You are on page 1of 11


IEEE Transactions on Power Delivery, Vol. 9, No. 4, October 1994


G. C. Montanaril, M. Logginil, A. Cavallinil, L. Pittil, D. Zaninelliz Senior Member Member Non Member Non Member Member

1Istituto di Elettrotecnica Industriale, Universith di Bologna 2 Dipartimento di Elettrotecnica, Politecnico di Milano Italy

A A . - This paper presents an arc-furnace model consisting of non-linear, time varying resistance where two different time-variation laws of arc length are considered. One consists of a periodic, sinusoidal law, the other of a band-limited white-noise law. The arc-furnace model is implemented by EMTP, referring to actual electric-plant configurations. Simulations are reported where the values of flicker sensation and short-term flicker severity, P , are determined according to UIE specifications. It results that the model based on sinusoidal time-variation law can be useful for worst-case approximations, while the model using white-noise law is able to fit flicker measurements made in electric plants supplying arc furnaces. The models are used to investigate the effect on flicker compensation of the insertion of series inductors at the supply side of the furnace transformer. It is shown that considerable reduction of Pm is obtained at the point of common coupling by series inductor installation at constant furnace active power. Keywords: Arc Furnace, EMTP simulation, Flicker.

l "
kcfurnaces used for steel production are a main cause of voltage fluctuations in electrical networks, which may give rise to the flicker effect. Voltage fluctuations, due to ramdom arc-length variations during scrap melting, have typical frequencies in the range 0.5-25Hz. Flicker consists of luminosity variations of lamps which may affect the human visual system, depending on their frequency and intensity. For example, voltage-amplitude variations of about 0.3% at frequency 10 Hz are sufficent to get Over the mean human perceptivity threshold [1-31. sincevoltage fluctuations are not limited to electric plants supplying arc furnaces, but may affect the HV network to a large extent, several MV and LV customers can be disturbed by flicker, so that e~ectricity-supp~y companies must take care of this problem. ~ ~ ~ the solutions able to reduce flicker are [4] i) to decrease furnace power, ii) to increase short-circuit power at the point of common coupling (PCC), iii) to install apparatus for flicker compensation. As is obvious, solution i) is not economically valid,
9 4 WM 0 8 6 - 9 PWRD A paper recommended and approved
by the IEEE Transmission and Distribution Committee of the IEEE Power Engineering Society for presentation at the IEEE/PES 1994 Winter Meeting, New York, New York, January 30 - February 3 , 1994. Manuscript submitted June 24, 1993; made available for Printing December 6, 1993.

except for short periods, while ii) depends on network management. Solution iii) is more easily available for the arc-furnace customer and preferable for the electricity supplier. Often, static-var-systems (SVS) are installed in plants feeding arc furnaces in order to compensate both voltage fluctuation and voltage distortion. However, this solution can be quite expensive, depending on the plant size, and, moreover, fluorescent-light flicker have been addressed to SVSnetwork interactions [4]. On-field experiences, as well as computer simulations, have proven the effectiveness of inserting series inductors at the supply side of the furnace transformer [5, 61, but in depth investigations on this topic are still lacking. The choice of the most convenient solution for flicker compensation requires availability of accurate arc-furnace models. This would allow simulating the electric plant supplying the arc furnace and studying the effects of compensation systems on arc and flicker behavior, as well as on voltage and current distortion factor, power factor, furnace active and non-active Dower. At mesent. electric arc is usually modelled by *voltage ggnerators which provide fundamental and harmonic voltages whose amplitudes are time-modulated to describe arc-length variations and, hence, network-voltage fluctuations. The modulation law can be sinusoidal at frequencies typical of flicker [6, 71. However, it has been shown in [81 that this Solution is not fully satisfactory being a linear representation of non-linear phenomena, which is unable to take into account the effect on electric-arc behavior of changes Of electric-plant (due to, insertion of series inductors or shunt filters). Therefore, an improved model was proposed in L81, where the arc is described by a non-linear, time-varying resistance. In this paper, the effect on flicker compensation of the installation of series inductors at the supply side of the furnace transformer is investigated, using EMTP simulation [9]. The electric arc is described by a non~ linear~ resistance ~ l and l two ~ different , time-variation laws (based on sinusoidal and white-noise functions) are considered. Comparisons of the proposed models with measurements made on north Italian plants are reported.

A typical circuit diagram of an electric plant supplying an arc furnace is shown in Fig. 1. The furnace is connected to bus 1, the PCC, by means of a HV/MV tranformer (TI) and is fed by a MV/LV tranformer (T2). The furnace-side of this transformer usually has adjustable voltage in order to vary the furnace power. XLSCis the short-circuit reactance at the p c c , xP the series reactance inserted for flicker compensation (both XLSc and X, are varied in the simulations; the reference value of the shortare circuit power at PCC is 3500 MVA). Xc and reactance and resistance of the connection line between

0885-8977/94/$04.00 0 1994 EEE

furnace electrodes and MVILV transformer, which generally make a significant contribution to the total impedance seen by the arc furnace. The values attributed to the plant parameters for the EMTP simulations presented in this paper are typical of arc-furnace plants installed in northern Italy. The HV/MV and MV/LV transformers (220121 kV and 21/0.9+0.6 kV) have rated power 95 MVA and 60 MVA, respectively, per cent values of short-circuit voltage and losses 12.5% and 0.5% (HV/MV), 10% and 0.5% (MVILV). The values of the lead reactance and resistance are &=3 10-3 Ohm and k = 3 10-4Ohm [7, 101. As mentioned above, simulation of the electric arc is realized by means of a non-linear model. The arc voltagecurrent characteristic,

c c



Fig. 1. Scheme of the reference electrical plant.


v , =



can be described by the following relationship

v ,

= Vat

+ n + I ,

















where V,, I, are arc voltage and current, Vat is the threshold value to which voltage tends when current increases, C and D are constants whose values (C,, D, and c b , Db) determine the difference between the increasing and decreasing-current parts of the v-i characteristic (eq.(2) is written for I,>O, but can be easily arranged for I,<O as well). Figure 2 shows the arc voltage-current characteristic obtained by the model of eq.(2), implemented in the TACS (Transient Analysis Control System) section of EMTP 191, with reference to an arc Cb=39000 W, length gving V,=200 V, C ~ ~ ~ 1 9 0 W, 000 D.=D,=~OOOA r8i. aIn t h Y e event th; (he arc length, 1, would not change with time (1= lo), the arc voltage-current characteristic would be time-invariant. Then, the furnace would not give rise to flicker at the PCC, but only to voltage and current harmonics due to intrinsic non-linearity of the arc characteristic. Simulating, by EMTP, .the electric plant feeding the arc furnace as well as the electric arc (according to eq.(2)), for different values of arc length, the characteristic curves of the arc furnace can be plotted for both single-phase and three-phase plant configurations. The characteristic curves generally consist of the graphs of adimensional quantities S/Ssc, PISsc, P,JSsc, Q/Ssc, PF, plotted as function of I&. S, P and Q are apparent, active and non-active power, respectively [ l l ] , P,, is arc active power, PF is the power factor, Ssc and Isc are the apparent power and current at bus 2 of Fig. 1 when the furnace operates in short-circuit conditions, I is the rated load current. In this paper, single-phase investigations are performed, since the main purpose is to propose arc-furnace models suitable for the study of the effect on flicker compensation of series inductors. However, the range of values of I/Isc where the EMTP simulations are realized assure continuous conduction both in single-phase and three-phase configurations. This would allow referring single-phase results to three-phase evaluations, even if quantitative estimates of voltage fluctuation are affected by the fact that the P/Ssc and Q/Ssc curves are slightly different for single and three-phase configurations (contrarily to adimensional apparent power) [8, 121. Three-phase estimates of voltage fluctuation, based on single-phase models, are also affected by the implicit assumption of balanced load which often is not the actual operating condition.

Fig. 2. Dynamic voltage-current Characteristicof the furnace electric arc obtained by the proposed modellization implemented by EMTP.

Figure 3 shows the characteristic curves of the electric arc fed by the circuit of Fig. 1, with the values of plant parameters previously reported (MV/LV transformer ratio 2110.6 kV). In order to bring the stationary arc-model to give rise to voltage fluctuations, cause of flicker, the voltage-current characteristic must undergo time variations which correspond to a time dependence of the arc length. This can be realized rewriting eq.(l) as [8]
V , = k Vao(Ia)

where V, is the arc voltage corresponding to the reference length, lo. Parameter k is the ratio of the threshold arc voltage corresponding to a length 1, V,,(I), to that relevant to the reference length, Va,(lo) [8, 121. Since the relationship between threshold voltage and arc length can be explained as
Vat = A

+ B


where 1 is the arc length (in centimeters), A is a constant taking into account the sum of anode and cathode voltage drops (A040 V), B represents the voltage drop per unit arc length (B = 10 Vlcm), then k is given by

In the TACS section of EMTP, k has been set as an input variable which allows realizing simulations with arc length either constant or time variable, once appropriate time-variation laws have been singled out.


The fast variations of the current absorbed by an arc furnace during melting time are connected with arc-length variations which are mainly caused by metal-scrap adjustments, electrodynamic forces and arc-electrode variable displacement.


k(t) =



( B D1/2)

(l+sinwt) (8)

1 A + B 1 0






0 . 8


0 . 9


Fig. 3. Characteristic curves of single-phase arc furnace obtained by EMTP simulation, with non-linear arc resistance. The voltage ratio of the furnace transformer is 2U0.6 kV.

The complex nature of these phenomena does not favor a physical approach to the study of arc-length variation. Therefore, flicker investigations have been performed on the basis of deterministic, [6, 81, or stochastic, [13], assumptions for the time-variation law. The deterministic approach considers that the arc length is subjected to sinusoidal time variations, with frequency suitably chosen in the range of those typical of flicker (i.e. 0.5-25 Hz). This assumption does not clearly represent a normal working condition for furnaces, but has the advantage of being easily manageable by computer simulation and require short-time simulations due to the periodicity of the law. The stochastic approach is supported by the observation that the arc-length time-variation can be considered a random phenomenon. In fact, extensive measurements of active and reactive power, voltage and current in plants feeding arc furnaces have shown that voltage fluctuations, as well as reactive power variations, at the PCC and the furnace bus, behave as a band-limited white noise, with time-varying amplitude [13]. Therefore, random-variation laws should be attributed to arc-length. Simulation times are longer than in the previous case, due to the the fact that the signal is not periodic. On the other hand, standards introduce weighted stochastic indices to evaluate flicker severity, such as PsT and PLT, which are based on 10 minutes, or more, of recording time [3]. The above-proposed non-linear model (eqns. (2)-(5)) is able to fit both deterministic and stochastic approaches providing that appropriate laws are attributed to arc length, I=l(t).

Sinusoidal law for time variation

The whole procedure for ATP modellization of the nonlinear arc resistance with time-varying sinusoidal law can be implemented in the TACS section of EMTP, as reported in [8]. Figure 4 shows the voltage and current waveforms at the arc-furnace bus, obtained by EMTP simulation on the basis of the proposed model, with sinusoidal time-variation law at frequency 10 Hz. The arc length varies in a wide range, corresponding to values of arc threshold voltage 40V I V,, I 2 4 0 V and continuous conduction. The furnace transformer has secondary voltage of 600 V. Under these conditions, the relative voltage variation at the PCC, DV/V, is 1.35%. Actually, the UIE flickermeter has an output which is representative of flicker sensation S(t). This quantity is obtained referring the voltage fluctuations to the thresholdperceptivity curve by means of a weighting filter having minimum attenuation at the frequency of 8.8 Hz which is about that of maximum eye sensitivity to light emitted by incandescent lamps [3]. One unit of output corresponds to the visual perceptivity threshold of flicker occurrence. The computer simulations based on the proposed model, with deterministic sinusoidal time-variation load, should conveniently provide flicker estimates which are directly comparable with the values given by the UIE flickermeter. For this purpose, the flickermeter can be implemented by EMTP, as shown in [8], so that the 8.8Hz equivalent voltage variation, DV,,/V, can be obtained. With reference to the simulation providing Fig. 4, the value 1.29% of DV, /V is thus derived (the small difference with respect to ?he value of DV/V above reported is due to the choice of the frequency of the sinusoidal law, which is very close to 8.8 Hz). Figures SA and 5B show the harmonic analysis of the voltage at bus 2 (Fig. l), obtained by simulations with time-varying (frequency of the sinusoidal law 10 Hz) and constant arc length. As can be seen comparing the figures, the non-linearity of the model causes, the presence of characteristic and non-characteristic harmonics (Fig. 5B) [14, 151, while interharmonics (i.e. non-multiple harmonics [161) are generated when the sinusoidal timevariation law is considered (fig. 5A). These side-bands of the multiple harmonics are always expected in the presence of voltage modulation due to arc-length time variation. The interharmonics responsible for the flicker effect are mainly those included in the sidebands of the

In order to approach periodic flicker behavior, simulations can be made attributing to arc length a sinusoidal law with frequency close to the most sensitive for flicker perceptivity. For example, the frequency of 10 Hz can be chosen, which lies in the center of the sensitivity range, close to the minimum of the flicker perceptivity threshold curve for sinusoidal voltage fluctuations [3]. With reference to eqns. (2)-(5), the arc-length timevariation law can be expressed as
l(t) = 1 ,

(D1/2) (l+sinwt)


where D1 is the maximum variation of arc length. The time variation of the arc voltage-current characteristic thus becomes, from (3), (5) and (6)


= k(t) Vao(Ia)


Fig. 4.Voltage and current waveforms at the arc-furnace bus (point 4 of fig. 1) obtained by arc-furnace simulation made by non-linear, time varying resistance.

2029 the flicker sensation, S(t), should be processed in order to




obtain the comprehensive quantity also available at the UIE flickermeter output, the so-called short-term flicker severity, PsT [3]. It consists of a weighted sum of percentiles of the cumulative probability distribution of S(t), having the purpose to provide objective information on the flicker-severity level independently of the type of flicker, its time law and evolution. PsT is thus defined as


= ( 0 . 0 3 1 4 Sgg.g%

0.28 Sgo%

0.08 S 5 0 ~ ) ~ /

0 . 0 5 2 5 Sgg% + 0.0657 S97% (11)













Fig. 5 . Harmonic analysis of the voltage at bus 2 of Fig. 1, for sinusoidal time-varying arc-length (A) and constant arc-length (B).

fundamental frequency, i.e. 50 Hz (on the other hand, the UIE flickermeter, [3] , prevailingly takes into account fundamental-voltage fluctuations).

White-ndse time variation

With reference to the proposed model, the time dependence of the arc-length can be expressed as

where the percentiles 50%, 90%, 97%, 99% and 99.9% of S(t) are considered. PsT estimates are based, according to UIE recommendations, on 10-minute observations (another quantity is also proposed in [3], that is, long-term flicker severity, P,, referred to two-hour observations), but the results here reported are relevant to one-minute simulations for the sake of computing-time saving. Figures 6A and 6B show the time behavior of non-active power (i.e. Q=(S2-P2)0.5 ) and the corresponding flicker sensation, S(t), at the PCC obtained by computer simulation based on the white-noise time variation of arc length. The plant parameters are the same as the previous simulations, pertinent to sinusoidal time law, with arcthreshold voltage, V , , varying in the range 40-240 V, where continuous conduction, as well as wide arc-length variations, are allowed. The value of PsT calculated for one-minute simulation is 1.6 (according to [3], PsT values exceeding 1 indicate flicker disturbance). The results of these simulations roved to reproduce quite-well real cases for both icker and voltage distortion-factor evaluations (considering also the approximations due to single-phase simulation). Indeed, measurements made on a plant with characteristics very close to those used for the simulations provided values of P , approaching 1.6 in the starting melting period. In simulations longer than one minute, the arc-length variation range could be changed with time, such as to describe the voltage-fluctuation decrease that generally occurs increasing the quantity of melted metal.

Model discussion
A direct comparison of the two kinds of arc-length timevariation laws presented above can be performed once the results obtained by the sinusoidal law, that is, DVeq/V, are converted into PsT. Indeed, PsT calculation is meaningless for a deterministic signal and, moreover, shorter times than 1 or 10 minutes are needed to evaluate flicker effcct for periodic signals. However, for the sake of comparison it can be observed that the probability distribution of a deterministic, sinusoidal signal is stepwise ([3]), hence from (11)
pST x
(0.5096 S & ) l I 2

where lo is the maximum arc length compatible with continuous conduction and r(t) is the law of arc-length variation with respect to the reference condition lo. Signal r(t) is a white noise with band in the frequency range where voltage fluctuations produce flicker. The time variation of the arc voltage-current characteristic has again expression (7), where k(t) becomes, for the white-noise time-variation law of eq. (9),


Even in this case, the procedure for EMTP simulation of variable arc length can be implemented in the TACS section resorting to three blocks, that is, a random-number generator, a pass-band filter with lower and upper cut frequencies 4 Hz and 14 Hz (according to [13]), and the third block where eq.(9) provides k(t) as output signal. With respect to the case of periodic time-variation laws (e.g. that sinusoidal previously accomplished), considerably longer simulation times are required for the random arc-length time law here considered. Moreover,

where SMm is the maximum value of flicker sensation. Under these premises, the value of PST derived from the simulations with sinusoidal time-variation law pertinent to Figures 4, 5 (providing DVeq/V= 1.29%) is 3.7. Therefore, the value of PsT obtained by the white-noise time-variation law is significantly lower (1.6) than that derived by the sinusoidal law (3.7), for the same plant and furnace working conditions. It can be argued that the sinusoidal law provides limit conditions for furnace operation, that is, a worst-case approximation, which enables determination of maximum flicker sensation caused at the PCC by a furnace of known













t [SI

Fig. 6. Time behavior of non-active power (A) and the corresponding per-unit flicker sensation (B) at the PCC obtained by computer simulation based on the white-noise time variation of arc length.

characteristics (similar results are expected by modulation with rectangular law). This approximation seems more effective than the criterion used in [12, 131, where DV/V is evaluated referring to the limit cases of not-feeded furnace and short-circuit conditions. Moreover, the proposed model allows to realize investigations on the effects of flicker-compensation systems in worst-case conditions, and provides the resulting flicker evaluations in terms of S(t) or PsT, according to UIE instrumentation. When the white-noise time-variation law is assumed for the arc length, real working conditions can be approached and a sort of average P , estimated for the studied plant. This analysis can be useful to evaluate flickercompensation strategies in the design stage of electric plants supplying arc furnaces or in existing plants where measurements show the need to resort to flickercompensation systems.


The calculations reported up to this point are relevant to secondary voltage of the MV/LV transformer of 600V, absence of the series inductor (i.e. Xp=O in Fig. 1) and short-circuit ratio (SCR) at PCC equal to 58 (the shortcircuit, ratio is defined as the ratio of the short-circuit power to the mean apparent power required by the load). In order to envisage the effects of the installation of series inductors in the plant feeding the furnace, the presence of a series reactance, X,, at the supply-side of the MV/LV furnace transformer can be considered, as shown by Fig. 1 (the equivalent reactor resistance is neglected). However, insertion of series inductors gives

rise to decrease of furnace power so that actions are to avoid significant reductions o f furnace productivity. Mainly, the transformer turns ratio of the MV/LV transformer can be changed, taking profit of the adjustable secondary voltage (varying from 600V to 9OOV, step 60V, in our simulations). This causes arc-length variations, too. Two design criteria are here compared in order to look in detail at the behavior of series inductors. One consists of keeping constant the power absorbed by the system series reactance-transformer-furnace with short-circuited electrodes (so that the SCR at PCC does not significantly change with inductor insertion). The other, which better conforms with the above requisites of furnace efficency, is to keep constant the mean active power absorbed by the furnace. In both cases, simulations with several values of series reactance have been realized, changing the furnaceside voltage of the MV/LV transformer in accordance to the design criteria. As a consequence of the assumed arcfurnace model, an increase of furnace-side voltage, due to insertion of series inductor, causes arc lengthening. According to on-field observations, [17], the maximum arc-length variations, D1, have been assumed independent of arc length in the range of values used for the simulations (i.e. corresponding to continuous conduction). Hence, longer arcs provide smaller variation of relative length D1/1. Both sinusoidal and white-noise laws for arclength time variation have been considered in the simulations (the former with frequency 10Hz). Tables 1 and 2 report the values of the series reactance, X , , inserted in the plant of the characteristics above described, together with the secondary voltage of the furnace transformer, the equivalent voltage variation, DV,,/V, and the corresponding PsT, for the two design criteria. The ratios of the series reactance to the total reactance, X,, as seen upstream the arc-furnace electrodes, are also given. The sinusoidal law of arc-length time variation is assumed. In Figure 7 the graph of voltage variation vs ratio Xp/Xt, relevant to the data of Table 2, is drawn. The simulation results show that by both design criteria the insertion of series inductors at the supply side of the furnace transformer can significantly reduce voltage fluctuations, and, therefore, flicker, at PCC. Taking advantage of the maximum voltage adjustment allowable at the secondary side of the furnace transformer, the voltage variation at PCC decreases to 80% and 60% with respect to the values determined in the absence of series reactance. Hence, significant compensation possibilities are provided by the design criterion of constant mean power absorbed by the furnace (Table 2), but also the other criterion, where the short-circuit power at bus 2 of Fig. 1 is practically kept constant, provides non-negligible flicker compensation. In the former case, however, higher seriesreactance values are required for the same transformer voltage. The dependence of the series-reactance compensation effect on the short-circuit ratio is depicted in Figure 8, referred to the criterion of constant furnace power and sinusoidal modulation. The surface shows that increasing SCR (SCR=58 corresponds to short-circuit power, Ssc, of 3500 MVA), the PsT decreases for any value of series reactance, as expected. In fact, a simple relationship, derived under the approximate assumption that varying the short-circuit power, the furnace current does not considerably change (being XLsc very small with respect to the total reactance of the line feeding the furnace) points out the inverse relationship between voltage variation (or PsT) and short-circuit power:

0.1 0.2





Fig. 7. Per cent voltage variations, DV, lV%, as a function of the ratio between series reactance and tota? reactance, Xp/X,, for a design based on the criterion of constant power absorbed by the system reactance-transformer-furnacewith short-circuited electrodes. Sinusoidal time-variation law.
( DVeq/V ) 2

( DVeq/V) 1 ( Ssca/Sscl)


where (DV,/V), and (DV,,/V)l correspond to Ssa and SscI, respectively. Similar results have been obtained for white-noise timevariation law of the arc length, even if the values of PST are smaller than those reported in Tables 1, 2 (valid for sinusoidal law). For example, in the case of constant mean active power absorbed by the furnace, PsT varies from 1.6, in the absence of series inductor, to 1.28 for xp/x,=3.29. Comparisons of these results with on-field measurements are not easily performable, referring to literature data. Electrical plants feeding arc furnaces generally show similar configurations, but various parameter values and, moreover, flicker measurements are not always reported in comparable conditions as measurement time and unit, melting process, or, otherwise, incomplete information on measurement conditions are given. However, in [17] some results of P, measurements are reported pertinent to plants with parameters similar to those used for our simulations. Indeed, the flicker severity is given by the maximum values of P,, measured for several times in a plant at different operating conditions, exceeded with probability 10% (that is, the 90% cumulative probability of the distribution of the peak PST values). Figures 9 and 10 report the behavior of the maximum PsT associated to exceeding probability 10% as function of the values of series reactance and short-circuit power, respectively. Both Figures show qualitative good agreement with the results of our simulations, summarized in Figs. 7, 8. As regards quantitative evaluations, under the above premises on comparison problems, it can be seen that Fig. 9 shows PsT variations with very close to those reported in Table 2 and Fig. 8. On t e other hand, the distribution of the maximum values of PsT could be reasonably compared with the simulation results relevant to sinusoidal time law, which provides, as previously mentioned, worst-case evaluations. Likewise, Fig. 10 does not remarkably differ from the intersections of the surface of Fig. 8 for Xp/Xt=O and Xp/Xt=0.52 Therefore, it can be concluded that the insertion of series inductors has significant effect on flicker compensation, and that the models here proposed can provide approximate, but meaningful, evaluations of this effect.

-. ...,,...

.... ._ i . . ....... .......... .<......

.... .../


0 0

Table 1. SBries-reactance values (Xp), ratios between series reactance and total feeding-line reactance (xp/xt), secondary voltages of the furnace transformer (vz), voltage variation (DVeq/V) and short-term flicker severity at the PCC (PsT) for design realized at constant power absorbed by the system reactancetransformer-furnace with short-circuited electrodes. Sinusoidal time-variation law of arc length.

II 11 11 1) (1 11

0.63 1.12

1.79 2.04

I O I 0.12 I 0.22 1 0.29 I 0.35 I 0.40


I 1 I


0.60 0.66 0.72 0.78 0.84 0.90


1.29 1.21
1.15 1.10



1.06 1.02


3.68 3.45 3.28 3.14 3.03 2.91

11 11 11 11 11 11
1 1

Table 2. Series-reactance .values (Xp), ratios between series reactance and total feeding-line reactance (Xp/Xt), secondary voltages of the furnace transformer (VZ), voltage variation (DVe,/V) and short-term flicker severity at the PCC (PsT) for design realized at constant mean power absorbed by the furnace. Sinusoidal time-variation law of arc length.

11 11 )I (1


0.98 1.73 2.37 2.86 3.29

1 I I I I

where Af is the amplitude of multiple harmonic voltages or currents, AI is the amplitude at the fundamental frequency (50 or 60Hz), N is normally lower than 50. It was recognized that insertion of protection reactances in electrical plants supplying static power converters generally causes reduction of voltage distortion at the supply bus [18, 191. In the case of arc furnaces, the opposite occurs for both design criteria (as shown by Tables 3, 4). Insertion of series reactances enhances arc length (due to feeding voltage adjustement), so that the current THD (THDI) increases. Consequently, the voltage THD (THD,) at bus 2, and then at the PCC, increases as the series-reactance value rises. However, only a slight increase is detected for the constant furnace-power criterion, which is likely the sought condition, for evident economical advantages (Table 3). It is interesting to observe that current THD does not depend on design criterion, but only on the arc-length variation range, determined by the value of furnace voltage (the same values of THDi are, in fact, obtained for V2=900V in Tables 3 and 4). On the other hand, insertion of capacitors and/or shunt filters can compensate for non-active power due to reactive and distortion powers [7], while they do not provide noticeable conttibution to flicker reduction. Use of filters is promoted to avoid dangerous and uncontrolled resonances that might occur when capacitor banks are used, and, in addition, to contribute to distortion compensation [20]. However, even the use of filters should be carefully regarded due to the almost-continuous harmonic spectra of voltage and current.

0.18 0.30 0.41 0.46 0.52

I I I 1

0.66 0.72 0.78 0.84 0.90


1.12 1.02 0.93 0.87 0.79

I 1 1 I I

3.20 2.91 2.66 2.48 2.26

11 11 11 1 11

The description of flicker behavior in electrical plants supplying arc furnaces by the models proposed in this paper seems quite satisfactory. The non-linear arc model provides voltage and current waveforms as well as arc characteristics which are similar to those observed in actual plants By means of suitable time-variation laws attributed to arc length, quite accurate evaluations have been made on the effects of the insertion of series inductors at the supply side of the furnace transformer, showing that the shortterm flicker severity, that is, the voltage variations cause of flicker, can be significantly reduced. Clearly, a limit value for inductor size is conditioned by arc stability, that is, the range of continuous conduction.


Arc furnaces are well-known harmonic voltage and current sources, owing to their non-linear characteristic. Multiple harmonics are injected in the feeding plant, due to waveform distortion with respect to the sinusoidal frame, besides non-multiple harmonics, or interharmonics, [16], which are caused by arc-length time variation [5, 7, lo]. Figure 5A, relevant to sinusoidal modulation, and Figure 11, pertinent to white-noise time-variation law, show that voltage and current in plants supplying arc furnaces have almost continuous frequency spectrum (the harmonic analysis of Fig. 1 is relevant to the entire simulation of one minute). However, calculation of distortion factor should conveniently separate the effect of interharmonics generation, which is computed by flicker measurements, by the waveform deviation from sinusoidal shape, which is well taken into account by the usual expression for total harmonic distortion (THD) calculation, recommended by IEEE 519 [14], that is

100 I









( 14 )

Fig. 11. Harmonic analysis of the arc-furnace current at the supplyside of the furnace transformer, for white-noise law of arc-length time variation. Xp=3.29 Ohm, transformer secondary voltage 900 V (bandwith for the measurement data 1 Hz).


Therefore, the use of series inductors, associate with capacitors andlor filters for non-active power compensation, seems to be, in certain plant conditions and after technical-economical evaluations, an alternative solution to installation of static var systems. Improvements of modelling accuracy would be achieved working on three-phase simulations, where imbalances of the arc-furnace operations as well as appropriate furnace characteristic curves can be taken into account and proper filter-effectiveness investigation realized.

Table 3. Series-reactance values (Xp), ratios between series reactance and total feeding-line reactance (xp/Xt), secondary voltages of the furnace transformer (Vz), per cent current and voltage total harmonic distortion at bus 2 of Fig. 1 (THDI% and THqr%, respectively) for design realized at constant power absorbed by the system reactance-transformer-furnace with short-circuited electrodes. Upper limit of summation for THD calculation N=ZO. Sinusoidal timevariation law of arc length.
I 1


1 I II


I1 11 11 11 11




I 1I

v2 (kV) 0.60 0.66 0.72 0.78 0.84 0.90





II 11 11


0.63 1.12 1.50 1.79 2.04


0.12 0.22 0.29 0.35 0.40



3.99 4.34 4.72 5.07 5.41 5.72

1.61 1.74 1.85 1.98 2.10 2.21

11 11 11 11

Table 4. Series-reactance values (Xp), ratios between series reactance and total feeding-line reactance (Xp/Xt), secondary voltages of the furnace transformer (V2), per cent current and voltage total harmonic distortion at bus 2 of Fig. 1 (THDI% and THDv%, respectively) for design realized at constant mean power absorbed by the furnace. Upper limit of summation for THD calculation N=20. Sinusoidal time-variation law of arc length.

i i xP
II II I1 11 11 11 11 11
( " )




(kv) 0.60 0.66 0.72 0.78 0.84 0.90

I I I 1 I I I




0.98 1.73 2.37 2.86 3.29

I o I 0.18 I 0.30 1 0.41 I 0.46 I 0.52


3.99 4.34 4.71 5.05 5.39 5.72

1 I

1.61 1.62 1.64 1.68 1.72 1.74

II II 11 )I (1 11 11 1)

J.J. Koenderink, A.J. Van Doorn, "Visibility of unpredictably flickering lights", Journal uf the Optical Soc. ofAmerica, Vol. 64, n. 11, November 1974. J.J. Koenderink, A.J. Van Doorn, "Detectability of power fluctuation of temporal visual noise", Vision &, Vol. 18, pp. 191-195, Pergamon Press, 1978. UIE Disturbances WG, Flicker measurements and evaluation, 1992.

[4] B. Bhargava, "Arc furnace masurements and control", IEEE Trans. on Power Del., Vol. 8, n. 1, pp. 400409, January 1993. [5] L. Bisiach, L. Campestrini, C. Malaguti, "Technical and operational experiences for mitigating interferences from high-capacity arc furnaces", Int. Con5 on, Large H igh-Voltage El. Svs. , CIGRE', Paris, France, September 1992. [6] M. Loggini, G.C. Montanari, L. Pitti, E. Tironi, D. Zanmelli, "The effect of series inductors for flicker reduction in electric power systems supplying arc furnaces" ,IEEEIM Ann. Meeting, Toronto, Canada, October 1993. [7] W.S. Vilcheck, D.A. Gonzalez, "Measurements and simulation-combined for state-of-the-art harmonic analysis" IEEEIIAS Ann. Meeting, pp. 1530-1534, Pittsburgh, USA, October 1988. [SI A. Cavallini, G.C. Montanari, L. Pitti, D. Zaninelli, "ATP simulation for arc-furnace flicker investigation", to be published in ETEP, 1993. [9] ATP Rule Book, Leuven EMTP Center, July 1987. [lo] S.R. Mendis, D.A., Gonzalez, "Harmonic and transient overvoltage analvses in arc furnace power systems", JEEE Trzns. on' Ind. Appl., Vol. 28; n.2, pp. 336-342, April 1992. R. Sasdelli, G.C. Montanari, "The compensable power: its definition for electrical systems in nonsinusoidal conditions", IEEE Trans. on Instr. and Meas., 1993. L. Di Stasi, Electric furnaces (in Italian), Patron ed., Padova, Italy, 1976." G. Manchur, C.C. Erven, "Development of a model for predicting flicker from electric arc furnaces", IEEE Truns. on Power Del., Vol. 7, n.1, pp. 416426, January 1992. IEEE Publ. 519, IEEE recommended practices and reauirements for harmonic contr01 in electric vower syitems , 199i. [15] IEC T C 33 (Secretariat) 148, Guide for a.c. harmonic filters for industrial avvlications , Ohober 1992. [16] A.E. Emanuel, J.A.-Orr, D. Cyganski, "Review of harmonics fundamentals and proposals for a standard terminology", 3rd ICHPS, pp. 1-7, Nashville, USA, September 1988. L. Campestrini, L. Lagostena, G. Sani, A. Bellon, R. Manara, E. Nazarri, "Flicker control in high power arc furnaces and cumulative flicker analysis in HV networks" , Int. Conf on Electricin, Distribution, Liege, Belgium, April 1991. G.C. Montanari, M. Loggini, "Voltage-distortion compensation in electrical plants supplying static power converters", IEEE Trans. on Ind. A & . , Vol. 23, n. 1, pp. 181-188, February 1987. G.C. Montanari, M. Loggini, "Filters and protection reactance for distortion compensation in low-voltage plants", IEEE IAS Annual Meeting, pp. 1488-1494, Pittsburgh, USA, October 1988. D.A. Gonzalez, J.C. McCall, "Design of filters to reduce harmonic distortion in industrial Dower systems", IEEE Trans. on Ind. Appl., Vol. 23,*n. 3, pp. 504-511, June 1987.


Gian Carlo Montanari (M'86, SM'91) was born in Bologna, Italy, on November 8, 1955. He received his Doctor's Degree in Electrical En ineering in 1979 from the University of 0 1 0 a. In 1983, he joined the University opBologna as researcher and has become rofessor of Electrical. technology in 19i6. He has worked since 1979 in the field of agmg and endurance of solid insulating materials and systems. He is also engaged in the fields of harmonic compensation in electrical power s stems, power electronics and statistics. He is an IEE6 Senior member and member of IEC 15B and IEC T C 33. Andrea Cavallini was born in Mirandola, Italy, on December 21, 1963. received his Doctor's De ree Electrical En ineering in 1990 from the University ofgBologna. At present he is Ph.D. student at the Institute of Industrial Electrotechnic o f . the University of Bologna/Italy. His interest fields are power s stems harmonics, reliability of electricar systems and power electronics.

7 ;

i (M'86) was born in Grosseto, Ita-y, on August 29, 1938. He received his Doctor's Degree in Electronic En ineering from the University of B o k g a . In 1970, he joined the same University as an assistant professor of Electrical technolo . At present, he is professor of Ingstrial e!ectrical a plications and works in the field of Karmonic compensation in electrical power systems and power electronics. He is an IEEE member and member of IEC T C 33.

i * was born in Arezzo, Ita1 , on O c ~ ~ ? l 5 , 1963. He receive2 .his Doctor's De ee in Electrical En 'neering in E 9 2 from the University of Sologna. He is present1 cooperating with the University of Boibgna, and is private consultant in electrical plant design, working in Areqo, via Nazario Sauro 32. His interest fields are power s stems harmonics, reliability of eyectrical systems and power electronics.


Dario Zaninelli (M'88) was born in Romano di Lombardia on April 3, 1959. He received. the. Ph.D. de ree in EJectrical En neerin at the Pofitecnico di Milano in B 8 9 anf b e e m e researcher at the Electrical Engineering De artment of the Politecnico d ! Milano on 1690. His areas of research include power system harmonics and power system analysis. He is an IEEE member.


S. Bhattacharya and W. Wong (ABB Transmission Technology Institute): We would like to congratulate the authors on their paper. The paper presents an arc furnace model. The authors use this model to study the effects on voltage flicker by adding a series inductor at the supply side of a fumace transformer. The paper presents two different time-variation laws that are used by the model. It also mentions the modulation with rectangular law and it provides similar results. Could the authors provide more information on the modulation with rectangular law and contrast its effect on the flicker sensation with respect to the white-noise time variation? In the text, the authors comment that the simulation results are difficult to compare with the on-field measurements. However in the conclusion section, the authors mention that the waveforms and the arc characteristics obtained from the model are similar to those observed in actual plants. Would the authors elaborate on the techniques (that overcome the comparison difficulties) they have used to compare the simulation results with the on-field measured waveforms? Would the authors also share a few on-field measured waveforms? Inserting a series inductor between the PCC point and MV/LV transformer reduces the voltage flicker. However, the voltage variation at the arc furnace increases. The authors suggest that the secondary voltage of the MV/LV transformer can be adjusted to compensate for the voltage variation. We foresee that without an on-load tap-changer the voltage regulation will be poor. For instance, if the secondary tap is set for the full-load operation, it will create overvoltage during the light-load condition. Do the authors assume that an on-load tap-changer is employed to adjust the secondary voltage? If so, would they comment on the frequency of operation of the tap-changer and the resulting wear-tear? To obtain an optimum performance and economy, an arc furnace requires a stable and steady voltage supply. Higher steady voltage provides shorter meltdown times, reduces energy cost and extends electrode life. Many installations effectively use static var compensator (SVC) to reduce flicker, to improve power factor and arc furnace efficiencies. The authors comment on the effect of the series inductor concept on the power factor compensation and the arc furnace efficiencies will be appreciated.
Manuscript received February 24, 1994.

and voltage variations can be considered, depending on the amount of data available for the studied plant. But let us answer in detail each question of the discussers. 1-Both sinusoidal and rectangular laws for arc-length modulation are deterministic laws, and should be regarded as worst-case approximations for flicker estimation. The difference in PsT values obtained by the two laws mainly depends on the behavior of demodulation and weighting filters of flickermeter [3]. In fact, on the basis of [3] the following, approximate expressions can be deducted, which relate the equivalent voltage variation, DVe,/V to the flicker sensation S(t):
S(t) = ( 100(DVeq/V)/0.25)2

and S(t) = (lOO(DV,,/V)/O.20)* for sinusoidal and rectangular laws, respectively. Hence, the perceptivity limit (S = 1) is given by DV,/V = 0.3 and 0.2 for sinusoidal and rectangular laws, respectively. Moreover, eq. (12) holds for both laws, since in both cases the probability distribution of the signal S(t) is stepwise. In conclusion, the rectangular modulation provides slightly higher values of S(t) and PsT than the sinusoidal one, so that both laws can be considered worst cases with respect to the white-noise time variation law, which should more closely simulate the actual arc-length variations. 2- As mentioned in the paper, quantitative comparisons of the proposed model with on-field measurements cannot be easily made, due to the single-phase simulation and the difficulties to know in any time the exact fumace working conditions. This led to the use of a stochastic model based on white-noise modulation. By this way, values of PsT close to the average measured during the starting melting periods of an arc furnace were obtained. Clearly, accuracy can increase resorting to three-phase simulations (which is the last achievement of the research) and to more accurate reproduction of the actual working conditions of the arc furnace. As regards the voltage and current waveforms, in steady conditions the model can well reproduce the typical waveforms of an electric arc, often displayed in literature. Considering a plant which supplies an arc furnace, there are so many different operating conditions, involving randow laws, that an accurate reproduction of actual waveforms can be seldom obtained by simulation based on white-noise law. However, the proposed model provides voltage and current waveforms which resemble those observed in plants during periods of arc operation, as it is also confirmed by the values of THD.As an example, Figs. C1 and C2 report the voltage and current waveforms measured and simulated at the bus 2 of the plant, when the series reactance (X,/X, = 0.3) is inserted. Simulation considers white-noise law with time variation close to that detected by the measure. Amounts of plots of active and non-active power could also be provided (see Fig. 6 of the paper), but their apparent fitting the simulations does not give anymore contribution to prove the model validity. 3- Plants where the solution of series inductors for flicker reduction is employed are not uncommon in North-Italy. On the basis of on-field measurements, the observations of the discusser seem appropriate. The capability of secondary-voltage regulation of the MV/LV transformer limits the size of the series inductor, since insertion of the series inductor forces to increase the secondary voltage (if the furnace active power must be kept constant),but reduces consequently,the tap-regulation range. Both simulation and experimental data show that increasing the arc voltage (i.e., lengthening the arc), arc stability decrease, but this can be properly taken into account by electrode control. Hence, the frequency of secondary-voltage adjustments should not significantly vary with respect to operation without series

G. C. Montanari, M. Loggini, A. Cavallini, L. Pitti, and D. Zaninelli: We thank the discussers for their congratulations and the stimulating questions. Before answering to each point raised by the discussers, we would like to point out that the proposed arc furnace model is a step forward in the investigation of flicker and distortion compensation in electrical plants supplying arc furnaces. The authors are well aware that the model does not allow a perfect simulation of furnace behavior during the whole melting cycle, but the comparison of previous models and experimental results lead to consider the proposed model a satisfactory compromise between the need to approach the problem of power qua1ity.h plants supplying arc furnaces and the difficulty to know the actual furnace working conditions at any time. For this reason, the model resorts to eqns. (6)-(9) where the evident simplification of steady power absorbed by the furnace (beyond the voltage fluctuations responsible by flicker) is made. Moreover, simulations are made resorting to the maximum range allowable by the condition of continuous conduction for arc length. However, the implemented model is flexible and, compatible with TACS size, more complex laws for furnace power


inductor. Indeed, it must be clarified that the series inductor is kept steadily inserted in the plant, so that the increase of the rate of commutation of tap changer is related only to the operating conditions of the furnace. In the paper, this .topic was not dealt with in depth, since the purpose was to show how the series inductor works in the plant. 4-Insertion of series inductor affects power factor, PF, due to the contrasting effects of increased line inductance (that lowers PF) and changed average working point of the furnace (the same furnace power is obtained by lower I/I, values and, hence, higher power factor, as shown by Fig. 3 of the paper). On the whole, power factor variations due to series-inductor insertion are relatively small, so that connection of filter or capacitor banks is needed. Three-phase simulations would be appropriate to investigate the influence of filters on THD, PF and PsT. The last development of the authors research, leading to three-phase plant simulation, show that filters reduce voltage THD (in the absence of significant resonance amplification, i.e., for well-designed filters), compensate for PF, but do not improve PST.For example, PsT vanes from 0.85 to 0.88 after filter insertion (tuned to 3rd harmonic), for a plant with xp/x, = 0.4, while voltage THD varies from 1.78 to 0.83. In general, it can be argued that the insertion of filters and series reactance can constitute an interesting solution, alternative and cheaper than SVC, in plants where problems of PSTand THD are not too dramatic, that is, the standard limits for these quantities are exceeded for a limited amount.

-21 11



- no1

Time (sec)



--- nL3

Fig. C1. Voltage and current waveforms measured at the PCC of the plant (B = voltage, A = current).

/ /


~, .. RIW



. L

-~ . . .




Fig. C2. Voltage and current waveforms obtained by simulation with whitenoise law at the PCC of the plant (B = voltage, A = current). Manuscript received April 11,1994.