You are on page 1of 5

2008 IEEE Electrical Power & Energy Conference

Dynamic Modeling of Induction Motor loads for


Transient Voltage Stability Studies

T. Aboul-Seoud J. Jatskevich
IEEE Student Member SM IEEE
The University of New Brunswick The University of British Columbia
Fredericton, NB, Canada Vancouver, BC, Canada
T.Aboul-Seoud@unb.ca jurij@ece.ubc.ca

Abstract— Due to the complex nature of the dynamic stability voltage varies between 0.95 and 1.05 pu [3], which is a narrow
studies, and the high cost of simulating large power systems in range compared to some severe fault cases.
real time, most of the voltage stability studies are limited to the
static techniques. The main disadvantage of the static modeling is Moreover, the induction motor static model doesn’t take
its inability to capture the dynamic behavior of induction motors into consideration the effect of the load inertia. The load inertia
which is a common load in the power system. This paper is a very significant factor when it comes to stability studies.
compares between the static and dynamic models of aggregated
This work is devoted to compare between the induction
induction motor loads. It demonstrates that in some disturbance
motor static model and dynamic model in transient voltage
cases, the static load modeling fails to predict the actual response
of induction motor loads. The studies show that while the static
stability studies. It explains the effect of the load inertia which
load model predicts a stable operation with both the conventional can’t be presented in the static model, and which has a
P-V curve and a P-V curve of the induction motor load at significant effect on the system stability.
different values of machine slip, the dynamic model shows
unstable operation for the same cases. Finally, the effect of the II. CASE SYSTEM DESCRIPTION
dynamic load inertia on system stability is studied. Although the
inertia constant of a dynamic load is significant in studying its The simplified network shown in Fig. 1 is used to simulate the
stability, static load model is unable to take it into consideration. responses of the aggregate industrial load. The system consists
Simulations will show that for similar rating motors with of a 10MW, 6.9kV Aggregate induction motor load connected
different inertia constant, the steady state condition varies. This to the grid via a 200km, 69kV double circuit OHTL. The
exposes another deficiency in static load models of induction transformers were modeled as a series RL circuit, ignoring the
motors. core resistance and the magnetization reactance effect, while
the OHTL was modeled as a П network. The length of the
Keywords: induction motors, load modeling, power system transmission line is 200km. At t=10 seconds, a three phase
dynamic stability, power system simulation. short circuit fault causes the loss of one of the double circuit
lines. This fault is cleared at t=30seconds.
I. INTRODUCTION
A. Static load model
Transient voltage stability studies and voltage collapse are
dynamic phenomena. However, since it is very expensive to A general form of the voltage dependent static load model
model power systems dynamically, due to the complex nature is algebraically described in (1) and (2). It is commonly known
of the problem and the different significant factors affecting it. as the exponential load model.
Voltage stability studies are usually limited to the power flow α
based static techniques [1]. ⎛V ⎞
P1 = Po ⎜⎜ 1 ⎟⎟ (1)
Most of the linear residential and commercial loads could ⎝ Vo ⎠
be successfully modeled as a static load since they are mainly β
resistive loads (heating, light), and hence they have no dynamic ⎛V ⎞
Q1 = Q o ⎜⎜ 1 ⎟⎟ (2)
response. However, this is not the case for industrial loads ⎝ Vo ⎠
which depends on induction motors and non-linear loads
(adjustable speed drives, arc furnaces, etc). Where, P1,Q1 are the active and reactive power consumed
by the load, Po ,Qo are the rated real and reactive load power,
Several models were developed to model various linear
loads statically. The most familiar model relates the load active while V1 is the terminal load voltage corresponding to
and reactive power exponentially [1]. This model was P1,Q1 and Vo is the rated voltage at which Po ,Qo are
improved by introducing the effect of frequency [2], [3]. calculated. The constants α and β are the active power-Voltage
However, even after this improvement, the model fails to
follow the actual experimental results precisely when the exponent, and the reactive power-Voltage exponent
respectively. Their standard values are given in the Appendix.

978-1-4244-2895-3/08/$25.00 ©2008 IEEE


Bus 3
Bus 1 Bus 4
A 3-phase fault Line 1 69/6.9
At t=10s 69/13.8kV Industrial

~ .
Line 2
IM Plant
(Load)
The Power Grid 200km
Bus 2 transmission line

Figure 1. Case System Layout

The P-V curve of the system was calculated twice; the first side. ψ qs ,ψ ds are the stator q-axis and d-axis per second flux
is the common PV curve [4] where the load is presented as a
linkage while ψ qr ′ ,ψ dr
′ are the rotor q-axis and d-axis per
constant power factor load. The second is calculated by
representing the load using the steady state induction motor second flux linkage referred to the stator side. Te ,TL are the
equivalent circuit at different slip values. d
electric and load torque respectively. p is the operator ,
dt
B. Dynamic load model while ω , ωo , ωr are the angular speed of the reference frame,
The dynamic model of the aggregate induction motor load the base angular speed and the rotor angular speed respectively.
is modeled using the five states model [5]. The output active The constant H is the inertia constant in seconds. The motor
and reactive powers are calculated using the algebraic was simulated in the stationary reference frame, thus ω = 0.
equations (9) and (10) respectively.
The induction motor parameters and the transmission line
ω p parameters are given in the Appendix.
v qs = rs iqs + ψ ds + ψ qs (3)
ωb ωb
III. SIMULATION RESULTS
ω p
vds = rs ids − ψ qs + ψ ds (4)
ωb ωb A. Static load model studies
The static model is much cheaper when it comes to
⎛ ω − ωr ⎞ p (5)
′ + ⎜⎜
′ = rr′iqr ⎟⎟ψ dr
′ + ψ qr
′ simulation cost. However, the saving comes on the expense of
v qr
⎝ ωb ⎠ ωb its accuracy and ability to estimate the actual response of
dynamic load such as induction motors. Fig. 2 shows that the
⎛ ω − ωr ⎞ p (6) P-V static response estimated the ability of the induction motor
′ − ⎜⎜
′ = rr′idr
vdr ⎟⎟ψ qr
′ + ψ dr
′ load to retain stability condition after the fault on the
⎝ ωb ⎠ ω b
conventional P-V curve of the system. Point A is the stable
(7) operating point of the motor before the fault, while point B is
⎛ 3 ⎞ ⎛ poles ⎞ ⎛⎜ 1 ⎞⎟
Te = ⎜ ⎟ ⋅ ⎜ ⎟ ⋅ ⎜ ⎟ ⋅ ψ qr
′ idr (
′ − ψ dr
′ iqr
′ ) the estimated stable operating point of the motor after the fault.
⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ ωb ⎠ Furthermore, fig. 3 shows that the static model will estimate a
(8) stable condition on the induction motor equivalent circuit P-V
⎛ω ⎞
Te = 2 H ⋅ p⎜⎜ r ⎟⎟ + TL curve as well. Point A is the stable operating point of the motor
⎝ ωb ⎠ before the fault, while point C is the estimated stable operating
point of the motor after the fault.
(9)
P=
3
2
(
v ds i ds + v qs i qs ) B. Dynamic load model studies
(10) Dynamic load models are the most accurate model of
Q=
3
2
(
vqs ids − vds iqs ) induction motors. In the studied case, the dynamic model
simulation estimates a motor crash as a result of the fault. Fig.
4 shows the change in the power delivered to bus 4, and the
r.m.s. voltage at bus 4. It can be noticed that at t=10s a rapid
decrease in bus 4 voltage started to occur leading to the motor
Where, P,Q are the active and reactive power consumed crash as the motor terminal voltage decreases below the motor
by the load, vds ,ids are the stator terminal direct axis voltage critical voltage.
and current while v qs ,iqs are the stator terminal q-axis voltage Fig. 5 shows that the P-V dynamic response estimated the
′ ,idr
and current. vdr ′ are the rotor direct axis voltage and current crash on the conventional P-V curve of the system.
referred to the stator side while v qs ′ ,iqs
′ are the rotor q-axis Furthermore, fig. 6 shows that the dynamic model will estimate
a crash on the induction motor P-V curve as well. Point A is the
voltage and current referred to the stator side. rs ,rr′ are the stable operating point of the motor before the fault in all cases.
stator and rotor per phase resistance as seen from the stator
From fig. 4, it is worth noting from the power response at 5
bus 4 that the motor doesn’t crash instantaneously after the Normal Condition
fault. This is due to the motor inertia that allows it to survive Faulty Condition
the crash if the fault is cleared before the critical clearing time. 4 Point A Dynamic Response

4 3

V(kV)
Point A
3.5
2
3
Point B
2.5 1
V(kV)

2
0
1.5 0 5 10 15 20 25
P(MW)
1
Normal Condition
0.5 Faulty Condition Figure 5. Dynamic response on a conventional P-V curve at H= 3s
Static Response
0 4
0 5 10 15 20 25
P(MW) Point A
3.5
Figure 2. Static response on a conventional P-V curve

4
Point A 3
V(kV)
3.5
Point C 2.5
3

2.5
Normal Condition
2
V(kV)

2 Faulty Condition
Dynamic Response
1.5 1.5
0 5 10 15
1 P(MW)
Normal Condition
Figure 6. Dynamic response on the induction motor equivalent circuit P-V
0.5 Faulty Condition
curve at H= 3s
Static Response
0
0 5 10 15
P(MW) C. Static models vs. Dynamic models
Figure 3. Static response on the induction motor equivalent circuit P-V curve The static model of the induction motor estimated that the
motor will be able to achieve a stable operating condition after
the fault as shown in fig 2 and fig. 3. However, the dynamic
10 model, which is more accurate, estimated a failure in the motor
P Bus4(MW)

stability that led to a crash as shown in fig. 5 and fig 6. This


demonstrates the fact that a static model of induction motors,
5
although offers a low simulation cost, fails to estimate the
operating condition successfully constantly.
0
0 5 10 15 20 25 30
Time(s) D. Effect of load inertia
4 One of the most important characteristics of power systems
is the inertia of the system. The higher the system inertia is, the
VBus4(kV)

3 higher the chance that the system will maintain a stable


operating condition after any disturbance. One of the main
2 disadvantages of the static load models is that it doesn’t take
the load inertia constant into consideration. To demonstrate the
0 5 10 15 20 25 30 significant effect of the inertia of the load, the system is
Time(s) simulated after tripling the induction motor inertia. The fault is
Figure 4. Power transferred to and root mean square voltage at bus 4 cleared at t=30s. Fig. 7 shows the power delivered to bus 4,
while fig. 8 demonstrates the voltage at bus 4. Comparing the
dynamic response of the induction motor at H=3s, which is the
nominal inertia, and H=9s, it could be concluded from fig. 7
and fig 8, that the motor steady state condition is different in 5
both cases. The motor was able to regain its pre-fault operating Point A Normal Condition
condition at H=9s while it fails to obtain this condition and Faulty Condition
ended in a crash condition at H=3s. It is worth mentioning that 4 Dynamic Response
the clearing time was chosen outside the normal range of the
power system fault clearing time just to demonstrate the effect 3

V(kV)
of inertia since the studied system is extensively simple
compared to the real power system. Point D
2
To establish a further discussion to explain the effect of
inertia on the motor stability, the P-V dynamic response for
H=9s is plotted on the conventional P-V curve in fig. 9 and on 1
the induction motor equivalent circuit P-V curve on fig. 10
Point A is the pre-fault as well as the final steady state 0
0 5 10 15 20 25
operating point. Point D is the fault clearing point. The
P(MW)
dynamic response follows the arrows in fig. 9 and fig. 10.
Comparing fig. 5 with fig. 9 and fig. 6 with fig 10 respectively, Figure 9. Dynamic response on the conventional P-V curve at H=9s
the effect of inertia could be identified. Although the clearing
time is similar in both cases, only in the case at which H=9s the 4
Point A
motor is capable of retaining its pre-fault condition.
20 3.5
H=3s.
H=9s.
15 3
V(kV)
2.5
(MW)

10
Point D
Normal Condition
Bus 4

5 2 Faulty Condition
P

Dynamic Response
0 1.5
0 5 10 15 20
P(MW)
-5 Figure 10. Dynamic response on the induction motor equivalent circuit P-V
0 10 20 30 40 50 curve at H=9s
Time (s)
Figure 7. Power transferred to bus 4 at H=3s and H=9s
IV. CONCLUSIONS
4 The high cost of simulating large power systems in real
time forced most of the voltage stability studies to be limited to
the static techniques. However, this comes on the expense of
3.5 the inability of the static model to capture the accurate dynamic
behavior of induction motors which is a common load in the
power system. This paper demonstrated that the static model of
VBus 4 (kV)

3
induction motor fails to successfully estimate the final
condition of the motor after a system disturbance. Although the
2.5 static model estimated that the motor will be able to continue
operation successfully after the fault, the dynamic simulation
H=3s. showed that the motor will crash. Moreover, one main
2 H=9s. disadvantage of the static model is that in addition to ignoring
the dynamics of the induction motors; it doesn’t take into
1.5 consideration the effect of the load inertia constant. Variation
0 10 20 30 40 50 in the inertia constant proved to have a significant effect on the
Time (s)
ability of the motor to retain stability after clearing the fault.

Figure 8. The root mean square voltage at bus 4 at H=3s and H=9s
APPENDIX REFERENCES
The studied industrial plant is modeled as an aggregate [1] Xu, W., Mansour, Y., “Voltage stability analysis using generic
induction motor which has the data given in table I. dynamic load models,’ in IEEE Trans. on Power Systems, vol. 9, no.
14, pp. 479-493, Feb. 1994.
[2] IEEE Task Force on Load Representation for Dynamic Performance
TABLE I. INDUCTION MOTOR PARAMETERS Ihara, “Standard load models for power flow and dynamic
performance simulation,” in IEEE Trans on Power Systems, vol.
Rating(MW) Voltage(V) Frequency Poles rs (Ω) 10, no. 3, pp: 1302 –1313, Aug. 1995.
(Hz)
10.0 6900 60 4 0.0584
[3] Louie, K. W., Marti, J. R., “A method to improve the performance
R’s (Ω) Xls (Ω) X’lr (Ω) XM (Ω) H (s)
of conventional static load models,” in IEEE Trans. on Power
Systems, vol. 20, no. 1, pp. 507-508, Feb. 2005.
0.0443 0.568 0.568 26.2646 3
[4] A. R. Bergen, V. Vittal, Power Systems Analysis, 2nd ed., Prentice Hall,
2000.
The transmission line is modeled using the medium
[5] P. C. Krause, O. Wasynczuk, S. D. Sudhoff, Analysis of Electric
transmission line П model. The line parameters are given in Machinery and Drive Systems, 2nd ed., IEEE Press: Wiley, 2002.
table II.

TABLE II. TRANSMISSION LINE PARAMETERS

Length(km) Voltage(V) rL (Ω/km) xL (Ω/km) yC (MΩ-1. km-1)


200 69,000 0.161 0.474 0.291

Such that:
rL: Transmission line resistance per unit length,
xL: Transmission line inductive reactance per unit length,
yC: Transmission line capacitive admittance per unit length.

The typical load parameters are given as follows [1]:


α = 0.72 − 1.30, β = 2.96 − 4.38 for residential loads;
α = 0.72 − 1.30 , β = 3.15 − 3.95 for commercial loads; and
α = 0.18, β = 6.0 for industrial loads.

You might also like