684

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS

VOL.

PAS-86, NO. 6

JUNE

1967

case of a fault current of a few amperes. It was also found that a noise voltage amounted to a few tens of volts in the case of a fault of a few tens of amperes. There is no appreciable noise caused by the arc, if the fault current is more than 200 amperes. 4) The noise voltage in a flashover of a tree or a suspension insulator string was not much different from that of an arc in the air gap with the same gap length. 5) Data presented will be useful in the design of teletransmission equipment and fault locators, also in understanding the operation of the equipment receiving a signal from a faulty transmission line.

ACKNOWLEDGMENT The results of measurements described in this paper were obtained at field investigations conducted by Tohoku, Tokyo, Chubu, Hokuriku, Kansai, and Kyushu Electric Power Companies in which the authors and their associates participated. The helpful cooperation given by the engineers in those companies is gratefully acknowledged as are the valuable suggestions made by S. Fujitaka, J. Tomiyama, and Y. Hirose. Thanks are also due E. Ashihara and AM. Takanashi for their assistance in carrying out the investigations and to Y. Wada for his help with the English language.

The Design and Testing

of

EHV

Shunt Reactors

MARTIN CHRISTOFFEL,
Abstract-The design and testing of high-voltage shunt reactors is discussed. The advantages of a gapped-core construction employing radial laminated limb sections with a relatively small number of air gaps, resulting in a simple mechanical construction, are described. A comparison is made between testing methods for transformers and reactors, special attention being paid to test power, loss measurement, measuring accuracy, and special tests. A bridge method for loss measurements is described. Among the various dielectric tests discussed, particular mention is made of the induced voltage test and its unrealistic representation of actual service conditions. It is suggested that this test be replaced by one utilizing a lightly damped oscillation produced by an impulse generator. The need for vibration measurements as a check on mechanical behavior is stressed.

SENIOR MEMBER, IEEE

gaps are necessary, which makes it difficult to construct a mechanically strong structure possessing a sufficiently high mechanical resonant frequency. A construction which avoids these difficulties is one in which the limb sections are radially laminated. This allows a relatively small number of large air gaps to be provided, without the danger of causing large eddy-current losses in the laminations. The international standards for testing transformers cannot readily be applied to the testing of high-voltage shunt reactors. Considerable difficulties arise such as, power required from the test plant, measuring accuracy necessary, and the need for special tests. This paper deals INTRODUCTION with the individual tests, and suggestions will be made IN EHV transmission systems operating at approxi- concerning changes which could be made in various tests, mately 400 kV and higher, increasing use is being made commonly used for testing transformers, in order that they of shunt reactors connected directly to the high-voltage may be more suitable for testing reactors. line. [1'] - There are two main types of construction used for such reactors. The first type has no iron core inside the REACTOR DESIGNS winding (coreless type); the second type has an iron core A reactor is an energy storage device which receives with air gaps (gapped-core type). The former differs someand returns it to, the supply at a rate, energy what in physical construction from a transformer and double Wm from of that the frequency supplied. this means that the designer can only use the experience gained in transformer construction to a limited extent. Wm = B2*V On the other hand, if a reactor is built using a core consistingf of parallel laminated limb sections, windings similar to those used in transformer construction are possible. The power rating is given by the relationship With this type of construction, however, many small air Pr = Wm CO Paper 31 TP 66-425, recommended and approved by the Trans- From the preceding equation for Win, it is evident that, formers Committee of the IEEE Power Group for presentation at for energy storage purposes, only nonmagnetic material the IEEE Summer Power Meeting, New Orleans, La., July 10-15, 1966. Manuscript submitted April 11, 1966; made available for where Ar = 1 is considered. For a specified rating, therefore, printing February 2, 1967. V is fixed. There are basically two possible The author is with Brown Boveri and Company, Ltd., Baden, the product B2solutions to this design problem. Switzerland.

CHRISTOFFEL:

EHV SHUNT REACTORS

685

Large Volume and Reduced Flux Density The air-cored reactor, which, of course, is normally fitted with magnetic yokes for the external return flux, corresponds to this solution. In this case the flux density and ampere-turns density are directly dependent upon each other, in accordance with the approximate equation
7-1

In large transformer designs the higher limit for the airgap induction is approximtely 0.2.. .0.25 V. s/m-2. Higher values are difficult to obtain because of eddy-current losses. For an economic air-cored reactor design, however, flux densities in the range 0.6 V s/M-2 are necessary. This leads to coils of large radial, and relatively small axial, dimensions, as well as to a winding arrangement that differs greatly from that used in transformers.
-

(a)

Small Volume and High Flux Density This solution leads to the gapped-core reactor. A corelimb of magnetic material is subdivided by means of a relatively large number of air gaps in which the magnetic energy is concentrated. The advantage of this type of construction lies in the fact that the flux and ampere-turns density vary more or less independently of each other. The following equation is approximately applicable
B
=

By a suitable choice of t/A it is possible to reach high flux densities-the upper limit is fixed solely by the specified linearity requirements of the reactance under overvoltage conditions-simultaneously providing ampere-turns densities and axial winding lengths that are usual in transformer construction. Particularly in the case of high-voltage shunt reactors, this advantage should not be underestimated, because the solution to insulation problems can be obtained from transformer experience. In the usual type of construction, the gapped core consists of a large number of parallel-laminated limb sections. In using parallel laminated core sheets, the individual air gaps must not be made too large, otherwise the stray magnetic flux, which would force its way sideways into the parallel laminations, would give rise to local heating and eddy-current losses.[6] The requirement of many small air gaps, however, is not very easy to fulfil. Spacers placed between the individual sections should possess a high modulus of elasticity in order that the lowest mechanical resonant frequency of the complete reactor lies as high as possible, in any case higher than double the system frequency. A solution which avoids these difficulties is described in the following paragraph.

(b) Fig. 1. (a) Schematic arrangement of a single limb high-voltage reactor core with radial laminated limb sections. (b) Sectional view of one of the limb sections showing the lamination arrangement (t, thickness of lamination; R, radius of limb; n, number of subdivisions).

rangement and center limbs consisting of radial-laminated sheets more than 20 years ago. This type of construction is now very well suited to reactors of large rating.][71-[1 Figure 1 shows the basic construction. All the air gaps are arranged in the central limb and the return magnetic path is completed by means of bracket-shaped ([) yokes buttjointed onto the upper and lower limb sections. A strong central bolt holds the central magnetic core together. As all the limb laminations are arranged radially, it is possible, without introducing difficulties, to make the air gaps between adjacent limb-sections much larger than with parallel laminated sections. Stray magnetic flux can enter a section radially unobstructed and without producing inadmissible eddycurrent losses. Figure 2 shows the magnetic core of a single-phase reactor for 50 mvar, 220/-\'2- kV. Ceramic spacers are used between the limb sections to provide the gaps. As this material possesses a very high modulus of elasticity, it is possible to obtain a lowest mechanical Reactor with Subdivided Limbs Consisting of Radialresonant frequency that is considerablv higher than the Laminated Sections exciting frequency, i.e., double the system frequency. Brown Boveri and Company started building single-phase The danger of vibration damage resulting from resonance shell-type transformers having a concentric winding ar- excitation therefore disappears.

686

IEEE TRANSACTIONS ON POWER APPARATUS AND SYSTEMS

JUNE 1967

Fig. 2. Single limb core: radial-laminated limb sections for highvoltage shunt reactor, 50 MVA, rated voltage 225/AV3 kV.

For very large ratings and high voltages it may be necessary to change to a 2-limb arrangement. Here too, the radial-laminated core-limb type of construction is possible. as Fig. 3 shows. The yokes are arranged horizontally in this case. Because of the precautions that must necessarily be taken to collect stray fluxes, the configuration of these yokes must be studied particularly closely. As a result of the predominating influence of the air gap, the impedance characteristic of the reactors described is linear to a voltage much higher than the rated value. In transmission networks using series condensers it is also occasionally necessary to ask for an impedance characteristic which is substantially linear even at high dynamic overvoltages. These stipulations can be met by an adequate design.[10] The radial-laminated construction can also be used for a 3-phase unit having a magnetic circuit similar to that of Fig. 3. An advantage gained is that the zero sequence reactance is only approximately half that of the positive sequence reactance. In many cases this can have a favorable influence on the power frequency overvoltages occurring during a single-phase ground fault.['2] Figure 4 shows a 165-MVA group of single-phase units for 735 kV. TESTS

The various testing specifications[3] are actually applicable to both transformers and shunt reactors. Nevertheless it has become apparent that these specifications cannot readily be used for large shunt reactors.[] Compared with the testing requirements for power transformers, there are three main differences to be mentioned, namely 1) the difficulty of measuring the losses with satisfactory accuracy 2) the need for special measurements, which on transformers are unusual and also unnecessary 3) the demands made upon available power and voltage sources during heat and induced voltage tests. These difficulties are dealt with in detail in the following three sections.
Fig. 3. Two-limb core: radial-laminated limb sections for highvoltage shunt reactor, 110 MVA, rated voltage 735/A/3 kV.

Fig. 4. A

excited to full rated voltage, full rated power must be supplied by the manufacturer's test plant. In testing highvoltage reactors there is the additional difficulty, because of the absence of a secondary low-voltage winding, of also having to supply this high power at high voltage. The problem can only be solved by means of high capital investment on the part of the manufacturer. [141-[16] In addition to 165-MVA shunt reactor group in service. Rated voltage the need for a larger generator, it is necessary to have a kV.
735

Impedance and Loss Measurements The total losses of a power transformer are determined in two separate tests, open-circuit and short-circuit. For modern power transformers, the test power requirements are less than one percent of rated power for the opencircuit test and approximately 20 percent, as an upper limit, for the short-circuit test. The situation is quite different in the case of reactors; here there is only one operating condition, the open-circuit test, in which the open-circuit current is simultaneously the full rated current. It follows automatically that, if the reactor is to be

CHRISTOFFEL:

EHV SHUNT REACTORS

687

Fig. 5. High-voltage condenser bank (450 kV) for testing large shunt reactors.

Fig. 6. Test circuit for determining impedance difference between two single-phase reactors.
s

TEST REACTOR LOSS-FREE GAS-FILLED CONDENSER

POTENTIAL TRANSFORMER

L,
L___
PRECISION CURRENT TRANSFORMER

P1R
MODIFIED

PROTECTIVE GAP

c
-

SCHERING 4 BRiDGE

SCREENING

Fig. 7. Test circuit for determining losses of shunt reactors using modified Schering bridge.

large condenser bank which must also be insulated for the full rated voltage. Figure 5 shows such an outdoor highvoltage installation suitable for voltages up to at least 450 kV. The use of a high-voltage step-up test transformer is here an advantage, as the condenser bank can then be connected in parallel with the test object and the delicate stability problems associated with series-resonant circuits,

always to a certain degree dangerous at such power ratings, can be avoided. The impedance of reactors can be obtained, using instrument transformers, from simultaneous measurements of both voltage and current up to rated voltage and, if necessary, higher. Accuracies better than one percent are, nevertheless, not possible. Occasionally a smaller tolerance on the difference in impedance between units of a 3-phase group is demanded and then this method is not accurate enough. A better method of determining these differences, and to great accuracy, is to use a bridge-method such as that shown in Fig. 6. The two reactors which are to be compared are connected in series and the voltage across each, reduced by means of a potential transformer, is fed to a normal potential transformer testing bridge. By this means, the impedance difference can be determined with a high degree of accuracy. In most cases this impedance comparison can only be made at reduced voltage; this is nevertheless no great disadvantage, as air-gapped highvoltage reactors exhibit a constant impedance up to a voltage well above the rated value. In measuring losses, difficulties occur, of which the most important concern measurement accuracy. For instrument transformer and wattmeters, the following equation for the relative measurement error applies: 1 AP = *8. 2.91 10-4 P cos 4 Whereas the power factor cos 0 for power transformers on short-circuit test is in the order of 3 X 10-2, the normal value for large shunt reactors is 3 X 10-3, i.e., a phase angle error of only one minute results in an error of ten percent. Under these conditions any corrections made to the measured values are of doubtful importance. Another, fundamentally different, method for determining the losses is the calorimetric method. In Sweden the appropriate measuring methods have been developed to a high degree of perfection. [171 Nevertheless, this test, because of the tine and preparatory work involved, can, at the most, only be carried out as a type test. A disadvantage associated with this method should not go unmentioned. The measurements taken give the losses for only one single voltage value, whereas it is the losses as a function of voltage that are really of interest. At an early date, and for the reasons stated, Brown Boveri searched for a more appropriate and expedient method. The idea of a bridge method suggested itself and the necessary considerations, together with a first practical use, are fully described in the literature.[18] This first solution has, in the meantime, been successfully employed many times. The not verv serious disadvantages encountered, namely those of screening difficulties and the dependence of the bridge balance on the fundamental frequency, have led in the meantime, and in cooperation with a firm specializing in measuring apparatus, to an even better solution. Figure 7 illustrates the general circuit arrangement. The reactor itself, represented by an inductance and a resistance corresponding to the losses, forms one arm of the bridge.

688s

IEEE TRANSACTIONS ON POWER

APPARATUS AND SYSTEMS

JUNE

1967

determined after disconnecting the reactor and using the winding resistance together with the well-known extrapolation method.['3] The correction to 75C results in only a small amount, which therefore has a very small effect on the final error. Dielectric Tests The dielectric testing of high-voltage reactors deserves detailed examination. Because graded insulation is nearly always used, the applied voltage test can only be used to test the insulation of the grounded (neutral) end of the winding. In general, the impulse test can be carried out exactly as for transformers. In fact, it is generally easier to obtain the required tail waveshape than with a transformer of the same size. This is because the transformer windings, not being impulsed, are usuallv loaded with low value resistances or even short-circuited. The result is that the transformer absorbs a higher impulse current than a shunt reactor of the same power and voltage. The induced voltage test(13] on transformers is usually carried out in such a way that the winding for the lowest voltage is excited in the open-circuit condition, either direct or through an intermediate transformer, from a generator whose frequency is higher than the rate reactor frequency. Saturation of the core is avoided by using a frequency higher by a factor between 3 and 7, i.e., a frequenev between 180 and 420 Hz. The resulting load is in most cases relatively small-at the most some thousands of kVA for high-voltage transformers-and capacitive. With highvoltage reactors there is only one possibility available, that of feeding the high-voltage winding directly. The capacitive component of the load in this case is negligibly small compared with the inductive component whose value is given by

Fig. 8. Comparison of loss measurement results by the method shown in Fig. 7 and those determined by the calorimetric method.

For the other high-voltage arm, loss-free compressed-gas condenser of approximately 100 pF, as manufactured by specialist firms for voltages up to 750 kV, is used. The key to the whole circuit lies in the mutual inductance M, which is fed from a precision current transformer. Not until quite recently has it been possible to make a mutual inductance meeting the high requirements demanded, especially the one of not being influenced by external stray fields. The remaining low-voltage arm is formed by a normal Schering-bridge. Whether the elements R4 and C4 are in series (as in Fig. 7) or parallel depends upon practical considerations. Various measures are necessary to ensure reliable measurements. The influence of the stray capacitance and insulation resistance of the connecting cable, between the high-voltage condenser and the bridge, must be eliminated by means of an automatic potential regulating device. Furthermore, a bridge balance is naturally only possible for the fundamental frequency. In itself, the use of a mutual inductance has the disadvantage that higher harmonics which may be present are amplified. The detectorusually an electronic galvanometer must therefore contain filters tuned to the fundamental. Figure 8 shows the loss curve of a single-phase shunt reactor for 110 mvar, 735/-V- kV. As a check, two points determined by the calorimetric method at somewhat more than half the rated voltage are also shown. The diff erences in the results determined by the two methods are very small. Measurement uncertaintv by the bridge method was three percent at the most, a value that is not appreciably improved upon when using the usual methods for shortcircuit measurements on power transformers. Because it is not possible to separate the total losses of a reactor exactly into the core loss and copper loss components, it is also not possible to correct exactly the losses measured at one particular temperature, to that of the guarantee temperature of 75C. For this reason, it pays to measure the losses at least at a temperature near this value. It is appropriate, therefore, to carry out these measurements at the time of the heat test. The average winding temperature at the time of the loss measurement is

Pr. =, u2~j. (fl

and this gives

ft .fr

0.7.. .2.3. Pr This means that even at the frequencies normally used, a test power is required that can be appreciably greater than the reactor rating, dependent on the combination and magnitude of the previous factors. It is actually easy, for the relatively short test-time involved, to compensate for this reactive power by means of a condenser bank, but there are additional difficuilties. The total losses of the complete testcircuit will be some percent of the total test power, so that the test generator must be driven by a very powerful motor, of some thousands of kW. If the parallel resonant circuit is chosen, a suitable high-voltage test transformer

CHRISTOFFEL: EHV SHUNT REACTORS

689

is necessary. On the other hand, if the series resonant circuit is favored, there are stability problems which, as already mentioned, present a certain amount of danger for the test object and equipment at these power exchanges In order that the test-circuit may be regulated with safety, the rating of the test generator should not be less than ten percent of the total test power required. Obviously, all these difficulties can be overcome by sufficient capital investment. But now is perhaps the right time to consider whether the usual transformer induced voltage test is a realistic replica of reactor operating conditions or not. The dynamic overvoltages possible in transmission systems at the highest voltages rarely reach the factor 1.5 and their duration is at the most some seconds. A decisive influence on the possible shape of the switching overvoltage is also exerted by the reactor itself. Almost always, lightly damped oscillations of a few hundreds of cycles, superimposed on the dynamic overvoltage, are involved. These considerations led to the following suggestions. Oscillating overvoltages are comparatively easy to produce using an impulse generator, as is shown in the simplified circuit of Fig. 9. The charging and discharging resistances, which must have high values, are not shown. The series resistance Rd controls the wave front time and also the damping of the oscillations, the frequency of which can be adjusted, within certain limits, by the loading condenser CL. It must also be mentioned that switching surge tests of the type being discussed for transformerstl9]'[20] are not possible, because in every case the relatively low inductance value of the reactor forces a damped oscillatory discharge of the impulse generator. With impulse generators of normal size, oscillations in the range 250-500 Hz are possible. It would perhaps be appropriate to carry out two tests, an induced voltage test and a switching surge test, more or less at the same time. For the first test the oscillations would be only lightly damped, the damping resistance Rd (Fig. 9) being small and the front time 10 Ms or less. In Fig. 10(a) the waveshape obtained during such a test is shown in which the amplitude of the first negative overshoot is -80 percent that of the peak applied voltage. The test level chosen would be that for switching surges. In the second test, by suitably increasing the value of Rd a front time of 100 ,us, as used for switching surge tests, would be obtained. In this way the well-known critical rate of voltage rise, necessary for testing the external insulation, would be met and the negative overshoot would also be very much smaller, as Fig. 10(b) shows. Here, too, the test level would be the same as for switching surges. What should finally be sufficient proof of a satisfactory design with respect to dynamic overvoltages is the fact that, in most specifications anyway, a record of the reactance values at voltages well above the rated value at normal service frequency is required. As the time necessary for these measurements is certainly greater than the duration of the dynamic overvoltages, it would also seem appriate to make the ionization measurements at the same time and under the same conditions.

Fig. 9. Circuit for producing damped oscillation by means of an impulse generator. Cg, impulse generator capacitance; Rd, series damping resistance; CL, total effective load capacitance; L, test reactor.

(a)

(b)
Fig. 10. Waveshapes obtained with circuit of Fig. 9. (a) Light damping, relative steep wavefront (1-10 ,us). (b) Heavier damping, less steep wavefront (O,100us).

Noise and Vibration Measurements Compared with transformers of the same rating, shunt reactors are distinguished by the many times larger exchange of magnetic energy. For this reason the designer is forced to devote much more attention to mechanical problems. As pointed out earlier, it is necessary to place the mechanical resonances in safe regions. A rigid core and coil assembly having a high resonant frequency is therefore desirable. A direct measurement of the mechanical resonant frequency is hardly realizable, which means that it is very important to check the mechanical behavior under service frequency conditions. Checks on the core and coils during the further development of a new design, however, belong

690

IEEE TRANSACTIONS ON POWER

APPARATUS AND SYSTEMS

JUNE

1967

II,

II

II II

II

II II,

tor. The frequency of this discharge would be between 250 and 500 Hz. Satisfactory proof of the ability to withstand dynamic overvoltages would be provided during the impedance measurements under overvoltage conditions. 5) As a check on the mechanical behavior of a highvoltage shunt reactor, it is necessary to carry out measurements of vibration in addition to those of noise level.

--

11i2
42

24~.

46 22 97 68 134 1197

Fig. 11. Representation of the results of vibration measurements on a 55-MVA reactor for 735/V/3 kV. The numbers represent vibration amplitudes (displacement) in jAm.

to development tests. Nevertheless for the final test, vibration measurements on the tank walls must definitely be made. A noise test serves in no way at all as a check on satisfactory mechanical behavior. There are, without doubt, certain relationships between noise level and vibration behavior, but a lower noise level is by no means a safe guarantee of sufficiently low vibration amplitudes. During the tests at full voltage it is practical to scan the tank with crystal accelerometers for places where the largest vibration amplitudes occur. Figure 11 shows the results of such measurements presented in the form used for a test report. There are, of course, no generally applicable limits to be laid down. It is suggested that the mechanical stresses at various points oni the tank wall be calculated on the basis of the vibration amplitudes measured at these points. Experience has shown that a stress of 6 kp cm-2 represents a safe upper limit.

APPENDIX NOMENCLATURE Wm magnetic energy B flux density volume of the magnetic energy storage device V V s volt seconds absolute permeability AoAr permeability of free space (1.25 X 10-1 H.m-') uo F MMF in ampere turns axial length of the winding sum of the air gaps in the core limb AP diff erence between actual and measured losses Cos 4 power factor phase angle error of instrument transformers and

Pr ft fr
u
0,

Pt

kp

kilopond (9.80665 Newton).

wattmeters test power rated power test frequency rated frequency overvoltage factor (ratio of the peak values of test and rated voltage) 2 7rf

CONCLUSIONS 1) Gapped-core reactors are particularly suited for the highest voltages; the reactor with radial laminated limbs is marked by a simple construction possessing a high resonant frequency. 2) Compared with transformers of the same rating, the testing of large high-voltage shunt reactors places specific demands with respect to accuracy, high test power, and the need for special measurements. 3) Wattmeter methods fail for the measurement of losses at rated voltage and frequency. By observing suitable precautions it is possible, using a bridge method, to obtain results whose accuracy is of the same magnitude as that obtained during the measurement of short-circuit losses on power transformers. 4) Induced voltage tests carried out on transformers basically can also be performed on reactors. Test plant requirements for such tests are, however, high. In addition, the test itself has no relation to service conditions. It is therefore suggested that the induced voltage test be replaced by a test in which a lightly damped oscillatory discharge occurs from an impulse generator into the reac-

[1] G. Jancke, R. Jenkins, B. Nordstrom, and L. Norlin, "The choice of shunt-reactors for the Swedish 400 kV system," Proc. CIGRE, Rept. 412, 1962. [2] A. A. Akopjan, U. S. Diskov, S. S. Rokojian, and V. P. Fotin, "Switching over-voltages on long distance 500 kv transmission lines and measures for limiting them," Proc. CIGRE, Rept. 405, 1962. [3] A. Edlinger, H. Glavitsch, and A. Ritter, "The use of high voltage reactors for the compensation of extra high voltage transmission lines," Proc. CIGRE, Rept. 402, 1964. [4] G. Jancke, "Grundsiitzliche Gesichtspunkte bei der Projektirung von Hochspannungs-iubertragungen," Bull SEV, vol. 55, pp. 420-424, May 1964. [51 A. Goldstein, "Quelques problemes de la transmission de 1'energie electrique a tre's haute tension (750 kV)," Bull. Soc. Roy. Belge Elect., vol. 81, no. 2, pp. 125-191, 1965. [6] G. Von Geijer, R. S. Jenkins, B. Sollergren, and R. Myklebust, "Some application, design and testing problems in conjunction with large shunt reactors," Proc. CIGRE, Rept. 118, 1964. [7] A. Meyerhans, "New designs of transformer and choke coils," Brown Boveri Rev., vol. 32, no. 3, pp. 91-100, 1945. [8] ,"Radial laminated transformers and reactors," Brown Boveri Rev., vol. 43, no. 6, pp. 187-205, 1956. [9] ,"A new design for shunt reactors of high rating," Brown Boveri Rev., vol. 45, no. 9, pp. 407-410, 1958. [10] A. Edlinger, "Transformers and shunt reactors for 750 kV," Brown Boveri Rev., vol. 51, no. 1/2, pp. 56-67, 1964. [11] A. Goldstein, "735 kv shunt reactors developed," Elect. World, p. 132, August 23, 1965. [12] M. Christoffel, "Zero sequence reactance of transformers and reactors," Brown Boveri Rev., vol. 52, no. 11/12, pp. 837-842, 1965. [13] "Test code for distribution, power and regulating transformers, and reactors other than current limitinz reactors," ASA Standard C57.12.90, 1958.

CHRISTOFFEL:

EHV SHUNT REACTORS

691

[14]G . W. Alexander, R. H. Hopkinson, and A. U. Welch, "Design shunt reactors," IEEE Trans, on Power and application of EHV Apparatus and Systems, vol. PAS-85, pp. 1247-1258, December 1966. [15] A. Goldstein, "735 kv reactors given extensive testing," Elect. World, pp. 137-138, October 4, 1965. [16] W. Erb and D. J. Kraaij, "Design and testing of reactors for 735 kv," Brown Boveri Rev., vol. 52, no. 11/12, pp. 864-875, 1965. [171 H. de Bourg, R. S. Jenkins, I. Slettenmarle, C. A. Tengstand, and C. E. Webster, "Calorimetric loss measurements on alternators and reactors," Proc. CIGRE, Rept. 119, 1964. [18] F. Deutsch, "Measuring the active power losses of large reactors," Brown Boveri Rev., vol. 47, no. 4, pp. 268-278, 1960. E. Brierley, N. Hylten-Cavallius, and D. H. [19] A. Xlgbrant, A. Ryder, "Switching surge testing of transformers," IEEE

in the coils, and increased heating precautions are taken.

coil vibration, unless special

Trans. on Power Apparatus and Systems, vol. PAS-85, pp. 5461, January 1966. [20] M. Christoffel, "Testing of transformers and reactors with switching surges," Brown Boveri Rev., vol. 52, no. 4/5, 1966.

Discussion
Frank W. Smith (Tennessee Valley Authority, Chattanooga, Tenn.): This paper is timely and is of particular interest to those of us who have been involved in preparing specifications for EHV shunt reactors. I would like to comment with regard to the induced test. As the author of this paper states, the induced voltage test does not truly represent any reasonable system operating condition. Nevertheless, it is our feeling that this is a valuable test which, over the years, has served a good purpose to the industry. The induced test not only subjects the transformer or reactor windings to a thorough dielectric stressing, but holds this stress over a significant )eriod of time. Thus, incipient trouble which might not show iiporl impulse or switching surge tests is very apt to be smoked out during this test. It is my personal opinion that the induced test is the most significant single dielectric test performed on transformers. While we fully appreciate the problems of the manufacturer in makiing this test on shunt reactors, we hope that it will riot be too lightly dismissed in favor of a less severe substitute. Shunt reactors are often continuously connected to the system, and the loss factor is high. It is, therefore, important that the user know the losses associated with this equipment and evaluate them properly. It is encouraging to see the efforts of the author in the direction of accurate measurement of reactor losses.
Manuscript received Jtuly 25, 1966.

The same argument proves the undesirability of using a small number of large gaps, as the resulting large wave-like pattern of transverse flux in the winding has the same unfortunate effects. It is true that flux fringing losses in the steel core blocks are less when all flux enters the sheets edgewise, but there is still considerable excess loss at gaps and this is greatly exaggerated by the use of that fringing at the sharp corner would large gaps. It will be if there were no saturation, actually the result in infinite flux corner saturates early in the cycle, and a wave of saturation moves up and down the edges of the steel sheets. The use of short blocks and gaps more than balance the effect of all-edge-oni punchings and, further, it reduces coil stray losses. Some undue concern with the subject of reactor tank vibration in the literature[7 of some has been generated by the early reactors, in which an elementary preliminary design calculation of magnetic force vs. magnetic shell stiffness would have predicted vibration. Few modern reactors are likely to have this vibration is problem. The important question is whether low enough to permit long life of insulation. Years of experience with large power transformers indicate that sound level is generally the best criterion of safe internal vibration, and that equipment that meets industry sound level specifications is dependable. Sound measurements in effect integrate most of the vibrational energy as it drives the air surrounding the tank. Sound level turns out to be essentially independent of the stiffness of the tank wall. It is true that stiffening may reduce the amplitude of wall vibration, but it does not change sound level or interinal vibration. Stiffening merely alters the vibrational patterns of an otherwise flexible wall; it may or may not reduce tank stress; to the degree that it reduces external vibration measurements, it may be mis-

seen densit.y

mention

unacceptable

internal

leading to the purchaser.

meaningful measurements of ionization or radio noise. A reactor would have to be in sad condition indeed to show corona at the
maximum power frequency voltage.

As to insulation testing, experience on several EHV reactors has proven the stability and ease of control of the series resonance method of induced potential test. It is only necessary to precompute accurately the number of capacitors required to match the reactor and generator reactance. This kind of homework is desirable for any inisulation testing. This is the only test which provides assurance of

A. U. Welch (General Electric Company, Pittsfield, Mass.): As might be inferred from Alexander et al.,['41 (referred to in the paper), we agree that the gapped-core reactor is best suited to the many requirements of EHV shuLnt service; our experience includes the design and manufacture of all the major types. An important virtue of the gapped-core reactor inot mentioned by the author is the opportunity it provides the designer to match accurately the core reluctance distribution with the coil ampere-turn distribution. This is demanded by the same reasons that demand accurate match of ampere-turn distribution in windings of transformers; poor distribution results in large orthogonal magnetic fields, stray losses, and local hot spots. The EHV insulation requirements preclude uniformity of ampere-turn distribution in any type of winding. Steel shell enclosed, coreless reactor coils have uniform reluctance; the consequent mismatch can produce severe local

M. Christoffel: The author is gratef ul for the valuable contributions to the discussions made by Mr. Smith anid Mr. Welch. It is encouraging to hear that the efforts made to measure the losses of shunt reactors with much greater accuracy are appreciated. With regard to the induced voltage test, I am quite in agreement with Mr. Welch that, at some expense, this test is technically possible. I also agree completely with MVIr. Smith that the test is a strenuous one. On the other hand I am, as before, of the opinion that it is Inot a realistic test and believe that in the fourth conclusion a good alternative has been suggested. With reference to the optimum size of the air gaps in core-type reactors, I am not in agreement with Mr. Welch. In the usual form, the of a core-type reactor consists of series core sections, each of which is parallel laminated as in normal transformer cores. With this however, the total air gap must be subdivided into a large number of individual air gaps. As a result of the parallel of the core sheets, the at the edges of the air gaps cuts across the surface of the sheets instead of entering the edges. The local eddy currents thus caused can only be kept within acceptable limits by a sufficient number of small air gaps. Saturation effects in the limb sections and additional stray losses in the windings play, on the contrary, absolutely no part. Our investigations have

limb construction, arrangement

flulx

Manuscript received Jtily 28,

191,56.

eManuscript received September 30, 1966.

692

IEEE TRANSACTIONS ON POWER

APPARATUS

AND SYSTEMS

VOL.

PAS-86, NO. 6

JUNE

1967

shown, and tests carried out have confirmied, that the last mentioned effect alone allows a much greater air gap. With the radial-laminiated type of construction of the central limb, the first type of eddycurrent effect disappears completely, which allows the optimum value for the length of the individual air-gaps to be chosenl with regard to the electrical and magnetic considerations. Concerning the vibrational effects, I am completely in agreement with Mr. Welch that the all-arounid noise level represents a good

yardstick for assessing the vibration of the core and coils. The same result can, however, be obtained more accurately with one or more built-in accelerometers, as is required from time to time in customers' specifications. On the other hand, I believe, even if the vibration of the core and coils is satisfactory, that it is essential to supplement the noise measurement with vibration measurements at various points on the tank wall, because it is not certain that inadmissable vibrations do not occur there.

Performance

of

Line Insulators

Under Rime
SENIOR MEMBER, IEEE

Ice

M. M. KHALIFA,

MEMBER, IEEE, AND

R. M. MORRIS,

Abstract-Single suspension insulator units of several types as well as multi-unit strings were studied in a large refrigerated chamber. Rime ice of various densities was deposited on these insulators. The increase of leakage currents, the reduction of flashover voltages, and the alteration of voltage distributions along insulator strings caused by the presence of the ice were measured. The effects of various protective coatings on the insulator performance were examined. The process of insulator flashover under rime ice is compared with that under the usual types of industrial and marine pollution at temperatures above freezing point.

INTRODUCTION

INSULATORS TESTED The tests were carried out on porcelaiin and glass insulators of the standard cap-and-pin type, on porcelain antifog anid on long-rod insulators, Fig. 1. The lengths of the leakage paths per insulator unit were 113/4, 17, and 96 inches for the standard, antifog, and long-rod insulators, respectively. The long-rod insulator was compared with a string of seven standard porcelain units, both having the same rated voltage of 63.5 kV (110/x/3). Some tests were also made on standard porcelain insulators with annular bands of semi-conducting glaze around the pins.
FORMATION AND PROPERTIES OF RIME ICE Rime ice is known to be formed in nature by the freezing of supercooled cloud or fog-water droplets on objects exposed to the cloudy or foggy atmosphere. It covers the objects by an amorphous coating containing air voids and having a density ranging between 20 and 50 lb/ft3. In the present tests, rime ice could be reproduced in a refrigerated chamber 15 by 25 by 12 feet high, where the

S OME TRANSMISSION line outages which have been experienced in winter seasons in the west coast areas of Canada and the U.S.A. have been caused by rime ice covering line insulators.1] In one instance, a reduction in the operating voltage from 315 to 280 kV was necessary to keep the lines in service under such conditions. Two possible reasons for the reduction of insulation strength under rime ice are examined here. One of these is the increase of the insulator leakage currents and the other is the disturbance of the patterns of voltage distribution along insulator strings. This paper describes a laboratory study of leakage currents, flashover voltages, and other performance characteristics of insulators as influenced by the density and the electrical conductivity of the rime ice. Although the field conditions were not exactly reproduced in the laboratory-because the rating of the transformer and the size of the sprayiing set were limitedessentially the same reduction of insulation strength as that experieinced in the field could be reproduced.

temperature could be closely controlled. After ensuring that the chamber contents were at about 0F, distilled or tap water was sprayed with heated compressed air onito the insulators from a special set of nozzles about 15 feet away. The water drops cooled on their way from the nozzles to the insulator surface where they collected as rime ice. The ice coating was made as uniform as possible by rotating the insulators periodically while sprayinig. The density of the ice formed could be controlled between 25 and 48 lb/ft3 by adjustiilg the temperature of the compressed air. The electrical conductivity of samples of the ice was measured Paper 31 TP 66-98, recommended and approved by the Transmis- and was found to increase with the ice density and temsion and Distribution Committee of the IEEE Power Group for pre- perature, Fig. 2. It was lower for ice made of distilled sentation at the IEEE Winter Power Meeting, New York, N.Y., January 30-February 4, 1966. Manuscript submitted August 16, than of tap water although the difference diminished as 1965; made available for printing November 24, 1965. be attributed to the The authors are with the Division of Radio and Electrical Engi- thawing was approached. This may traces of foreign materials carried by the distilled water neering, National Research Council of Canada, Ottawa, Canada.