Date post: | 07-Feb-2016 |
Category: |
Documents |
Upload: | shin-mb-moh |
View: | 6 times |
Download: | 0 times |
Energy Conversion and Management 82 (2014) 11–26
Contents lists available at ScienceDirect
Energy Conversion and Management
journal homepage: www.elsevier .com/ locate /enconman
Modeling, control and fault diagnosis of an isolated wind energyconversion system with a self-excited induction generator subjectto electrical faults
http://dx.doi.org/10.1016/j.enconman.2014.02.0680196-8904/� 2014 Elsevier Ltd. All rights reserved.
⇑ Corresponding author.E-mail addresses: [email protected], [email protected] (I. Attoui), omeiri.
[email protected] (A. Omeiri).
Issam Attoui a,⇑, Amar Omeiri b
a Welding and NDT Research Centre, BP 64, Cheraga, Algeriab Electrical Laboratory LEA, Badji Mokhtar-Annaba University, BP 12, Annaba 23000, Algeria
a r t i c l e i n f o a b s t r a c t
Article history:Received 25 November 2013Accepted 27 February 2014Available online 19 March 2014
Keywords:Wind energy conversion systemsFault detection and diagnosisSEIGFractional-Order Controllers
In this paper, a contribution to modeling and fault diagnosis of rotor and stator faults of a Self-ExcitedInduction Generator (SEIG) in an Isolated Wind Energy Conversion System (IWECS) is proposed. In orderto control the speed of the wind turbine, while basing on the linear model of wind turbine system about aspecified operating point, a new Fractional-Order Controller (FOC) with a simple and practical designmethod is proposed. The FOC ensures the stability of the nonlinear system in both healthy and faulty con-ditions. Furthermore, in order to detect the stator and rotor faults in the squirrel-cage self-excited induc-tion generator, an on-line fault diagnostic technique based on the spectral analysis of stator currents ofthe squirrel-cage SEIG by a Fast Fourier Transform (FFT) algorithm is used. Additionally, a generalizedmodel of the squirrel-cage SEIG is developed to simulate both the rotor and stator faults taking iron loss,main flux and cross flux saturation into account. The efficiencies of generalized model, control strategyand diagnostic procedure are illustrated with simulation results.
� 2014 Elsevier Ltd. All rights reserved.
1. Introduction
The recent technological developments of the wind turbinesystems focus on maintenance cost reduction [1–3], productioncapacity improvement [4,5], control stability and operationalreliability [6–8]. The use of an induction machine as a generatorin almost all types of these systems [9,10] as in isolated windenergy conversion system justifies the importance of thesupervision of their normal operations. A variety of faults can oc-cur within the induction machine [11], such as the rotor cagemalfunctions and the stator phase unbalance, that may resultin a complete breakdown if the progress of the fault is notdetected.
Generally, the fault diagnosis method consists of creating thereal fault into the physical system, and evaluating its effect on dif-ferent measured variables. Such approach can be dangerous for thegenerator and may lead to the destruction of the wind turbine.Therefore, adequate models of the induction machine for studyingthe behavior of the wind energy conversion system are needed and
remain an effective tool to predict the performance of the IWECSunder fault conditions.
The classical dynamic models incorporating the effect of satura-tion in the magnetic circuit of the SEIG which are based on station-ary reference frame d–q axes theory as in [12–16] suppose thatboth stator and rotor windings are symmetric. In these models,the equivalent resistance matrix is diagonal and the equivalentinductance matrix is symmetric. These models are simple for sim-ulation but they cannot reflect any asymmetries due to stator orrotor faults.
In the literature, other analytical models of induction ma-chines are still the most common choices for the emulation ofstator and rotor faults [17–19]. The simplest method to simulatea stator fault is to insert an additional resistance in series to onephase stator winding in order to cause a stator phase unbalance.While emulating a rotor fault, the classical model takes into ac-count the individual conductors in the rotor cage using R–L ser-ies circuits, with current loops defined by two adjacent rotorbars connected by portions to the end ring. When a broken rotorbar fault is considered, the corresponding bar resistance value isassumed to be high value to force the current in the bar to zero[20]. Therefore, for squirrel-cage self-excited induction generator,a model with Nr rotor bars takes into account the nonlinearity ofthe magnetic circuit characteristic of the machine. This approach
i j+1i j Re
Lb
Lb
Rb
LeRe
LeRe
12 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
leads to a very complicated transient model. Such a model isquite complex and its computer simulation becomes very long.So, it will not be used in this work.
In this paper, a new model based on stationary frame d–q of theSEIG incorporating the effect of stator and rotor asymmetries (dueto faults) is developed. The effects of iron losses, main flux andcross flux saturation are also incorporated in the model. The modelsupposes that stator resistances are not equal. This situation isequivalent to a dissymmetry winding due to faults in the SEIG sta-tor such as the inter-turn short circuit. The rotor is decomposed toNr loops. For a healthy rotor, the rotor loops are identical and havethe same parameters that make this model similar to the classicald–q model, but when a rotor fault occurs, some loops are affected.In this condition, the equivalent resistance of the rotor in the twoaxes d–q model is not diagonal, making this model more general-ized than the classical d–q model.
Fault diagnosis allows the detection and the identification ofany deviation of the operating parameters from their normal or ex-pected values [21,22]. Most popular methods of induction machinecondition monitoring use the steady-state spectral components ofthe current variables [23,24]. In this paper, in order to diagnosisthe stator and rotor faults in the SEIG, we use a diagnostic proce-dure based on the stator current analysis using FFT algorithm thathas the ability to identify and isolate certain frequency compo-nents of interest [25].
Recently, application of fractional calculus in the control areais increasingly used [26–30]. In comparison with the classical or-der controllers, the fractional order controllers have a potentialto improve the control performance [27], and increase thesystem robustness because of extra real parameters involved[31]. In this paper, according to the imposed three tuning con-straints, to guarantee control performance and the robustnessto the loop gain variations, a new fractional order controller(Fractional Order Proportional, Integral, Derivative and Integratororder Derivative PIaDlD) is proposed and designed for the fixed-speed operation of wind turbines by adjusting the blade-pitchangle.
The paper is organized as follows. The model of the systemrepresented by Fig. 1 (SEIG, actuator and turbine) is developedin Section 2. The pitch control design, based on proposedfractional-order PIaDlD controller, is represented in Section 3.The diagnosis procedure is represented in Section 4. The simula-tion results are given in Section 5. The conclusion is given inSection 6.
2. Modeling of the system
2.1. Modeling of the squirrel-cage SEIG
2.1.1. Stator modeling taking into account stator faultThe dynamic model for the stator windings of the three-phase
squirrel cage induction generator is developed and the relevantvolt–ampere equations are:
Fixed capacitor bank
Wind direction
RL load
CC
C
Gear box
Induction generator
Fig. 1. Basic schema: SEIG with a capacitor excitation system driven by a windturbine.
�Vs ¼ RsIs þd/s
dtð1Þ
Vs ¼ ½Vsa;Vsb;Vsc�T is the stator voltage vector; Is ¼ ½isa; isb; isc�T is thestator current vector; Rs ¼ diag½Rsa;Rsb;Rsc� is an 3 by 3 stator resis-tance matrix; /s ¼ ½/sa;/sb;/sc�
T is a stator flux vector.With the Park transformation (Ps), the voltage equations for the
stator windings can be written as:
�Vsdq¼ PsRsP�1s IsdqþPs
d P�1s /sdq
� �dt
¼RSDQ Isdqþd/sdq
dtþxs
0 �11 0
� �/sdq
ð2Þ
where RSDQ is the equivalent resistance matrix and is given by:
RSDQ ¼ PsRsP�1s ¼
Rds Rsdq
Rsdq Rqs
� �;
Ps ¼ffiffiffi23
rsinðhÞ sin h� 2p
3
� �sin hþ 2p
3
� �cosðhÞ cos h� 2p
3
� �cos hþ 2p
3
� �" #ð3Þ
To emulate a stator phase unbalance, we insert an additionalresistance Rd in series to one phase stator winding (Rsa = Rs + Rd).
2.1.2. Rotor modeling taking into account rotor faultA squirrel-cage rotor is often modeled by NR meshes, as shown
in Fig. 2. Each mesh is substituted by an equivalent circuit repre-senting the resistive and inductive nature of the cage [32,33].
From the rotor cage equivalent circuit represented by Fig. 2, theelectric equation of the j-th loop can be defined as:
0 ¼ ðRbðjÞ þ Rbðj�1Þ þ 2ReÞij � Rbðj�1Þij�1 � RbðjÞijþ1 þd/rj
dtð4Þ
Expression of /rj is given by:
/rj ¼ ðLrp þ 2ðLb þ LeÞÞij þXNr
k ¼ 1k–j
Mrrirk þX3
k¼1
Mrjskisk � Lbiðj�1Þ � Lbi jþ1ð Þ
ð5Þ
Thus, the rotor electric equations can be written as:
0 ¼ RIr þddt
/r
/r ¼ ½/r1;/r2; . . . ;/rNr�
Ir ¼ ½ir1; ir2; . . . ; irNr �
ð6Þ
where R is the equivalent rotor loop resistance matrix, its expres-sion is given by:
Le
Le
Re
Le
Re
LbRb
i j-1
Rb
Le
Re
LeRe Le
Re
Le
Re
Lb
Rb
Le
Re
Le
Re
i 1
Lb
RbLeRe
Lb
Rb
LeRe
Le
Re
Fig. 2. Rotor cage equivalent circuit.
R ¼
Rb1 þ 2Re þ RbðNrÞ �Rb1 0 0 . . . �RbðNrÞ
�Rb1 Rb2 þ 2Re þ Rb1 �Rb2 0 . . . 00 �Rbðk�1Þ Rbðk�1Þ þ 2Re þ RbðkÞ �RbðkÞ . . . 0
. . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
�RbðNrÞ 0 0 . . . �RbðNr�1Þ RbðNr�1Þ þ 2Re þ RbðNr Þ
2666666664
3777777775ð7Þ
I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26 13
Rb and Re are: the bar resistance and the end-ring segment resis-tance, respectively.
A change of variables which formulates transformation of the Nr
rotor variables to the arbitrary two axes reference may be done bythe use of Kr transformation matrix given by:
Kr ¼
ffiffiffiffiffiffi2Nr
scos hs � hr � Ppar
2
� �. . . cos hs � hr � Pp 2j�1ð Þar
2
� �sin hs � hr � Ppar
2
� �. . . sin hs � hr � Ppð2j�1Þar
2
� �264
375ð8Þ
with Pp is the number of pole-pairs, ar = 2p/Nr and hs, hr are the twoaxes reference angular velocity.
By the use of Kr, the relation between (d,q) components of rotorcurrents and rotor current loops can be written as:
idqr ¼ Krir ð9Þ
Consequently, the rotor electric Eq. (4) becomes:
0 ¼ mf KrRK�1r Irdq þ
d/rdq
dtþ ðxs �xrÞ
0 �11 0
� �/rdq ð10Þ
While comparing with the traditional Park model of the rotor’svoltage equations in (11),
Vrdq ¼ RRDQ Irdq þd/rdq
dtþ xs �xrð Þ
0 �11 0
� �/rdq ð11Þ
we obtain the expression of rotor resistance:
RRDQ ¼ mf KrRK�1r
� �¼
Rdr Rrdq
Rrdq Rqr
� �ð12Þ
The values of RRDQ must be considered with a multiplicativeequivalence factor (mf) given by:
mf ¼Rr
ffiffiffiffi2
Nr
q2 Re
Nrþ 2Rbð1� cosðPparÞÞ
ð13Þ
The rotor resistance in Park transformation is calculated accord-ing to the bar and the end-ring segment resistances. A sharp vari-ation of rotor resistance is directly related to a defect in the rotorbars.
2.1.3. Stator and rotor flux modelingThe stator and rotor fluxes in the dq reference frame are given,
respectively, by:
/sd ¼ lsisd þ /md
/sq ¼ lsisq þ /mq;
/rd ¼ lr ird þ k/md
/rq ¼ lr irq þ k/mqð14Þ
where /md and /mq are, respectively, the direct and quadrate mag-netizing air–gap fluxes and are defined as:
/md ¼ Mdimd þMdqimq þ /d0
/mq ¼ Mqimq þMdqimd þ /q0ð15Þ
where Md, Mq and Mdq are new inductive parameters.With [34] one gets:
Md ¼@/md
@imd; Mq ¼
@/mq
@imq; Mdq ¼
@/md
@imq¼@/mq
@imdð16Þ
The analytical expressions of main and cross flux saturation de-pended coefficients are expressed as follows [34]:
Md ¼ Lm þ@Lm
@imdimd ¼ Lm þ
dLm
djImji2md
Im
Mq ¼ Lm þdLm
dIm
i2mq
Im; Mdq ¼
dLm
djImjimdimq
Im
ð17Þ
Therefore, the (d–q) components of the [im]123 system takingiron losses into account satisfy:
imd ¼Rm
Rm þ Lmpðisd þ kirdÞ
imq ¼Rm
Rm þ Lmpðisq þ kirqÞ
ð18Þ
and
i2m ¼ i2
md þ i2mq
� �=2 ð19Þ
where k (k = 1 in this work) is the transformation report of the SEIG.The iron losses are represented by means of an equivalent iron
loss resistance Rm connected in parallel with the magnetizinginductance Lm [35]. The variation in Rm is modeled by the followingcurve fit
Rm ¼ Vph þ 1200 ð20Þ
where Vph is the phase RMS voltage.
2.1.4. The generalized two axes SEIG modelThe generalized two axes model of a squirrel-cage self-excited
induction generator to simulate both the rotor and stator faults,taking into account iron loss, main flux and cross flux saturation,is presented by:
½v� ¼ ½RT �½i� þ ½LT �ddt½i� þ ½GT �½i� ð21Þ
The matrices of Eq. (21) are defined as:
½v� ¼ ½�Vds � VqsVdrVqr�T ; ½i� ¼ ½idsiqsidriqr �T ;
½RT � ¼ ½RsdqRrdq�T ; Rsdq ¼ PsRsP�1s ; Rrdq ¼ mf KrRK�1
r :
½LT � ¼
ls þMdNEW MdqNEW MdNEW MdqNEW
MdqNEW ls þMqNEW MdqNEW MqNEW
MdNEW MdqNEW lr þMdNEW MdqNEW
MdqNEW MqNEW MdqNEW lr þMqNEW
2666437775
½GT � ¼
0 �xsðls þ LmNEWÞ 0 �xsLmNEW
xsðls þ LmNEWÞ 0 xsLmNEW 00 �ðxs �xrÞLmNEW 0 �ðxs �xrÞðls þ LmNEW Þ
ðxs �xrÞLmNEW 0 ðxs �xrÞðls þ LmNEWÞ 0
2666437775
14 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
where
MdqNEW ¼ MdqRm
Rm þ Lmp; MdNEW ¼ Md
Rm
Rm þ Lmp;
MqNEW ¼ MqRm
Rm þ Lmp; LmNEW ¼ Lm
Rm
Rm þ Lmp
The definition of electromagnetic torque is
TSEIG ¼32
PpRm
Zðiqsidr � idsiqrÞdt �
ZRm
LmTSEIG
dt
¼ 32
PpRmLm
Rm þ Lmpðiqsidr � idsiqrÞ ð22Þ
Fig. 3. Block diagram of the linear wind turbine model.
2.2. Load (RL) and Fixed Capacitor Bank model
Assume that the load is an RL (per phase value) series circuitconnected to the stator winding. The voltages and currents equa-tions in this case will be given by
ddt Vds ¼ 1
C ðids � idLÞ þxsVqs
ddt idL ¼ 1
LchðVds � RLidLÞ þxsiqL
ddt Vqs ¼ 1
C ðiqs � iqLÞ �xsVds
ddt iqL ¼ 1
LchðVqs � RLiqLÞ �xsidch
8>>>>><>>>>>:ð23Þ
where iqL and idL are the q- and d-axis load currents.Under no-load conditions; the minimum capacitance value
needed for the onset of self-excitation is approximated as [36]:
Cmin �1
x2r Ln
m
ð24Þ
Any change in load or rotor speed may result in a loss of excita-tion. Hence, it is recommended to use about 25% overestimatedcapacitance values C = 1.25Cmin [13].
2.3. Model of wind turbine
The mathematic expressions of wind turbine are expressed asfollows [37,38]:
Taer ¼ GTg ¼ Pw=Xt ¼ GPw=Xmec ¼12
Cpðk; bÞqARt3w=Xt ð25Þ
Cpðk; bÞ ¼ 0:5176116ki� 0:4b� 5
e�21ki þ 0:0068k ð26Þ
1ki¼ 1
kþ 0:08b� 0:035
b3 þ 1ð27Þ
k ¼ XtRtw
ð28Þ
where Pw is the aerodynamic power captured from the wind; q isthe density of air; AR is the surface covered by the wind wheel; Cp
is the power coefficient; tw is the wind speed; k is the tip speed ra-tio (TSR); b is the pitch angle; Xt is the wind turbine mechanicalangular velocity; Taer is the wind turbine output torque; Tg is thedriving torque of the generator, G is the gear ratio and Xmec is thegenerator mechanical angular velocity.
Because the torque coefficient is related to the power coeffi-cient, CP, through the following relation:
Cpðk;bÞ ¼ kCqðk; bÞ ð29Þ
The manipulation of the torque coefficient using k and b will re-sult in the manipulation of the power produced by the turbine.
The fundamental dynamics of the wind turbine are captured bythe following simple mathematical model:
Tg � TSEIG ¼ JtdXt
dtð30Þ
2.4. Wind turbine model linearization
In order to design a robust linear system controller such as PIa-
DlD, it is first necessary to consider an appropriate linear dynamicmodel of the system. This requires that the non-linear turbinedynamics are to be linearized about a specified operating point.But, because the real turbine model is nonlinear, the deviation ofthe turbine operating point provokes variations in the linearizedwind turbine model parameters. This requires that the system con-troller is to be robust to parameter variation.
The linearization of the turbine Eq. (30) would yield [39,40]:
JtD _Xt ¼ cDXt þ nDtw þ dDb ð31Þ
Where the linearization coefficients are given by:
c ¼ @Tm@Xt
���op¼ @
@XtðJt
_XtÞ���op¼ 0:5qARt3
wop@@Xt
Cpðk;bÞXt
h i���op
n ¼ @Tm@tx
���op¼ @
@twJt
_Xt
� ����op¼ 0:5qAR
1Xtop
@@tx
Cpðk;bÞ � t3w
� ��op
d ¼ @Tm@b
���op¼ @
@b Jt_Xt
� ����op¼ 0:5qAR
t3wop
Xtop
@@b ½Cpðk;bÞ�
��op
8>>>>><>>>>>:ð32Þ
where kop = RXtop/txop
Here, DX, Dtx, and Db represent deviations from the chosenoperating point, Xtop, txop, and bop. The rotational speed operatingpoint, Xtop, is selected to be the desired constant speed of the tur-bine, 450 rpm (47.1 rad/s). The blade-pitch and wind speed operat-ing points are selected as (bop = 16� and txop = 11 m/s).
After Laplace transformation, the Eq. (31) becomes:
JtpDXt ¼ cDXt þ nDtx þ dDbðpÞ ð33Þ
Let
D ¼ cJt
ð34Þ
The turbine rotor shaft speed can be represented as
Reference Speed (Ωt-ref)
Actuator+ Rate limiter
+ MDZ
Pitch Angle Limits0°: 90°
tNon-Linear
Turbine +
SEIG
Wind Speed
-
+ΔΩt Δβ
++PIαDµDController
Reference Pitch (βref)
Ω
Fig. 4. Simulation of the block diagram of controlled system.
I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26 15
DXt ¼nJt
DtxðpÞ þdJt
DbðpÞ� �
1p� D
ð35Þ
Eq. (35) describes the linearized model of the wind turbine.Such model is represented by the block diagram shown in Fig. 3.
For the control of the wind turbine speed, the factor Dtx(p)�(n/Jt) is seen as a constant perturbation and the parameters (�1/D),(�d/JtD) are the time constant and static gain respectively. Whenthe turbine operating point deviates with a constant wind turbinespeed, the static gain parameter gets the most significant variation(Appendix C). For this reason, a robust control strategy to gain vari-ations of the system should be considered. In this work, the frac-tional order PIaDlD controller is proposed.
2.5. Model of actuator
The actuator turbine blade adjustment, which may be repre-sented by the transfer function given in (36) where ba(p) andbo(p) are the Laplace transform of the pitch angle input and outputrespectively, tm is the time constant.
b0ðpÞbaðpÞ
¼ 1tmpþ 1
ð36Þ
2.6. Model of wind speed
The wind speed can be modeled as a deterministic sum of sev-eral harmonics, as in [41]:
tx ðtÞ ¼ 11þ 0:2 sinð0:1047 tÞ þ 2 sinð0:2665 tÞþ sinð1:2930 tÞ þ 0:2 sinð3:6645 tÞ ð37Þ
3. Control system
The system of the studied device including speed control isshown in Fig. 4.
The wind turbine control algorithm has focused on maintainingconstant rotor speed. In this case, the input is generally the blade-pitch angle (b) and the output is the rotor speed (Xr).
The PIaDlD controller is used for controlling the rotor speed,where DXt(p) represents the input rotor speed (error signal), andDb(p) represents the output pitch angle change.
A dead zone (MDZ) is added when the command pitch rate isless than a certain value (0.1 deg/s in this study) to eliminate thenoise in the command signal and reduce actuator motion for alonger lifetime [42]. The pitch rate commanded by the actuator isphysically limited to 10� degrees per second according to manufac-turer recommendations.
3.1. Fractional order controllers
Fractional calculus gives a generalization of ordinary differenti-ation and integration to arbitrary (non-integer) order [28].
The fractional-order differentiator can be denoted by a generaloperator aDa
t [43], given by:
aDat ¼
da
dta; RðaÞ > 0
1; RðaÞ ¼ 0R ta ðdsÞ
�a; RðaÞ < 0
8>><>>: ð38Þ
where a is the order of derivative or integrals, R(a) is the real part ofthe a.
The called Riemann–Liouville definition of fractional derivativesand integrals is
aD�at f ðtÞ ¼ 1
CðaÞ
Z t
aðt � sÞa�1f ðsÞds ð39Þ
aDat f ðtÞ ¼ 1
Cðn� aÞdn
dtn
Z t
a
f ðsÞdsðt � sÞk�nþ1 ds
" #ð40Þ
where:
CðxÞ ¼Z 1
0yx�1e�ydy ð41Þ
is the Euler’s Gamma function, a and t are the limits of the opera-tion, and a is the number identifying the fractional order. In this pa-per, a is assumed as a real number that satisfies the restrictions0 < a 6 1. Also, it is assumed that a = 0. The following conventionis used: 0D�a
t � D�at .
In this work, the Oustaloup’s approximation method [44] isused. Assuming the frequency range to fit is selected as (xb, xh),the Oustaloup algorithm is based on the approximation of a func-tion of the form:
HðpÞ ¼ pa; a 2 Rþ ð42Þ
by a rational function:
bHðpÞ ¼ CYN
k¼�N
pþx0kpþxk
ð43Þ
where the zeros, poles and the gain can be evaluated from
x0k ¼ xbxh
xb
kþNþ12ð1�aÞ
2Nþ1
; xk ¼ xbxh
xb
kþNþ12ð1þaÞ
2Nþ1
; C ¼ xh
xb
�a2 YN
k¼�N
xk
x0k
The differential equation of the proposed fractional-order PIaDl-
D controller, 0 < l and k < 2, in time domain, is given by:
uðtÞ ¼ Kp eðtÞ þ KiD�at eðtÞ þ Kd1Dl
t eðtÞ þ Kd2DteðtÞ� �
ð44Þ
where Kp is a proportional constant, Ki is an integration constant.Kd1 and Kd2 are the derivation constants. Taking l = 1, a = 1 andKd2 = 0 in (44), a classical PID controller is obtained. Hence, using La-place transforms the transfer functions of the proposed fractionalorder PIaDlD controller and classical PID controller are respectivelygiven by:
CðpÞ ¼ Kp 1þ Ki
pa þ Kd1pl þ Kd2p
ð45Þ
GðpÞ ¼ Kp 1þ Ki
pþ Kdp
ð46Þ
16 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
The fractional-order PIaDlD controller is more flexible than theclassical PID controller, because it has one more adjustableparameter.
3.2. Design specifications
Assuming that the gain crossover frequency is xc and the phasemargin is /m, for the system stability and robustness, three speci-fications concerned with the phase and the gain of the open-looptransfer function are proposed as follows [45,46],
(i) The phase margin specification
Arg½GðjxcÞ� ¼ Arg½CðjxcÞPðjxcÞ� ¼ �pþ /m
(ii) The gain specification at the crossover frequency
jGðjxcÞjdB ¼ jCðjxcÞPðjxcÞjdB ¼ 0
(iii) Robustness to variation in the gain of the plant demands thatthe phase derivative w.r.t. the frequency is zero, i.e., thephase Bode plot is flat, at the gain crossover frequency. Itmeans that the system is more robust to gain changes andthe overshoots of the response are almost the same.
dðArgjGðjxÞjÞdx
x¼xc
¼ 0
3.3. FO-PIDD controller design procedure to wind turbine speedcontrol
The block diagram of the wind turbine speed Xt control is givenby Fig. 5:
with
FðpÞ ¼ d=Jt
p� D
1
tmpþ 1
¼ K
1tmpþ 1
1
� 1D pþ 1
!and K ¼ ð�d=DJtÞ ð47Þ
Note: D < 0 and d < 0.The open-loop transfer function G(p) of the fractional order PIa-
DlD controller for the wind turbine speed control is that,
GðpÞ ¼ CðpÞFðpÞ
¼ Kp 1þ Ki
paþ Kd1pl þ Kd2p
:
d=Jtð Þðtmpþ 1Þðp� DÞ
ð48Þ
The proposed procedure of the FO-PIDD controller design issummarized in two steps as:
3.3.1. First step: poles compensationThe first step of the proposed FO-PIDD controller design proce-
dure consists in poles compensating of the actuator system by thezeros of the regulator.
The compensation of the time-constant tm leads to choose
Kd1 ¼ Kitm
Kd2 ¼ tm
l ¼ 1� a
8><>: ð49Þ
We obtain in open loop, the transfer function G(p)
Fig. 5. Block diagram of the wind turbine speed control.
GðpÞ ¼ CðpÞFðpÞ
¼ Kp 1þ Ki
pa
� ðtmpþ 1Þ � ðd=JtÞ
ðtmpþ 1Þðp� DÞ
ð50Þ
3.3.2. Seconde step: determination of Kp, Ki and a according toimposed three tuning constraints
According to the open loop transfer function G(p) from (50), wecan get its frequency response as follows,
GðjxÞ ¼ Kp 1þ Ki
ðjxÞa
ðtmðjxÞ þ 1Þ
� d=Jt
ðtmðjxÞ þ 1ÞððjxÞ � DÞ
¼ KpK 1þ Ki
ðjxÞa
� 1� jx
D þ 1
!ð51Þ
with
K ¼ � dJt
1D¼ � d
c
According to Specification (i), the phase of G(jx) can be ex-pressed as,
ArgjGðjxcÞj ¼ � arctanKix�a
c sinðap=2Þ1þ Kix�a
c sinðap=2Þ þ arctanðxc=DÞ
¼ �pþ /m ð52Þ
From (52), the relationship between Ki and a can be established asfollows,
Ki ¼� tan½� arctanðxc=DÞ þ /m�
x�ac sinðkp=2Þ þx�a
c cosðkp=2Þ tan½� arctanðxc=DÞ þ /m�ð53Þ
According to specification (iii) about the robustness to gain vari-ations in the plant,
d Arg G jxð Þj jð Þdx
x¼xc
¼ Kikxa�1c sin kp=2ð Þ
x2ac þ 2Kixa
c cosðkp=2Þ þ K2i
þ 1=D
1þ ðxc=DÞ2
¼ 0 ð54Þ
From (54), we can establish an another equation about Ki in the fol-lowing form,
Cx�2ac K2
i þ rKi þ C ¼ 0 ð55Þ
that is
Ki ¼�r�
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffir2 � 4C2x�2a
c
q2Cx�2a
cð56Þ
where
C ¼ � 1=D
1þ ðxc=DÞ2and r ¼ 2Cx�a
c cosðap=2Þ � kx�a�1c sinðap=2Þ
ð57Þ
According to Specification (ii), we can establish an equation aboutKp,
KpKffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ Kix�a
c cosðap=2Þ� 2 þ Kix�a
c sinðap=2Þ� 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiD2 þ ðxcÞ2
q0B@
1CA ¼ 1 ð58Þ
Clearly, we can solve Eqs. 53, 56 and 58 to get a, Ki and Kp.
NOEnd
Activate Alarms/Warnings
YES Operator ControlShutdown Equipment
NOYES
Fault component detection
Fault componentsdetection at
frequencies equal to 3fs (stator fault)and/or (1±2s) for
(rotor fault). Faulty (Stator or rotor) Healthy
Stator current data ias (t).
Begin
Data vector x(t-0.5T:t+0.5T)
t = t + T
FFT computation
FFT is computed on the stator
current data x(t)
s, r
Move the temporal window
Ω Ω
Fig. 7. Flowchart of the fault diagnosis procedure.
I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26 17
Using a graphical method [33], the used procedure for thedetermination of the a, Ki and Kp is summarized as.
(1) Given xc = 30 (rad/s), the gain crossover frequency.(2) Given um = 70�, the desired phase margin.(3) Plot curve 1, Ki w.r.t a, according to (50), and plot curve 2, Ki
w.r.t a, according to (53).Fig. 6a shows the two curves.
(4) Obtain the values of a and Ki from the intersection point onthe above two curves.a = 0.2947, Ki = 10.77, (Kd1 = tm = 0.25, Kd2 = Kitm = 2.6925and l = 1 � a = 0.7053).
(5) Calculate the Kp from (55), Kp = �3.6750.
Then, we can fix the open-loop frequency response by means ofFO-PIDD controller. The proposed FO-PIDD controller is,
CðpÞ ¼ �3:5420 1þ 10:77p0:2947 þ 2:6925p0:7053 þ 0:25p
ð59Þ
So, the Bode diagrams of designed system can be plotted as inFig. 7b and c, we can see that the phase Bode plot is flat, at the gaincrossover frequency, and the all three specifications are satisfied. Itmeans that the system is more robust to gain changes. But the threeSpecifications cannot be satisfied simultaneously for classical PID orFO-PI (a = 0.1089, Ki = �1.341, Kp = �659.94, Kd1 = Kd2 = 0) control-lers with the same design method (2nd step).
4. Implemented diagnosis procedure
During the past years, the use of induction machines in theindustry has shown that broken rotor bars and stator phase unbal-ance can cause serious mechanical damage to the insulation and aconsequential winding failure may follow, resulting in a costly re-pair and a considerable loss in production [47]. This section focuseson the application of stator current signature analysis to diagnose
Fig. 6. (a) Ki versus a; (b) bode plot of th
rotor and stator faults in three-phase squirrel-cage SEIG for isolatedwind energy conversion system. The procedure is an on-line mon-itoring technique that can reduce unexpected failures, downtime,and both maintenance and operational costs.
The implementation procedure can be illustrated with the flow-chart in Fig. 7. This procedure is based on stator current analysis byFFT algorithm that gives a frequency representation of the tempo-ral signal. TheFFT results from the healthy and faulty machine arefirst compared to identify the harmonics caused by the fault. These
e open loop transfer function G(p).
18 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
harmonics are then compared to the theoretical fault harmonics(Eqs. (57) and (58)). Recognition of fault type is based on differ-ences in spectral signatures on the line current of the faulty/healthy machine and theoretical fault spectra signatures.
4.1. Faulty rotor operating condition
As far as the rotor fault is concerned, a broken rotor bar is pro-duced by the assumed linear change of the value of the resistanceRbF named fault bar resistance. This asymmetry produces a chain offault components in the stator windings. The firsts are the wellknown harmonic components (fsa) at frequencies:
fsa ¼ ð1� 2sÞfs ð57Þ
where s is the slip and fs is the stator voltage frequency.
4.2. Faulty stator operating condition
As far as the stator fault is concerned, a simulation of the SEIGhas been performed with an additional resistance of the same va-lue of one stator phase resistance connected in series with one sta-tor phase winding. An unbalance in the stator produces in thestator current a component at frequency:
fsb ¼ kfs ð58Þ
with k = 3, 5, 7 . . ..
Fig. 8. (a) Wind speed profile; (b) rotational speed of the SEIG in c
5. Results and discussion
To verify the efficiencies of the generalized model, the controlstrategy and the diagnostic procedure, three simulation cases ofthe Isolated Wind Energy Conversion System (WECS) are pre-sented in Fig. 1 with the specified parameters given in theappendix A. The first case is the simulation of the SEIG modeland the regulation of the rotational speed Xt of the wind turbinein healthy conditions with wind speed fluctuations. The secondcase presents the same problems but the wind turbine is infaulty conditions with broken rotor bar. The third case presentsalso the same problems but the wind turbine is in faulty condi-tions with stator and rotor faults.
Case 1: The IWECS Simulation results in healthy conditions.The time-series traces of wind speed profile, rotor speed,
blade-pitch angle are represented in Fig. 8. The pitch rate doesnot exceed ±10 deg/s and does not less than a 0.1 deg/s. How-ever, the goal of maintaining constant rotational speed is metsatisfactorily as the wind speed fluctuates between 7.6 m/sand 14.4 m/s. Fig. 9 shows the dynamic characteristics of differ-ent parameters of the induction generator when driven by thewind turbine. At t = 0 s a capacitor with capacitance of 55 lFis connected at the stator terminals of the induction generatorwithout any load, and voltage is generated because of the avail-able rotor speed at t = 0 s. The generated voltage Vas, shown inFig. 9a, is expressed as an instantaneous value and rms value.At t = 3 s a RL load (R = 250 X, L = 200e�3H) is connected. Thegenerated voltage (Vas), magnetizing current (Im) and stator cur-
ase 1; (c) blade-pitch angle (b) and its reference (ba) in case 1.
Fig. 9. Case 1: (a) instantaneous value and rms value of the generated voltage Vas, (b) stator current ias; (c) load current ila; (d) magnetizing inductance Lm; (e) magnetizingcurrent Im; (f) electromagnetic torque TSEIG.
I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26 19
rent (ias) decrease. However, the load current (iLa), magnetizinginductance (Lm) and induced electromagnetic torque (TSEIG)increase.
At t = 7 s the load resistor is decreased to 200 X. The generatedvoltage, magnetizing current and stator current decrease further.
The load current and magnetizing inductance increase. However,the electromagnetic torque decreases.
At t = 11 s and t = 22 s the RL load is increased to 300 X, 150e�3and 300 X, 150e�3, respectively. The generated voltage, magnetiz-ing current (Im), stator current (ias) and induced electromagnetic
Fig. 10. Case 2: (a) wind speed profile; (b) rotational speed of the SEIG; (c) blade-pitch angle (b) and its reference (ba); instantaneous value and rms value of the generatedvoltage Vas.
20 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
torque increase. However, the load current (iLa) and magnetizinginductance (Lm) decrease.
Case 2: The IWECS simulation results in faulty conditions withbroken rotor bar.
The versus time curves of rotor speed, blade-pitch angle anddynamic characteristics of different parameters of the inductiongenerator when driven by the wind turbine are plotted in Figs.10 and 11, respectively with applied constant RL load(R = 250 X, L = 200e�3H) at t0 = 3 s. It is considered that the fail-ure of a bar starts at the moment t1 = 7 s. The oscillations of thedynamic characteristics of different parameters of the inductiongenerator justify the presence of a break rotor bar defect inthe machine.
Case 3: The IWECS simulation results in faulty conditions withstator and rotor faults.
The versus time curves of rotor speed, blade-pitch angle and dy-namic characteristics of different parameters of the induction gen-erator when driven by the wind turbine are plotted in Figs. 12 and13, respectively with applied constant RL load (R = 250 X,L = 200e�3H) at t = 3 s. It is considered that the stator unbalanceand break rotor bar starts at the moment t0 = 7 s and t1 = 15 s,respectively. At t0 = 7 s all currents, torque and generated voltageincrease. The oscillations of the dynamic characteristics of the dif-ferent parameters of the induction generator shown in Figs. 12 and13, respectively justify the presence of a stator unbalance at t = 7 sand the broken rotor bar at t = 15 s in the machine.
Fig. 11. Case 2: (a) stator current; (b) load current; (c) magnetizing inductance; (d) magnetizing current; (e) electromagnetic torque of the SEIG.
I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26 21
Fig. 14a shows the FFT of the stator current for the SEIG inhealthy condition. The spectrum is completely free of any currentcomponents around the main supply frequency, fs, and conse-quently, the expected faulty current frequency components, dueto broken rotor bars or stator unbalance, are empty. The SEIG thusshows no signs of broken rotor bars or stator unbalance.
Fig. 14b shows the FFT of the stator current for the SEIG case ofstator unbalance. The presence of the fault component at fre-quency 3fs in search band is evident. The component is a sign ofstator asymmetry.
While as Fig. 14c shows the FFT of the stator current in caseof rotor unbalance. The fault components are distributed sym-metrically around fs in search band. The components are a signof rotor asymmetry. While as Fig. 14d shows the FFT of the sta-tor current in case of stator and rotor asymmetry. The presenceof the fault components at frequencies 3fs, (1 ± 2 s) fs and3(1 ± 2 s) fs is evident. The rotor fault components are distributedsymmetrically around fs and 3fs. The components are a sign ofrotor and stator asymmetry.
6. Conclusion
In this paper, a new model of the squirrel-cage SEIG is devel-oped to simulate both the rotor and stator faults taking iron losses,main flux and cross flux saturation into account. This proposed
model can be utilized to study the behavior of the wind energyconversion systems with SEIG in healthy and faulty conditions,and provides an interesting tool to test the fault diagnosis proce-dures’ effectiveness.
The spectrum analysis of the stator currents shows an indica-tion of the SEIG health condition. It aims at detecting and diagnos-ing fault type (stator unbalance or/and bar broken) of the SEIG inwind turbine system.
A fractional-order PIDD controller with a practical and simpledesign method is proposed to control the wind turbine speed inhealthy and faulty conditions. Simulation results show that theclosed-loop wind turbine system can achieve favorable dynamicperformance and robustness. This control strategy can be also usedfor different wind turbine systems that use the blade-pitch anglecontrol.
The simulation results presented in this paper validate the pro-posed model, the chosen diagnostic method and the proposed con-trol scheme.
Appendix A
A.1. Simulation of dynamic model of SEIG
Rearranging Eq. (21) of the SEIG gives a 2nd order differentialequation represented by
Fig. 12. Case 3: (a) rotational speed of the SEIG; (b) zoom of the rotational speed of the SEIG; (c) blade-pitch angle (b) and its reference (ba); (d) zoom of the blade-pitch angle(b) and its reference (ba); (d) instantaneous value and rms value of the generated voltage Vas.
22 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
Fig. 13. Case 3: (a) stator current ias; (b) load current ila; (c) magnetizing inductance Lm; (d) magnetizing current Im; (e) electromagnetic torque of the SEIG TSEIG.
I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26 23
p2½i� ¼ ½A0�p½i� þ ½A1�½i� þ ½B0�p½v � þ ½B1�½v �
where
½A0� ¼
� RmLmþ MdRm
lsLmþ Rds
ls
� �� Rdqs
lsþ MdqRm
lsLm�xs
� ��MdRm
lsLm
� Rdqs
lsþ MdqRm
lsLmþxs
� �� Rm
Lmþ MqRm
lsLmþ Rqs
ls
� ��MdqRm
lsLm
�MdRmlr Lm
�MdqRm
lr Lm� Rm
Lmþ MdRm
lr Lmþ R
l
��MdqRm
lr Lm�MqRm
lr Lm� Rdqr
lrþ MdqRm
lr Lmþxs
�
2666666664
½v� ¼ ½VdsVqsVdrVqr�T ; ½i� ¼ ½idsiqsidriqr�T
�MdqRm
lsLm
�MqRm
lsLm
drr
�� Rdqr
lrþ MdqRm
lr Lm�xs þxr
� ��xr
�� Rm
Lmþ MqRm
lr Lmþ Rqr
lr
� �
3777777775
Fig. 14. (a) Stator current FFT in healthy condition; (b) stator current FFT with stator fault; (c) stator current FFT with rotor fault; (d) stator current FFT with stator and rotorfaults.
24 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
Fig. 15. Variations in the static gain (DG) and in the time constant (DT).
A1½ � ¼
� RmRdslsLm
� �� RmRdqs
lsLm� Rm
Lmxs � Rm
lsxs
� �0 Rmxs
ls
� RmRdqs
lsLmþ Rm
Lmxs þ Rm
lsxs
� �� RmRqs
lsLm
� �� Rmxs
ls0
0 Rmðxs�xrÞlr
� RrdRmlr Lm
� �� RdqrRm
lr Lm� xs �xrð Þ Rm
Lmþ Rm
lr
� �� �� Rmðxs�xrÞ
lr0 � RdqrRm
lr Lmþ ðxs �xrÞ Rm
Lmþ Rm
lr
� �� �� RrqRm
lr Lm
� �
2666666664
3777777775
I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26 25
½B0� ¼
� 1ls
0 0 0
0 � 1ls
0 0
0 0 1lr
0
0 0 0 1lr
2666664
3777775; ½B1� ¼
� RmlsLm
0 0 0
0 � RmlsLm
0 0
0 0 Rmlr Lm
0
0 0 0 Rmlr Lm
2666664
3777775
A.2. Results of linearization of the wind turbine equation (MatLabCode)
% MatLab Code.
ru = 1.225.R = 1.6.Ar = pi � R � R.syms wt vw beta Ar R ru.lambda = wt � R/vw.lambdai = 1/((1/(lambda + (0.08 � beta))) � (0.035/
(1 + (beta3)))).Cp = (0.5176 � ((116/ lambdai) � (0.4 � beta) � 5) � (exp(�21/
lambdai))) + (0.0068 � lambda).CpY = 0.5 � ru � Ar � (vw3) � (Cp/wt).
gamma_op = diff (CpY,wt).xi_op = diff (CpY,vw).delta_op = diff (CpY,beta).
Appendix B
B.1. SEIG parameters
2.2 kW, 3-phase, 4-pole, 50 Hz, 415 V, 4.5 A, star connected,1440 rpm, Rs = 3.735 X, Rr = 2.91 X, Xls = 4.727 X, Xlr = 4.727 X;Re = 72e�6 X; Rb = 150e�6 X; Number of rotor bar Nr = 16. Baseimpedance = 53.24 X. Base speed = 1500 rpm, J = 0.0842 kg m2.
The magnetizing inductance Lm is related to the magnetizingcurrent in the following manner:
Lm ¼ 0:3177 for Im 6 0:75¼ 0:3502� 0:0349Im � 0:0017Im for Im < Im 6 4:25¼ 0:17667 for Im > 4:25
B.2. Wind turbine
Nominal power = 2.2 kW, Base wind speed = 11 m/s,J = 1.25 kg m2; R = 1.6 m; G = 3.18, tm = 0.25 s, nominal time con-stant = 1.334, nominal static gain = �2.5453.
Appendix C
By assuming that the wind turbine speed is kept constant in thedesigned control law, the variations in the static gain (DG = j(�d/Jt-
D) � (�dopt/JtDopt)j) and in the time constant (DT = j � 1/D + 1/Doptj)of the linear wind turbine model about their nominal operatingpoint according to the blade-pitch angle for various values of thewind speed are represented in Fig. 15a and b, respectively. The sta-tic gain parameter has the most significant variation.
References
[1] Kamel Rashad M. Effect of wind generation system types on Micro-Grid (MG)fault performance during both standalone and grid connected modes. EnergyConvers Manage 2014;79:232–45.
[2] Feili Hamid Reza, Akar Navid, et al. Risk analysis of geothermal power plantsusing Failure Modes and Effects Analysis (FMEA) technique. Energy ConversManage 2013;72:69–76.
[3] Byon Eunshin, Ntaimo Lewis, Ding Yu. Optimal maintenance strategies forwind turbine systems under stochastic weather conditions. IEEE Trans Reliab2010;59(2):393–404.
[4] Haque MH. Evaluation of power flow solutions with fixed speed wind turbinegenerating systems. Energy Convers Manage 2014;79:511–8.
[5] Ouchbel T, Zougga S, et al. Power maximization of an asynchronous windturbine with a variable speed feeding a centrifugal pump. Energy ConversManage 2014;78:976–84.
[6] Wang Li, Truong Dinh-Nhon. Dynamic stability improvement of four parallel-operated PMSG-based offshore wind turbine generators fed to a power systemusing a STATCOM. IEEE Trans Power Deliv 2013;28(1):111–9.
26 I. Attoui, A. Omeiri / Energy Conversion and Management 82 (2014) 11–26
[7] Kamel Rashad M. Maintaining stability of standalone Micro-Grid by employingelectrical and mechanical fault ride through techniques upon fixed speed windgeneration systems. Energy Convers Manage 2013;74:149–61.
[8] Odgaard Peter Fogh, Stoustrup Jakob. Fault-tolerant control of wind turbines: abenchmark model. IEEE Trans Control Syst Technol 2013;21(4):1168–82.
[9] Bansal RC, Bhatti TS, Kothari DP. A bibliographical survey on inductiongenerators for application of nonconventional energy systems. IEEE TransEnergy Convers 2003;18(3):433–9.
[10] Khan PKS, Chatterjee JK. Three-phase induction generators: a discussion onperformance. Electr Mach Power Syst 1998;27(8):813–32.
[11] Benbouzid MEH. A review of induction motors signature analysis as a mediumfor faults detection. IEEE Trans Ind Electron 2000;47(5):984–93.
[12] Singh GK. Modeling and experimental analysis of a self excited six-phaseinduction generator for stand-alone renewable energy generation. RenewEnergy 2008;33(7):1605–21.
[13] Bašic M et al. Dynamic and pole-zero analysis of self-excited inductiongenerator using a novel model with iron losses. Electr Power Energy Syst2012;42(1):105–18.
[14] Chauhan Yogesh K, Jain Sanjay K, Singh Bhim. Static volt ampere reactivecompensator for self-excited induction generator feeding dynamic load. ElectrPower Compo Syst 2008;36(10):1080–101.
[15] Mahato SN, Singh SP, Sharma MP. Dynamic behavior of a single-phase self-excited induction generator using a three-phase machine feeding single-phasedynamic load. Electr Power Energy Syst 2013;47:1–12.
[16] Bodson M, Kiselychnyk O. Analysis of triggered self-excitation in inductiongenerators and experimental validation. IEEE Trans Energy Convers2012;27(2):238–49.
[17] Baccarini LMR et al. Fault induction dynamic model, suitable for computersimulation: simulation results and experimental validation. Mech Syst SignalProcess 2010;24(1):300–11.
[18] Luo X et al. Multiple coupled circuit modeling of induction machines. IEEETrans Ind Appl 1995;31(2):311–8.
[19] Cunha CCM, Lyra ROC, Filho BC. Simulation and analysis of induction machineswith rotor asymmetries. IEEE Trans Ind Appl 2005;41(1):18–24.
[20] Ben Sasi AY et al. A validated model for the prediction of rotor bar failure insquirrel-cage motors using instantaneous angular speed. Mech Syst SignalProcess 2006;20(7):1572–89.
[21] Márquez FPG et al. Condition monitoring of wind turbines: techniques andmethods. Renew Energy 2012;46:169–78.
[22] Venkatasubramanian V et al. A review of process fault detection and diagnosis.Part I: Quantitative model-based methods. Comput Chem Eng2003;27(3):293–311.
[23] Pineda-Sanchez M, Riera-Guasp M, Perez-Cruz J, Puche-Panadero R. Transientmotor current signature analysis via modulus of the continuous complexwavelet: a pattern approach. Energy Convers Manage 2013;73:26–36.
[24] Benbouzid ME. A review of induction motors signature analysis as a mediumfor faults detection. IEEE Trans Industrial Electron 2000;47(5):984–93.
[25] Jardine AKS, Lin D, Banjevic D. A review on machinery diagnostics andprognostics implementing condition-based maintenance. Mech Syst SignalProcess 2006;20(7):1483–510.
[26] Melício R, Mendes VMF, Catalão JPS. Fractional-order control and simulation ofwind energy systems with PMSG/full-power converter topology. EnergyConvers Manage 2010;51:1250–8.
[27] Li HS, Luo Ying, Chen YQ. A Fractional Order Proportional and Derivative(FOPD) motion controller: tuning rule and experiments. IEEE Trans ControlSyst Technol 2010;18(2):516–20.
[28] Luoa Ying, Chenc Yang Quan, Wangc Chun Yang, Pi You Guo. Tuning fractionalorder proportional integral controllers for fractional order systems. J ProcessControl 2010;20(7):823–31.
[29] Das Saptarshi, Pan Indranil, Das Shantanu. Fractional order fuzzy control ofnuclear reactor power with thermal-hydraulic effects in the presence ofrandom network induced delay and sensor noise having long rangedependence. Energy Convers Manage 2013;68:200–18.
[30] Li W, Hori Y, Peng FZ. Vibration suppression using single neuron-based PIfuzzy controller and fractional-order disturbance observer. IEEE Trans IndElectron 2007;54(1):117–26.
[31] Oustaloup A, Sabatier J, Lanusse P. From fractional robustness to CRONEcontrol. Fract Calc Appl Anal 1999;2(1):1–30.
[32] Alfredo RM, Lipo TA. Complex vector model of the squirrel-cage inductionmachine including instantaneous rotor bar currents. IEEE Trans Ind Appl1999;35(6):1332–40.
[33] Toliyat HA, Lipo TA. Transient analysis of cage induction machines understator, rotor bar, and end ring faults. IEEE Trans Energy Convers1995;10(2):241–7.
[34] Hallenius KE, Vas P, Brown JE. The analysis of a saturated self-excitedasynchronous generator. IEEE Trans Energy Convers 1991;6(2):336–44.
[35] Deraz SA, Abdel Kader FE. A new control strategy for a stand-alone self-excitedinduction generator driven by a variable speed wind turbine. Renew Energy2013;51:263–73.
[36] Ahmed T, Nishida K, Nakaoka M. Advanced control of PWM converter withvariable-speed induction generator. IEEE Trans Ind Appl 2006;42(4):934–45.
[37] Brekken TKA. A novel control scheme for a doubly-fed induction windgenerator under unbalanced grid voltage conditions. Ph.D. Thesis. Universityof Minnesota; 2005.
[38] Siegfried H. Grid integration of wind energy conversion systems. Wiley; 1998.[39] Hilloowala RM. Control and interface of renewable energy systems. Ph.D.
thesis. The University of New Brunswick, Canada; 1992.[40] Boukhezzar B, Siguerdidjane H. Comparison between linear and nonlinear
control strategies for variable speed wind turbines. Control Eng Pract2010;18(12):1357–68.
[41] Mirecki A, Roboam X, Richardeau F. Architecture complexity and energyefficiency of small wind turbines. IEEE Trans Ind Electron 2007;54(1):660–70.
[42] Kamel Rashad M, Chaouachi A, Nagasaka Ken. Enhancement of Micro-Gridperformance during Islanding mode using storage batteries and new fuzzylogic pitch angle controller. Energy Convers Manage 2011;52:2204–16.
[43] Bode HW. Network analysis and feedback amplifier design. New York: VanNostrand; 1945.
[44] Oustaloup A et al. Frequency-band complex non-integer differentiator:characterization and synthesis. IEEE Trans Circ Syst I: Fund Theor Appl2000;47(1):25–39.
[45] Franklin G, Powell J, Naeini A. Feedback control of dynamic systems. Reading,MA: Addison-Wesley;; 1986. 34.
[46] Chen YQ, Moore KL. Relay feedback tuning of robust PID controllers with iso-damping property. IEEE Trans Syst Man Cybernet Part B 2005;35(1):23–31.
[47] William T Thomson, Fenger Mark. Current signature analysis to detectinduction motor faults. IEEE Ind Appl Mag 2001;7(4):26–34.