You are on page 1of 5

Modeling Electric Arc Furnace to Estimate Flicker

Levels in Power Systems Using ATP


G.R. Estrada-Villa, E.A. Cano-Plata and A.J. Ustariz-Farfan
Department of Electrical, Electronic and Computer Engineering
Universidad Nacional de Colombia - Sede Manizales
Manizales, Colombia
AbstractElectric arc furnace and its influence in the power
system are studied in this paper. Two models that allow to
evaluate its behavior during steady state are shown. These
models has different complexity level; the first model takes into
account the main nonlinearities of the load; the other assumes
that current-voltage characteristic in the arc is piecewise linear.
Electromagnetic transients program ATP was used to simulate
power system and load. Simulated results are compared with
actual measurements. Power system was modeled including the
point of common connection and electric utility grid.
Keywords: Arc furnace model, Current-voltage characteristics,
Power system simulation, Power quality, Voltage fluctuations.

I.

INTRODUCTION

The use of nonlinear loads and sensitive devices has


resulted in concerns about power quality [1].
The electric arc furnace is usually the largest single load
encountered in a distribution network; this one consumes huge
amounts of real and reactive power. It is known that a single
load of this magnitude is certainly a potential source of power
quality problems, like harmonics, voltage fluctuations and
unbalance, which are found in an utility grid [2][4].
There are several solutions for power quality problems that
could be incorporated in the design stage or added later. It is
necessary anticipate the flicker level to determine the solution
able to reduce the voltage fluctuation. That being the case, it is
important obtaining arc furnace time response. The most
effective and simple way to obtain the load response is
measuring directly voltage and current [2][4], but in new
projects it is impossible. Hence, electric utilities, and users with
this type of load, should have a method that allows them assess
the problem. It is important that the electric arc furnace can be
served without causing problems to other existing or future
loads in the grid [2], [3].
From an electromagnetic view point, the arc furnace
behavior is not strictly periodic, it cant be analyzed with
precision by Fourier series and harmonics analysis [5].
Sometimes it is considered as a transient load, and the flicker is
a greater problem than the harmonics [6].

This work was supported in part by Universidad Nacional de Colombia Manizales.

978-1-4673-6487-4/14/$31.00 2014 IEEE

From a power quality point of view, this paper analyses the


arc furnace modeling problem and focuses in compare two
kind of models and determine the error when the flicker level is
calculated.
The first model type uses electric arc equations to depicted
the phenomenon. In this model, main nonlinear features of the
furnace systems are represented. Arc furnace is depicted using
the Mayr's model and an iterative method is used to add a
random signal in this approximation.
Another model type, based on field measured data, is
reviewed in this paper. This model is a deeper simplification
based on a piece-wise linear approximation of current-voltage
characteristic, and pays special attention to the current and
power in the arc because they are closely related to the flicker
study.
Both models use system data that are readily accessible to
planning engineer, and the amount of flicker created by the
load is known via calculation using the criteria provided by
IEC 61000-3-7 [7] and IEEE 1453 [8]
II.

ARC FURNACE CIRCUIT

The main elements of arc furnace circuit are shown in Fig.


1. They are: transformer, power cable (RLC component) and
arc model. The power system simulation was made using the
electromagnetic transients program ATP [9].

Fig. 1 Arc furnace circuit.

Reference[4], shows a description and modeling of the


different circuit elements. In this, the study of reactance and
power cable has been emphasized. However, the model can be
simplified to get a fast simulation process.
Fig. 2 shows the voltage waveform in actual arc furnace
measured at the PCC.

i1 =

Vig

(3)

R1

i2 = Vex Vig (1

R2
)
R1

(4)

iR 0 i i
1
1

R
v = iR2 + Vig 1 2 i1 i i2
R1

( i i ) e pro t / a d i ( t ) < 0
1 2
dt

Fig. 2 Voltage waveform

All system parameters were obtained from measures on an


actual arc furnace located in Colombia, rated at 27MVA and
connected to a 13.8/0.6 kV transformer at 60 Hz of power
frequency.
A. Model type I: Electric arc furnace equations
Using equations to simulate the electric arc behavior, a
complete arc furnace model was made taking into account the
main nonlinear features of the load.
The electric arc model is derived from Mary's model and is
represented by a time domain controlled voltage source using
MODELS in ATP [5], [9]. In this way, an easy model for
computer implementation is obtained.
Mary's model shows that a arc current:

ia = 2I sin t

(1)

where

(5)

V
V
ex and ig are extinction and ignition voltage
i

respectively, 1 and 2 its respective currents, and the thermal


inertia is modeled using the last part of the Mayrs model, whit

time constant a .
Electromagnetic forces in the electrodes and the random
oscillation effect in the power cable, were replaced by a
R

dynamic pattern in the resistance parameter 1 and 2 with


values between 0.0001 and 0.0009 using a band-limited
(4-16Hz) white noise variation.
Fig. 3 shows arc furnace characteristic simulated using
TACS from ATP [9]. The action of ignition, extinction, and reignition the electric arc, from a electrical perspective, is the
source of wide power swings which translate into potentially
large voltage swings resulting in flicker. This phenomenon is
shown Fig. 3.

causes an arc voltage that is expressed as follows:


va =

2V0 sin t
sin(2t + a )
1 + (2 a )2

where V0 =

pr0
;
2I

tan

(2)
a

1
2 a

a is the arc time-

constant, and r 0 the arc column power at moment of


interruption [4], [5]. The ATP-MODELS code is shown in
appendix VIII A.

Fig. 3 Arc furnace current-voltage characteristic

B. Model type II: Piece-wise linear approximation


Reference [4], provides an approximation using the
approach proposed in [6]. This model is based on a piece-wise
linear approximation of the current-voltage characteristic rated
of the load. In this case, the current-voltage characteristic is
modified in such a way that power consumed is that specified
by the user and thermal inertia is an exponential expression.
The model proposed by [4] has three regions:

III.

FLICKER IN POWER SYSTEM

A. Flicker Definition
Voltage fluctuations on electric power systems give rise to
noticeable illumination changes from lighting equipment. This
phenomenon is often referred as light flicker, voltage flicker or
just flicker [8].

The IEC Flickermeter estimates the level of sensation on


light flicker, starting from the voltage fluctuations that cause it
[10]. The measure of severity based on an observation period
of ten-minute is designates short-term flicker severity (Pst).
This value is the standard output of the IEC flickermeter [8].
Pst is obtained from a statistical analysis of the instantaneous flicker level. The following relationship is used [8]:

Symmetric RL coupled line model was used to represent


transmission lines and power cable. Internal network (BUS2) is
a subterraneous cable and was modeled with RLC component.
Given that predominant effect in the subterraneous cable is
capacitive, this was simulated neglecting the inductive effect.
The example system include capacitor bank.
TABLE I.

Pst

= k0.1P0.1 + k1 P1 s + k3P3s + k10P10 s + k50 P50 s

Parameter

(6)

where k0.1, k1, k3, k10 and k50 are constants defined by IEC a
whose values are 0.0341, 0.0525, 0.0657, 0.28 and 0.08
respectively, and P0.1, P1, P3, P10 and P50 are the instantaneous
flicker level values exceeded for 0.1%, 1%, 3%, 10% and 50%
of the time during the observation period. The suffix s in the
formula indicates that the smoothed value should be used;
these are obtained using [8]:

P +P +P
P +P +P +P +P
P50 s = 30 50 80 , P10s = 6 8 10 13 17
3
5
(7)
P2.2 + P3 + P4
P0.7 + P1 + P1.5
,
P3 s =
P1 s =
3
3
The term P0.1 cant change abruptly and no smoothing
needed for this percentile [8], [10].
IV.

SYSTEM PARAMETERS

ARC FURNACE MODEL VALIDATION

The system shown in Fig. 4 represents an actual arc furnace


located in Colombia. This system was used to know the
performance of both methods.

Z+ = 0.2494+j0.7252
Zo = 0.6766+j2.3655
R = 0.0123
C = 0.0227 F

BUS1
BUS2

T1

Z% = 10.27% on 27 MVA
bases, X/R = 10

T2

Z% = 6.2% on bases,
X/R = 40

CAPACITOR BANK
POWER CABLE

C = 38.05

R = 0.657
Z+ = 0.00025+j0.003121
Zo = 0.0566+j0.90549

A. Simulation using Model I


Short term flicker severity calculations were performed
using the method provided by [11]. Pst value simulates at the
PCC is shown in TABLE II.
The waveform shown in Fig. 5, simulated in ATP, was
taken at the low-voltage terminals of furnace transformer, and
shows the voltage signal distorted due to the furnace operation.

Fig. 5 Simulated voltage in ATP using Model I

Fig. 4 Single line diagram of power system

To observe the effect of electric distortions caused by the


load, the power system was simulated including the point of
common connection (PCC) and electric utility grid. System
parameters are provided in TABLE I.

B. Simulation using Model II


Pst calculations were performed using the method provided
by [11]. The results are shown in TABLE II.
Fig. 6 shows the voltage signal distorted due to the furnace
operation. The waveform, simulated in ATP, was taken at the
low-voltage terminals of furnace transformer.

TABLE II.

COMPARISON OF SIMULATION RESULTS AND


MEASUREMENTS AT PCC

Pst phase a

Pst phase b

Pst phase c

Model I

2.2658

1.3771

0.9272

Model II
Measurements Pst95%

1.2566

1.2580

1.2561

1.41

1.25

1.23

The model inaccuracies are shown in TABLE II. and the


possible reasons are discussed in subchapter V.
Fig. 6 Simulated voltage in ATP using Model II

C. Flicker measurements
The flicker measurements were taken at the PCC in the
power system in Fig. 4 using an IEC compliant flickermeter.
Fig. 7 shows Pst values measured at the PCC in a timeline
of two hours. Abrupt changes indicate the start or stop in
furnace operation.

Model I shows a bigger error in highest measured value


(60.7%). Though the error in the phase b and c is lower
(10.16% and 24.62% respectively) that in phase a, is high
enough to consider the model unreliable.
Although the difference between the highest measured
value and the simulated using Model II exceeds 11%, is lower
than in the other case and shows better accuracy, for this reason
is used to estimate the voltage fluctuation level in the others
points. Flicker levels at PCC, BUS1 and the low-voltage
terminals of furnace transformer (l-v T2) were simulated and
results are shown in TABLE III.
TABLE III.

Fig. 7 Flicker Pst measured at the PCC

Fig. 8 shows the cumulative probability distribution of Pst


measurements taken at the PCC.

SIMULATION RESULTS OF MODEL II

Pst phase a

Pst phase b

Pst phase c

BUS1

1.1548

1.1559

1.1541

PCC

1.2566

1.2580

1.2561

l-v T2

13.5929

27.5034

13.6131

The low-voltage terminals of furnace transformer showed


the highest Pst values; however, diminish upstream of arc
furnace. The flicker level is greatly attenuates when it
propagates through both transformers T2 and T1 and reaches
the 115 kV bus (BUS1).
V.

FURTHER ANALYSIS

Though Model II shows a higher accuracy, its important


discuss the possible reasons for the differences among the
phases.

Fig. 8 Cumulative probability of Pst measured at the PCC

Field measurements and simulation results are shown in


TABLE II. Although there were not available flicker measures
of one full week, reference was made to the Pst95% to validate
both methods by following the indices reported in [10].

In the example power system, any other load different to


the furnace was ignored. Further, system voltage was assumed
balanced, but field measurements show a small imbalance at
the PCC. Transmission lines were modeled using their
symmetrical component sequence impedances, and by
definition, this assumes the same mutual coupling between
each phase of a given transmission line. When shield wires are
involved the amount of error introduced by this is on the order
of 1% or more [12]. Model parameters for each furnace phase
are equal, so its possible that similar or duplicate results are
obtained. Accordingly, the modeling assumptions are the main
cause of error between measured values and simulated values.
For planning purposes, typical flicker level values used to
evaluate the load impact in the power systems are listed in [8].
These values provide planning criteria that should be used to

assess serviceability of flicker generating loads such as arc


furnace. However, what if a load that does not exceed the
permitted limit is connected to a network that already has this
problem? and who bears the responsibility for implementing
the corrections?
VI.

CONCLUSIONS AND RECOMMENDATIONS

It was observed that the flicker level decreased passing


through both transformers. Therefore, it's possible to reduce the
flicker propagation in power systems by varying some
transformer parameters.
Lack of measures at other points of the network, did not
allow to validate the model behavior for flicker propagation
studies in a full utility grid. Even so, based on the results
obtained at PCC, Model II performance can be considered
satisfactory for flicker level studies.
The imprecision in the results is not because faults in the
model, but rather the parameters assumptions. Some
simplifications can achieve a quick and easy study, but may
increase the error.
To achieve a better understanding of electrical systems, this
kind of studies should have a physical support of related
phenomenon.
VII.

REFERENCES

[1]

E. A. Cano Plata and H. Tacca, Power quality assessment and


load identification, in Ninth International Conference on
Harmonics and Quality of Power. Proceedings (Cat. No.00EX441),
2000, vol. 3, pp. 840845.

[2]

G. W. Chang, Y. J. Liu, and C. . Chen, Modeling voltage-current


characteristics of an electric arc furnace based on actual recorded
data: A comparison of classic and advanced models, in Power and
Energy Society General Meeting - Conversion and Delivery of
Electrical Energy in the 21st Century, 2008 IEEE, 2008, pp. 16.

[3]

R. Horton, T. A. Haskew, and R. F. Burch IV, A Time-Domain


AC Electric Arc Furnace Model for Flicker Planning Studies,
Power Deliv. IEEE Trans., vol. 24, no. 3, pp. 14501457, 2009.

[4]

E. A. Cano-Plata and H. E. Tacca, Arc furnace modeling in ATPEMTP, in The 6th International Conference on Power Systems
Transients (IPST 05), 2005.

[5]

A. E. Emanuel and J. A. Orr, An improved method of simulation


of the arc voltage-current characteristic, in Harmonics and Quality
of Power, 2000. Proceedings. Ninth International Conference on,
2000, vol. 1, no. 2, pp. 148154.

[6]

S. Varadan, E. B. Makram, and A. A. Girgis, A new time domain


voltage source model for an arc furnace using EMTP, Power Deliv.
IEEE Trans., vol. 11, no. 3, pp. 16851691, 1996.

[7]

Assessment of Emission Limits for Fluctuating Loads in MV and


HV Power Systems, IEC 61000-3-7.

[8]

IEEE Power Engineering Society, IEEE Recommended Practice


for Measurement and Limits of Voltage Fluctuations and Associated
Light Flicker on AC Power Systems, IEEE Std 1453-2004
(Adoption CEI/IEC 61000-4-151997+A12003), 2005.

[9]

Leuven Center, Electromagnetic transients program (EMTP)/ATP


Reference Manual. Belgium, 1987.

[10]

P. Caramia, G. Carpinelli, and P. Verde, Power Quality Indices in


Liberalized Markets. Chichester, UK: John Wiley & Sons, Ltd,
2009.

[11]

B. E. M. C. Publication, Electromagnetic compatibility (EMC) Part 4-30: Testing and measurement techniques - Power quality
measurement methods, IEC 61000-4-30, 2003.

[12]

P. M. Anderson, Analysis of Faulted Power Systems, IEEE


Press, 1995.

VIII.

APPENDIX

A. ATP-MODELS data file for the implementation of Mary's


model
MODEL MAYR
CONST
maxconduc {VAL:1.E6}
DATA
AP
{DFLT:1.0}
BP
{DFLT:1.0}
AQ
{DFLT:0.1}
BQ
{DFLT:1.0}
imax {DFLT:30000}
INPUT
curr
VAR
ten
power
tau
derc
deri
conduc
arcres
driving
OUTPUT
ten
HISTORY
curr, conduc {dflt:1000}
INIT
CONDUC:=1000
arcres:=recip(conduc)
tau:=AQ*conduc**BQ
power:=AP*conduc**BP
ENDINIT
EXEC
conduc:=curr**2/AP
tau:=AQ*conduc**BQ
power:=AP*conduc**BP
driving:=curr**2/power
DIFFEQ( 1.0|D0 + tau|D1 ) |conduc:=driving
arcres:=recip(conduc)
ten:=arcres*curr
derc:=deriv(conduc)*1.0e-6
deri:=abs(deriv(curr)*1.0e-6)
ENDEXEC
ENDMODEL

You might also like