Industrial Power Transformers -- Testing transformers [part 2]

Home | Articles | Forum | Glossary | Books


Impulse generators

The production of voltage impulses is achieved by the discharge of a capacitor or number of capacitors into a wave forming circuit and the voltage impulse so produced is applied to the object under test. For conducting high-voltage impulse tests a multistage generator as shown in FIG. 15, a modified version of Marx's original circuit, is now used. This consists of a number of capacitors initially charged in parallel and discharged in series by the sequential ? ring of the interstage spark gaps.

A simple single-stage impulse generator is shown diagrammatically in FIG. 16. The generator consists of a capacitor C which is charged by direct current and discharged through a sphere gap G. A resistor Rc limits the charging current whilst the resistors Rt and Rf control the wave shape of the surge voltage produced by the generator. The output voltage of the generator can be increased by adding more stages and frequently up to 20 stages are employed for this purpose. Additional stages are shown in FIG. 17 and as will be seen from this diagram all stages are so arranged that the capacitors C1, C2, C3, etc. are charged in parallel. When the stage voltage reaches the required level V the first gap G1 discharges and the voltage V is momentarily applied to one electrode of the capacitor C2. The other electrode of C2 is immediately raised to 2V and the second gap G2 discharges. This process is repeated throughout all stages of the generator and if there are n stages the resultant voltage appearing at the output terminal is nV. This output is the surge voltage which is applied to the test object.

FIG. 15 Impulse generator having an open-circuit test voltage of 3.6 MV and stored energy of 100 kWs. Each of the 18 stages has an output of 200 kV (Peebles Transformers)

FIG. 16 Single-stage impulse generator

Impulse voltage measurement

There are a number of devices available for the measurement of impulse volt age, the two most common methods being as follows.

The first is the sphere gap. Details of this method and the required gap settings are given in IEC 60052:2002. This method has the disadvantages of requiring a large number of voltage applications to obtain the 50 percent flashover value and of giving no indication of the shape of the voltage wave.

FIG. 17 Impulse generator: The second method of measurement requires a voltage divider and a high speed oscilloscope. These oscilloscopes use sealed-off tubes with accelerating voltages of 10-25 kV or continuously evacuated tubes with accelerating voltages of up to 60 kV. Besides giving the amplitude of the voltage wave, the oscilloscope can also be used to provide a photographic record from which the wave front time and the time to half value on the tail of the wave can be determined.

The full peak voltage cannot be applied directly to the deflecting plates of the oscilloscope as the input voltage to these instruments is usually limited to 1 or 2 kV. The necessary reduction in voltage is obtained by means of the voltage divider. The ratio of the divider can be determined accurately and hence by suitable calibration and measurement at the low-voltage tapping point, the amplitude of the impulse voltage can be ascertained.

Impulse tests on transformers

The withstand impulse voltages to be applied to a transformer under test are specified in EN 60076-3 and the test voltages are required to be applied in the following order:

(1) One calibration shot at between 50 and 75 percent of the standard insulation level.

(2) Three full-wave shots at the standard level.

The application of voltages 1 and 2 comprises a standard impulse-type test and they are applied successively to each line terminal of the transformer. If during any application, flashover of a bushing gap occurs, that particular application shall be disregarded and repeated.

Where chopped waves are specified, the test sequence is as follows:

(a) One reduced full wave, at 50-75 percent of the test level.

(b) One full wave at the test level.

Table 3 Standard rod gap spacing for critical flashover on 1.2/50 _s wave

(c) One or more reduced chopped waves.

(d) Two chopped waves at the test level.

(e) Two full waves at the test level.

For oil-immersed transformers the test voltage is normally of negative polarity since this reduces the risk of erratic external flashover.

The time interval between successive applications of voltage should be as short as possible.

These tests employ the 1.2/50 µs wave shape and the chopped waves can be obtained by setting the gap in parallel with the transformer under test. Values of rod gap setting are given in Table 3.

The rod gap spacings given in Table 3 are for standard atmospheric conditions, that is:

barometric pressure (p) 760 mm temperature (t) 20ºC humidity 11 g of water vapor per m3 (11g/m3 x 65 percent relative humidity at 20ºC).

For other atmospheric conditions a correction should be made to the rod gap spacing as follows. The spacing should be corrected in an inverse proportion to the relative air density d, at the test room where

d p t

_ _ 0 386 273

The gap spacing should be increased by 1.0 percent for each 1 g/m3 that the humidity is below the standard value and vice versa.

In some cases when testing large transformers, particularly those having comparatively few winding turns, the impedance may be so low that the standard wave shape of 1.2/50 µs cannot be obtained from the impulse generator even with a number of stages connected in parallel. It is permissible in such cases for a shorter wave shape than the standard to be agreed between the purchaser and the manufacturer. When the LV winding cannot be subjected to lightning overvoltages, by agreement between the manufacturer and the purchaser this winding may be impulse tested with surges transferred from the HV winding. Alternatively the non-tested terminals may be earthed through resistors but the value should not exceed 500 Ohm. Voltage oscillograms are recorded for all shots and, in addition, as part of the fault detection technique, oscillographic records can be taken of one or more of the following:

(a) The current flowing in the earthed end of the winding under test.

(b) The total current flowing to earth through a shunt connected between the tank insulated from earth and the earthing system.

(c) The transformed voltage appearing across another winding.

These records are additional to those obtained of the applied surge voltage and the method adopted from either (a), (b) or (c) is chosen by the transformer manufacturer in agreement with the purchaser according to which is the most appropriate and effective for the particular transformer under test.

During an impulse test the transformer tank is earthed, either directly or through a shunt which may be used for current measurement. The winding under test has one terminal connected to the impulse generator whilst the other end is connected to earth. In the case of star-connected windings having no neutral point brought out to a separate terminal, or in the case of delta-connected windings, it is usual to connect the two remaining terminals together and earth via a measuring shunt unless otherwise agreed between the manufacturer and the purchaser. It is essential that all line terminals and windings not being tested shall also be earthed directly or through a suitable resistance in order to limit the voltage to not more than 75 percent of the rated lightning impulse withstand voltage.

Where arcing gaps are fitted to bushings they should be set to the maximum permissible gap in order to prevent flashover during testing.

The general arrangement of the various pieces of equipment employed for an impulse test on a transformer is shown diagrammatically in FIG. 18.

FIG. 18 General arrangement of equipment for an impulse test (diagrammatic only)

FIG. 19 Connections used for fault detection when impulse testing transformers. Note: Terminals of windings not under test shall be earthed either directly or through resistors. Each phase should be damped by a suitable resistor.

Fault detection during impulse tests

Detection of a breakdown in the major insulation of a transformer usually presents no problem as comparison of the voltage oscillograms with that obtained during the calibration shot at reduced voltage level gives clear indication of this type of breakdown. The principal indications are as follows:

(1) Any change of wave shape as shown by comparison with the full-wave voltage oscillograms taken before and after the chopped-wave shots.

(2) Any difference in the chopped-wave voltage oscillograms, up to the time of chopping, by comparison with the full-wave oscillograms.

(3) The presence of a chopped wave in the oscillogram of any application of voltage for which no external flashover was observed.

A breakdown between turns or between sections of a coil is, however, not always readily detected by examination of the voltage oscillograms and it is to facilitate the detection of this type of fault that current or other oscillograms are recorded. A comparison can then be made of the current oscillograms obtained from the full-wave shots and the calibrating oscillograms obtained at reduced voltage.

The differential method of recording neutral current is occasionally used and may be sensitive to single turn faults. All neutral current detection methods lose sensitivity when short-circuited windings are magnetically coupled to the winding being tested. Connections for this and other typical methods of fault detection are shown in Figs 19(a)-19(e).

In all cases the current flows to earth through a non-inductive shunt resistor or resistor/capacitor combination and the voltage appearing across this impedance is applied to the deflection plates of an oscilloscope.

Another indication is the detection of any audible noise within the transformer tank at the instant of applying an impulse voltage. This has given rise to a completely different method of fault detection known as the electro acoustic probe, which records pressure vibrations caused by discharges in the oil when a fault occurs. The mechanical vibration set up in the oil is detected by a microphone suspended below the oil surface. The electrical oscillation produced by the microphone is amplified and applied to an oscilloscope, from which a photographic record is obtained. Alternatively acoustic devices may be attached to the external surfaces of the tank to detect these discharges.

Fault location

The location of the fault after an indication of breakdown is often a long and tedious procedure which may involve the complete dismantling of the transformer and even then an interturn or interlayer fault may escape detection.

Any indication of the approximate position in the winding of the breakdown will help to reduce the time spent in locating the fault.

Current oscillograms may give an indication of this position by a burst of high-frequency oscillations or a divergence from the 'no-fault' wave shape.

Since the speed of propagation of the wave through a winding is about 150 m/µs, the time interval between the entry of the wave into the winding and the fault indication can be used to obtain the approximate position of the fault, provided the breakdown has occurred before a reflection from the end of the winding has taken place. The location of faults by examination of current oscillograms is much facilitated by recording the traces against a number of different time bases. Distortion of the voltage oscillogram may also help in the location of a fault but it generally requires a large fault current to distort the voltage wave and the breakdown is then usually obvious.

FIG. 20 illustrates a typical set of voltage and neutral current oscillograms associated with an impulse withstand test, and FIG. 21 those obtained with increasing impulse voltage levels up to breakdown, which is clearly shown in FIG. 21(f).

A wave of negative polarity and having a wave shape of 1.06/48 µs was employed for all tests. The voltage calibration corresponds to 107.4 kV and the time corresponds to 10 µs on oscillograms (a), (b), (d) and (e), and to 1 µs on (c).

FIG. 20

FIG. 21

Switching impulse test

Surges generated by lightning strikes have very steep rise times which cause transformer windings to appear as a string of distributed capacitance rather than the inductance which is presented to a power-frequency voltage. Surges generated by system switching do not have such rapid rise times -- times of 20 µs are typical -- and at this frequency the transformer winding behaves much as it would do at 50 Hz. The voltage is evenly distributed, flux is established in the core and voltages are induced in other windings in proportion to the turns ratio. The magnitude of switching surges, though generally lower than lightning surges, is considerably greater than the normal system voltage (perhaps 1.5 times or twice), so that the overpotential test is not an adequate test for this condition. Switching surge tests are therefore carried out on all transformers which might be subjected to switching surges in service. The test is a routine test for windings rated at 300 kV and above.

The impulses are applied either directly from the impulse voltage source to a line terminal of the winding under test, or to a LV winding so that the test voltage is inductively transferred to the winding under test. The specified test voltage must appear between phase and neutral and the neutral is to be earthed.

In a three-phase transformer the voltage developed between phases during the test is normally 1.5 times the voltage between phase and neutral. The test volt age is normally of negative polarity because this reduces the risk of external flashover in the test circuit.

The voltages developed across different windings of the transformer are approximately proportional to their effective number of turns, and the maximum voltage will be determined by the winding with the highest-voltage rating.

FIG. 22 Typical switching impulse wave shape.

The voltage impulse shall have a virtual front time of at least 100 µs, a time above 90 percent of the specified amplitude of at least 200 µs, and a total duration to the first zero of at least 500 µs. FIG. 22 shows a typical switching impulse wave shape.

The front time is selected by the manufacturer in agreement with the purchaser so that the voltage distribution along the winding under test will be essentially uniform. Its value is usually less than 250 µs. During the test, considerable flux is developed in the magnetic circuit. The impulse voltage can be sustained up to the instant when the core reaches saturation and the magnetizing impedance of the transformer becomes considerably reduced. The maximum possible impulse duration can be increased by introducing remanence of opposite polarity before each full voltage test impulse. This is accomplished by applying lower-voltage impulses of similar shape but of opposite polarity or by temporary connection to a DC source of supply.

The test sequence consists of one calibration impulse at a voltage level between 50 and 75 percent of the full test voltage, and three subsequent impulses at full voltage. Oscillograph records are taken of at least the impulse wave shape on the line terminal being tested. If the oscillographic recording should fail that application is disregarded and a further application made.

During the test the transformer must be on no load and this presents sufficient impedance; windings not being tested are earthed at one point but not short circuited. The test is successful if there is no collapse of the voltage as indicated by the oscillograms but it should be noted that due to the influence of magnetic saturation successive oscillograms may differ in wave shape.

Digital data collection systems

With the increasing use of computers in all areas of technology at the present time, it must be inevitable that these should be applied to the gathering and processing of transformer impulse testing data. Accordingly manufacturers of high-speed oscilloscopes which have been almost exclusively used hitherto as the means of recording of impulse waves have in recent years turned their attention to the production of software enabling voltage and current signals to be digitized in such a manner as to enable them to be recorded, analyzed and printed out by computer. Some such systems have been in use by some transformer manufacturers since the mid-1980s. Many transformer engineers, however, have been cautious in their acceptance of this new technology. Because of the very rapid rates of change involved in transformer impulse waves it is necessary to utilize exceedingly high sampling rates in order to accurately represent them other wise there is a danger that some high-frequency elements might be significantly distorted or even lost entirely. It is possible for software to record a voltage wave and compute the front and tail times, but if the wave shape departs from the ideal depicted in FIG. 14 by being 'peaky,' for example, as shown in FIG. 23, then the software will arrive at very different front and tail times than an operator who would use his judgment in taking measurements from oscilloscope records. On the credit side, the digital software can be made to perform comparisons between test impulses and the reference record so as to provide a plot of difference versus time, but even when performing this function start time and sampling discrepancies can lead to differences being identified which do not exist.

FIG. 23 'Peaky' impulse voltage record

Low-voltage surge tests

The insulation of a transformer must be proportioned to the surge voltages which will appear at the various points throughout the windings. High-voltage surge tests on a completed transformer are costly and take a great deal of time.

In addition, these are pass or fail tests and they do not give indication of margins and failures can be expensive. In order to obtain the maximum possible amount of information it is desirable to have electrical contact with the maximum number of points on the winding.

Furthermore, for high-voltage transformers the core and windings must be immersed in oil and mounted in the tank. This condition does not facilitate the collection of data. Tests have shown that the surge voltage distribution in a winding is independent of the magnitude of the applied voltage and that the same results may be obtained by applying a reduced surge voltage, of the order of a few hundred volts.

These tests are made with a recurrent surge generator which consists of a capacitor charged to a suitable voltage and discharged by means of a thyratron into a circuit which is designed to generate the required low-voltage surge of the standard wave shape. The charge and discharge sequence is repeated at such a rate as will allow the effect of each applied surge to have totally decayed before application of the subsequent one. Fifty times per second is usually found to be convenient. The output voltage from the recurrent surge generator is applied to the terminal of the transformer winding under investigation, in a similar manner to that in which a high-voltage surge test would be conducted, whilst the surge voltage appearing at any point of the winding can be measured and displayed on the screen of an oscilloscope.

The time base is arranged so that it is synchronized with the recurrent discharge of the capacitor. By this means it is possible to obtain a standing picture on the screen of the applied voltage and of the voltage appearing at points along the winding, together with a time calibration wave which can be viewed directly by the operator or photographed for permanent record and later analysis.

In order to increase the usefulness of the recurrent surge oscilloscope for development and research investigations, facilities to vary the wavefront and wave tail, to produce chopped waves, and to give variable time sweeps and timing waves, are incorporated in the equipment.

Temperature rise test: oil-immersed transformers When a test for temperature rise is specified it is necessary to measure the temperature rise of the oil and the windings at continuous full load, and the various methods of conducting this test are as follows:

(a) short-circuit equivalent test,

(b) back-to-back test,

(c) delta/delta test,

(d) open-circuit test.

The temperature rise limits are valid for all tappings; except in special cases, the temperature rise test needs be carried out on only one tapping.

Method (a)

The general procedure under this method is as follows.

One winding of the transformer is short circuited and a voltage applied to the other winding of such a value that the power input is equal to the total normal full-load losses of the transformer at the temperature corresponding to continuous full load. Hence it is necessary first of all to measure the iron and copper losses as described earlier in this section. As these measurements are generally taken with the transformer at ambient temperature, the next step is to calculate the value of the copper loss at the temperature corresponding to continuous full load.

Assuming the copper loss has been measured at 15ºC, the copper loss at the continuous full-load temperature will be equal to the measured copper loss increased by a percentage equal to 0.4 times the anticipated temperature rise.

This calculation assumes the copper-loss varies directly as the resistance of the windings. This is not quite true, however, since a portion of the copper loss consists of eddy-current loss, and this portion will decrease as the resistance of the windings increases. The inaccuracy is slight, however, and has the advantage that it tends to increase the power supplied and consequently to shorten the test. Before commencing the test it is desirable to calculate also the approximate current required in order to avoid an excessive current density.

At the commencement of the test this will be given by

and at the end of the test by normal current iron loss hot copper loss.

However, to ensure greater accuracy, the test is made by measuring the power input, which is finally increased to include the hot copper loss, though the cur rent obtained by the above calculation indicates how much the winding will be overloaded from the current density point of view. In general it will be seen that this test is most suitable when the copper loss is high compared with the iron loss, and conversely discretion is needed when dealing with transformers having relatively high iron losses.

When the normal temperature rise is approached the copper loss should be measured and any necessary current adjustment should then be made in order to correct the power input to obtain the true losses under normal full-load conditions, that is as regards current and temperature rise.

The short-circuit equivalent test should not be adopted when the ratio of copper loss to iron loss is less than two to one; for loss ratios below the figure mentioned the open-circuit test is preferable.

FIG. 24 Single-phase short-circuit equivalent

Single-phase transformers

The LV winding is short circuited and the HV winding connected to a single phase supply with an ammeter, voltmeter and wattmeter in circuit, as shown in FIG. 24. The current in the HV winding is adjusted until the power input is equal to the sum of the calculated hot copper loss and the iron loss. The current required is in excess of the full-load current, and the voltage across the phases is higher than the impedance voltage in order to compensate for the inclusion of the iron loss with the copper loss.

Three-phase transformers

The various means of utilizing this test for three-phase transformers are shown in Figs 25-27.

FIG. 25 Three-phase short-circuit equivalent

FIG. 25 shows a star/star-connected transformer ready for the test, the HV windings of the transformer being connected to a low-voltage three-phase supply, and the LV windings being short circuited. Links are provided in the supply leads to phases A and C, and the various instruments are connected to a double-pole changeover switch such that by closing the switch in either phase and opening the corresponding link, the ammeter and wattmeter current coil will be in series with that phase, and the voltmeter and wattmeter voltage coil will be connected between the same phase and phase B. The three-phase supply switch is first closed and the double-pole switch then closed in phase A, the link in A then being opened. The supply voltage is increased until the current shown by the ammeter is slightly in excess of the full-load HV current. This current may be calculated as previously explained. The wattmeter reading is then noted.

The link in phase A is next closed and the double-pole switch changed over to phase C, the link in this phase being then opened, and the wattmeter reading again noted. This process is repeated until, after making the necessary adjustments, the algebraic sum of the two wattmeter readings is equal to the sum of the iron and hot copper losses.

FIG. 26 Single-phase 'series HV' short-circuit equivalent

FIG. 26 shows an alternative method of connecting up a star/star transformer for test. The LV windings in this case are short circuited through the neutral, the HV being temporarily 'series' connected. The two open ends of the HV windings are then connected to a single-phase supply through a watt meter and ammeter. The current is adjusted until the power input is equal to the sum of the iron and hot copper losses. This current is somewhat higher than the normal full-load line current if the transformer is normally star connected, and somewhat higher than the normal full-load line current divided by 3 if the transformer is normally delta connected. The corresponding value of the applied single-phase voltage required will be somewhat higher than 3 times the transformer impedance voltage per phase. Of course, this method can only be used if the HV star connection is capable of being temporarily opened.

FIG. 27 shows a further method in which the LV windings are connected in closed delta, and the HV in open delta. [This must not be confused with the so-called open delta or vee connection for giving a three-phase supply from two single-phase transformers.]

It is, of course, only possible to use this method provided the HV delta connection can be opened. The method is applicable to any three-phase transformer whatever the normal interphase connections, and temporary connections are made as necessary. The test should be confined to transformers of low and medium impedances, however, and it should not be used for transformers of high impedance. For the latter the short-circuit equivalent test illustrated by FIG. 26 is recommended. The HV windings are connected to a single-phase supply, and the same procedure as described for FIG. 27 is followed. The current and voltage required will be the same as given for FIG. 26.

Method (b)

In this method, known as the back-to-back (or Sumpner) test, the transformer is excited at normal voltage and the full-load current is circulated by means of an auxiliary transformer.

FIG. 27 Single-phase 'open delta HV' short-circuit equivalent

Single-phase transformers

FIG. 28 shows the method of connection for single-phase transformers.

The transformers (two identical units are required) are placed not less than 1 m apart with the HV sides adjacent. The HV windings are then connected in opposition through an ammeter. The LV winding of one transformer is connected to a single-phase supply, and the other is connected in parallel with it, but the LV winding of a suitable auxiliary transformer is included in this circuit. The HV winding of the auxiliary transformer is either supplied from a separate source as shown in FIG. 28 or is placed in parallel across the other mains with a variable resistor in series with it.

Normal LV voltage at the correct frequency is then applied to the LV windings in parallel, and the supply voltage to the HV winding of the auxiliary transformer is adjusted at correct frequency until the ammeter in the HV circuit of the transformer under test reads the normal full-load current. If the variable resistor connector is used for the auxiliary transformer, its resistance is adjusted until the ammeter in the HV circuit of the transformer under test indicates the normal HV full-load current.

It should be noted that in this method no wattmeter is used, as the actual full load conditions, that is normal excitation and full-load current, are reproduced.

The copper and iron losses must therefore be those which would normally occur, and there is consequently no need to measure them during this test.

The machine supplying the LV windings in parallel must be capable of giving the normal LV voltage of the transformer under test and twice the no-load current, and it is this circuit that supplies the iron losses.

FIG. 28 Single-phase back to back.

The LV winding of the auxiliary transformer must supply twice the impedance voltage of the transformer under test at the normal LV full-load current, and when the method shown in FIG. 28 is used, the machine supplying the auxiliary transformer must be capable of giving a voltage equal to the ratio of transformation of the auxiliary transformer multiplied by twice the impedance voltage of the transformer under test, and a current equal to the LV current of the transformer under test divided by the ratio of transformation of the auxiliary transformer. This circuit supplies the copper losses to the transformers under test.

There is a further method of making a back-to-back test on two similar single phase transformers which is possible when the transformers are provided with suitable tappings. The transformers are connected as shown in FIG. 29 which is similar to the previous method except that the auxiliary transformer is omitted and the current circulation is obtained by cutting out a portion of the HV winding of one of the transformers.

FIG. 29 Single-phase back to back.

It will be evident that the percentage difference between the numbers of turns in the two HV windings should be approximately equal to the sum of the percentage impedances. For example, if the transformers are provided with plus and minus 2.5 and 5 percent tappings and the impedance of each is 3.75 percent, this test could be made by using the _5 percent tapping on one transformer and the _2.5 percent tapping on the other transformer.

An ammeter is connected in the HV side, as in the previous test, and the supply to the LV windings in parallel is given at the normal voltage and frequency. If it is found that with the best available tappings the ammeter does not indicate exactly the correct full-load HV current, the supply voltage may be varied slightly up or down and the power input adjusted as already described for method (a), that is the short-circuit equivalent test. When it is necessary to raise the supply voltage above normal in order to obtain the correct power input, it is evident that the transformers have a greater iron loss and lower cop per loss than would be the case under normal full loading and excitation. The converse, of course, holds true when it is necessary to lower the supply voltage below normal in order to obtain the correct power input.

It should be noted that the tappings are assumed to be on the HV winding as this arrangement is more common, but the test may be made equally well if the tappings are on the LV winding.

FIG. 30 Three-phase back to back

Three-phase transformers

The diagram of connections for the test on three-phase transformers is shown in FIG. 30 which corresponds to FIG. 28 for single-phase transformers. The diagram shows two star/star-connected transformers, but the external connections are the same for any other combination of interphase connections. The ammeter on the HV side of the transformers under test is, for the sake of simplicity, shown permanently connected in the middle phase, but it would actually be arranged for connecting in any phase by means of changeover switches. The same remark applies also to the voltmeter across the supply. The method of procedure is the same as described for single phase transformers connected as in FIG. 28.

FIG. 31 indicates the connections for two star/star transformers, using the voltage adjusting tapping method, though these would be the same irrespective of the normal interphase connections, temporary connections being made as desired.

The general procedure is identical with that outlined for the single-phase transformers shown in FIG. 29. The LV windings of the two transformers are connected in parallel and excited at the normal voltage while the HV windings are connected in opposition, but at the same time suitable tappings are selected to give the voltage difference necessary to provide the circulating full-load current.

When these methods of testing are used it will be found that one transformer has a temperature rise higher than that of the other. This is due to the fact that the copper loss is supplied by means of a common circulating current, whereas the iron loss is supplied to the two transformers in parallel. The no-load current is out of phase with the circulating current, but not actually in quadrature with it, and consequently the phasor sum of the no-load and circulating currents in one LV winding is greater than the corresponding sum in the other LV winding.

The back-to-back tests illustrated by Figs 28-31 inclusive may, of course, be applied to delta/star and to star/interconnected-star transformers.

Two alternative forms of three-phase, back-to-back temperature rise tests are illustrated in Figs 32, 33a and 33b. The arrangement shown in FIG. 32 may be applied to three-phase transformers of any type, of any combination of primary and secondary connections, and of any impedance, it only being necessary that the two transformers under test are identical. As shown in the diagram, an auxiliary booster transformer is used for providing the circulating current passing through the windings of the transformers under test, and the normal excitation supply is applied to the center points of the secondary winding of the booster transformer. In the event of no centre points being accessible on the booster windings, the normal excitation supply may be applied to the terminals of either transformer, in which case one transformer would have a slightly lower voltage across its terminals than the other, due to the impedance drop in the secondary windings of the booster transformer.

FIG. 31 Three-phase back to back FIG. 32 Three-phase back to back, employing three-phase excitation and current circulation

Where the normal excitation is applied to the center points of the booster transformer, the supply voltage should be slightly higher than the rated voltage of the transformers under test in order to compensate for the impedance drop in the secondary winding of the booster transformer.

The copper losses are supplied from the three-phase source which provides the necessary circulating currents via the primary windings of the booster transformer, while the iron losses are supplied from the three-phase source which supplies the normal excitation to the transformers. The primary windings of the booster transformer are supplied at a voltage which is approximately equal to the sum of the impedance voltages of the two transformers under test multiplied by the booster transformer ratio.

FIG. 33(a) Three-phase back to back on two delta/ star-connected transformers

This method has an advantage that it is not necessary to make any temporary connections inside the transformers, nor it is necessary to reinforce any connection temporarily to carry any special heavy test currents.

Figures 33(a) and 33(b) illustrate a type of test which is applicable to three-phase delta/star and star/star transformers. The LV windings are connected back to back, and current is circulated in them from a single-phase sup ply. The LV windings are excited at their normal rated voltage, so that this method also simulates very closely the heating conditions which arise in the ordinary course of operation.

With this connection the neutral leads on the star sides must be reinforced to carry 3 times the normal full-load current. Circulating current is supplied from a single-phase source, so that the currents in all three limbs are equal and in phase. The leakage flux between windings returns partly through the tank walls, and for this reason the method indicated by FIG. 33 should not be used for transformers where the impedance exceeds 5 percent. Otherwise it is quite a satisfactory method of conducting a load test.

FIG. 33(b) Three-phase back to back on two star/star-connected transformers

Method (c)

This method, known as the delta/delta test, is applicable to single- as well as three-phase transformers where the single-phase transformers can be connected up as a three-phase group.

FIG. 34 shows the diagram of connections employed. The LV windings are connected in closed delta, and supplied from a three-phase source.

The HV windings are connected in open delta [This must not be confused with the so-called open delta or vee connection for giving a three-phase supply from two single-phase transformers.] and include an ammeter.

Voltmeters are connected between phases in the LV circuit. Three-phase volt age at the correct frequency is applied to the LV windings and is adjusted until it equals the normal LV voltage. Single-phase current is supplied separately to the HV windings and is adjusted to the normal HV full-load current.

This method may be used whatever the normal internal connections of the transformer, temporary connections being made if necessary. The voltages and currents required under this test for various normal interphase connections are given in Tables 4 and 5.

FIG. 34 Delta/delta connection

Table 4 Application of voltage or current

Table 5 Application of voltage or current

FIG. 35 Delta/delta connection.

If the normal HV voltage is of the order of 11 000 V and above, the method shown in FIG. 35 is safest. In this method the HV winding is simply closed delta connected, the LV being connected in open delta. A three-phase voltage equal to the normal LV phase voltage is applied to the LV winding at the correct frequency, and the LV copper loss current is supplied single phase.

Method (d)

If it happens that a transformer possesses a low ratio of copper loss to iron loss it is generally impossible to conduct a temperature rise test by the short-circuit method.

This is because the required power input necessitates an excessive current in the windings on the supply side of the transformer, so that a prohibitively high cur rent density would be reached. In such cases it may be possible to test the transformer on open circuit, the normal losses being dissipated in the iron circuit.

If a supply at a frequency considerably below the normal rated frequency of the transformer is available, a condition may be obtained whereby the total losses are dissipated at a test voltage and current in the neighborhood of the normal rated voltage and current of the transformer. If, however, a lower frequency supply is not available, the transformer may be run at the normal rated frequency with a supply voltage greater than the normal rated voltage, and of such a value that the total losses are dissipated in the iron circuit.

Assuming that the iron loss varies as the square of the voltage, the required voltage under these conditions is given by the formula:

Either side of the transformer may be supplied according to which is the more convenient. The method can be applied to both single-phase and poly-phase transformers.

It is important that instruments connected in HV circuits should be earthed; alternatively voltmeters and ammeters should be operated through voltage and current transformers respectively.

Temperature readings

The top oil temperature of the transformer under test is measured by means of a thermometer so placed that its bulb is immersed just below the upper surface of the oil in the transformer tank.

When bulb thermometers are employed in places where there is a varying magnetic field, those containing alcohol should be employed in preference to the mercury type, in which Eddy currents may produce sufficient heat to yield misleading results.

Surface temperatures

When measuring the temperature of a surface, such as a core or a winding, the thermometer bulb should be wrapped in a single layer of tin foil at least 0.025 mm thick and then secured to the surface. The exposed part of the wrapped bulb should then be covered with a pad of insulating material without unduly shielding the test surface from normal cooling.

Cooling air

The cooling air temperature should be measured by means of several thermometers placed at different points around the transformer at a distance of 1-2 m from the cooling surface, and at a level approximately midway up the transformer cooling surface. The thermometers should be protected from draughts and abnormal heat radiation. In the case where forced air cooling is employed and there is a well-defined flow of air towards the coolers then the thermometers should be placed in this cooling stream.

To avoid errors due to the time lag between variations in the temperature of the transformer and that of the cooling air, the thermometers may be immersed in a cup containing a suitable liquid, such as oil, having a time constant of about 2 hours.

The temperature of the cooling air for the test is taken as the average of the thermometer readings taken at equal intervals during the last quarter of the test period.

The carrying out of temperature rise tests is an activity which has been very much simplified in recent years by the use of electronic data-logging equipment, although the measurement of temperatures using thermometers as described above remains a totally acceptable method, it is likely that most manufacturers would now replace these with thermocouples monitored by electronic temperature measuring equipment.

FIG. 36 Typical time/temperature rise curve

The temperature rise test of a transformer should be of such duration that sufficient evidence is available to show that the temperature rise would not exceed the guaranteed limits if the test were prolonged (see EN 60076-2). One way of determining this is by taking readings of the top oil temperature at regular intervals and plotting a curve on linear co-ordinate paper. FIG. 36 illustrates a typical time/temperature rise curve obtained from a test. Alternatively, the temperature test may be continued until the temperature rise does not exceed 1ºC per hour during 4 consecutive hourly readings.

In addition to ascertaining the temperature rise of the oil, it is usual to calculate the temperature rise of the windings from measurements of the increase of resistance. To do this, it is necessary to measure the resistance of the windings before the test (R1) noting the temperature of the windings at the time of the reading, and to measure the resistance (R2) at the close of the test. Over the normal working temperature range the resistance of copper is directly proportional to its temperature above _235ºC.

Top oil temperature

The top oil temperature rise is obtained by subtracting the cooling medium temperature from the measured top oil temperature.

If the total losses cannot be supplied then a value not less than 80 percent may be used and the measured top oil temperature rise corrected using the following correction factor.

where x _ 0.8 for AN circulation

and x _ 1.0 for AF or WF circulation

To obtain an accurate value of the temperature rise of the windings the temperature T1 must be the temperature at which the resistance of the windings is R1. Due care must be taken in the measurement of T1, particularly in the case of large transformers because even if a transformer is left unenergized for days, the oil temperature usually varies from the top to the bottom of the tank, so that the top oil temperature may differ from the mean temperature of the windings by some degrees.

Dry-type transformers

When measuring the cold winding resistance the winding temperature should be approximately equal to that of the surrounding medium. This is confirmed by mounting at least three thermometers on the surface of the winding. The winding resistance and temperature should be measured simultaneously.

Temperature rise tests on dry-type transformers should be performed with the core excited at normal flux density; so that two loading methods are avail able, either a direct load test or a back-to-back test.

The test may be carried out at a current not less than 90 percent of the rated current, or at a current supplying the total losses of the transformer. When the winding test current I1 is lower than the rated current IN, the temperature rise


of the windings, measured by the resistance method, after reading steady state conditions should be corrected to that for the rated load condition, ??N, using the following formula:

where q = 1.6 for AN transformers and q = 1.8 for AF transformers

Winding temperature

The temperature of a winding at the end of the test period is usually calculated from its resistance Rc at that time and resistance R measured at a known temperature Tc, usually the ambient temperature. Care must be taken in measuring Tc particularly for large transformers, because even though a transformer has been de-energized for several days a temperature gradient of several degrees may exist between the top and bottom of the tank, so that the top oil temperature differs from the mean winding temperature.

Over the normal working temperature range the temperature Th corresponding to the resistance Rh may be obtained from the formulae:

At the end of the temperature rise test, when the power supply to the transformer is shut off, the temperature of windings is appreciably higher than the mean tempera ture of the cooling medium, which is the oil around the windings in the case of oil immersed transformers or the surrounding air in the case of dry-type transformers.

Consequently, the windings cool in an exponential manner towards the cooling medium temperature, the thermal time constant of this phase of the cooling being that of the windings only, and of short duration, for example 5-20 minutes.

The winding resistance Rh may be obtained by one of two methods.

(a) Without interruption of the supply, by the superposition method where a DC measuring current is superposed on to the load current. (b) By taking resistance measurements after switching off, using a Kelvin bridge, having allowed the inductive effect of the windings to disappear. Fans and water pumps must be stopped but oil pumps are left running. A correction must then be applied for the delay between shut-down and the commencement of measurement. The correction is calculated by plotting a resistance/time curve for the cooling winding using either linear or log/linear scales and extrapolating back to the time of shut-down.

It is usually more accurate to preset the bridge before each reading and to note the time at which the bridge meter reads zero.

FIG. 37 Extrapolation method for winding resistance at shut-down

(i) Linear scales

FIG. 37 illustrates this method in which decreases in resistance corresponding to equal intervals of time are projected horizontally at the appropriate points of the ordinate to give a straight line L. The resistance at the instant of shut-down is derived by plotting the resistance projections for equal intervals back to 0 time from this line.

FIG. 38 illustrates a typical curve for the HV winding of a 1000 kVA transformer which was plotted using projection intervals of 1 minute.

(ii) Log/linear scales

The difference ?R_ between the measured resistance and the resistance R_, corresponding to the temperature to which the winding is cooling after switching off the supply, is plotted with ?R_ as the logarithmic axis and time as the linear axis. The resistance R_ is chosen in such a way that a straight line is obtained. The resistance at 0 time is then equal to R_ _ ?R0 _, where ?R0 _ is found by extrapolating the line back to 0 time.

FIG. 38 Cooling curve for a 1000 kVA transformer plotted with linear scales

Electronic data-loggers

As in the case of measurement of temperature, mentioned above, resistance measurements after shut-down can now be recorded by means of an electronic data logger. Winding resistance is computed by the voltage/current method but because it is only necessary to make one initial connection to the windings, a higher-driving voltage can be used than would be the case for manual measurements, which speeds up current stabilization. FIG. 39 shows a typical circuit used for measurements using a data-logger. The series resistors limit the current flow to around 20-30 A and by inputting the appropriate voltages across windings and standard shunts the resistances can be computed. These equipments can be set up to take a series of resistance measurements over a pre-determined time period, plot a resistance/time curve and extrapolate this back to shut-down automatically to provide the values of winding resistances at the instant of shut-down.

FIG. 39 Method of connecting three-phase star/delta transformer for inputting resistance measurements to data-logger

Winding temperature rise

The winding temperature rise is obtained by subtracting the external cooling medium temperature from the average winding temperature measured by one of the methods described above. In these cases a correction must be applied to the winding temperature rise using the following correction factor:

where y _ 1.6 for ON and OF oil circulation

and y _ 2.0 for OD oil circulation

Duration of temperature rise tests

In general, temperature rise tests last from 6 to 15 hours. They may be shortened, if necessary, by overloading the transformer at the commencement of the test and then reverting to full-load losses as the final temperature is approached, but this method should only be adopted in special cases for if insufficient time is allowed for the windings to attain their correct steady temperature, errors will be introduced.

As an alternative method it is possible, in the case of separate radiators or coolers, to restrict the normal oil flow and so accelerate the temperature rise of the oil in the early stages of the test. Further information is given in EN 60076-2 regarding the measurement of oil and winding temperatures at the end of a temperature rise test.

Noise level tests

Reference should be made to Section 6.3 for details of measurement of transformer noise.

FIG. 40 Typical transformer test certificate

Test certificate

At each stage in the testing of a transformer the results are recorded on the testing department's records and subsequently these are transferred to an official test certificate for transmission to the customer. Typical test certificates are shown in Figs 40 and 41.

FIG. 41 Typical transformer temperature rise test report [not shown]


In the introduction to this section it was suggested that there are tests over and above those described in EN 60076 which can be considered for important transformers for which it is required to have the highest level of confidence in their integrity and suitability for service in a demanding situation. Such additional testing will, in itself, add to the first cost of the transformer and it might be that the manufacturer will wish to design into the transformer some additional safety factors which will also add to the first cost. However, this will add to the confidence in the integrity of the unit which was one of the objects of the exercise. Transformers which might appropriately be included in this category for special treatment would be all of those operating at 275 and 400 kV as well as strategically important lower-voltage transformers, possibly supplying a steel smelter or other important process plant. It is clearly the responsibility of the user to decide whether his transformer is to be considered important or not.

What additional testing might be carried out fit this is a question which was posed by CEGB in the early 1970s. At this time there was great concern expressed at the highest level within that organization at the high failure rate of large generator transformers. At one time a CEGB internal report predicted that, on a purely statistical basis derived from the observed incidence of failures in the organization's existing generator transformer population, one of the largest generator transformers was expected to fail every 0.7 years! Concern was expressed by management that many of the observed failures occurred in early life and the question was asked as to why works testing had not detected incipient weakness in these transformers. Not surprisingly, management demanded that as a matter of urgency measures should be put in hand to remedy the situation and, logically, one arm of the strategy was to devise and implement a regime of more effective testing. The next problem, then, was to set about establishing this more effective testing.

To do this, it is reasonable to start by considering how the transformer is likely to fail. There are, of course, many failure mechanisms for something as involved as a large transformer, but from an assessment of the failures experienced it could be concluded that these are likely fall into one of three classes:

(1) Insulation will break down under the influence of the applied voltage stress.

(2) Insulation will be prematurely aged, due to overheating.

(3) Windings will suffer mechanical failure, due to inability to withstand the applied forces.

Since failure mechanisms are often complex, some of these were difficult to classify, being possibly due to a combination of more than one of the above causes. Overheating, for example, especially if not too severe, often will not itself cause failure, but will reduce the mechanical strength of the insulation, so that when the transformer is subjected to some mechanical shock, such as a system fault close to the terminals, it will then fail. It is possible, too, that inadequate mechanical strength, on occasions, allowed movement of conductors which reduced electrical clearance so that electrical breakdown actually caused failure. Common amongst many of the failure modes was an area of localized overheating due to poor joints, high leakage flux or inadequate local cooling.

Even though failure modes are not always straightforward, the study pro vided a basis for objective discussion of appropriate methods of testing and the next step was to consider existing tests and identify their shortcomings in the light of the experience gained.

next >>

Top of Page PREV.   NEXT   Guide Index HOME