 considering thatthe resistance cannot be negative
Step 2 Establish the EPL according to (31) to calculate thefitness of each particle and acquire the local optimal solution119875119894best and the global optimal solution 119866
119894best of the particle119909119894 Put 119866
119894best and the smaller fitness value into the EPL asexcellent particles Judge whether the particle completes theiterations and if it does the iteration will end
Step 3 Supposing the EPL is made up of 1198731particles and
random initialization 119872 sensitive particles calculate thefitness of sensitive particle 119894 by (31) and 119865 by (36) If 119865 gt 1198651then reinitialize according to (37) and to Step 2
Step 4 The velocity and position of each particle are updatedby the following equation
V119905+1119894119889
= 120596V119905119894119889+ 1198881times rand
1(sdot) times (119901
119894119889minus 119909
119905
119894119889)
+ 1198882times rand
2(sdot) times (119901
119892119889minus 119909
119905
119892119889)
119909
119905+1
119894119889= 119909
119905
119894119889+ V119905+1119894119889
(39)
where 120596 is used as an inertia weight and the initial value of120596 is set within the range [0 1] 120596 is updated by the followingequation
120596 = 120596max minus120596max minus 120596min
119879
sdot 119905 (40)
where120596max and120596min denote themaximum inertia weight andthe minimum 119879 is the maximum number of iterations and 119905is the current iteration number at the present number
Step 5 Regulate the particle concentration Select119873 particlesfrom the (119873 + 119872) particles generated in Steps 1 and 2 Theprobability of each particle is selected according to (38)
Step 6 Use the excellent particles in EPL to replace theparticles of larger fitness In this way a new generation ofparticle swarm is generated and then go to Step 2
Step 7 End the iteration
5 Results and Analysis
Supposing the initial voltage of the stator 119880119904= 690V the
initial voltage of the rotor 119880119903= 220V when the voltage swell
is 03 and the rotor speed is 1800 rmin the fitness curvesof the ordinary PSO algorithm the changing inertia weightPSO algorithm and the EPL of dynamic PSO algorithm areas Figure 8
From Figure 8 the ordinary PSO algorithm convergesfinally but the convergence speed is slow and the convergencetime is long Comparing with the ordinary PSO algorithmthe changing inertia weight PSO algorithm increases the
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Mathematical Problems in Engineering
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520
024046064078087093097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(a) Without virtual resistance
minus60 minus50 minus40 minus30 minus20 minus10 0minus20minus15minus10
minus505
101520 024046064078087093
097
0992
024046064078087093097
0992
1020304050
Root locus
Real axis (sminus1)
Imag
inar
y ax
is (sminus1)
(b) With virtual resistance
Figure 7 The root locus of the disturbance transfer function with and without virtual resistance
V1199031198893
=
120596119903119877119886119871119898[119887 cos (120596
119904119905) minus 119886 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
V1199031199023
=
120596119903119877119886119871119898[119886 cos (120596
119904119905) + 119887 sin (120596
119904119905)]
119899119871119904[(119889 minus 120590)
2
+ 120596
2
119904]
119890
minus120590119905
(30)
where119898 119886 and 119887 are the same as the above description 120596slipis the slip frequency and 120596slip = 119904 sdot 120596119904 119904 denotes the slip
Obviously there are two objective functions the currentand voltage This paper adopts the method of importantobjective and satisfied constraints to simply the objectivefunctions Let the current objective function as the firstoptimization goal when the grid voltage swell is less than03 pu Let the voltage objective function as the first goalwhen the grid voltage swell is greater than 03 pu Meanwhilemake the satisfaction function as a constraint condition Theconstraint value is described as follows
max 1198911(119877119886) le 1205721
max 1198912(119877119886) le 1205722
119869 = 12057311198911(119877119886) + 12057321198912(119877119886)
(31)
where 1205721 1205722are the satisfaction function values and equal to
the maximum current allowed to flow through the converterWhen the swell amplitude reduces 120573
1decreases and 120573
2
increases Thus we can seek out the best solution
42 The EPL of Dynamic PSO Algorithm In particle swarmoptimization (PSO) algorithm the particles are optimizedby changing the speed and position of the particles As thealgorithm proceeds the effective distance between particleswill be diminished and the particle concentration increasedThen particles with small Euclidean distance will iteraterepeatedly prone to premature convergence and lead thePSO algorithm to the local best solution which affect thespeed of convergence and the precision of algorithm Henceit is necessary to adjust the particle concentration in PSOalgorithm The particle concentration is associated with the
degree of similarity between particles The degree of similar-ity between particles is decided by two aspects the Euclideandistance between two particles 119889(119909
119894 119909119895) and the difference
between their fitness function value |fit(119909119894) minus fit(119909
119895)| If 119909
119894=
(1199091198941 1199091198942 119909
119894119898) and 119909
119895= (1199091198951 1199091198952 119909
119895119898) are noted as two
particles of PSO algorithm the Euclidean distance betweenthem is
119889 (119909119894 119909119895) = radic
119898
sum
119896=1
(119909119894119896minus 119909119895119896)
2
(32)
When 119889(119909119894 119909119895) le 119863 and |fit(119909
119894) minus fit(119909
119895)| le 119891 the
two particles 119909119894and 119909
119895can be regarded as similarity 119863 is
the threshold value of distance and 119891 is the threshold valueof fitness The concentration of the particle 119894 is defined asfollows
119886119894=
119899
119873
(33)
where 119899 is the particle number which is similar to 119894119873 is thetotal particle number in the particle swarm
The probability of retained particles is
119875 (119909119894) =
fit (119909119894) 119886119894
sum
119873
119895=1[fit (119909
119895) 119886119894]
(34)
The probability of retained particles is determined by theconcentration and the fitness of particle 119894 The probabilityof retained particles is lower if the particles concentrationis larger and the probability is higher if the fitness valueis larger This method controls the evolution of the highparticle concentration and also prevents the occurrence ofldquoprematurerdquo phenomenon Concurrently the particles withlarger fitness value can evolve fully which can keep thepopulation diversity
The particles with larger fitness value inevitably containsome key information Hence this paper adopts probabilisticmethod to keep some excellent particles and then theseexcellent particles iterate into the next generation withlarge probability It not only avoids the blind search of thealgorithm but also can improve the convergence speed
Mathematical Problems in Engineering 7
The steps to establish the excellent particle library (EPL)are as follows after each iteration of particle swarm optimiza-tion (PSO) the PSO can produce two best solutions (denotedas fitness) which are 119875
119894best denoted as pbest and 119866best denotedas gbest 119875
119894best is the fitness which has been achieved tillcurrent iteration and 119866best is the global best solution Let thecapacity of the EPL be 119876 make the 119875
119894best sort in terms of thefitness of the particles from small to large and select119866best andthe first119876minus1 of119875
119894best to replace the original excellentmemoryparticles in the EPL Thus in processing a new generationof particles the particles of smaller fitness in PSO will bereplaced by more excellent particles Hence the fitness of thewhole particle groups is improved and the convergence speedof the algorithm is accelerated
In a static environment on the basis of local and globaloptimal solutions the particles can finally find the optimalvalue in the solution space through iterative formula Butbecause the rotor voltage and current of doubly fed inductiongenerator change at any time the dynamic improvement isintroduced to adapt to the changes of the current environ-ment
The achievement of dynamic PSO is as follows first ofall the solution space is divided into 119899
1uniform subspaces
In the subspaces 1198992particles sensitive to the environment
are initialized randomlyThen the fitness of sensitive particlesdenoted as 119891
119894is calculated separately in every iteration and
the difference of the adjacent two particles denoted as Δ119891119894is
calculated by (33) In the end all the absolute values ofΔ119891119894are
summed denoted as 119865 by (34) Consider
Δ119891119894= 119891119894(119896 + 1) minus 119891
119894(119896) (35)
119865 =
119899
sum
119894=1
1003816
1003816
1003816
1003816
Δ119891119894
1003816
1003816
1003816
1003816
(119899 = 1198991times 1198992) (36)
If 119865 is not equal to 0 it can be thought that the externalenvironment has been changed We can set a threshold valuedenoted as 119865
119900 When 119865 is greater than 119865
119900 the dynamic
response is triggered The response mechanism is used toreinitialize the position and velocity of particles proportion-ally which is described as follows
If 119865 gt 119865119900 then
V (119894) = rand (119872) times Vmax119909 (119894) = rand (119872) times 119909max
(37)
where rand(119872) are119872 dimensions random numbers in [0 1]119909119894= (119909
1198941 1199091198942 119909
119894119872) and V
119894= (V
1198941 V1198942 V
119894119872) are
denoted as the position and velocity of the 119894th particle selectedfrom the population of the new initialization119872 denotes thedimension 119883max denotes the maximum position and 119881maxdenotes the maximum velocity
43 Steps of the EPL of Dynamic PSO Algorithm Probabilityto be selected is related to the concentration of the particlesBoth the current and voltage in the objective function are tofind the minimum values so opposite to (34) it is replacedby (38) as follows
119901 (119909119894) =
1 (119886119894fit (119909119894))
sum
119873
119895=1[1 (119886
119894fit (119909119895))]
(38)
The detailed algorithm simulation steps are as follows
Step 1 Initialize the particle parameters which include thesearching space the particle numbers119873 initial position andvelocity All of them are random numbers meanwhile theposition and velocity are between 0 and 1 considering thatthe resistance cannot be negative
Step 2 Establish the EPL according to (31) to calculate thefitness of each particle and acquire the local optimal solution119875119894best and the global optimal solution 119866
119894best of the particle119909119894 Put 119866
119894best and the smaller fitness value into the EPL asexcellent particles Judge whether the particle completes theiterations and if it does the iteration will end
Step 3 Supposing the EPL is made up of 1198731particles and
random initialization 119872 sensitive particles calculate thefitness of sensitive particle 119894 by (31) and 119865 by (36) If 119865 gt 1198651then reinitialize according to (37) and to Step 2
Step 4 The velocity and position of each particle are updatedby the following equation
V119905+1119894119889
= 120596V119905119894119889+ 1198881times rand
1(sdot) times (119901
119894119889minus 119909
119905
119894119889)
+ 1198882times rand
2(sdot) times (119901
119892119889minus 119909
119905
119892119889)
119909
119905+1
119894119889= 119909
119905
119894119889+ V119905+1119894119889
(39)
where 120596 is used as an inertia weight and the initial value of120596 is set within the range [0 1] 120596 is updated by the followingequation
120596 = 120596max minus120596max minus 120596min
119879
sdot 119905 (40)
where120596max and120596min denote themaximum inertia weight andthe minimum 119879 is the maximum number of iterations and 119905is the current iteration number at the present number
Step 5 Regulate the particle concentration Select119873 particlesfrom the (119873 + 119872) particles generated in Steps 1 and 2 Theprobability of each particle is selected according to (38)
Step 6 Use the excellent particles in EPL to replace theparticles of larger fitness In this way a new generation ofparticle swarm is generated and then go to Step 2
Step 7 End the iteration
5 Results and Analysis
Supposing the initial voltage of the stator 119880119904= 690V the
initial voltage of the rotor 119880119903= 220V when the voltage swell
is 03 and the rotor speed is 1800 rmin the fitness curvesof the ordinary PSO algorithm the changing inertia weightPSO algorithm and the EPL of dynamic PSO algorithm areas Figure 8
From Figure 8 the ordinary PSO algorithm convergesfinally but the convergence speed is slow and the convergencetime is long Comparing with the ordinary PSO algorithmthe changing inertia weight PSO algorithm increases the
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 7
The steps to establish the excellent particle library (EPL)are as follows after each iteration of particle swarm optimiza-tion (PSO) the PSO can produce two best solutions (denotedas fitness) which are 119875
119894best denoted as pbest and 119866best denotedas gbest 119875
119894best is the fitness which has been achieved tillcurrent iteration and 119866best is the global best solution Let thecapacity of the EPL be 119876 make the 119875
119894best sort in terms of thefitness of the particles from small to large and select119866best andthe first119876minus1 of119875
119894best to replace the original excellentmemoryparticles in the EPL Thus in processing a new generationof particles the particles of smaller fitness in PSO will bereplaced by more excellent particles Hence the fitness of thewhole particle groups is improved and the convergence speedof the algorithm is accelerated
In a static environment on the basis of local and globaloptimal solutions the particles can finally find the optimalvalue in the solution space through iterative formula Butbecause the rotor voltage and current of doubly fed inductiongenerator change at any time the dynamic improvement isintroduced to adapt to the changes of the current environ-ment
The achievement of dynamic PSO is as follows first ofall the solution space is divided into 119899
1uniform subspaces
In the subspaces 1198992particles sensitive to the environment
are initialized randomlyThen the fitness of sensitive particlesdenoted as 119891
119894is calculated separately in every iteration and
the difference of the adjacent two particles denoted as Δ119891119894is
calculated by (33) In the end all the absolute values ofΔ119891119894are
summed denoted as 119865 by (34) Consider
Δ119891119894= 119891119894(119896 + 1) minus 119891
119894(119896) (35)
119865 =
119899
sum
119894=1
1003816
1003816
1003816
1003816
Δ119891119894
1003816
1003816
1003816
1003816
(119899 = 1198991times 1198992) (36)
If 119865 is not equal to 0 it can be thought that the externalenvironment has been changed We can set a threshold valuedenoted as 119865
119900 When 119865 is greater than 119865
119900 the dynamic
response is triggered The response mechanism is used toreinitialize the position and velocity of particles proportion-ally which is described as follows
If 119865 gt 119865119900 then
V (119894) = rand (119872) times Vmax119909 (119894) = rand (119872) times 119909max
(37)
where rand(119872) are119872 dimensions random numbers in [0 1]119909119894= (119909
1198941 1199091198942 119909
119894119872) and V
119894= (V
1198941 V1198942 V
119894119872) are
denoted as the position and velocity of the 119894th particle selectedfrom the population of the new initialization119872 denotes thedimension 119883max denotes the maximum position and 119881maxdenotes the maximum velocity
43 Steps of the EPL of Dynamic PSO Algorithm Probabilityto be selected is related to the concentration of the particlesBoth the current and voltage in the objective function are tofind the minimum values so opposite to (34) it is replacedby (38) as follows
119901 (119909119894) =
1 (119886119894fit (119909119894))
sum
119873
119895=1[1 (119886
119894fit (119909119895))]
(38)
The detailed algorithm simulation steps are as follows
Step 1 Initialize the particle parameters which include thesearching space the particle numbers119873 initial position andvelocity All of them are random numbers meanwhile theposition and velocity are between 0 and 1 considering thatthe resistance cannot be negative
Step 2 Establish the EPL according to (31) to calculate thefitness of each particle and acquire the local optimal solution119875119894best and the global optimal solution 119866
119894best of the particle119909119894 Put 119866
119894best and the smaller fitness value into the EPL asexcellent particles Judge whether the particle completes theiterations and if it does the iteration will end
Step 3 Supposing the EPL is made up of 1198731particles and
random initialization 119872 sensitive particles calculate thefitness of sensitive particle 119894 by (31) and 119865 by (36) If 119865 gt 1198651then reinitialize according to (37) and to Step 2
Step 4 The velocity and position of each particle are updatedby the following equation
V119905+1119894119889
= 120596V119905119894119889+ 1198881times rand
1(sdot) times (119901
119894119889minus 119909
119905
119894119889)
+ 1198882times rand
2(sdot) times (119901
119892119889minus 119909
119905
119892119889)
119909
119905+1
119894119889= 119909
119905
119894119889+ V119905+1119894119889
(39)
where 120596 is used as an inertia weight and the initial value of120596 is set within the range [0 1] 120596 is updated by the followingequation
120596 = 120596max minus120596max minus 120596min
119879
sdot 119905 (40)
where120596max and120596min denote themaximum inertia weight andthe minimum 119879 is the maximum number of iterations and 119905is the current iteration number at the present number
Step 5 Regulate the particle concentration Select119873 particlesfrom the (119873 + 119872) particles generated in Steps 1 and 2 Theprobability of each particle is selected according to (38)
Step 6 Use the excellent particles in EPL to replace theparticles of larger fitness In this way a new generation ofparticle swarm is generated and then go to Step 2
Step 7 End the iteration
5 Results and Analysis
Supposing the initial voltage of the stator 119880119904= 690V the
initial voltage of the rotor 119880119903= 220V when the voltage swell
is 03 and the rotor speed is 1800 rmin the fitness curvesof the ordinary PSO algorithm the changing inertia weightPSO algorithm and the EPL of dynamic PSO algorithm areas Figure 8
From Figure 8 the ordinary PSO algorithm convergesfinally but the convergence speed is slow and the convergencetime is long Comparing with the ordinary PSO algorithmthe changing inertia weight PSO algorithm increases the
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Mathematical Problems in Engineering
0 20 40 60 80 100 120 140 160 180 200700
800
900
1000
1100
1200
1300
Iteration numbers
Fitn
ess
Ordinary PSOChanging inertia weight PSOThe EPL of dynamic PSO
Fitness and iteration numbers = 200
Figure 8 Fitness curve
convergence speed of the algorithm and shortens the con-vergence time The excellent particle library of dynamic PSOalgorithm is on the establishment of excellent particle libraryand a dynamic optimization is added Considering that therotor voltage changes along with the voltage swell the newparticles which can respond to the external environmentchange are used in the EPL of dynamic PSO algorithmThus local optimum of the algorithm will be avoided becauseof the similarity and unnecessary search between particlesand the convergence speed and convergence precision of thealgorithm are guaranteed
Through the comparison of three kinds of PSO algorithmthis paper put forward the EPL of the dynamic PSO algorithmbymodifying the distance between the particles adjusting theparticle concentration avoiding the repeated computation inthe process of PSO algorithm and saving a lot of time tomakethe algorithm fast convergence
In conclusion the EPL of dynamic PSO algorithm issuitable for calculating the virtual resistance values in dif-ferent rotor speeds and different voltage swells In practicalapplication the virtual resistance can be selected accordingto the current rotor speed and voltage swell
6 Simulation Results
Simulations were performed using MATLAB based on thesystem configuration shown in Figure 2 The current studyuses 2MW DFIG as the prototype to conduct simulationstudies to verify the correctness of the aforementionedtheoretical analysisThemain parameters of DFIG in the sim-ulation are as follows the stator inductance 119871
119904= 00125H
the rotor inductance 119871119903= 00126H the mutual inductance
119871119898= 00123H the stator winding resistance 119877
119904= 00043Ω
and the rotor winding resistance 119877119903= 00041Ω
Figures 9ndash11 are simulationwaveforms of rotor current 119902-axis component 119894
119903119902 rotor voltage V
119903119902 A-phase rotor current
119894119903119886
that used damping respectively when the amplitude ofgrid voltage swell are 13 pu and the fault duration is 100msIn Figure 9 the three-phase grid voltage swells to 130 of its
normal value for 100ms and thus induces big dc componentin the stator flux linkage which will result in the high rotorcurrent the oscillation amplitude of the rotor current in thegrid voltage is higher than the amplitude of the rotor currentin the grid voltage swells because the point of grid voltageswells is uncertain It can be seen from Figure 9 to Figure 11that the use of damping control has amore obvious inhibitioneffect on the rotor current and the rotor voltage does notexceed the upper limit voltage when failure amplitude ismaximumWhen the grid voltage swells with the rising of thevalue of virtual resistance the value of rotor current decreasesand the value of rotor voltage increases It can play a role ofsuppression of rotor current and at the same time the rise ofthe rotor voltage would threaten the safety of the converter
7 Experimental Results
Experimental tests of the virtual resistance control strategywere performed on a laboratory setup of the 11 Kw DFIGsystem to further verify the effectiveness of the proposedstrategy A 15 Kw AC motor driving the DFIG at desiredspeed was used to emulate wind turbine A TMS320LF28335DSP was employed to control the rotor side converter andthe grid side converter The switching frequency for bothconverters was set at 5 kHz with a sampling frequency of10 kHz PM100CLA120was used for the switching deviceThewaveforms are acquired by a DPO4032 digital oscilloscope
The main parameters of DFIG are as follows the statorresistance 119877
119904= 02858Ω the rotor resistance 119877
119903= 02983Ω
the stator leakage reactance 119871119904= 0001323H the rotor
leakage reactance 119871119903= 0001781H and excitation reactance
119871119898= 00676HFigures 12ndash14 show the experimental waveforms of the
119902 axis rotor current component 119894119903119902 119902 axis rotor voltage
component V119903119902 and A-phase rotor current 119894
119903119886as the DFIG
undergoes subsynchronization synchronization and super-synchronization When the grid voltage three-phase sym-metrical swell is 30 the fault duration is 100ms andthe virtual resistance is 05 pu and 15 pu Increasing thevirtual resistancemore significantly inhibits the rotor currentoscillation caused by the grid voltage swell on the controlsystem When the grid voltage recovers the oscillationamplitude of the rotor current becomes higher than whenthe grid voltage swells because of the fault duration andthe point of grid voltage recovery If the point is cut offby the fault and the electromagnetic oscillation caused bythe grid voltage swell has not yet stopped the oscillationcaused by grid recovery will have been superimposed onthe previous one The oscillation amplitude of the rotorcurrent under the super-synchronized operation is higherthan that in the subsynchronized operation Because the rotorside induction voltage in the supersynchronized operation ishigher than that in the subsynchronized operation under gridvoltage swell the generator power is greater with the highermotor speed Controlling the virtual resistance increasesthe rotor voltage but reduces the rotor current oscillationTherefore the selection of virtual resistance should considerthe inhibition of the converter rotor voltage
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 9
minus2
0
2
Vs
(pu)
051
15
i rq
(pu)
minus020
0204
Vrq
(pu)
2 25 3 35 4 45 5 55 6minus2
0
2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
2 25 3 35 4 45 5 55 6
t (s)
0
2
Vs
(pu)
05
1
15
i rq
(pu)
00204
Vrq
(pu)
0
2
i ra
(pu)
minus2
minus2
minus02
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 9 Rotor current voltage with variable damping control in subsynchronization
02
012
002
2 25 3 35 4 45 5 55 60
1
2
minus2Vs
(pu)
i rq
(pu)
minus02Vrq
(pu)
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
02
012
002
2 25 3 35 4 45 5 55 6012
t (s)
Vs
(pu)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 10 Rotor current voltage with variable damping control in synchronization
minus04
0
2
0
1
2
002
2 25 3 35 4 45 5 55 6
0
2
minus2
i rq
(pu)
minus02
Vrq
(pu)
Vsq
(pu)
minus2
t (s)
i ra
(pu)
(a) Rotor current and voltage with 05 pu virtual resistance
0
2
0
1
2
002
0
2
Vs
(pu)
2 25 3 35 4 45 5 55 6
t (s)
i rq
(pu)
Vrq
(pu)
i ra
(pu)
minus2
minus2
minus02minus04
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 11 Rotor current and voltage with variable damping control in supersynchronization
8 Conclusion
This paper analyzes the DFIG electromagnetic transientprocess under a grid voltage swell introduces active dampingcontrol to DFIG rotor excitation control based on passivedamping control and proposes an improved control strategyfor variable damping and the acquisition of virtual resistances
on the basis of the EPL of dynamic PSO algorithmDue to therapidity and accuracy of the algorithm the virtual resistancevalue is more accurate The control method based on activedamping inhibited the rotor current oscillation caused bythe grid voltage swell reduced the effect of electromagnetictorque oscillation in the system during HVRT and reducedthe crowbar motion and its adverse effects Variable damping
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
10 Mathematical Problems in Engineering
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 12 Rotor current and voltage with variable damping coefficient in subsynchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 13 Rotor current and voltage with variable damping coefficient in synchronization
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(a) Rotor current and voltage with 05 pu virtual resistance
ug (300Vdiv)
ird (5Adiv)
Vrq (100Vdiv)
t (200msdiv)
ira (10Adiv)
(b) Rotor current and voltage with 15 pu virtual resistance
Figure 14 Rotor current and voltage with variable damping coefficient in supersynchronization
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Mathematical Problems in Engineering 11
control over the effect of different rotation speeds and therange of grid voltage swell in the system improved the HVRTcontrol of DFIG
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgment
The project was supported by National Natural ScienceFoundation of China (51277050)
References
[1] H Nian Y Song P Zhou and Y He ldquoImproved direct powercontrol of a wind turbine driven doubly fed induction generatorduring transient grid voltage unbalancerdquo IEEE Transactions onEnergy Conversion vol 26 no 3 pp 976ndash986 2011
[2] J Hu Y He L Xu and B W Williams ldquoImproved control ofDFIG systems during network unbalance using PI-R currentregulatorsrdquo IEEE Transactions on Industrial Electronics vol 56no 2 pp 439ndash451 2009
[3] J Lopez P Sanchis X Roboam and L Marroyo ldquoDynamicbehavior of the doubly fed induction generator during three-phase voltage dipsrdquo IEEE Transactions on Energy Conversionvol 22 no 3 pp 709ndash717 2007
[4] S Xiao G Yang H Zhou and H Geng ldquoAn LVRT controlstrategy based on flux linkage tracking for DFIG-basedWECSrdquoIEEE Transactions on Industrial Electronics vol 60 no 7 pp2820ndash2832 2013
[5] H Geng C Liu and G Yang ldquoLVRT capability of DFIG-based WECS under asymmetrical grid fault conditionrdquo IEEETransactions on Industrial Electronics vol 60 no 6 pp 2495ndash2509 2013
[6] J P da Costa H Pinheiro T Degner and G Arnold ldquoRobustcontroller for DFIGs of grid-connected wind turbinesrdquo IEEETransactions on Industrial Electronics vol 58 no 9 pp 4023ndash4038 2011
[7] O Abdel-Baqi and A Nasiri ldquoA dynamic LVRT solution fordoubly fed induction generatorsrdquo IEEE Transactions on PowerElectronics vol 25 no 1 pp 193ndash196 2010
[8] S Muller M Deicke and R W de Doncker ldquoDoubly fedinduction generator systems for wind turbinesrdquo IEEE IndustryApplications Magazine vol 8 no 3 pp 26ndash33 2002
[9] H Zhou G Yang and JWang ldquoModeling analysis and controlfor the rectifier of hybrid HVdc systems for DFIG-based windfarmsrdquo IEEE Transactions on Energy Conversion vol 26 no 1pp 340ndash353 2011
[10] B C RabeloWHofmann J L da Silva R G deOliveira and SR Silva ldquoReactive power control design in doubly fed inductiongenerators for wind turbinesrdquo IEEE Transactions on IndustrialElectronics vol 56 no 10 pp 4154ndash4162 2009
[11] G Iwanski and W Koczara ldquoDFIG-based power generationsystem with UPS function for variable-speed applicationsrdquoIEEE Transactions on Industrial Electronics vol 55 no 8 pp3047ndash3054 2008
[12] N Patin E Monmasson and J-P Louis ldquoModeling and controlof a cascaded doubly fed induction generator dedicated to
isolated gridsrdquo IEEE Transactions on Industrial Electronics vol56 no 10 pp 4207ndash4219 2009
[13] A Luna F K de Araujo Lima D Santos P RodrıguezE H Watanabe and S Arnaltes ldquoSimplified modeling of aDFIG for transient studies in wind power applicationsrdquo IEEETransactions on Industrial Electronics vol 58 no 1 pp 9ndash202011
[14] K Rothenhagen and F W Fuchs ldquoCurrent sensor fault detec-tion isolation and reconfiguration for doubly fed inductiongeneratorsrdquo IEEE Transactions on Industrial Electronics vol 56no 10 pp 4239ndash4245 2009
[15] F Bonnet P E Vidal and M Pietrzak-David ldquoDual directtorque control of doubly fed induction machinerdquo IEEE Trans-actions on Industrial Electronics vol 54 no 5 pp 2482ndash24902007
[16] J Lopez E Gubıa E Olea J Ruiz and L Marroyo ldquoRidethrough of wind turbines with doubly fed induction generatorunder symmetrical voltage dipsrdquo IEEE Transactions on Indus-trial Electronics vol 56 no 10 pp 4246ndash4254 2009
[17] Z Xie Q Shi H Song X Zhang and S Yang ldquoHigh voltageride through control strategy of doubly fed induction windgenerators based on active resistancerdquo in Proceedings of theIEEE 7th International Power Electronics and Motion ControlConference (IPEMC 12) pp 2193ndash2196 IEEE Harbin ChinaJune 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of