Load Modeling For Voltage Sag Studies

Load Modeling for Voltage Sag Studies J.A. Martinez, Member, IEEE, J. Martin-Arnedo, and J.V. Milanovic, Senior Member, lEEE

Abstract-. This paper investigates the influence of load

modeling on voltage sag characteristics and compares the results of voltage sag calculation with different load models. It is shown that some voltage sag characteristics are strongly influenced by the models chosen for representing power system loads. Therefore, depending on the aims of the voltage sag study, load models need to he carefully chosen. Index Terms- Load Modeling, Power Distribution, Power Quality, Simulation. 1. INTRODUCTION

A’

simulation tool based on a time-domain technique can be a powerful mean for analyzing the main characteristics of voltage sags originated in transmission and distribution networks. Many voltage sag studies are presently performed with the ElectroMagnetic Transients Program (EMTP) and like [I], 121. The built-in capabilities available in these programs can he used to duplicate very accurately most transients in power systems [3]. To date there is a significant experience on the application of EMTP-like tools in power quality studies [4]. However, there is a lack of experience for representing loads in power quality studies for which a very detailed model is not advisable. Consider, for instance, the simulation of voltage sags in a large power system. A detailed representation of the load at every node of the system is theoretically possible, but it is also time consuming. In addition, the derivation of such load models can he a very complex task, and not always possible. Therefore, simple load models are needed. Although load modeling is not a critical issue in many voltage sag studies, this aspect can he very important for some calculations, for instance the calculation of a voltage sag index based on the lost energy [ 5 ] , [6].

simulations using the Alternative Transients Program (ATP) is proposed in this paper. It is a static representation that includes a separate voltage dependence for real and reactive power demands [7]. A more accurate representation can be achieved by using generic dynamic load models or a combination of induction motor and static load models (i.e., hybrid approach). The paper has been organized as follows. Section I1 presents and analyzes a static load model. Section I11 presents a comparison of the results obtained with a static load model and an induction motor model in voltage sag calculations. Section IV describes a detailed study with several load models. They are used for voltage sag calculations in a power system with several voltage levels. The study will be useful to compare the severity of voltage sags originated at each level. The main conclusions of the study are given in Section V. 11. A VOLTAGE-DEPENDENT LOADMODEI A realistic load model has to include voltage and frequency dependence, as well as a dynamic behavior. This last aspect can have a strong influence on the characteristics of the voltage sags at some nodes. The frequency dependence is not usually a concern in voltage sag calculations. The models used in this work are aimed at representing loads for voltage sag calculations in transmission and distrihution networks. The goal is to use a model that can represent the interaction between a load and the system during the event that causes the voltage sag. The study included in this section is focused on the voltage dependence of static loads and illustrates the importance of load representation for voltage sag calculations. A power demand that incorporates voltage dependence can he expressed as follows n”

The most common approach for representing loads when using an EMTP-like tool is a constant impedance. This representation is a very crude approach of an actual demand and cannot represent accurately most situations, mainly when the dynamic behavior of the load or the energy supplied by the system are important issues. A new model for representing loads in voltage sag Juan A. Maninrr and Jacinta Manin-Ameda are with the Deparrament CEnginyetia El&ctrica,Univrnilal Politknica dc Catalunyu. 08028 Barcelona. Spain. laciao Manin-&do is preparing his Ph.D. with a grant from the Spanish Ministerio de Ciencia y Tecnolo@a. Iovica V. Milanovic is with the Dept. of Electtical Engineering and Elcctronicr. UMIST. P.O. Box 88. Manchester. M60 IQD, UK.

0-7803-7989-6/03/$17.00 02003 IEEE

where Pi,, and Qi,o,are the rated real and reactive power at nominal voltage, and V, is the p.u. voltage. On the other hand, “P

“4

aik = 1 k=O

bit = 1

; k =O

There is no built-in model in ATP for representing the load model described by equations ( I ) and ( 2 ) . Fig. 1 shows the implementation used in this work. The voltage dependence is modeled by means of a controlled current source whose value is adjusted during the initial steps of the simulation. The actual transient simulation can start only after the convergence to the rated values has been achieved.

2508

i

_L ~

- 1

Fig. I . ATP rrpmcntation of a static load model.

This approach has a good behavior in steady state calculations. The results, however, should he considered as an approximation of the real load model. The above equations assume that there could be a part of a power demand that is voltage-independent. In fact, this is the approach implemented in the majority of load flow programs. By using this approach, the power demand remains the same irrespectively of the values of bus voltages. This is not a realistic model for voltage sag calculation, as it would mean that even for very low rctained voltages, the demand will be the same as that prior to the sag. A VI dependence means that the load behaves as a constant current source, while a V2 dependence means that a load behaves as a constant impedance. Fig. 2a shows a single-line diagram of a very simple system used to test the performance of these models. The two plots included in this figure illustrate their behavior. It can be easily deduced that the per unit energy lost during a voltage sag will be greater than the per unit voltage drop if the load hehaves as a constant impedance model. As a conclusion, an accurate calculation, not only for steady state analysis, hut also for voltage sag calculations should be based on an accurate knowledge of the actual demand performance, since any of the models mentioned above would be a very crude representation for most loads.

Ill. LQADMODELCOMPAP~SON To illustrate the differences between the above static load

model and a model that takes into account a dynamic behavior of the load, two simple test systems have been used. The configurations are shown in Fig. 3a and 4% respectively. The first system is a two-feeder medium voltage (MV) network that feeds two constant impedance loads (V2 model). The MV side of the substation transformer is grounded through a zigzag reactance of 75 R per phase. The second system is a very simple low voltage (LV) network feeding a constant impedance load in parallel with an induction motor. In both cases, a short circuit is originated in one feeder (points F1 and F2 in Fig. 3a and 4a, respectively). The plots of Fig. 3 and 4 show the behavior of both models following a single-line-toground fault and a three-phase fault. Note that voltages and active powers have similar responses when a static load model

0.5

04

4 0

10

5

15

20

25

30

Sn [MVAI

h) Vollage at the load node

A ’

5

0

4,0

5

io

15

20

25

30

Sn[MVA>

c) Power demand 81 the load node Fig. 2. Steady-state analysis of volrageedependenlload d e l s (Power factor = 0.8 lag&).

is used, but they can he very different, depending on the fault, when the load is represented by an induction motor. In addition, voltage responses derived from both models are very different. Another conclusion from the comparison of both models is that the energy to he supplied by the system is very different. One can also observe that in the first case voltage swells and sags appear simultaneously as the result of a single-line-to-ground fault. This is due to the substation transformer group connection and the value of the grounding impedance. The severity of the voltage sag decreases at low voltage levels due to the presence of distribution transformers, as it will he shown in Section IV.

IV. LOADMODEL ANALYSIS A. Test System

Fig. 5 shows the diagram of the multi-bus test system used for the analysis of the influence of load models on voltage sags. The MV side of the substation transformer is also grounded through a zig-zag reactance of 75 Q pcr phase. This example attempts to illustrate how different load models can affect voltage sag characteristics.

2509

q

24 kw (lag) pf=O.XO

3ROV

1M HVyuivalrnl : 110 kV, 1500 MVA. XIR = 10 Subslalion Iransfomr: I10125 kV. 20 MVA. 8%. Yd Lines:Zli.=0.61 +j0.39,2,=0.76+jl.S6Lukm

-

a) D i a p m of the test system D

Equivalent impedance:Z W= 0.05 + j0.2S. 2, = 0.10 + j0.SO Qikm Line : ZLE= 0.60 + jO.lO, 2, = 1.50 + jO.20 Wkm

a) Diagram of Ihe test system

-

9

~.~ 40

-

0

en

120 Tinntmrl

160

1w

mo

1

en

120

nm lmsl

160

m

sm

TimI

0

la0

N,

b) Singlc-line-to-@ound fault - Voltages at mlor terminals

b) Single-line-to-groundfaull -Voltages ill load terminals

40

SO kW pf=0.93 (lag)

100

200

m

sm

,a0

T i m Imal

rm

c) Single-line-to-pound fault - Active powers

C) Single-line-io-pound fault - Active powers

Phase A

250

-

m

-E

150

3

100

0

P

50 I

On 40

m

160

120

lime (mr) d) Threephase fault - Voltage at load tcrminals

200

d) Three-phase fault - Voltage at motor leminals

Phase A

1.0,

Phase A

30

2

20

d

a

g

-

10

L

a

P

O

-10

0.0

40

en

tm

160

TlmC(r4

e ) Three-phase fault

~

m

1W

yw)

sm

l3w

Ticmlmrl

e) Three-phasefault - Active power

Active power

Fig. 4. Induction m l o r behavior.

Fig. 3. Constant impedance load model behavior.

2510

17M

HV equivalen!: I IO kV, 1500 MVA, WR = 10 Substation transformer: 110125 kV. 20 MVA. 8% Yd Distribution transfomr: 25 kVl415 V. 1 MVA. 470, Dyn Liner:Zla=0.61 +j0.39.&=0.76+j1.56~m Fig. 5 . D i a p m of the lest systcm.

B. Modeling Guidelines The present study assumes that voltage sags are caused only by faults. In gencral, transients associated with faults can be classified as low frequency or slow front transients. Since the goal is to obtain voltage sag characteristics, a simplified representation of the fault and the power components will suffice. All models used in this study are linear and havc lumped parameters. The representation of some important components, such as protective devices or mitigation equipment, has not been included. C. Load Models Four different low voltage load models have been analyzed. Their main characteristics are listed below. Constant impedance. The following parameters are used to represent this model: rated voltage (rms line-to-line) = 415 V; rated apparent power = 840 kVA; power factor = 0.95 (lagging). Voltage dependent load. Active and reactive power of every load represented by this model are calculated according to the following expressions P = P,, (0.2t 0.3V t OW2) (3)

D. Simulatioii Resirlts The study was performed by simulating faults at all MV nodes and measuring the voltage sag characteristics at MV and LV nodes. Voltage sags caused by a11 types of faults were analyzed. A software application was developed to classify the recorded waveforms and depict simulation results. Fig. 6 , 7 and 8 show a selection of results obtained by considering some load models detailed above. These results correspond to variables measured at MV and LV sides of the transformer located at BUS06 when the faults were applied at BUSOZ. Note that active and reactive power plots, shown in Fig. 8, correspond to three-phase powers; they were measured at the LV side of distribution transformers. The main conclusions derived from these results can he summarized as follows: When the voltage sags are caused by single-line-to-ground faults, the severity of the sag depends on the voltage level at which it is measured and the transformer connections. For the system tested in this work, it is evident that the voltage sag severity is lower at the LV side of the distribution transformers. As expected, thc load model has a strong influence on the voltage sag characteristics, as well as on the value of the lost energy. When the bus load or a part of it is represented as an induction motor, the duration of the sag is longer than the fault duration and the shape of the sag is nonrectangular. If the load is assumed to be static however, the sag will have a rectangular shape. Different static load models can also have some influence on simulation results. This can he deduced by comparing the rcsults obtained with a voltage dependent load, Fig 6, with those obtained with a constant impedance load, Fig. 3.

E. Discussion

Q=Q ,, (0.2 t 0.1V t 0.3V2 t O X 3 t 0.2V4) (4) where Vis the per unit rms voltage. The rated powers are the same as those for the constant impedance model. Induction motor load. It is represented by a single induction motor whose parameters have been adjusted to obtain a power demand of 800 kW and 400 kVAr, at a rated voltage of 415 V Hybrid load model. It consists of a combination of constant impedance and induction motor loads. The constant impedance demand is set to 420 kW, power factor = 0.95 (lagging), while the IM power is set to 400kW1200 kVAr. 2511

The energy lost during voltage sags can be a good index to assess or benchmark systcm performance 151. However, the calculation of the energy must be carefully made. Some voltage sag energy-based indices have been proposed for which the calculation is made with the per unit voltage drop for each phase, see for instance [ 5 ] , 161. From the results derived for single-line-to-ground faults, one can deduced that such values will he very different if the voltages are measured at MV or LV sides of transformers. If the transformer losses are neglected, only an index based on the actual lost energy can give similar results by using variables measured at both transformer sides. For instance

where P,,re.xa8and P,,, are the active power prior to the sag and during the sag, respectively 181. A very important issue related to these results is the model chosen for representing transformers. It is obvious that distribution transformers can become saturated during voltage swells. This effect has not been included in any transformer model in this study, as all models are linear. Another important aspect of transformer modeling is the model chosen to represent the core topology. The models used in this work are based on a three single-phase

c

--* 5

-

1.2

0.8

9

1

OC

0.0 200

am Imp ,MO nm(nr) a) Voltage dependent load Single-line-to-ground fault 400

200

6w

c

B

0,o

2W

400

em

cm

n m (m)

0.0

,WO

ZW

b) Voltage dependent load - mer-phase fault

ZW

1 2

-

-

ua 6w m Imp nm (nm) c) Induction motor - Single-line-to-pound fault B

UIO

6w

“(nu)

8m

200

ua

BM T i m (m)

Ea0

e) Hybrid load

~

sm

hmlm)

Imp

b

-

c

2m

,WO

d) Induction mtor - Three-phasefault

0.1

-

Em

-

-

0.0

ua

BM (m)

-

Ti-

1 m

0.0

200

6w

b) Voltage dependent load - Three-phase fault 1.1

. ,

400

a) Voltage dcpendcnt load - Single-line-lo-poundhull

~

,WO

400

cm n- im)

m

,WO

d) Induction motor - Three-phasefault

I

2m

1200

400

WO

*W T i m (nu)

,WO

1200

e ) Hybrid load - Three-phasd fault

Thrm~phasefault

Fig. 6 . Lovd d e l analysis - Pcr unit vdwgei at BUS06 (Vanahles are masured at MV side).

Fig. 7. Laad model analysis -Per unit voltages ut Load 5 (Variables are m a r u e d at LV side).

2512

transformer hank. It is well known that this is no1 a good representation of some core designs, i.e. a three-legged stacked core. The main goal of the study, however, was to illustrate the differences that can result from the application o l different load models in voltage sag calculations; therefore, transformer modeling as such is not of a paramount importance. V. CONCLUSIONS

0.0 200

UXI

6m

am

,om

Tim tm1

1200

a) Voltage dependmi load - Single-line-lo-pound fault

zw

-

uy)

MD

am

Tim ("4

Imo

Load reprcsentation is an important subject for which timedomain tools have significant advantages. This paper has illustrated the importance of load modeling for voltage sag analysis using a widely used simulation package. The paper emphasizes the importance of load modeling for an accurate determination of voltage sag characteristics and the subsequent assessment of equipment sensitivity to voltage sags. It is shown that, depending on the load model, different values of energy based voltage sag indices can he obtained.

1200

-

VI.

b) Voltage dependent load - Three-phase fault AOUWtMm

R-stive

111

(MVAr)

121

131 141 0.0 2W

400

w4

6w

T"

c) Induction motor load

,Ow

I

[61

Single-line-to-pound hull

~

rsi

1 m

(M)

171

2,01 2,o

i'\

1.5

181

REFERENCES

J.A. Maninei and J. Martin-Amedo, "Vallage sag analysis using an elecuonwgnetic trunsienls pro@m" IEEE PES Winrer Meeting,January 27-31. 2002. New York. J.A. Maniner-Velasco and J. Manin-Ameda. "Stochaslic predinion of vollap dips using an electromag?etic transients program" 141h PSCC, June 24-28.2002. Seville. H.W. D o m l , ElectroMognetir Trarisientr Pro,q,nni. Rqfererice Monuul (EMTP Theory Book). Bonnrvillr Power Administration. Ponland. 1986. J.A. Murtincr-Velasco (Ed.), Computer A,tulyris of Electric POIW.PT Sytem Tmnrienrr. IEEE Press. 1997. IEEE Voltage Quality Working Group, "Recommended practice for the c s t a b l i i h n t of voltage sag indices." IEEE PISM, Dnft.March 2001. R.S. Thallnm and G.T. Heydt. "Power acceptability and voltage sag indices in the three phase sense," 2000 IEEE PES S u m r Meeting, July 16-20.2000. Seattle. P.A. Gnadt and J.S. Lawler (Eds.), Automatic Electric Utiliry Diitriburim Syrtenrr, Prentice Hall, 1990. J.A. Maninez and J. Manin-Amedo, "Voltage sag stirhastic prediction using an elcctromagnetic transients propam" submitted for publication in IEEE Truris.on P o w r Delivery.

1.0

V11. BIOGRAPHIES 0.0

4.5

4M

rwa

MD

1x4

Juan A. Martine2 was ban in Barcelona (Spin). He is Profesor XTUlar at the D r p v w r r n t #Enginyea El&micaofthe UniveaiW Politi.oliw. de Cutalunya. His teachjnr and r e m c h intcmls include Transmission and DisVibtUion Power System Analysis and EM"

applications.

Jacinto M a r t i n - A d o was bom in Barcelona (Spin). He is cumntly a W.D. candidate at the Univmitat Polilecnica de Cawlmya. His research interrsls include Power Wily SNdies using EMF-rype tools and Transient Analysis of Power systcm.

I

4 . 4

University of Newcastle, Australia. His employment experience includes "Energoproject-MDD consulting and engineering company. as well as the Universities of Belgrak in Yugoslavia and Newcastle and Tasmania in Auslralia. In January 1998 he joined the Depannlenl of Elrcuical Engineering and Electronics at UMIST. UK where he is currently a Senior Lecturer.

0.1 -0.5

zw

Jovica V. Milsnovic (M95, SM98) received his DiplJng. and his M.Sc. degree from the University of Belgrade, Yugoslavia, and his Ph.D. degree from the

4M

6W

,000

Ti-tm)

e) Hybrid load

~

Threephase h u h

Fig. 8. Load d e l analysis - Active and reactive powers - b d 5

2513