Industrial Power Transformers -- Testing transformers [part 1]

Home | Articles | Forum | Glossary | Books


Unlike many items of power system electrical plant (e.g. switchgear and motors) most transformers are still virtually handmade, little or no mass production is employed in manufacture and each is produced very much as a one-off. This means that the user cannot rely on extensive type testing of pre-production prototypes to satisfy himself that the design and manufacture renders the transformer fit for service, but must have such proving as is considered necessary carried out on the transformer itself. From a series of works tests, which might at most be spread over a few days, it is necessary to ascertain that the transformer will be suitable for 30 years or more in service. It is therefore logical that this testing should be complemented by a system of QA procedures which operate on each individual unit and throughout the whole design and manufacturing process.

The final tests, with which this section mainly deals, are checks on all QA procedures carried out throughout the period of manufacture. The stringency and thoroughness of these tests are of vital importance. This section gives a detail description of the various methods employed. To obtain accurate results it is essential that low power factor wattmeters, precision grade ammeters, voltmeters, and class 0.1 (see European Norm (EN) 60044) current and volt age transformers are used. These instruments should be checked at intervals not exceeding 12 months to ensure that the requisite accuracy is maintained.

The above comments might be less true for small distribution transformers where a degree of standardization, automation and mass production technology is tending to appear in some production areas, notably in the manufacture of cores, insulation components and resin encapsulated windings for dry types.

Distribution transformers and dry-type transformers will be considered further in Section 7 and most of the following comments concerning QA and testing refer to larger transformers which are still manufactured by 'conventional' methods.

Details of operation of QA systems are beyond the scope of this volume and are covered adequately elsewhere, for example, by EN ISO 9000 series, Quality management systems, but it must be pointed out that testing alone will not demonstrate that the transformer is fully compliant with all the requirements which may be placed upon it. Many factors which will have a strong bearing on the service life of a large high-voltage transformer are very dependent on attention to detail in the design and manufacture and the need for a high standard of QA and a culture of quality consciousness in the manufacturer's works cannot be emphasized too strongly.

Tests during manufacture

As part of the manufacturer's QA system some testing will of necessity be carried out during manufacture and it is appropriate to consider the most import ant of these in some detail. These are.

Core-plate checks. Incoming core plate is checked for thickness and quality of insulation coating. A sample of the material is cut and built up into a small loop known as an Epstein Square from which a measurement of specific loss is made.

Such a procedure is described in EN 60404-2 Magnetic materials. Methods of measurement of the magnetic properties of electrical sheet steel and strip by means of an Epstein frame. Core-plate insulation resistance should be checked to ensure that the transformer manufacturer's specified values are achieved. EN 10282 gives two alternative methods for carrying out this measurement. The actual method to be used should be agreed between purchaser and supplier.

Core-frame insulation resistance. This is checked by Megger and by application of a 2 kV r.m.s. or 3 kV DC test voltage on completion of erection of the core. These checks are repeated following replacement of the top yoke after fitting the windings. A similar test is applied to any electrostatic shield and across any insulated breaks in the core frames.

Many authorities consider that for large transformers a test of the core and core-frame insulation resistance at 2 kV r.m.s. or 3 kV DC is not sufficiently searching. Modern processing techniques will enable only a very small physical dimension of pressboard to achieve this level under the ideal conditions within the manufacturer's works. The core and the windings supported from it can have a very large mass so that relatively minor shocks suffered during transport can easily lead to damage or dislocation of components so that the small clearances necessary to withstand the test voltage are lost, with the result that core and core-frame insulation which was satisfactory in the factory gives a low insulation resistance reading when tested on site. For this reason some utilities, for example National Grid in the UK, specify increased insulation test requirements for the core/frame/tank for transformers operating at 275 and 400 kV to that appropriate to 3.3 kV class, that is, in the dry state prior to oil filling the test voltage becomes 8 kV r.m.s. and immediately prior to despatch but whilst still oil filled, these tests must be repeated at 16 kV r.m.s.

Core loss measurement. If there are any novel features associated with a core design or if the manufacturer has any other reason to doubt whether the guar anteed core loss will be achieved, then this can be measured by the application of temporary turns to allow the core to be excited at normal flux density before the windings are fitted.

Winding copper checks. If continuously transposed conductor is to be used for any of the windings, strand-to-strand checks of the enamel insulation should be carried out directly the conductor is received in the works.

Tank tests. The first tank of any new design should be checked for stiffness and vacuum withstand capability. For 275 and 400 kV transformers, a vacuum equivalent to 25 mbar absolute pressure should be applied. This need only be held long enough to take the necessary readings and verify that the vacuum is indeed being held, which might take up to 2 hours for a large tank.

After release of the vacuum, the permanent deflection of the tank sides should be measured and should not exceed specified limits, depending on length.

Typically a permanent deflection of up to 13 mm would be considered reason able. Following this test, a further test for the purpose of checking mechanical withstand capability should be carried out. Typically a pressure equivalent to 3 mbar absolute should be applied for 8 hours.

For transformers rated 132 kV and below a more modest vacuum test equivalent to 330 mbar absolute pressure should be applied. The permissible permanent deflections following this test should be similar to those allowed for 275 and 400 kV transformer tanks reduced pro rata for smaller tanks.

Wherever practicable, all tanks should be checked for leak-tightness by filling with a fluid of lower viscosity than transformer oil, usually white spirit, and applying a pressure of 700 mbar, or the normal pressure plus 350 mbar, whichever is the greater, for 24 hours. All welds are painted for this test with a flat white paint which aids detection of any leaks.


Final works tests for a transformer fall into three categories:

(1) Tests to prove that the transformer has been built correctly. These include ratio, polarity, resistance and tapchange operation.

(2) Tests to prove guarantees. These are losses, impedance, temperature rise and noise level.

(3) Tests to prove that the transformer will be satisfactory in service for at least 30 years. The tests in this category are the most important and the most difficult to frame: they include all the dielectric or overvoltage tests and load-current runs.

All the tests in the first two categories can be found in EN 60076.

EN 60076 also describes dielectric tests and load-current runs, so it is largely possible to meet all of the three requirements by testing to this International Standard (EN). However for large important transformers it is desirable to go beyond the requirements of the standard if it is required to gain maximum reassurance in the third category and this aspect will be discussed later.

Firstly however it is appropriate to consider the testing requirements set out in EN 60076.

Testing to the European Norm

Routine tests

All transformers are subjected to the following tests:

(1) Voltage ratio and polarity.

(2) Winding resistance.

(3) Impedance voltage, short-circuit impedance and load loss.

(4) Dielectric tests:

- Separate source AC voltage.

- Induced overvoltage.

- Lightning impulse tests.

(5) No-load losses and current.

(6) On-load tapchangers, where appropriate.

Type tests

Type tests are tests made on a transformer which is representative of other transformers to demonstrate that they comply with specified requirements not covered by routine tests:

(1) Temperature rise test.

(2) Dielectric type tests.

Special tests: Special tests are tests, other than routine or type tests, agreed between manufacturer and purchaser, for example:

(1) Dielectric special tests.

(2) Zero-sequence impedance on three-phase transformers.

(3) Short-circuit test.

(4) Harmonics on the no-load current.

(5) Power taken by fan and oil-pump motors.

(6) Determination of sound levels.

(7) Determination of capacitances between windings and earth, and between windings.

(8) Determination of transient voltage transfer between windings.

(9) Tests intended to be repeated in the field to confirm no damage during shipment, for example frequency response analysis (FRA).

The requirement for type or special tests to be performed, or for any tests to be performed in the presence of the purchaser or his representative, must be determined for particular contacts.

These tests are briefly described for three-phase transformers in the following text. The procedure is generally similar for single-phase units.

Voltage ratio and polarity test

Measurements are made on every transformer to ensure that the turns ratio of the windings, tapping positions and winding connections are correct. The EN tolerance at no load on the principal tapping for a specified pair of main windings is the smaller of either:

(a) _0.5 percent of the declared ratio or (b) a percentage of the declared ratio equal to one-tenth of the actual percent age impedance voltage at rated current.

The voltage ratio for other tappings is to be agreed between manufacturer and purchaser, but shall not be less than the lesser of the values in (a) and (b) above.

For further pairs of windings the tolerance is as for other tappings given above.

These measurements are usually carried out both during assembly of the core and windings, while all the connections are accessible, and finally when the transformer is fully assembled with terminals and tapchanging mechanism.

In order to obtain the required accuracy it is usual to use a ratiometer rather than to energize the transformer from a low-voltage supply and measure the HV and LV voltages.

Ratiometer method

The diagram of connections for this test is shown in FIG. 1. The ratiometer is designed to give a measurement accuracy of 0.1 percent over a ratio range up to 1110:1. The ratiometer is used in a 'bridge' circuit where the voltages of the windings of the transformer under test are balanced against the voltages developed across the fixed and variable resistors of the ratiometer. Adjustment of the calibrated variable resistor until zero deflection is obtained on the galvanometer then gives the ratio to unity of the transformer windings from the ratio of the resistors. This method also confirms the polarity of the windings since a zero reading would not be obtained if one of the winding connections was reversed.

With this type of ratiometer the test can be performed at normal mains sup ply voltage without loss of accuracy, limiting the highest voltage present during the test to the mains supply voltage.

One disadvantage in the use of low-voltage supplies for ratio measurements is that shorted turns in windings with a high number of turns, or windings that have parallel connections, can be very difficult to detect. A method of overcoming this is to supply the LV winding with a voltage which will produce about 1000 V in the HV and then scan the winding with a sensitive flux meter while monitoring the supply current. A shorted turn will then appear as a marked change in the leakage flux without any corresponding change in current. This check must, of course, be carried out before the transformer is installed in the tank.

FIG. 1 No-load voltage ratio test

Polarity of windings and phasor group connections

Polarity and interphase connections may be checked whilst measuring the ratio by the ratiometer method but care must be taken to study the diagram of connections and the phasor diagram for the transformer before connecting up for test. A ratiometer may not always be available and this is usually the case on site so that the polarity must be checked by voltmeter. The primary and secondary windings are connected together at one point as indicated in FIG. 2. An low-voltage three-phase supply is then applied to the HV terminals. Voltage measurements are then taken between various pairs of terminals as indicated in the diagram and the readings obtained should be the phasor sum of the separate voltages of each winding under consideration.

FIG. 2 Diagrams for checking polarity by voltmeter

Load loss test and impedance test

FIG. 3 Copper loss and impedance voltage test: two-wattmeter method

These two tests are carried out simultaneously, and the connections are shown in FIG. 3. The two-wattmeter method is no longer accepted by the European Norm as a means of measuring the load (copper) loss of a three-phase transformer for the reason given below. A description of the method is included for completeness. One instrument is normally used, the connections from which are changed over from any one phase of the transformer to any other by means of a double-pole switch. Closing the double-pole switch on phase A places the ammeter and the current coil of the wattmeter in series with that phase. The wattmeter voltage coil and the voltmeter are connected across phases A and B, both leads from the wattmeter voltage coil being taken direct to the transformer terminals. When the double-pole switch is subsequently closed on phase C, the ammeter and the wattmeter current coil are in series with that phase, and the wattmeter voltage coil and the voltmeter will be connected across phases B and C, the one voltage coil lead being changed from phase A to phase C. Voltage is applied to the HV windings, the LV being short circuited. The links in phases A and C are closed and a low voltage is applied to the HV windings, the initial value being a fraction of the calculated impedance voltage. The double pole switch is then closed and the link on phase A opened. The applied voltage is gradually increased until the ammeter in the HV circuit indicates the normal full-load current when wattmeter, ammeter and voltmeter readings are noted.

The link in phase A is then closed, the double-pole switch changed over, the link in C opened and the wattmeter voltage coil connection changed over from phase A to C. Wattmeter, ammeter and voltmeter readings are again taken.

These readings complete the test and the total copper loss is the algebraic sum of the two wattmeter readings. The impedance voltage is given by the volt meter reading obtained across either phase. The copper loss would be the same if measured on the LV side, but it is more convenient to supply the HV winding. It is important that a copper-loss test should be carried out at the frequency for which the transformer is designed, as the frequency affects the eddy-current copper-loss component, though not affecting the I^2R losses.

The connections given and procedure outlined are exactly the same what ever the interphase connections of the transformer windings. For single-phase transformers the total copper loss is given by a single wattmeter reading only, and similarly for the impedance voltage. In many cases in practice it is necessary to employ instrument transformers whilst conducting the tests described earlier, and in such cases the reference to the changing of current and voltage coils when making wattmeter connections refers to the secondary circuits of any instrument transformers employed in the test.

For power transformers having normal impedance values the flux density in the core during the short-circuit test is very small, and the iron loss may there fore be neglected. The losses as shown by the wattmeter readings may thus be taken as the true copper loss, subject to any instrument corrections that may be necessary. In the case, however, of high-reactance transformers the core loss may be appreciable. In order to determine the true copper loss on such a transformer the power input should be measured under short-circuit conditions and then with the short-circuiting connection removed (i.e. under open-circuit conditions) the core loss should be measured with an applied voltage equal to the measured impedance voltage. This second test will give the iron loss at the impedance voltage, and the true copper loss will be obtained by the difference between these two loss measurements.

When making the copper-loss test it must be remembered that the ohmic resistance of the LV winding may be very small, and therefore the resistance of the short-circuiting links may considerably affect the loss. Care must be taken to see that the cross-sectional area of the short-circuiting links is adequate to carry the test current, and that good contact is obtained at all joints.

To obtain a true measurement it is essential that the voltage coil of the watt meter be connected directly across the HV windings, and the necessary correction made to the instrument reading.

The temperature of the windings at which the test is carried out must be measured accurately and also the test must be completed as quickly as possible to ensure that the winding temperature does not change during the test.

Should several copper-loss and impedance tests be required on a transformer (i.e. on various tappings) then it is advisable to carry out these tests at reduced currents, in no case at less than half the rated current, and correct the results to rated values of current.

A disadvantage of the two-wattmeter method of measurement is that at the low power factors encountered this will produce two large readings, one positive and one negative which when summated algebraically produce a small difference with a relatively large error and this is the reason that the European Norm now specifies that three wattmeters should be used.

The three-wattmeter method can also be adopted for copper-loss measurement with advantage where the test supply is unbalanced. This test is essentially the same as the two-wattmeter method where one winding is short circuited and a three-phase supply is applied to the other winding, but in this case the wattmeter current coil is connected to carry the current in each phase whilst the voltage coil is connected across the terminals of that phase and neutral. The sum of the three readings taken on each phase successively is the total copper loss of the transformer. During this test the current in each phase can be corrected to the required value before noting wattmeter readings. On large transformers where the impedance of the transformer causes a low power factor it is essential that wattmeters designed for such duty are employed.

The copper loss and impedance are normally guaranteed at 75ºC but in fact both are normally measured at test room temperature and the results obtained corrected to 75ºC on the assumption that the direct load loss (I^2R) varies with temperature as the variation in resistance, and the stray load loss varies with the temperature inversely as the variation in resistance.

The tolerance allowed by EN 60076 on impedance was reduced when the document was issued in 1997. This probably reflects the greater accuracy with which modern computer programs are able to determine leakage flux pat terns. This is now _7.5 percent of the declared value for a two-winding transformer when the impedance value is 10 percent. When the impedance value is 10 percent the tolerance is _10 percent. These tolerances apply on principal tapping for two-winding transformers and for a specified pair of main windings of multi-winding transformers. On other tappings the tolerances are _10 percent when the impedance value is 10 percent and _15 percent when the impedance value is 10 percent. The copper loss at 75ºC and the iron loss are each individually subject to a tolerance of _15 percent but iron plus copper losses in total must not exceed _10 percent of the guaranteed value.

The test connections for a three-phase, interconnected-star earthing transformer are shown in FIG. 4. The single-phase current I in the supply lines is equal to the earth fault current and the current in each phase winding is one-third of the line current. Under these loading conditions the wattmeter indicates the total copper loss in the earthing transformer windings at this particular current while the voltmeter gives the impedance voltage from line to neutral. The copper loss measured in this test occurs only under system earth fault conditions. Normally earthing transformers have a short-time rating (i.e. for 30 seconds) and it may be necessary to conduct the test at a reduced value of current, and to omit the measurement of the copper loss, thus testing impedance only.

FIG. 4 Copper loss and impedance voltage test for a three-phase interconnected-star neutral earthing transformer.

At the same time as the copper loss is being measured on the three-phase interconnected-star earthing transformer, the zero-phase sequence impedance Z0 and resistance R0 can be obtained as follows:

Z V I 0 3 per phase (ohms) _ where I is the current in the neutral during the test and

R I 0 2 3 per phase (ohms) power (watts) _ _

All other tests on earthing transformers are carried out in the same way as for power transformers.

Insulation resistance test

Insulation resistance tests are carried out on all windings, core and any core clamping bolts. The standard Megger testing equipment is used, the 'line' terminal of which is connected to the winding or core bolt under test. When making the test on the windings, so long as the phases are connected together, either by the neutral lead in the case of the star connection or the interphase connections in the case of the delta, it is only necessary to make one connection between the Megger and the windings. The HV and LV windings are, of course, tested separately, and in either case the procedure is identical. In the case of core bolts, should there be any, each bolt is tested separately.

Should it be required to determine exactly the insulation resistance of each separate winding to earth or between each separate winding, then the guard of the Megger should be used. For example, to measure the insulation resistance of the HV winding to earth the line terminal of the Megger is connected to one of the HV terminals, the earth terminal to the transformer tank and the guard terminal to the LV winding. By connecting the windings and the instrument in this way any leak age current from the HV winding to the LV windings is not included in the instrument reading and thus a true measurement of the HV insulation to earth is obtained.

Resistance of windings

The DC resistances of both HV and LV windings can be measured simply by the voltmeter/ammeter method, and this information provides the data necessary to permit the separation of I^2R and eddy-current losses in the windings.

This is necessary in order that transformer performances may be calculated at any specified temperature.

The voltmeter/ammeter method is not entirely satisfactory and a more accurate method such as measurement with the Wheatstone or Kelvin double bridge should be employed. It is essential that the temperature of the windings is accurately measured, remembering that at test room ambient temperature the temperature at the top of the winding can differ from the temperature at the bottom of the winding. Care also must be taken to ensure that the direct current circulating in the windings has settled down before measurements are made. In some cases this may take several minutes depending upon the winding inductance unless series swamping resistors are employed. If resistance of the winding is required ultimately for temperature rise purposes then the 'settling down' time when measuring the cold winding resistance should be noted and again employed when measuring hot resistances taken at the end of the load test.

Iron-loss test and no-load current test

FIG. 5 Iron-loss and no-load current test: two-wattmeter method.

These two tests are also carried out simultaneously and the connections are shown in FIG. 5. This diagram is similar to FIG. 3, except that in FIG. 5 voltage is applied to the LV windings with the HV open circuited, and one wattmeter voltage coil lead is connected to the transformer side of the current coil. The two-wattmeter method is adopted in precisely the same way as described for the copper-loss test, the double-pole switch being first closed on phase A. The rated LV voltage at the specified frequency (both of which have previously been adjusted to the correct values) is first applied to the LV windings, and then readjusted if necessary, the links being closed in phases A and C. The double-pole switch is then closed, the link opened in phase A and wattmeter, ammeter and voltmeter readings are noted. The wattmeter is then changed over to phase C, and one voltmeter connection changed from phase A to phase C. Wattmeter, ammeter and voltmeter readings are again noted. These readings complete the test, and the total iron loss is the algebraic sum of the two wattmeter readings. The no-load current is given by the ammeter reading obtained in each phase. The iron loss would be the same if measured on the HV side, but the application of voltage to the LV winding is more convenient.

The no-load current would, however, be different, and when checking a test certificate, note should be taken of the winding on which this test has been carried out. The same comments apply with regard to accuracy of the two wattmeter method as made in relation to load loss measurements except that the power factor for no-load loss is not quite so low as for load loss. Many manufacturers would thus prefer to use three wattmeters.

The connections given and procedure outlined are exactly the same what ever the interphase connections of the transformer windings. For single-phase transformers the iron loss is obtained simply by one wattmeter reading.

For all transformers except those having low-voltage primary and secondary windings this test is conducted with the transformer in its tank immersed in oil.

If the LV voltage is in excess of 1000 V, instrument transformers will be required and the remark made earlier equally applies.

In making this test it is generally advisable to supply to the LV winding for two reasons: firstly, the LV voltage is more easily obtained, and secondly, the no-load current is sufficiently large for convenient reading.

The supply voltage can be varied either by varying the excitation of the alternator or by using an induction regulator. A variable resistor in series with the transformer winding should not be used for voltage adjustment because of the effect upon the voltage wave shape and the transformer iron loss.

The iron loss will be the same if measured on either winding, but the value of the no-load current will be in inverse proportion to the ratio of the turns.

This no-load loss actually comprises the iron loss including stray losses due to the exciting current, the dielectric loss and the I 2 R loss due to the exciting current.

In practice the loss due to the resistance of the windings may be neglected.

It is sometimes more convenient to measure the iron loss by the three wattmeter method, particularly when the LV voltages are of a high order. In all cases low power factor wattmeters must be used.

If only one wattmeter is available a possible method of connection is shown in FIG. 6.

The test is conducted as follows:

The double-pole switch a_ is closed and the link opened, switches b_ and c_ being open with their corresponding links closed. The voltmeter switch is put on to the contact a_. The supply voltage is adjusted until the voltmeter reads the correct phase voltage. The frequency being adjusted to the correct value, ammeter, voltmeter and wattmeter readings are taken. The link on the double pole switch a_ is then closed and the switch opened. Switch b_ is closed and the corresponding link opened, while the voltmeter switch is moved to con tact b_. Any slight adjustment of the voltage that may be necessary should be made and the meter readings again noted. This operation is again repeated for phase C.

The algebraic sum of the three wattmeter readings will then give the total iron loss. The EN 60076 tolerance on iron loss is _15 percent but the combined iron loss plus copper loss must not exceed _10 percent of the declared value; the tolerance on no-load current is _30 percent of the declared value.

It is essential that the supply voltage waveform is approximately sinusoidal and that the test is carried out at the rated frequency of the transformer under test.

For normal transformers, except three-phase transformers without a delta connected winding, the voltage should be set by an instrument actuated by the mean value of the voltage wave between lines but scaled to read the r.m.s. value of the sinusoidal wave having the same mean value.

FIG. 6 Iron-loss and no-load current test: three-wattmeter method.

For three-phase transformers without a delta-connected winding the no-load losses should be measured at an r.m.s. voltage indicated by a normal instrument actuated by the r.m.s. value of the voltage wave, and the waveform of the supply voltage between lines should not contain more than 5 percent as a sum of the fifth and seventh harmonics.

FIG. 7 Graphical method of separating hysteresis and eddy-current losses.

In all cases when testing iron losses, the rating of the alternator must be considerably in excess of the input to the transformer under test.

In the routine testing of transformers it is not necessary to separate the components of hysteresis and eddy-current loss of the magnetic circuit, but for investigational purposes or for any iron-loss correction, which may be necessary on account of non-sinusoidal applied voltage, such procedure may be required. The losses may be separated graphically or by calculation, making use of test results at various frequencies. Generally, loss tests at a minimum of three frequencies, say, at 25, 50 and 60 Hz, are sufficient for the purpose.

All the tests are carried out in the standard manner already indicated, and at a constant flux density, the value of the latter usually being that corresponding to the normal excitation condition of the transformer. The two methods are then as follows:

(1) Graphical

This method is illustrated in FIG. 7. The measured losses are converted into total energy loss per cycle by dividing the total power by the frequency, and the results are then plotted against the respective frequencies. The resulting graph should be a straight line, intercepting the vertical axis as shown. The ratio of the ordinate value at the vertical axis (i.e. at zero frequency) to the ordinate value at any other frequency gives the ratio of hysteresis loss per cycle to total measured iron loss per cycle at that frequency, and the hysteresis loss per cycle in watts can then be determined.

(2) By calculation

The hysteresis loss component varies directly with frequency, while the eddy current loss component varies with the square of the frequency. Having measured the total iron loss at two frequencies, adjusting the applied voltage to maintain constant flux density, the loss component can be separated by substitution of the total loss values into simultaneous equations derived from the relationship given in Eq. (eqn. 1).

Pf = fPh + f 2 Pe (eqn. 1)

where Ph and Pe are the hysteresis and eddy-current losses respectively. If the iron-loss tests have been made at 25 and 50 Hz and the total iron losses are Pf25 and Pf50 respectively, then, from Eq. (eqn. 1),

Pf50 = 50Ph + 2500Pe (eqn. 2)

Pf25 = 25Ph + 625Pe (eqn. 3)

Multiplying Eq. (eqn. 3) by 2 and subtracting the result from Eq. (eqn. 2) eliminates Ph and so enables Pe to be determined, as then

(eqn. 4)

Substitution of the value of Pe in Eq (eqn. 2) or (eqn. 3) then enables Ph to be deter mined at the relevant frequency.

Dielectric tests: windings

The insulation of the HV and LV windings of all transformers is tested before leaving the factory. These tests consist of:

(a) induced overvoltage withstand test, (b) separate source voltage withstand test, (c) lightning impulse withstand tests when required, (d) switching impulse withstand test when required.

Viewed simplistically the induced overvoltage test tests the insulation between turns, and between windings and terminals of different phases; the separate source voltage withstand test tests the insulation between the complete windings of different voltage systems, and to earth, and lightning and switching impulse tests represent simulations of conditions which might stress the insulation in a different way than the stress occurring under steady state power frequency conditions. Certainly when they were first introduced a good many years ago, this was the reason for their introduction. It is however, also the case that, in most cases, lightning impulse tests can prove to be very searching tests of the insulation between turns and, for series reactors, lightning impulse tests represent the only means of testing interturn insulation.

Following the philosophy set out in the previous paragraph, when high-voltage transformers were first designed and manufactured, insulation test levels were arbitrarily set at twice normal volts. This represented a convenient factor of safety over rated conditions and ensured that equipment in service was never likely to be stressed to a level approaching that to which it had been tested.

This simplistic approach falls down however when transformers are considered that have non-uniform insulation as described in Section 4.3. With this system, the neutral ends of star-connected HV windings that are intended to be connected solidly to earth have a very much reduced separate source test level.

For example, the HV winding of a 132 kV transformer having uniform insulation might be tested at a separate source voltage level of 230 kV. The same voltage class of transformer with non-uniform insulation could have a separate source test voltage of only 38 kV. Clearly, this test level is entirely inadequate for the major insulation, that is the insulation near to the line ends of the 132 kV windings. In this situation, therefore, the induced overvoltage test must become the means of testing this insulation between line ends and earth as well as between the line ends of the separate phases.

The use of non-uniform insulation therefore causes the situation to become very much more complicated. If it is required to test the insulation at the line end at 230 kV to earth and the neutral end of the winding must be solidly connected to earth, this means that during the induced overvoltage test 230 kV must be developed across an HV phase winding. Even if this test voltage is induced in the winding when connected on, say, a _10 percent tapping, this still represents a volts per turn of

FIG. 8 Phase-by-phase test on a three-phase transformer

FIG. 9 Phase-by-phase test on a three-phase transformer

Furthermore, if the induced overvoltage test were to be done as a three-phase test, this would represent a line voltage during the test of 230_/3 =398 kV.

For many years, in the UK, non-uniform insulation has been the norm. The induced overvoltage test has traditionally been performed single phase by means of an arrangement as shown in FIG. 8, which, for the transformer in the example, raises the line terminal to 230 kV with the neutral earthed. This requires that the appropriate test voltage must be applied to one phase of the LV winding. If this winding is connected in delta, which is generally the case, the same voltage is automatically applied to the other two phases connected in series with the opposite polarity. The non-tested HV phases therefore develop voltages of 115 kV below earth with respect to the phase under test, so that adjacent HV phase terminals are subjected to a test voltage of (230 +115) = 345 kV, which is greater than twice normal but still less than the voltage that would occur if the test were done three phase. This high voltage between phases led to difficulties when the UK tried to agree a common test method with other parties to the IEC and various methods of carrying out the single phase test were developed which enabled the voltage to be limited to a some what lower value, for example, by connecting the two non-tested HV line terminals to earth as shown in FIG. 9. This arrangement, however, leads to a voltage equivalent to one-third of the test voltage appearing between the HV neutral and earth, that is 76.7 kV for a 132 kV transformer, and this is greater than the test voltage to earth specified for the neutral. Arrangements of this type were described in the now superseded IEC 76-3:1980.

A better compromise has been reached for the 2001 issue of IEC 60076-3 which has become the European Norm. Induced overvoltage testing carried out in accordance with this document results in lower test voltages between phases than specified hitherto in the UK, but supporters of these new procedures would argue that the induced overvoltage testing has in most instances been made more effective in identifying shortcomings within the insulation by the incorporation of more sophisticated techniques, namely partial discharge measurement, as described below. For transformers with non-uniform insulation it has become necessary to carry out the short-duration AC withstand test as a two part test to ensure that all sections of the insulation are tested to an adequate standard whilst ensuring that the voltages between phases are never excessive.

Unfortunately, the other effect of compromise has been to produce a set of testing requirements which are complex and not readily summarized. The following paragraphs will nevertheless make an attempt at a summary but for the precise requirements the reader is referred to the document itself.

To each winding of a transformer is assigned a value of highest voltage for equipment, given the symbol Um. The dielectric test withstand voltages, both induced and separate source, for a winding is determined by Um. If this requirement leads to conflict of test values for different windings within a transformer, the value determined for the winding with the highest Um applies.

For windings with non-uniform insulation different values of Um are assigned to the line and neutral terminals. This has the effect of making the separate source test voltage for a winding equal to the test voltage appropriate to that for the lower, neutral end, value of Um. The value of Um for the line terminal of a winding determines the requirements for transient overvoltage testing of the winding and also determines how the induced overvoltage test should be performed. For Um up to 72.5 kV the measurement of partial discharge during the induced overvoltage test is not considered necessary and lightning impulse testing is recommended only as a type test. The 'standard' induced overvoltage test is termed the 'short duration' (ACSD) test, and a longer induced overvoltage test termed the 'long duration' (ACLD) test is available as a more searching option for the higher-voltage classes.

The relationship between Um and the recommended dielectric test voltages generally applicable for Europe is set out in Table 1. In North America test voltages differ from those used in Europe for Um 170 kV and the applicable values are given in Table 2.

Carrying out the dielectric tests

The exact procedure for carrying out the induced overvoltage test varies according to the class of transformer as determined by its value of Um. For those classes of uniformly insulated three-phase transformers for which a short duration test (ACSD) is specified, the test is done three phase, and serves as a phase-to-phase test. Phase-to-earth withstand is tested by applying a separate source test at the same level. The simplest option for the induced overvoltage test is that applicable to transformers for which Um 72.5 kV. For this class of transformer, the impulse test is recommended only as a type test.

For 72.5 Um 170 kV, for which the option of uniform or non-uniform insulation exists, the impulse test becomes a routine test. As the rated voltage increases to 170 Um 300 kV, for both uniform and non-uniform insulation, a long-duration induced overvoltage test is recommended rather than the short-duration test, to enable a more exacting test of partial discharge to be carried out, but the short-duration test can be retained, in which case a switching impulse test must be added to the range of dielectric tests to be carried out.

Finally for Um 300 kV both long-duration induced overvoltage and switching impulse tests are recommended.

Table 1 Rated withstand voltages for transformer windings with highest voltage for equipment Um 170 kV - Series I based on European practice.

Note: Dotted lines may require additional phase-to-phase withstand tests to prove that the required phase-to-phase withstand voltages are met.

Table 2 Rated withstand voltages for transformer windings with highest voltage for equipment Um 169 kV - Series II based on North American practice

Partial discharge measurement

It is appropriate at this point to examine the process of partial discharge measurement in more detail. Partial discharge measurement aims to identify weakness or 'incipient breakdown,' which might indicate that defects exist within the insulation structure that could result in unacceptable performance in service.

The disadvantage of reliance on this approach, as seen by the 'traditionalists,' is that although the induced overvoltage test is the principal means of testing the insulation at the line end for transformers with non-uniform insulation; including that between line lead and tank, core and core frame, as well as between HV and LV windings, HV and taps and across the tapping range; it is also primarily designed as the means of testing insulation between turns. Since, even in quite large transformers, the volts per turn is rarely more than 200, and on many occasions considerably less, then under induced overvoltage conditions the voltage between turns will still be quite modest so that reducing this further is counter productive. It has also to be recognized that under factory test conditions the insulation should be in a far better state and the oil more highly 'polished' than it is ever likely to be again during service, hence it is to be expected that its electrical withstand strength will be greatly superior than it is ever likely to be again.

Considering more closely the nature of partial discharges, a partial discharge is an electrical discharge that only partially bridges the insulation between conductors. Such a discharge is generally considered to take place as a precursor to total insulation failure but may exist for a long period of time, possibly years, before total breakdown occurs. In some circumstances the existence of the discharge will modify the stress distribution so as to initially reduce the tendency to total breakdown. In time, however, total breakdown will always result, often because the discharge itself leads to chemical breakdown of the insulation which reduces its electrical strength. Clearly, in a healthy transformer under normal operating conditions the only acceptable level of partial discharge is nil. 'Normal operating conditions' means any non-fault condition which is likely to occur in operation, for example, system overvoltages which may occur following a reduction in system load until corrected by tapchanger operation or operator intervention where necessary. It should be noted also that since many electrical systems frequently experience continuous overvoltages of up to 10 percent there should be no partial discharge present with this level of overvoltage.

Detection of partial discharge relies on the fact that in a transformer, these cause transient changes of voltage to earth at every available winding terminal.

The actual charge transferred at the location of a partial discharge cannot be measured directly. The preferred measure of the intensity of a partial discharge is the apparent charge 'q' as defined in IEC 60270:2000. The specified provisional acceptance values of apparent charge (the actual values are detailed in the descriptions, below, of the tests themselves) are based on practical partial discharge measurements made on transformers which have passed traditional power-frequency dielectric tests.

The measuring equipment is connected to the terminals by matched coaxial cables. The measuring impedance in its simplest form is the matching impedance of the cable, which may, in turn, be the input impedance of the measuring instrument. The signal-to-noise ratio of the complete measuring system may be improved by the use of tuned circuits, pulse transformers and amplifiers between the test terminals and the cable. The circuit must present a fairly constant impedance to the test terminals over the frequency range used for the partial discharge measurements.

When measuring partial discharge between the line terminal of a winding and the earthed tank a measuring impedance Zm is connected between the bushing tapping and the earthed flange. Calibration of the measuring circuit is carried out by injecting a series of known charges at the calibration terminals from a calibration signal generator.

FIG. 10(a) shows a measurement and calibration circuit of this type where the calibration generator consists of a pulse generator and a series capacitor C0 of approximately 50 pF. Where the calibration terminals present a capacitance much greater than C0 the injected charge will be:

q0 _ U0 _ C0 where U0 is the voltage step.

FIG. 10(b) illustrates an arrangement where a bushing tapping is not available and the measuring impedance, with protective spark gap, is connected to the LV terminal of a partial-discharge-free HV coupling capacitor C, whose value is large compared with C0. There are two types of measuring instrument in use: (a) narrow band and (b) wide band.

FIG. 10 Partial discharge calibration and measurement circuits. (a) Using a condenser bushing capacitance tap and (b) using a high-voltage coupling capacitor.

Precautions must be taken to eliminate interference from radio broadcast stations, spurious partial discharges from other sources in the surrounding area, the power supply source and the terminal bushings. These include the fitting of electrostatic shielding on the outside of the transformer and oscillographic monitoring of the test. If a transformer exhibits unacceptable partial discharge levels then, because visible traces of partial discharge are not usually found, attempts must be made to identify the source without removing the transformer from its tank. It may be useful to consider the following possibilities:

(a) Partial discharge in the insulation system may be caused by insufficient drying or oil impregnation. Reprocessing or a period of rest, followed by repetition of the test may therefore be effective.

(b) A particular partial discharge gives rise to different values of apparent charge at different terminals of the transformer and the comparison of simultaneous indications at different terminals may give information about the location of the partial discharge source.

(c) Acoustic or ultrasonic detection of the physical location of the source within the tank.

The reader is referred to EN 60076-3 for additional information.

Use of partial discharge measurement as a diagnostic tool

Partial discharge is likely to occur at points in the insulation structure where the stress level is high. Assuming that the designer of the transformer has been careful to ensure that such high stress points to not occur as a feature of his design, that is he has been careful to ensure that there are no stress raisers at winding ends or where leads pass close to core frames or edges of tank pockets, then possible locations for discharge could be where oil impregnation has not been fully effective, possibly in bulky lead insulation, or maybe a physical defect in a laminated-wood cleat bar. Partial discharge requires that the stress in such locations reaches a certain critical level; it also requires time to become established. It is only harmful if, once established due to the application of overvoltage, it does not become extinguished when the over voltage is removed. Use of partial discharge measurement as a diagnostic tool relies, therefore, on detection and measurement of partial discharge levels as the induced overvoltage test is applied; the levels at which the discharge inception occurs as voltage is increased; the levels at which it is extinguished as the test voltage is reduced, and whether there is any tendency for the latter to lag the former.

Induced overvoltage withstand test

Because of the options of uniform and non-uniform insulation, as identified above, as well as variations in the detail of the test in accordance with the volt age class of the transformer, the user of EN 60076-3 must select the correct option to establish the precise requirements for the test.

The HV windings are left open circuit, the test voltage being applied to the LV windings. The test voltage may be measured on the LV side of the transformer under test, either directly or using a voltage transformer, or the peak value of the voltage induced in the HV winding can be measured using an electrostatic voltmeter or a suitable voltage divider.

The connections given and procedure outlined for the voltage tests are exactly the same for single-phase and three-phase transformers whatever the interphase connections of the windings. The tests are carried out with the transformer assembled as for service except for the optional fitting of cooling and supervisory equipment.

During the test the supply frequency is increased, usually to at least twice the rated frequency, to avoid over-fluxing the core. Care must be taken to ensure that excessive voltages do not occur across the windings. Any winding not having non-uniform insulation may be earthed at any convenient point during the test. Windings having non-uniform insulation should be earthed at a point that will ensure the required test voltage appearing between each line terminal and earth, the test being repeated under other earthing conditions when necessary to ensure the application of the specified test voltage to all parts of the winding.

The test should be commenced at a voltage not greater than one-third of the test value and increased to the test value as rapidly as is consistent with measurement. At the end of the test the voltage should be reduced rapidly to less than one-third of the test value before switching off.

The duration of the test should be 60 seconds at any frequency up to and including twice the rated frequency. When the frequency exceeds twice the rated frequency the duration of the test should be equal to:

120 x [rated frequency / test frequency] seconds but not less than 15 seconds

The test levels and methods must be agreed at the time of placing a contract.

Simplest tested are transformers for which Um < 72.5 kV and having uniform insulation, which is usually the case for all transformers of this voltage class. For these transformers the standard does not recommend the use of partial discharge measurement. The test, therefore, has no hold points and consists simply of taking the transformer to the test voltage as rapidly as possible, holding this for the required time and then reducing the voltage as rapidly as possible.

For Um > 72.5 kV partial discharge measurements are to be made and the procedure is more complex.

For non-uniformly insulated transformers two tests must be carried out, one with the rated withstand voltages applied between phase and earth, and the other with the rated withstand voltage applied between phases. Partial discharge must be measured for each test.

FIG. 11 Time sequence for the application of test voltage with respect to earth.

The time sequence for application of the test voltage, whether this is a test to earth or a test between phases is as shown in FIG. 11. The test voltage is to be switched on at a level not greater than one-third of U2, raised to 1.1 Um_/3 and held there for a duration of 5 minutes. This is to demonstrate that the unit is discharge free at the highest voltage that is likely to be experienced in service.

At the end of this 5 minute period the voltage is raised to U2 the partial discharge evaluation level, and held there for a further 5 minute period after which the voltage is raised to the specified test level and held there for the required test duration. Immediately after the test time the voltage is reduced again with out interruption to U2, held there for a duration of at least 5 minutes for further measurement of partial discharge, then reduced to 1.1Um/_3 and held there for a further period of 5 minutes in order to demonstrate that the unit remains discharge free at its likely highest service voltage. Following this final check the unit is reduced to a value of below one-third of U2 before switching off.

The partial discharge evaluation level, U2, is 1.3 Um_/3 for a phase-to-earth test and 1.3Um for a phase-to-phase test.

The test is successful if:

• No collapse of the test voltage occurs.

• The continuous level of apparent charge at voltage U2 during the second 5 minute period does not exceed 300 pC on all measuring terminals.

• The partial discharge behavior does not show a continuous rising tendency.

• The continuous level of apparent charge does not exceed 100 pC at the 1.1 Um_/3 voltage level.

The standard suggests that a failure to meet the partial discharge criteria should lead to consultation between purchaser and supplier with a view to establishing a strategy for further investigations. As indicated in the description of partial discharge measurement as a diagnostic tool above, any unusual or unexplained features of the partial discharge observed during the test should be investigated further. For example, if the test has been performed as a single phase test, any significant differences in partial discharge between phases, even if the absolute magnitude is within the acceptance limits, should be investigated and explained.

FIG. 12 Time sequence for the application of test voltage for induced ACLD tests.

One possible way to carry out further investigation is to perform an AC long duration test (ACLD). The sequence for this test is shown in FIG. 12. The procedure is generally as for the short-duration test except that the hold period immediately following the application of the test voltage may be 60 minutes, for Um 300 kV, or 30 minutes for lower-voltage classes, with continuous monitoring of partial discharge during these extended periods. The criteria for evaluating the partial discharge are similar to those for the short-duration test but the increased duration of the application of enhanced voltage will increase the likelihood of detecting incipient failure or conversely, should the partial discharge remain unchanged, confidence in the integrity of the insulation will be increased.

Separate source voltage withstand test

The terminal ends of the winding under test are connected to one HV terminal of the testing transformer, the other terminal being earthed. All the other winding ends, core, frame and tank are earthed. FIG. 13 shows the connections for testing the HV windings of a transformer.

The test should be commenced at a voltage not greater than one-third of the test value and increased to the test value as rapidly as is consistent with measurement. At the end of the test the voltage should be reduced rapidly to less than one-third of the test value before switching off.

The full test voltage is applied for 60 seconds, the peak value being measured and this divided by _2 must be equal to the test value. In the case of transformers having considerable electrostatic capacitance, the peak value of the test voltage is determined by means of an electrostatic voltmeter or a suit able voltage divider.

The value of test voltage to be applied depends on a number of factors which include whether the transformer windings are (i) air or oil insulated and (ii) uniformly or non-uniformly insulated.

The test voltage applied to dry-type transformers for use at altitudes between 1000 and 3000 m above sea level, but tested at normal altitudes, is to be increased by 6.25 percent for each 500 m by which the working altitude exceeds 1000 m.

This does not apply to sealed dry-type or oil-immersed transformers, but it may be necessary to select a bushing designed for a higher insulation level than that of the windings.

Lightning impulse testing of transformers

Lightning impulse test levels

Lightning impulse voltage test levels have been chosen after many years' study of surges on supply systems. These levels are based on uniform and non-uniform insulation. Impulse voltage withstand test levels for transformers have been standardized in EN 60076-3 and values appropriate to the highest system volt ages are given in Tables 1 and 2.

FIG. 13 Voltage tests: separate source test

Transformers to be impulse tested are completely erected with all fittings in position, including the bushings, so that, in addition to applying the surge volt age to the windings, the test is applied simultaneously to all ancillary equipment such as tapchangers, etc., together with a test on clearances between bushings and to earth.

Impulse voltage wave shapes

A double exponential wave of the form ? _ V(eat _ eßt ) is used for laboratory impulse tests. This wave shape is further defined by the nominal duration of the wavefront and the total time to half value of the tail, both times being given in microseconds and measured from the start of the wave. British and IEC Standards for impulse testing are under revision with the ultimate objective of a EN setting out all the requirements. In the interim these are covered by a combination of IEC 60060 and BS 923. These define the standard wave shape as being 1.2/50 µs and give the methods by which the duration of the front and tail can be obtained. The nominal wavefront is 1.25 times the time interval between points on the wave front at 10 and 90 percent of the peak voltage; a straight line drawn through the same two points cuts the time axis (? _ 0) at O1 the nominal start of the wave. The time to half value of the wave tail is the total time taken for the impulse voltage to rise to peak value and fall to half peak value, measured from the start as previously defined. The tolerances allowed on these values are _30 percent on the wavefront, and _20 percent on the wave tail. A typical wave shape, the method of measuring it and the tolerance allowed are shown in FIG. 14.

FIG. 14 Standard impulse voltage wave shape: 1.2/50 µs.

Nominal wavefront O1X1 _ 1.2 µs, tolerance _30%. Nominal wave tail O1X2 _ 50 µs, tolerance _20%

Another waveform used in transformer impulse testing is the 'chopped wave' which simulates an incoming surge chopped by a flashover of the co-ordination gaps close to the transformer. During this test a triggered-type chopping gap with adjustable timing is used, although a rod gap is permit ted, to produce a chopping of the voltage after 2-6 µs. An impulse chopped on the tail is a special test and when made it is combined with full-wave tests.

The peak value of the chopped impulse is nowadays specified to be 1.1 times that for a full-wave impulse, however before the introduction of triggered chopping gaps which can be relied upon to operate within the required tolerances it was customary to specify that the chopped-wave tests, which relied on the operation of rod gaps to provide the chopping, should be carried out using a wave having 1.15 times the full-wave peak value and some authorities have continued to specify this level. It is also recommended that the over swing to opposite polarity following the chop should be limited to 30 percent of the amplitude of the chopped impulse, and the insertion of impedance in the chopped circuit is permitted to enable this to be achieved.

The clearances of the electrodes from floor, walls and earthed metal in all directions must be adequate. A chopped-wave shape is also shown in FIG. 14 and can be compared with the 1.2/50 µs wave shape.

next >>

Top of Page

PREV.   NEXT   Guide Index HOME