Professional Documents
Culture Documents
1 s2.0 S1474667016437419 Main
1 s2.0 S1474667016437419 Main
1 s2.0 S1474667016437419 Main
Abstract: In this paper, a dynamic state space model of a DFIG with stator inter-turn short circuit fault is
proposed. This model can quantitatively describe the fault at any level in any single phase. Both healthy
and faulty conditions can be simulated by using this model. Base on this model, an observer based fault
detection and diagnosis (FDD) scheme is developed, which can not only provide a rapid detection when
fault occurs but also give an accurate diagnosis of the fault position and level. In order to ensure the
observer stability under a wide rang of the speed variation of DFIGs, a time varying Kalman like gain
supplied with measured rotor speed is applied to the observers. Moreover, an exponential adaptive
observer is employed to provide a desirable estimation of fault level. The simulation results demonstrate
the effectiveness of this approach in detecting and diagnosing the faults under both stationary and speed
varying operations. The latter is particularly important for fault detection of wind turbine DFIGs.
Keywords: DFIG, inter-turn short circuit, state space model, FDD, observer.
T ∂Lsr ∂L
Te = P ( I abc _s I abc _ r + µ f xT i f sr I abc _ r ) (3)
∂θ r ∂θ r
Fig. 1. Stator winding configuration with the inter-turn short Transforming the model equations (1) and (2) into a
circuit fault in phase ‘a’. stationary d-q reference frame (stator fixed d-q reference
frame), a DFIG model in d-q variables is derived as follows,
2.1 Model in Natural abc Reference Frame with the fault position parameter f x defined as
T T
A DFIG model in natural abc reference frame is derived to T 1 3 1 3
describe the inter-turn short circuit fault at any level in any f a = [1 0] , f b = − , fc = − − .
single phase of stator. In this model, the fault position 2 2 2 2
parameter f x is defined as below for three cases that fault The voltage and flux linkage model equations are given as
occurs in phase ‘a’, ‘b’ and ‘c’, respectively.
Vdq _ s Rs′ 2
02×2 − µ rs f x I dq _ s d λdq _ s λdq _ s
T T
f a = [1 0 0] , f b = [ 0 1 0] , f c = [ 0 0 1] .
T Vdq _ r = 3 I dq _ r + λdq _ r + [ Ω] λdq _ r
v 02×2 Rr′ 02×1 dt λ λ
f µ rs f xT 01×2 − µ rs i f f f
The fault level parameter µ denotes the fraction of the
shorted winding. The voltage and flux linkage model (4)
equations in matrix form are given as ′
2
λdq _ s Lss Lsr′ − µ Lss′ f x I
3 dq _ s
Vabc _ s Rs 03×3 − µ Rs f x I abc _ s λabc _ s λ = 2 I (5)
Vabc _ r = 03×3 R 03×1 I abc _ r + d λabc _ r (1) λdq _ r Lrs′ Lrr′ − µ Lrs′ f x dq _ r
v µ R f T 0 r − µ R i dt λ f T T
3 i f
f s x 1×3 s fx f f µ Lss′ f x µ Lsr′ f x − µ LDD
λabc _ s Lss Lsr − µ Lss f x I abc _ s among which, the modified resistance and inductance
λabc _ r = Lrs Lrr −µ L f I (2) matrices are given as
λ µ L f T µ L f T − µ Lrs x i abc _ r
f ss x DD f
Rs′ = s , Rr′ = r .
sr x
r 0 r 0
Among these equations, the resistance and inductance 0 rs 0 rr
matrices are given as
Lss′ = s
L 0
, L′ = r
L 0 ′ Lm 0
, L = , Lrs′ = Lsr′ T .
0 Ls rr 0 Lr sr 0 Lm
rs 0 0 rr 0 0
Rs = 0 rs 0 , Rr = 0 rr 0 . The stator, rotor and faulty circuit ‘as2’ self inductances for
0 0 rs 0 0 rr this model are defined as
1014
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011
{x& = ( Ac + Aωr ωr ) x + ( Bc + Bθ θ )u
y = Cx
(7)
2006). Actually the inter-turn short circuit fault can be
considered as an actuator fault as the model described in (7).
Hence in this section a close-loop state observer is
where, the state, input variables and output measurements are (Luenberger observer) is applied to detect fault. However, it
T T
x = isd isq ird irq µ i f , u = usd usq urd urq 0 , and is impossible to design the observer directly based on model
T (7) as the fault position parameter f x is unknown. Therefore,
y = isd isq ird irq .
model (7) needs to be modified firstly. By analyzing the
model equation (7), it can be seen that the fault current i f is
For notation simplification, by defining
governed by the following dynamics.
µ
θ= (8) d r 1 3µ
2µ − 3 (µ i f ) = − s µi f − f xT Vdq _ s (9)
dt Lls Lls 2µ − 3
D = Ls Lr − L2m , and J = 0 1 , the coefficient matrices can After the system enter into steady state, equation (9) becomes
−1 0
be reduced into 1 3µ
µi f = − f xT Vdq _ s (10)
rs 2µ − 3
′ ′ ′ ′ 2
− Lrr Rs Lsr Rr 3 rs Lr f x 2 By substituting it into (7), the model can be reduced into
1 1 0 2× 2 0 2× 2 − rs f x
Ac = 2 + 3
d Idq _ s I V 1 2µ 1 −Lrr′ Rs′ + L2m Jωr
D Lsr′ Rs′ − Lss′ Rr′ − rs Lm f x Lls 0 2× 2 0 2× 2 0 2×1 = AH dq _ s + BH dq _ s + fx fxTVdq _ s
3 01× 2 01× 2 − rs dt Idq _ r Idq _ r Vdq _ r rs 2µ − 3 D Lsr′ R′ − Ls Lm Jωr
01× 2 01× 2 0
I V 1 2µ f f TV
= AH dq _ s + BH dq _ s + AH x x dq _ s
2 2 2 I
dq _ r V
dq _ r rs 2µ − 3 02×1
Lm J Lr Lm J − 3 Lm Jf x L′ −Lsr′ 02×1
1 1 rr (11)
Aωr = 2 Bc = −Lsr′ Lss′ 02×1
D − Ls Lm J − Ls Lr J Ls Lm Jf x D 0 0 The parameter matrices in above model are
3 1×2 1×2 0
0
1× 2 0 0
1× 2
AH = Ac _ H + Aωr _ H ωr (12)
−2 f x f xT 02× 2 02×1 1 0 0 0 0
1 − Lrr′ Rs′ Lsr′ Rr′ 1 L2m J Lr Lm J
C = 0 0
1
Bθ = 02×2 02× 2 02×1 1 0 0 Ac _ H = Aωr _ H =
Lls T 0 0 1 0 0 D Lsr′ Rs′ − Lss′ Rr′ D − Ls Lm J − Ls Lr J
−3 f x 01×2 0 0 0 0 1 0
1 Lrr′ − Lsr′
BH = C = I 4× 4 .
D − Lsr′ Lss′
1015
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011
Based on this modified model, a well known close loop state 3.3 Step 3: Fault Level Estimation
observer (Luenberger observer) is constructed as
After fault position has been located, in order to estimate the
x&ˆ = AH xˆ + BH u + K ( y − Cxˆ ) fault level online, a well known online parameter estimation
yˆ = C x (13)
H method (adaptive observer) is applied. Since the parameter
where, estimated states, input signals and prediction outputs f x is known after the fault has been located, the model (7)
are defined respectively as can be used for observer design with a little modification as
xˆ = [ Iˆdq _ s , Iˆdq _ r ]T , u = [Vdq _ s , Vdq _ r ]T yˆ = [ Iˆdq _ s , Iˆdq _ r ]T
Comparing (13) and (11), the estimation error equations are
{xy& == CxA(t) x + B u + Ψ(t)θ
c (18)
T
where, A(t ) = Ac + Aωr ωr (t ) , Ψ (t ) = Bθ Vsd (t ), Vsq (t ) , with
d eIdq _ s eIdq _ s 1 2µ f f V T
1016
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011
4. SIMULATION RESULTS applied at 2 sec and develops into 4% at 3sec. The FDD
results are given in Fig. 5. As it shown, the fault can be
A transient behaviour of DFIG from the normal condition to rapidly detected by observing the norm of stator current
a 1% inter-turn short circuit fault at t=2 is simulated by using residuals. After the fault has been detected, the feedback loop
the proposed state space model. The parameters of the of the state observer is opened at 2.2 sec. The fault can be
simulated DFIG model are reported in Appendix. The located by comparing eI d _ s and eI q _ s referring to the relations
simulation results are given in Fig. 3. As it shown, the stator
and rotor currents both decrease when short circuit fault described in (17). At 2.5 sec, the adaptive observer is
occurs. The stator current becomes unbalanced and the activated and produces an accurate estimation of the fault
current in phase ‘a’ is bigger the other two. The asymmetry level µ . The stator current residual signatures for different
of the stator currents causes a fluctuation in the rotor speed at cases that fault occurs in phase ‘a’, ‘b’ and ‘c’ are given in
high frequency. The steady state stator currents for the cases Fig. 6. This figure shows that the simulation results are in
that fault occurs in phase ‘a’, ‘b’ and ‘c’ is shown in Fig. 4. It good agreement with the relations described in (17).
can be seen that the current in faulted phase is the biggest.
1017
18th IFAC World Congress (IFAC'11)
Milano (Italy) August 28 - September 2, 2011
synchronous region to super-synchronous region is varying gain is able to provide a good estimation of level
performed by applying a stepwise load torque ( Tload changes under a wide rang of speed variation.
from -230Nm to -300Nm at t=2s) to the generator running
under decoupled DQ control. A 1% inter-turn short circuit is REFERENCES
applied at t=1s. The maximum power extraction control Amirat Y., Benbouzid M.E.H., Al-Ahmar E., Bensaker B.
strategy is employed to adjust the active power setting point, and Turri S. (2009). A brief status on condition
and the reactive power setting point is set as zero. The monitoring and fault diagnosis in wind energy conversion
generator output active and reactive powers in this whole systems. Renewable and Sustainable Engergy Reviews,
process are shown in Fig. 7. The corresponding rotor speed 13(9), pp. 2629-2636.
variation is presented in Fig. 8. It can be seen that the short Arkan M., Kostic-Perovic D. and Unsworth P.J. (2005).
circuit fault causes a minor decrease in the rotor speed and Modelling and simulation of induction motors with inter-
the active power output. The fault level estimation is given in turn faults for diagnostics. Electric Power System
Fig. 9. It shows that the estimation of µ is not affected by Research, 75(1), pp. 57-66.
the speed variation. Besancon G., De Leon-Morales J., and Huerta Guevara O.
(2006). On adaptive observers for state affine systems.
International Journal of Control, 79(6), pp. 581-591.
Douglas H., Pillay P., Barendse P. (2003). The detection of
inter-turn stator faults in doubly-fed induction generators.
In Proceeding of IEEE Industry Applications Conference,
2, pp. 1097-1102.
Gritli Y., Stefani A., Filippetti F. and Chatti A. (2009). Stator
fault analysis based on wavelet technique for wind
turbines equipped with DFIG. In Proceeding of IEEE
Fig. 7. Output active and reactive power. International Conference on Clean Electrical Power,
Italy, pp. 485-491.
Holdsworth L., Wu X.G., Ekanayake J.B. and Jenkins N.
(2003). Comparison of fixed speed and doubly-fed
induction wind turbines during power system disturbance.
IEE Proc.-Gener.Transm. Distrib., 150(3), pp. 343-352.
Isermann R. (2006). Fault Diagnosis Systems: An
Introduction from Fault Detection to Fault Tolerance.
Springer-Verlag, Berlin.
Joksimovic M. G. and Penman J. (2000). The detection of I
Fig. 8. Rotor speed. nter-turn short circuits in the stator windings of operating
motors. IEEE Trans. Ind. Electron., 47(5), pp. 1078-1084.
Lu B., Li Y., Wu X., and Yang Z. (2009). A review of recent
advances in wind turbine condition monitoring and fault
diagnosis. In Proceeding of IEEE PEMWA, Lincoln.
Tallam R. M., Habetler T. G., Harley R.G. (2002). Transient
model for induction machines with stator winding turn
faults. IEEE Trans. on Application, 38(3), pp. 632-637.
Zhang Q. (2002). Adaptive observer for multiple-input-
multiple output (MIMO) linear time-varying systems.
Fig. 9. Fault level estimation under varying speed operation. IEEE Transaction on Automatic Control, 47(3), pp. 525-
529.
CONCLUSIONS Appendix
In this paper, a state space model of DFIGs with stator inter- Machine Parameters:
turn short circuit fault has been proposed. This model rated stator frequency: f s = 50 Hz;
allowed for the simulations of different fault scenarios and rated stator phase voltage: Vs = 130 V;
operation conditions. Moreover, the fault position and level
mutual inductance: Lm = 44.2 mH;
can be quantitatively described by two model parameters.
Based on this model, an observer based FDD scheme has stator leakage inductance: Lsσ = 673.97 µH;
been proposed, which was completed within three steps using rotor leakage inductance: Lrσ = 490.60 µH;
three different observers. The validity of this scheme on stator resistance: Rs = 45 mΩ;
diagnosing the fault at different levels in different phases has
been demonstrated through simulation tests. This scheme has rotor resistance: Rr = 66.5 mΩ;
been also tested under speed varying operation. It was shown machine inertia: J = 0.4 kg/m2;
that the applied exponential adaptive observer with a time number of pole pairs: P = 2.
1018