Missouri University of Science and Technology

Scholars' Mine

Electrical and Computer Engineering Faculty Electrical and Computer Engineering

Research & Creative Works

01 Jan 2004

Power Flow and Stability Models for Induction Generators Used in

Wind Turbines
Kiran K. Nandigam

Badrul H. Chowdhury
Missouri University of Science and Technology, [email protected]

Follow this and additional works at: https://scholarsmine.mst.edu/ele_comeng_facwork

Part of the Electrical and Computer Engineering Commons

Recommended Citation
K. K. Nandigam and B. H. Chowdhury, "Power Flow and Stability Models for Induction Generators Used in
Wind Turbines," Proceedings of the IEEE Power Engineering Society General Meeting, 2004, Institute of
Electrical and Electronics Engineers (IEEE), Jan 2004.
The definitive version is available at https://doi.org/10.1109/PES.2004.1373229

This Article - Conference proceedings is brought to you for free and open access by Scholars' Mine. It has been
accepted for inclusion in Electrical and Computer Engineering Faculty Research & Creative Works by an authorized
administrator of Scholars' Mine. This work is protected by U. S. Copyright Law. Unauthorized use including
reproduction for redistribution requires the permission of the copyright holder. For more information, please
contact [email protected].
 1

Power Flow and Stability Models for Induction

Generators Used in Wind Turbines
Kiran Nandigam1, Student member, IEEE Badrul H. Chowdhury, Senior member, IEEE

equipping their power generating units with induction

Abstract -- Model initialization of the induction generator generators and the behavior of these machines are a far cry
for wind power generation is investigated. Incorrect from the conventional synchronous machines in both the
initialization may lead to numerical instability problems, not to
steady state and dynamic domains. Consequently, power
mention erroneous results from dynamic simulations. The
machine model is studied from the perspective of slip and flow and transient stability models for these asynchronous
reactive power calculation and tested for various cases. The machines have to be determined prior to assessing their
dynamic reactive power calculation in well-known software behavior on the power system.
package is discussed and the deviation of dynamic reactive For representing the induction generator model in
power from the steady state reactive power is highlighted. stability simulations, model initialization needs to be carried
out as the initial conditions need to be matched before the
Index terms -- Dynamic induction machine model, load flow simulations. This paper discusses the model initialization
studies with induction generator, dynamic stability, model process, deviation of reactive power values in steady state
initialization.
and dynamic analysis for system studies. A technique to
address the discrepancy between power flow and dynamic
I. INTRODUCTION
simulations is discussed. Simulations are performed with

W ind turbines are quietly becoming the fastest-

growing energy source in the world. A record 6,868
megawatts (MW) of new wind power capacity was
PTI’s PSS/E software [3] due to its wide use and popularity.

installed worldwide in 2002 bringing the total wind power II. INDUCTION GENERATOR MODEL
capacity was 31,000 MW [1]. Total installed US wind
energy capacity stood at 5,326 MW in October 2003 with In synchronously rotating frame, in terms of the d and
California leading the way with about 2000 MW, followed q components, the stator and rotor flux linkages of a
by Texas with almost 1100 MW [2]. To get an idea of the squirrel cage induction machine can be expressed by [4]:
rapid pace of wind power deployment in the US alone, one ψ ds = Lss ids + Lm idr (1)
only has to compare the current rate of installation with
previous years. In the one year period from 1998 to 1999, ψ qs = Lss iqs + Lm iqr (2)
the increase was 1848 MW to 2511 MW – a 36% increase. ψ dr = Lrr i dr + Lm ids (3)
In contrast, from 2000 to 2001 the increase was 2578 MW
to 4275 MW – a 66% increase! ψ qr = Lrr i qr + Lm iqs (4)
Despite this renewed interest in wind power, this form of where
renewable energy still generates less than 1 percent of the Lm = 3 / 2 LaA
nation's electricity. Nonetheless, most planners in the
electric utility business project wind power to have a larger
The stator and rotor voltages in terms of the d and q
role as they struggle to meet soaring electrical demands and
components are
face uncertainties in electricity prices. The good news is
vds = Rs ids − wsψ qs + pψ ds (5)
that with the development of bigger, more sophisticated
turbines, the cost of wind-generated electricity, once seen as v qs = R s i qs + wsψ ds + pψ qs (6)
prohibitive, is now nearly competitive with that of its rivals,
all but eliminating what was once a major barrier. The bad
v dr = R s i dr − swsψ qr + pψ dr (7)
news is that the present day power system control structure v qr = Rr iqr + swsψ dr + pψ qr (8)
may not be as yet ready to handle large scale penetration of
wind power. Reactive power requirements as well as power III. MODELING
quality issues continue to be hurdles and solution strategies
are rather complex and expensive. Besides, system studies For constructing a dynamic induction machine model, in
using standard power flow and stability tools are also not the stator voltage equations, the stator transients are usually
neglected which correspond to ignoring the dc component
well defined. Most wind turbine manufacturers are now
in the stator transient currents and hence only the
1
Kiran Nandigam (email: [email protected]) and Badrul H. Chowdhury
fundamental frequency components are represented. This
(email: [email protected]) are with the Electrical & Computer Engineering assumption works well when the induction machine slip is
Department, University of Missouri-Rolla, Rolla, MO 65409-0040
 2

small and when the ac network frequency deviations are A. Induction Generator Model Used as a Wind Generator
zero. The most common way to treat induction generator-
Upon neglecting the stator transients and on short based wind generations in power systems is to run power
circuiting the rotor, the stator and rotor voltages become: flow and dynamic simulations iteratively to find a mismatch
v ds = R s i ds − w sψ qs (13) between the two program runs [5].
v qs = Rs iqs + wsψ ds (14) The model that was investigated for representing a
dynamic induction model in PSS/E is the CIMTR1 model,
v dr = 0 = R r i dr − swsψ qr + pψ dr (15) which is basically an induction generator model that
considers only rotor flux transients and neglects the stator
v qr = 0 = R r i qr − swsψ dr + pψ qr (16) transients. This model represents a fixed speed induction
For getting an induction machine transient model, the generator in the real world.
rotor currents are eliminated and a relationship is expressed CIMTR1 model can be used to represent both a single
between the stator current and voltage behind the transient cage and a double cage machine. In this paper, dynamic
reactance. From Eq. (3) simulations have been done in PSS/E for both single cage
ψ dr − Lm i ds and double cage machine. In the PSS/E power flow studies,
idr = (17) a generator with a positive electrical power should be used
Lrr for representing a wind generator. As there are no sub-
Now substituting the above equation in (1) transient reactances and sub-transient time constants
Lm ⎛ L2 ⎞ associated with a single cage machine, either T ′′ (sub
ψ ds = ψ dr + ⎜⎜ Lss − m ⎟⎟ids (18) transient time constant) or X ′′ (sub transient reactance)
Lrr ⎝ Lrr ⎠ should be set equal to zero and ZSORCE (generator
Similarly, dynamic impedance ) in the power flow model should be set
equal to X ′ (transient reactance). In case of a double cage
Lm ⎛ L2 ⎞
ψ qs = ψ qr + ⎜⎜ Lss − m ⎟⎟i qs (19) machine, ZSORCE should be set equal to X ′′ while
Lrr ⎝ Lrr ⎠ providing appropriate values for T ′′ and X ′′ . A double cage
machine will have time constants much smaller than T ′′
For representing the stator voltages in terms of the d and
and, hence, a time step much smaller than the normal value
q components, we substitute ψ qs and ψ ds terms in Eqs. of ½ cycle are necessary.
(13) and (14): B. Reactive Power Calculations
v ds = Rs ids − X s/ iqs + v d/ (20) The wind generating plants are usually modeled as
v qs = R s i qs + X s/ i ds + v q/ (21) collector systems that comprise of a large number of
individual units that are interconnected with other units in
w s Lm radial or parallel arrangements .Power flow and transient
where v d' = − ψ qr (22) stability models for induction generator models are not
Lrr
always well understood.
wL In this paper, several different scenarios cases have been
v q' = s m ψ dr (23)
investigated to gain an insight into how the asynchronous
Lrr
wind generating machines behave during system studies.
⎛ L2m ⎞ Fig. 1 shows three wind collected buses connected to a
X = '
s ws ⎜⎜ Lss − ⎟
⎟
(24) larger power system. Studies are performed on this system.
⎝ Lrr ⎠
'
In Eq. (24), X s is the transient reactance of the induction windbus1
machine. 1
The three state equations of the transient single-cage
Rest of the
induction machine can be represented by eliminating the windbus2
Power
rotor currents and expressing the rotor flux linkages in 2 System
' '
terms of v and v :
d q windbus3
1
[
p (vd' ) = − ' v d' + ( X s − X s' )iqs + pθ r v q'
T0
] (25)
3

1
[
p(v q' ) = − ' v q' − ( X s − X s' )i ds − pθ r v d'] (26)
Fig. 1. Three wind collector buses as part of a large power system.
T0
1
p(ω r ) = (Te − Tm ) (27) The dynamic induction generator model is tested for its
2H behavior with regard to different sets of active power and
voltage changes. The wind collector buses are located at
windbus1, windbus2 and windbus3 buses. Dynamic
simulations were first run with voltage at windbus1
 3

scheduled at 1.02 pu. Unlike a synchronous generator, an windbus1 0.98 5 -30.4 -0.1781 -0.1331
induction generator does not have the capability to control
windbus2 1.02 33 0 0.2067 -0.2145
its terminal voltage. Therefore, the CIMTR1 model consists
of two internal variables called Var (L) and Var (L+1) that windbus3 1.02 33 0 0.2067 -0.2145
are used for calculating the reactive power output of the
machine. Var (L) is the artificial shunt admittance Fig. 2 shows the relationships between bus voltage and
corresponding to the Mvar difference in the initial reactive power outputs In the plots, VAR101P1, VAR101P2
condition. This fictitious admittance is placed at the wind and VAR101P3 represent Var (L) for windbus1 at 33 MW,
generator bus so as to match the dynamic reactive power 15 MW and 5 MW respectively. Similarly, VAR104P1,
with the steady state reactive power and var (L+1) is the VAR104P2, VAR104P3 represent Var (L) for windbus2
motor dynamic reactive power. and windbus3 at 33MW, 15 MW and 5 MW respectively.
Thus, the total reactive power output of the induction Furthermore, VAR102P1, VAR102P2, VAR102P3
generator is calculated using the formula given below: represent Var (L+1) for windbus1 at 33MW, 15 MW and 5
Qe = Var ( L ) x V + Var ( L + 1) xSb MVar
2 MW respectively, while VAR105P1, VAR105P2,
(28)
VAR105P3 represent Var (L+1) for windbus2 and
The value of Qe then should match the steady state reactive windbus3 at 33MW, 15 MW and 5 MW respectively.
power in the power flow solution.
Dynamic simulations were next carried out by varying
the windbus1 bus voltage from 1.05 pu to 0.98 pu with
active power at windbus1 being progressively decreased
from 33 MW to 5 MW. The behavior of the model with
respect to reactive power outputs and the two different
variables var (L) and var (L+1) were observed. Table 1
shows this relationship.

TABLE I
VARIATION OF INTERNAL VARIABLES DUE TO CHANGES IN WIND BUS
COLLECTOR VOLTAGE

Bus V MW MVar Var(L) Var(L+1)

windbus1 1.05 33 10.9 0.3041 -0.2238
windbus2 1.02 33 -10.1 0.1096 -0.2145
windbus3 1.02 33 -10.1 0.1096 -0.2145 Fig. 2. Variation of internal dynamic variable Var (L) – the fictitious shunt
windbus1 1 33 -15.3 0.0561 -0.2090 admittance – with bus voltage.

windbus2 1.02 33 -2.5 0.1819 -0.2145

windbus3 1.02 33 -2.5 0.1819 -0.2145
windbus1 0.98 33 -30.2 -0.1029 -0.20272
windbus2 1.02 33 -0.3 0.2029 -0.2145
windbus3 1.02 33 -0.3 0.2029 -0.2145
windbus1 1.05 15 10.2 0.2516 -0.1751
windbus2 1.02 33 -12.1 0.0895 -0.2146
windbus3 1.02 33 -12.1 0.0895 -0.2146
windbus1 1 15 -17.4 -0.0171 -0.1570
windbus2 1.02 33 -4.2 0.1660 -0.2145
windbus3 1.02 33 -4.2 0.1660 -0.2148
windbus1 0.98 15 -30.5 -0.1629 -0.1487
windbus2 1.02 33 -0.1 0.2050 -0.2145
Fig. 3. Variation of internal dynamic variable Var (L+1) – the motor dynamic
windbus3 1.02 33 -0.1 0.2050 -0.2145 reactive power.
windbus1 1.05 5 11.6 0.2513 -0.1596
From Fig. 3, one can observe that as the voltage at
windbus2 1.02 33 -9.9 0.1110 -0.2146
windbus1 decreases from 1.05 pu to 0.98 pu, Var (L) – the
windbus3 1.02 33 -9.9 0.1110 -0.2146 fictitious shunt admittance - decreases at windbus1 while
windbus1 1 5 -11.2 0.0322 -0.1445 the shunt admittances at the other buses more or less remain
windbus2 1.02 33 -3.4 0.1732 -0.2145 constant. This is to be expected because changes in voltage
are brought about only at windbus1. In Fig. 3, one can
windbus3 1.02 33 -3.4 0.1732 -0.2145
 4

observe that as the voltage at windbus1 decreases from 1.05

pu to 0.98 pu, Var (L+1) - the motor dynamic reactive In the process of calculating the reactive power, we
power - decreases at windbus1 while the reactive powers at obtain two values of slip. The lower value of slip is
the other buses remain perfectly constant. Again, this is to considered for further analysis as it lies in the stable region
be expected because changes in voltage are brought about of active power versus slip characteristics of the generator.
only at windbus1. On substituting the system parameters and the active power
of -0.5 per unit at a terminal voltage of 1.0089 pu, a reactive
C. Reactive Power Initialization without Full Scale
power of 0.3651 per unit is obtained. Hence, there is a
Dynamic Simulations
mismatch of 0.0076.
As obvious, it would be more beneficial to be able to In the general case, this amount of mismatch is
calculate the correct reactive power output without resorting represented by an artificial shunt admittance. In PSS/E this
to full scale dynamic simulations. The set of equations (28) shunt admittance is represented by Var (L).
through (37) are useful for estimating the reactive power On varying the reactive loads at the induction generator
output from induction generators [6]. buses, the reactive power mismatches vary widely. Table II
shows the different reactive loads (Q), the slip (S) of the
K1 = X r + X m (28) induction generator at voltage (V) and the reactive power,
x A + xB + C = 0
2
(29) Qc, that is calculated using Eqs. (28) to (37). The
mismatches are in absolute values.
rr TABLE II
where x =
s CALCULATION OF MISMATCH WITH VARYING INITIAL CONDICTIONS

A = P (rs 2 + K32 ) − V 2rs (30) V Q, pu QC, pu Slip Mismatch (abs)

B = 2 P(rs K 2 + K 3 K 4 ) − V ( K 2 + K1K 3 )
2
(31) 1.00245 0 0.3682 0.0079 0.3682
1.0089 0.35 0.3576 0.0081 0.0076
C = P ( K 22 + K 42 ) − V 2 K 1 K 4 (32) 0.9974 0.6 0.35 0.0083 0.25
where 0.988 0.8 0.3438 0.0085 0.4562
K 2 = − X S K1 − X r X m (33)
K3 = X m + X s (34) IV. CONCLUSIONS
The model initialization of the induction generator has
K 4 = rs K 1 (35)
been discussed with reference to adopted methods. The
V 2{xT 1 + K1T2} mismatch between the steady state reactive power and
P= (36) dynamic reactive power calculations is added as a fictitious
T3
admittance. Methods are available that may be used to
QC = −V {K1T1 − xT2 }
2
(37) calculate this mismatch without resorting to full scale
T3 dynamics. This allows convenient system impact studies
where with wind generators in the system.
T 1 = xrs − X S K 1 − X r X m
T 2 = x( X m + X S ) + rs K1 V. APPENDIX
TABLE A.1
T 3 = T12 + T22 BUS DATA FOR THE 3-MACHINE, 9-BUS SYSTEM
To illustrate the method, a 3 machine, 9 bus system,
shown in Fig. 4, is tested for the initialization process. This Bus V Pg, pu Qg Pl Ql Type
test system has an induction generator at bus 8 which is 1 1.04 0 0 0 0 Slack
modeled as a negative load. It is assigned a reactive load of 2 1.025 1.63 0 0 0 PV
0.35 per unit in the power flow set up. The power flow data
3 1.025 0.85 0 0 0 PV
for this system is shown in the Appendix.
4 1.0 0 0 0 0 PQ
5 1.0 0 0 0.9 0.3 PQ
6 1.0 0 0 0 0 PQ
2 3 7 1.0 0 0 1 0.35 PQ
8 8 1.0 0 0 -0.5 0.35 PQ
6 9 1.0 0 0 1.25 0.5 PQ
9
7
IG

4 5
1

Fig. 4. A 3 machine, 9 bus test system.

 5

H inertia constant
TABLE A.2 Te electrical torque
LINE DATA FOR THE 3-MACHINE, 9-BUS SYSTEM
Tm motor torque
From To R, pu X, pu B, pu X r rotor reactance
1 4 0 0.0576 0 X m magnetizing reactance
4 5 0.017 0.092 0.158 X s stator reactance
5 6 0.039 0.17 0.358
3 6 0 0.0586 0 VII. REFERENCES
6 7 0.0119 0.1008 0.209 [1] P. Fairley, “Steady As She Blows,” IEEE Spectrum,
7 8 0.0085 0.072 0.149 vol. 40(8), August 2003, pp 35-39
8 2 0 0.0625 0 [2] American Wind Energy Association,
8 9 0.032 0.161 0.306 http://www.awea.org/projects/index.html
9 4 0.01 0.085 0.176 [3] Power Technologies, Inc., PSS/E-28 Program Operations
Manual, November 2001, Schenectady, NY.
TABLE A.3
[4] P. Kundur, Power System Stability and Control, New York,
INDUCTION GENERATOR DATA US: McGraw-Hill, Inc., 1994.
[5] J.G. Slootweg, H. Polinder, W.L. Kling, “Initialization of
Wind Turbine Models in Power System,” Proc. IEEE Porto
rs xs Xm rr xr Power Tech, 10-13 Sept. 2001.
0.0574 0.0769 2.9061 0.0238 0.0709 [6] D. Ruiz-Vega, T. I. Olivares and D. O. Salinas, “An Approach
to the Initialization of Dynamic Induction Motor Models”,
IEEE Transactions on Power Systems, vol. 17, no. 3, pp. 747
VI. LIST OF SYMBOLS -751, August 2002.

ψ ds d-axis stator flux linkage VIII. BIOGRAPHIES

ψ qs q-axis stator flux linkage Kiran Nandigam received his B.E degree in Electrical and
ψ dr d-axis rotor flux linkage Electronics Engineering from Andhra University, Visakhapatnam,
India in 2000. He is presently an MS student at the Department of
ψ qr q-axis rotor flux linkage
Electrical and Computer Engineering, University of Missouri-Rolla.
ids d-axis stator current His research interests are in power system modeling, stability and
idr d-axis rotor current control.

vds d-axis stator current

vqs q-axis stator current Badrul H. Chowdhury obtained his M.S. and Ph.D. degrees in
Electrical Engineering from Virginia Tech, Blacksburg, VA in 1983
vdr d-axis rotor current
and 1987 respectively. He is currently a Professor in the Electrical
vqr q-axis rotor current & Computer Engineering department of the University of Missouri-
Rs stator resistance Rolla. From 1987 to 1998 he was with the University of Wyoming’s
Electrical Engineering department where he reached the rank of
Rr stator resistance Professor. Dr. Chowdhury’s research interests are in power system
ws angular velocity of the stator field modeling, analysis and control; power electronics and drives.
vd' d-axis transient reactance
v q' q-axis transient reactance