Pushing past the limitations of power factor testing

Introduction
Line-frequency power factor tests – also known as tan δ tests – are among the most widely performed field tests on power assets. It is not always completely understood, however, that these tests have a serious limitation – they are averaging tests. In other words, they deliver a result that represents the average condition of the total insulation under test, however big or small that might be.

This averaging affects the sensitivity of power factor tests to problems, particularly in assets like power transformers where the insulation system is large and the problem affects only a small part of the system. It also means that power factor tests are unable to discriminate between severe, localized problems that demand immediate attention and widespread general deterioration that may only require regular monitoring.

There are two principal ways of addressing the sensitivity problem of line frequency power factor testing. One way is to reduce the amount of insulation under test by (electrically) dividing the asset under test into as many sections as possible. The other is to perform the test over a range of frequencies rather than just at line frequency.

Looking at the first approach of dividing the asset under test into sections, the technique adopted will depend on the type and configuration of the asset. With transformers, for example, some of the possibilities for sectionalizing the insulation beyond the typical segmentation (e.g., CH, CHL and CL on a two-winding unit), are:

• Use a core ground, if easily accessible, as a test point; many transformer manufacturers now bring the core ground out through the tank wall for test convenience.
• Use the DETC (de-energized tap changer) set between on-tap positions. Note, however, that many transformer owners are reluctant to allow the DETC to be moved, so this approach is unlikely to be suitable for routine testing. It can nevertheless be a useful option when investigating a known problem.
• „„ Use the cross-check test. This has been popularised in North America for testing potential (voltage) transformers, but can also be applied to power transformers with a delta-wye (delta-star) configuration.

The second method of overcoming the limitations of line-frequency power factor testing is that of using a multiple frequency test, wherein one option is to use a power factor test set capable of operating over a wide frequency range. Test sets in the Megger Delta 4000 family support testing at frequencies from 1 Hz to 500 Hz. Another option is to carry out dielectric frequency response (DFR) testing – also known as frequency domain spectroscopy – using a test set of the type
included in Megger’s IDAX family.

Multiple frequency insulation testing will be covered in a future article in this series, but the remainder of this article will discuss how Southern California Edison (SCE) has successfully adopted methods based on electrically dividing the insulation of power transformers into even further sections than typical to facilitate diagnosis of suspect transformers, specifically relying on the crosscheck and the open DETC methods.

Background to SCE experience
In early 2013, SCE Distribution Apparatus Repair began using a “per-phase” power factor protocol on all medium-power transformers sent in for evaluation and refurbishment, further segmenting CH and CHL components into 3 parts each. It was found that in 5% of these 95 and 150 kV BIL class units 10 years old or less, significant differences existed in “per-phase” CHL values (e.g., CHL1; CHL2; and CHL3). While GST test-derived values did not seem to significantly change with age, UST test-derived values (i.e., CHL) drifted upwards significantly as insulation began to deteriorate. “Per-phase” CHL differentials of up to 0.25% were found. Some units had incipient anomalies, while others appeared to have been subjected to unbalanced loads.

Physical arrangement of components – Core form 3 phase winding assemblies 
The winding assemblies of the subject SCE transformers were of core form construction that uses a three-leg configuration (versus, for example, the fiveleg configuration often used in the shell type core).

As illustrated in Fig. 1 and Fig. 2, the three individual coil packages are concentrically arranged around the vertical members of the core. The individual coil assemblies are constructed so that the low-voltage winding is located on the inside of the assembly. This arrangement isolates the primary coil packages from the vertical (ground potential) core legs and facilitates efficient phase/tap lead routing.

Basis of the cross-energisation process
When a standard overall power factor test is performed on a single-phase winding, whereby the winding terminals are short-circuited, the full applied test voltage is impressed across all points of the energized winding. When the “per-phase” crosscheck test is performed, the H1 terminal is energized
and the return lead is connected to the H2 terminal to guard the excitation current flowing through the winding while measuring the leakage current flowing from the insulated primary winding to the secondary  winding or ground. In this scenario, a voltage gradient is established across the winding, whereby the voltage at H1 is at 100% of the applied voltage and the potential at H2 is essentially zero.

(1) - For an overall CH test of the winding as given in Fig. 3,

CH Capacitance Current = ICA + ICB + ICC + ICD + ICE + ICF + ICG + ICH + ICI

(2) - If a single “per-phase” test is performed by energizing H1 and guarding H2, test capacitive current equals:

ICA + (7/8)ICB + (3/4)ICC + (5/8)ICD + (1/2)ICE + (3/8)ICF + (1/4)ICG + (1/8)ICH + (0)ICI

If it is assumed that the condition of insulation is homogenous for the entire winding, mathematically it may be assumed, from the second equation, that the measured capacitance and watt losses in the crosscheck will be very close to 50% of the measured capacitance and watt losses in the overall test. If the test is repeated by energizing the H2 terminal, the sums of the watt losses and capacitances from both cross excitation tests will nearly exactly match the values from the overall test.

Application
If a high loss anomaly exists on coil G of the transformer shown in Fig. 3, it will be detected in the overall test. The overall test is therefore useful, but will not provide information about the location of the anomaly. When the cross-check test is performed by energizing the winding from the Coil A side, the voltage at Coil G will be dissipated to a level where the measured losses from the anomaly at coil G will be very low. Conversely if the test is performed from the Coil I side, the power factor at coil G will be measured at 75% of applied voltage and losses measured at coil G will be significant. An anomaly located in the center of the winding will result in equal watt losses as measured from each side.

Physical geometry of Delta Wye configurations
The cross-check method works most reliably on a Delta-Wye configured transformer with a three leg core as given in Fig. 4.

Fig. 5 provides an illustration of a three-phase coreform construction discussed previously, and given in Fig. 2, but with the physical connections of the delta-wye winding superimposed. The low-voltage coil packages are concentrically arranged around the vertical core legs, an interwinding barrier with oil ducts is installed around the outside edge of the low voltage package, and the primary winding package is installed tightly around the interwinding barrier. The lead from the outside layer of the primary winding in the center winding assembly is connected to the H2 bushing. The center winding assembly is designated as B phase. The leads from the outside layer of the other 2 winding packages are connected to the H1 and H3 terminals. The outer lead connected to the H1 terminal is designated as A phase, while the outer lead connected to the H3 terminal is designated as C phase.

A jumper is connected from the A phase lead on the inside of the winding to the lead connected to the H3 terminal. A second jumper is connected from the B phase lead on the inside of the winding to the lead connected to the H1 bushing. The final delta jumper is connected from the C phase lead on the inside of the winding to the lead connected to the H2 terminal. Not shown in Figure 5 is the gang operated de-energized tap changer (DETC) on each phase.

Theory of the GST cross-check test for CH anomalies
The delta-wye (or delta-star) “per-phase” crosscheck test is performed by applying a test voltage on one primary terminal while guarding the other two primary terminals to shunt the excitation current around the metering circuit. This creates a voltage gradient whereby 100% of the applied voltage is applied at the energized primary terminal and essentially zero voltage exists at the return lead connections.

Because of the voltage gradient, most of the current passing through the metering circuit is from the region in proximity to the energized terminal. If the assumption is made that the condition of the winding insulation is homogenous, the third of the winding segments nearest to the energized terminal accounts for approximately 60% of the metered current. The center third of the windings account for 33%, and the third of the windings adjacent to the return leads account for less than 8%.

Any anomaly in the winding insulation which is in the portion of the windings adjacent to the voltage
source results in a higher power factor than when the other terminals are energized. In cases where tests indicate the anomaly is located near the center of the winding, the DETC may be opened to sectionalize the winding to determine where the anomaly is located in relation to the DETC. A significant anomaly in a winding is detectable from either end of the winding, but in cases where the anomaly is small and located at the extreme end of one winding, the power factor deviation as seen from the other end may be only several hundredths of a percent. On this basis the location of an anomaly can be determined in terms of its phase relationship and by its location in the winding relative to the terminals.

The effect of physical geometry on UST and GST measurements
Generally, even though three CH values are generated as well as three CHL results, it is not known to be possible to calculate the current and loss for each phase alone. Rather, if a problem is isolated to a
single phase, this problem should be evident to varying degrees on two of the three cross-check
measurements but completely absent in the third test in which the questionable phase is short-circuited. Robert Breazeal of SCE suggests that there is more to consider: “Properly interpreting UST test data requires taking into account the physical geometry of the individual coil assemblies. The primary winding is located on the outer portion of the coil assembly. In the delta primary configuration, each terminal lead splits with one leg connected to the outside end of a primary coil, and the other leg connected to the inside end of the primary winding on a different phase. Because the HV/LV barrier is physically located on the inside edge of the primary coil, the applied voltage which enters the primary winding from the outside edge of the primary is 95% dissipated by the time the applied voltage reaches the barrier. Conversely, the voltage which is applied on the inside edge of the primary coil on the opposite leg is essentially at 100% as seen from the barrier because the inside primary lead enters the winding at the barrier.

The significance of this arrangement is that when perphase UST measurements are performed, virtually all of UST current and watts are derived from a single winding despite the fact that two windings are energized. In GST measurements, a CH anomaly may be detectable from both directions depending on the location and nature of the anomaly, but a CHL anomaly detected in a UST test will be nearly invisible as seen from the outer portion of the winding. When the open DETC test is performed, the winding that contributes the smaller portion of the current/watts is removed from the circuit. Even if a significant problem exists on the winding which is dropped out, the net change from the standard per-phase UST test for that phase will only be a few hundredths of a percent.”

Effect of CL ground faults on “Per-phase” test data (Cross-check and DETC open)
SCE’s Distribution Asset Repair shop asserts that in performing failure analysis on transformers, the “per-phase” protocol has proven to be a preferred method of insulation assessment. This, however, has been challenging at times because, in many cases, when a winding has failed, standard “per-phase” testing is impossible due to low primary winding insulation value and/or high excitation current from shorted turns. In cases involving a secondary fault, “perphase” test data is seemingly difficult to interpret due to what appear to be erratic values.

In Fig. 6, a ground fault is shown to exist in the vicinity of the X1 bushing of the A phase secondary
winding. The GST test voltage is applied to the H1 terminal while guarding H2 and H3 (and X0). In this
CH1 test, the watts and power factor will be elevated well above what would be consistent with the CH1 value in the absence of the CL fault. This is because a physical process that is present in the measurement, but typically concealed, falls out of balance and becomes visible, that is: during the CH1 cross-check test, voltage is induced in low voltage windings X1 – X0 and X2 – X0, because flux is moving in opposite directions in A and B core legs during this test, the voltage induced from X0 to X1, with respect to X0 (where a measuring return lead is placed), is 180° out of phase with the voltage induced from X0 to X2. Given healthy winding insulation conditions, the dielectric current and loss associated with the induced voltage across X1 – X0 (which is unavoidably included in the measurement given the test’s circuit configuration) tends to cancel with that which is also measured and results from the voltage induced in X2 – X0. But in this example, the X1 – X0 loss contribution
with a low resistance faulted path to ground is much higher than the X2 – X0 contribution so the influence is seen in the CH1 test results.

When the CH3 test is performed by energizing the H3 terminal and guarding terminals H1 and H2 (and X0), the test results will be abnormally low. During thistest, voltage is induced in windings X1 – X0 and X3 – X0 and the preceding explanation similarly applies here, albeit with a net current contribution seen by the meter that is of the opposite polarity. When the CH2 test is performed by energizing H2 and guarding terminals H1 and H3 (and X0), voltage is induced in LV windings X2 – X0 and X3 – X0. As neither of these windings is affected by the ground fault, their resultant current and loss contributions that are unavoidable seen by the meter remain more or less equal but of opposite polarity and will therefore cancel each other. Consequently, the resulting data will accurately reflect the CH value of the insulation in proximity of the H2 terminal.

During the “per-phase” CHL tests, similar extraneous contributions due to induced voltages on the LV
windings are included in the measurement. These too tend to stay invisible (since they cancel with
each other) until there is a problem in a LV winding which upsets balance in the contributions from each LV winding. In such a case, the test results will be affected much like those of the CH “per-phase” tests.

Table 1 contains data from “per-phase” testing performed on a medium-power transformer in good condition. In this example, a 2.7 MΩ resistor was connected from the X1 terminal to ground in order to simulate a ground fault on the outside edge of the A phase secondary winding. The test specimen in this case is a 750 kVA unit with a 12000Δ primary and a 4160Y/2400 secondary.

Upon examination of the “per-phase” data, it is evident that the values for the “per-phase” crosscheck
tests closely follow the pattern outlined previously. Conversely, the open DETC “per-phase” test data exhibits no influence of the ground fault on the UST and GST measurements. 

During open DETC “per-phase” tests, each low voltage winding terminal (X1, X2, X3, and X0) is guarded so the low voltage windings are effectively short-circuited. Therefore very little voltage is present on the low voltage windings during tests. The extraneous contributions that may or may not affect the “per-phase” cross-check tests are not factors in an open DETC “per-phase” test. As a result the open DETC test data accurately reflects the condition of the CH and CHL insulation. 

As the location of the secondary ground fault is moved closer to the X0 end of the secondary winding, the resultant loss contribution decreases. If the ground fault resistor is moved from the X1 terminal to the X0 terminal in the above scenario, the resultant “per-phase” data will be identical to the baseline per-phase data for the test specimen before the resistor is installed.

The concluding part of this article will appear in the next issue of Electrical Tester and will include an overview of supplementary tests carried out by SCE, discussion of transformer configurations and limitations, and a case study involving a 69 kV substation transformer with a secondary ground fault.
A further article, scheduled for a future issue, will look at multiple frequency insulation test techniques
for transformers, and explore the benefits of this approach.