Duisburg-Essen University
Duisburg, Germany
E-mail: istvan.erlich@uni-due.de
E-mail: jose.rueda@uni-duisburg-essen.de
AbstractA methodology for parameter estimation of equivalent wind power plant is proposed in this paper. The parameter algorithm is formulated as an optimization problem
which miminize the output between the real system and the
mathematical model. The parameters are updated using NewtonRaphson method which depends on the information of trajectory
sensitivity functions. A generic equivalent of wind power plant
was used as a structure of the model, which is valid for DoubleFed Induction Generator (DFIG) and full Converter Based Wind
Turbine. In addition, the procedure to obtain of the trajectory
sensitivity functions of the model is presented in this paper. A
discussion of the estimated results is also analysed. In general,
the convergence happened in a few seconds for most of the cases
studied.
KeywordsTrajectory sensitivity, parameter estimation, wind
power plant, wind generator, Newton-Raphson method.
N OMENCLATURE
r Resistance in [p.u].
x Reatance in [p.u].
I NTRODUCTION
In this section, the trajectory sensitivity algorithm to estimate parameters of a dynamical system is presented. Trajectory
f (x(t), p, u(t))
g(x(t), p, u(t))
(2)
1
2
To
(
0
y T y
) ( )dt (k)
p
p
p=p
(6)
(1)
J(p) =
To
III.
dt pi
x
pi
pi
y(t)
g(x(t), p, u(t)) x(t)
g(x(t), p, u(t))
=
+
pi
x
pi
pi
(3)
J(p)
=
p
To y T
(yr y)dt = 0
(4)
Newtons method can be used to solve the nonlinear equation (4). Starting from an initial parameter guess p(0) = po ,
the parameter fitting at the k th iteration is given by:
p(k+1) = p(k) hopt 1 G(p)
p=p(k)
(5)
Figure 1.
IAc
PT ref
VT
IRe
Kvc(VT ref VT ) +
(7)
QT ref
VT ref
(8)
=
=
IAc
IRe
(9)
(10)
v
T
Vq
v
T
Pe
v
T
Qe
1
(VP AS Vd )
Tv2
1
2 (VQAS Vq )
Tv
Vt Vtmin
IV.
otherwise
Vt < Vtmin
IQref = q
min(IRe , Imax )
2
2
IP ref = Imax
IQref
v
X(VT d T
Vd
r2 + X 2
Tv
Tv
v
+ VT q T
Vq ) r(VT d Vq VT q V
VQA
Z t
KI
P
P
)+
(IP ref
)dt
VT
TI 0
VT
Z t
KI
Q
Q
IQref ) +
IQref )dt
(
KI (
VT
TI 0 VT
KI (IP ref
r2 + X 2
Vd
Tv ,
TVvq =
Vq
Tv ,
TPve =
Pe
Tv
(21)
(22)
and TQve =
VP A
(20)
where TVvd =
Qe
Tv .
IP ref = q
min(IAc , Imax )
2
2
IP
IQref = Imax
ref
(19)
Tv
Tv
Tv
v
r(VT d T
V + VT q Vq ) + X(VT d Vq VT q V
Else:
if
1 Tv
Tv Vd
1 Tv
Tv Vq
(11)
(12)
=
=
(13)
(14)
Vq
i
1 h
VP AS Vd
Tv
i
1 h
VP AS Vq
Tv
Figure 2.
(15)
(16)
(18)
Pe
Qe
(17)
Real
System
yr = (P, Q)
Input
(VT,qv P, Q)
Output
The comparison of outputs of the real system and mathematical model at the beginning and at the end of the estimation
process for case 1 is shown in Figs 4 - 7. After the convergence
of the parameter, the active and the reactive power of both
system are very similar.
Mathematical
Model
Initial
guess
0.511
0.026
12.62
2.040
0.026
0.237
Final
Values
0.3891
0.0379
10.398
1.999
0.0332
0.1973
Nominal
values
0.3930
0.0380
10.516
2.000
0.0332
0.1972
Error
(%)
-0.992
-0.329
-1.116
-0.005
-0.036
0.035
Parameters
2
1.5
1
0.5
0
0.2
0.4
0.6
0.8
Time (s.)
Figure 4.
Real
Model
1.2
New
Figure 3.
Real
Model
2.5
1
0.8
0.6
0.4
0.2
0.2
0.4
0.6
0.8
Time (s.)
Figure 5.
Real
Model
Sensitivity
Equations
y = (Pe, Qe)
y
p
0.5
0.5
Table II.
Param.
Tv
TI
KI
Kvc
R
X
Initial
guess
0.350
0.026
9.000
2.040
0.020
0.180
Final
Values
0.3848
0.0378
10.286
1.999
0.0332
0.1972
Nominal
values
0.3930
0.0380
10.516
2.000
0.0332
0.1972
Error
(%)
-2.0891
-0.4798
-2.1892
-0.0083
-0.0251
-0.0186
0.2
0.4
0.6
0.8
Time (s.)
Figure 6.
R EFERENCES
Real
Model
0.5
0.5
0.2
0.4
0.6
0.8
Time (s.)
Figure 7.
C ONCLUSIONS
ACKNOWLEDGMENTS