Transformer Insulation Diagnostics Beyond Line- ...

Presented By:
Dr. Diego M. Robalino
Megger North America
TechCon 2018

If you want to learn more, please visit our website.

Abstract

Line-frequency power factor / dissipation factor testing technique has been used throughout the electrical industry and has triggered the attention of those who are directly involved in electrical assets operation, maintenance and management. The power factor/dissipation factor testing technique, as a tool for insulation diagnostics, has evolved beyond the initial line-frequency analysis and this evolution should be well understood for field implementation.

In this paper, the author describes advanced features incorporated into the power factor testing technique for field insulation diagnostics. The Individual Temperature Correction (ITC) algorithm to normalize power factor measurements based on insulation condition assessment and not generic tables; and, the voltage dependence detection (VDD) algorithm as an evaluation tool to identify non-linearity of the dielectric system.

This paper provides practical guidelines to facilitate the decision making process. The advanced features described herein, will allow prioritization of maintenance activities and better understanding of results.

Introduction

The process to deliver electric energy from generation to transmission and distribution systems involves the use and application of complex engineering design and operation strategies. The path for electric energy involves a number of components with dedicated functionality and specific technical characteristics. In general, rotating machines, cables, power, distribution and instrument transformers, protection and control devices are designed, built and commissioned to perform in a reliable and safe manner during the entire expected service life.

Low voltage (LV), medium voltage (MV), high voltage (HV) and extra-high voltage (EHV) components must pass a series of testing procedures before reaching the field. The test should resemble normal operation condition and potential abnormal conditions to simulate the mechanical, dielectric, electric and thermal stress expected during the service life of a specific component.

Factory and field-testing procedures have been developed to assure standardized practices around the entire technical community to deliver a high quality final product to the end-user and minimize the risk of failure during its service life.

Undoubtedly, a good understanding of the condition of the insulation in the electrical system is paramount. The risk of failure of electrical equipment in the field due to dielectric breakdown increases as the aging of the assets in the field increases. Aging is not the only factor, for power and distribution fluid-filled transformers, moisture, oxygen, oil-aging byproducts and particles of different origin are agents of degradation, shortening transformer service life under thermal, electric, electromagnetic and electrodynamic stresses [1].

The electrical power industry in North America has relied on single-frequency (line-frequency) power factor/ dissipation factor and capacitance testing to assess the insulation condition of power and distribution transformers for almost a century. Industry specialists, researchers and instrument manufacturers have worked together to provide advanced features to improve, expand and better interpret the information obtained from line-frequency power factor testing on fluid-filled power and distribution transformers.

I. LINE-FREQUENCY POWER FACTOR

A perfect or ideal insulation system (represented by a capacitor) would have no other component of current except the capacitive component. However, no dielectric materials, not even vacuums, are perfect. When an AC voltage is applied across an insulation system, while most of the resulting current that flows through the insulation is capacitive (representing the energy being stored by the insulation), loss (resistive) current results as well. This loss component, IR, which is in phase with the applied AC voltage, is associated with the insulation dielectric losses. The total resulting current that flows through the insulation, IT, is the vector sum of the capacitive current, IC, and loss current, IR, as given in Figure 1.

Dielectric loss is the energy lost or released as heat when an electrostatic field is present across an insulation system. Losses can be broadly classified as conductive losses, arising from leakage current, or polarization losses. The total losses (IR ∙ V) in a dielectric are equal to the combined polarization losses and conductive losses present.

Dielectrics perform best when they are clean, dry, relatively void-free, and utilized within a certain temperature range. Adversaries to a dielectric’s continued good health are heat, moisture and oxygen. Continuous degradation of the insulation system is observed when power factor/dissipation factor test reports higher dielectric loses. Normal transformer service or aging show a slow increase of these dielectric losses and that is expected to see during normal and/or routine maintenance testing. Nevertheless, a rapid increase of losses is an indication of an active failure condition that may lead to dielectric breakdown.

Power Factor (PF) as expressed in (1) is the cosine of  (the complementary angle of the “loss angle”) while Dissipation Factor (DF) as expressed in (2) is the tangent of . As described in [2], the normal in-service and new PF limit for mineral-oil-filled power transformers < 230 kV is 0.5% PF at 20 °C, and the normal and new limit for transformers  230 kV is 0.4%. To help reduce the risk of catastrophic failure, the limit for serviceability of all mineral-oil-filled transformers is 1.0% PF at 20 °C. PF values between 0.5% and 1.0% at 20 °C require additional testing and investigation to confirm that a problem is not worsening.

II. SPECIFICS OF THE LINE-FREQUENCY POWER FACTOR

A. Line-frequency reference only

Line-frequency power factor is commonly used to detect potential aging and degradation of insulation due to thermal, chemical, mechanical or electrical stress. Trending analysis for interpretation of power factor/dissipation factor results relies on measurements at one specific voltage and one single frequency and “assumes” proper temperature correction of this measured value.

Increasing levels of some contaminants might not provide notable change in line-frequency power factor and depending on the conductivity of the material, the effect of contamination may be observed at a frequency different from 50/60Hz.

To overcome this limitation, a dielectric frequency sweep in the range between 1 – 500 Hz is required.

B. Thermal dependence

On this point and from IEEE Std. C57.12.90-2015, the following is extracted as published: “Table 4. NOTE 3— b) Experience has shown that the variation in power factor with temperature is substantial and erratic so that no single correction curve will fit all cases…”

The standards clearly indicate that measurements of losses in the dielectric material are sensitive to temperature variation as well as dielectric condition changes. Therefore, changes in the dielectric condition should imply changes in the thermal behavior of dielectric parameters and table correction factors selected based on nameplate information are merely an average reference not specific for the asset under test.

To overcome this limitation, an algorithm to estimate the individual temperature correction is required.

C. Voltage Dependence

The voltage dependence phenomenon in solid insulation during line-frequency PF/DF test is well described in [3] and [4]. The losses measured in the insulation system when a line-frequency (50/60Hz) signal is applied, is a composite of dielectric losses which are constant with voltage and power loss due to discharges. Mathematically, this can be expressed as the total conductance of the system in (3)

Based on (3), if no discharges occur, the amount of charge increase on the electrodes or conductors as a result of internal discharges (Qi) equals zero. To visualize this effect, a dielectric frequency sweep between 500Hz and 1Hz is performed at different voltage levels on epoxy type MV equipment and results are presented in Figure 2.

It is clear from Figure 2 that the voltage dependence of an insulation system is detected by measurement at different voltage levels. In the specific case of oil-impregnated insulation, the expected condition is to measure PF/DF at line frequency at a single voltage and that the value under ideal conditions will not change. A deviation from this statement will indicate that the ideal condition has been altered and potential degradation is taking place within the insulation system of the transformer. Figure 3 shows an ideal condition of the insulation inside the transformer in a tip-up test.

Historically, power factor test at line frequency has been carried out at one test voltage value. To overcome this limitation, a technical approach to determine voltage dependency is required.

III. TRANSFORMER INSULATION DIAGNOSTICS BEYOND LINE-FREQUENCY POWER FACTOR

A. Narrow band dielectric frequency response

It is not a big surprise to find, on different objects, similar line-frequency power factor values. This finding should trigger the “curiosity” of the testing specialist to further investigate this coincidence. The approach provided nowadays to simply visualize the dielectric response in a short period of time (~3 minutes) is to perform a narrow band dielectric frequency response (NBDFR) sweep. NBDFR is typically carried out in the range between 500Hz and 1Hz and for the most part, the sweep is obtained by applying a low voltage AC signal to the insulation system.

The benefit of performing NBDFR in a complex insulation system is difficult to rank. First, NBDFR provides direct visualization of the dielectric material condition and, in the case of bushings; it provides a comparative analysis signature. Figure 4 describes six different examples of insulation conditions where line-frequency power factor is very similar amongst different objects but the NBDFR sweeps show important differentiation.

The research presented in [5] shows the importance of the narrow band dielectric response. The specific application on oil-impregnated paper (OIP) HV bushings allows a better assessment of the insulation at 1Hz. It is recommended to plot the narrow band dielectric response in a logarithmic scale (as shown in Figure 4), otherwise the smallest decade (1 – 10Hz) is not clearly visualized and important information will be omitted.

NBDFR from 1 to 500Hz has already been implemented in state-of-the-art power factor test sets and its use has more applications as discussed in the next section.

B. Individual Temperature Correction (ITC)

As mentioned in the previous section, NBDFR has an additional application when used for advanced insulation diagnostics.

The implementation of mathematical algorithms allows moving from the dielectric response in the frequency domain into the dielectric response in the thermal domain [6]. Figure 5 shows how the dielectric response in the frequency domain and in the thermal domain change based on the condition of the insulation system and, therefore, the use of tables and average correction factor values might be misleading and not accurate.

The obtained NBDFR from the inter-winding insulation is also used to determine the individual temperature correction (ITC) of the line-frequency power factor measured at temperatures different from 20°C.

The mathematical approach used to correlate line-frequency power factor with temperature is an Arrhenius-based equation (4). This approach describes the ‘frequency shift’ factor, which is dependent on the temperature difference between the temperature of the measurement T2 and the reference temperature T1 (expressed in Kelvin).

The equation considers an exponential function related to temperature (T), Boltzmann constant (kB) and the activation energy value (Ex,y) of the material. It has been found that the shape of the dielectric response (PF/DF versus frequency) does not change very drastically with temperature for quite a large group of solid dielectric materials; rather, as temperature changes, the response (a spectral shape) shifts with respect to frequency while remaining intact. This means that a PF/DF value measured at line-frequency and at one specific temperature, is the exact same PF/DF that would be measured at the reference temperature (typically 20°C) at a frequency different from line-frequency.

Therefore, using NBDFR, the individual temperature correction (ITC) factor can be estimated for the line-frequency power factor based on the real insulation condition [6].

C. Voltage Dependence Detection (VDD)

The effect of voltage over an insulation system is not observed by a single-voltage, single-frequency power factor test. The voltage dependency effect has been observed and studied by different researchers and end-users who faced power factor voltage dependency on insulation systems which should not be voltage dependent as it is the case of oil-paper insulation in power and distribution transformers. The developed feature to identify voltage dependency on the insulation system is the percentage voltage dependence factor (%VDF).

To perform a line-frequency power factor test, the instrument applies a perfect sinusoidal voltage to the insulation system. Thus, it is expected (ideally) that a perfect sinusoidal current, representing IT will be measured. In fact, this is not always the case. As an example, the distortion in the measured current signal is represented in Figure 7.

The deviation of the measured signal from its fundamental implies that the insulation system has lost its linearity [7]. Therefore, the %VDF can be calculated in a similar way as the total harmonic distortion (THD) of the measured current signal as presented in (5):

Voltage dependence detection is easily demonstrated while testing solid insulation specimens when performing a tip-up test as shown in Figure 7. The ideal oil-paper insulation presented in Figure 3 compared to the plot in Figure 7 clearly support the need to have implemented voltage dependence detection algorithms in the power factor testing instrument.

Figure 7 shows voltage dependency on a solid insulation specimen. The %VDF follows the PF voltage dependency. If the insulation shows deterioration, PD activity or contamination, %VDF provides a warning flag to the end user “suggesting” to add a tip-up test to the series of tests carried out on that specific transformer.

%VDF (percentage voltage dependence factor) is a very small value, typically below 0.1% for linear %PF response.

IV. ADVANCED DIAGNOSTICS BY MEANS OF DIELECTRIC FREQUENCY RESPONSE

Frequency Domain Spectroscopy (FDS) also known as Dielectric Frequency Response (DFR) is an advanced application of the power factor test to determine the condition of the insulation system in power and distribution transformers.

In the frequency domain, the dielectric spectrum is obtained applying an AC excitation voltage between 140Vrms and 1400Vrms to the insulation system. The frequency spectrum is typically obtained between 1 kHz and 1 mHz as shown in Figure 8.

The response is analyzed based on mathematical modeling and comparative analysis against a well-documented materials’ database to determine primarily:

For more information, please visit Hengfeng.

• Percentage moisture concentration in the solid insulation, and;
• Conductivity of the liquid insulation

In addition, DFR provides information regarding:

1. Presence of contaminants by non-typical responses, and;
2. A solution to convert the frequency domain response into a thermal domain response of the insulation system as a plot of %PF or %DF vs. insulation temperature as shown in Figure 9.

The IEEE transformers committee is finalizing the work to generate the guidelines for DFR application on power and distribution transformers. The document is expected to be published in 2018 under the designation IEEE Std C57.161.

V. SUMMARY

As presented in this paper, condition assessment of transformer insulation is not a simple task. The limited information obtained by a single line-frequency power factor value may now be expanded with newly developed features that certainly ease the condition assessment of complex insulation systems in power and distribution transformers.

References

[1] V. Sokolov, V. Bulgakova, Z. Berler, “Assessment of Power Transformer Insulation Condition”. Proceedings of the IEEE Electrical Insulation Conference & Electrical Manufacturing Coil Winding Conference, 2001

[2] IEEE Guide for Diagnostic Field Testing of Fluid-Filled Power Transformers, Regulators, and Reactors. IEEE Std C57.152TM-2013.

[3] IEEE Std. 286-2000 (R2012). “IEEE Recommended practice for measurement of power factor tip-up of Electric machinery Stator Coil insulation”.

[4] T. W. Dakin, “The Relation of Capacitance Increase with High Voltages to Internal Electric Discharges and Discharging Void Volume”. AIEE Transactions. Part III, Issue 3, 1959

[5] D. Robalino, I. Guener, P. Werelius, “Analysis of HV Bushing Insulation by Dielectric Frequency Response”. Proceedings from the IEEE EIC Conference 2016.

[6] D. Robalino, “Individual Temperature Compensation – benefits of dielectric response measurements”. Transformers magazine, Vol. 2, Issue 3. 2015.

To explain VLF tan delta, we must first explain the reason it was developed and how the concept and equipment used to perform this test came about. Tan delta is still a form of high-potential (hipot) testing but utilizes the advantages of both AC and DC testing while eliminating many, but not all, of the disadvantages. The very-low-frequency (VLF) portion is explained in a short synopsis of AC and DC hipot testing that covers the good, bad, and ugly of each test.   

Most test technicians have been required to perform dielectric voltage withstand testing, also known as high-potential testing, on cables. They usually discovered the differences between AC and DC testing the hard way, whether performing acceptance or maintenance testing.

Both hipot tests have natural advantages and disadvantages that are apparent depending on the equipment under test. The difference between AC and DC voltage and current becomes clear after a few hard-learned testing situations.

Every young test tech asks why the test voltages for DC and AC hipot testing are not the same. My immediate answer is a question: “Are 120V DC and 120V AC the same?” The initial answer is “Yes” until the tech sees me cringe and thinks about the question. Once I ask, “Do the words root mean square (RMS) and peak voltage mean anything to you?” the answer becomes intuitively obvious and the reason for the difference in voltage falls into place. They now understand that for the stress level of DC voltages and AC voltages to be equivalent, DC must be raised to the peak value of AC, which equates to 1.414 of the AC value. The tech often asks a follow-up question: “How do you determine the value of voltage to use for a specific test?”

My answer is less from an engineering perspective than from an on-the-job training point of view. There are very accurate ways to determine the exact answer, but this rule of thumb works very well in most cases: The AC value equals 2 times 120% of nominal line voltage + 1000 V, applied for a unit of time. The DC test voltage would just be the peak AC equivalent. The test time for most standards, including products covered under IEC 60950, is 1 minute.

DC Hipot

DC hipot testing measures the insulation resistance of cables by applying high voltage in steps or ramping up the voltage between the center conductor and shield. The leakage current is measured, and resistance is calculated using Ohms Law to obtain the insulation resistance. One negative aspect is that the cable conductor, shield, and insulation form a capacitor that initially acts as a short circuit while charging the capacitance of the cable — as we all have seen on our analog test set current meter — by bouncing the needle off the right-hand side of the meter case. This requires understanding how to properly run the hipot and why this issue occurs. The instantaneous current surge can damage service-aged cables during the test due to the instantaneous over-voltage applied. Therefore, the test tech must ramp the voltage up in small increments to keep the charging current minimal and to avoid tripping off the test set during the charging interval. 

Since the applied DC voltage on the test specimen is higher than for an AC hipot test, it is more prone to flash-over and corona discharges. The tech must utilize corona balls and bagging techniques to minimize the problem during the test.

When performing a DC hipot test, the capacitance of the circuit will maintain the applied voltage; therefore, a discharge circuit or other means to protect the test tech from possible lethal shock is required.

The DC hipot test is typically recommended only for new cables, but it is sometimes used on service-aged cables. One of the advantages of a DC test is that the leakage-current trip points can be set to a much lower value than an AC test voltage. That makes it possible to separate out specimens that have marginal insulation that would have been passed by an AC tester. There is some evidence that HVDC testing of aged extruded dielectric (particularly XLPE) cable is harmful and may result in failure of the cable when re-energized with AC or shortly after.

A simple circuit and graph (Figure 1) plots the charging of a capacitor using DC and shows initial voltage and currents from no charge to fully charged after five time constants (TC) where  TC  = RxC.

AC Hipot

The AC hipot test, on the other hand, is performed by applying high-voltage AC on the conductors to be tested and measuring the current. Test results show a much higher current reading because the cable now performs both as a capacitor and a resistor in parallel. The center conductor acts as one plate of the capacitor and the outside layer represents the opposite plate separated by the dielectric insulation in the middle. AC current flows easily on a capacitor due to reversal of polarity, whereas DC current decreases as the effective capacitance of the cable charges to the test voltage. The AC voltage applied on the test conductor is lower compared to DC, but it requires considerably higher amounts of power because at 60 hertz, capacitive reactance is very low. Most test set power supplies have limitations that adversely affect the size of the specimen that can be tested and force the use of DC testing.

AC hipot tests include capacitive current (IC) and resistive current (IR) in parallel (Figure 2). The capacitive current is out of phase, leads the voltage by 90 degrees, and is typically large relative to the leakage (resistive) current. Thus, the charging current masks the leakage current and the test is only Go/No-Go; it is not diagnostic.

The problem with AC hipot testing is the ratio of the resistive component and the reactive component. Leakage current in the specimen under test caused by defects can be easily overlooked (Figure 3).

Hipot Summation

Technicians face many challenges in AC and DC hipot testing that must be understood to accurately diagnose the condition of insulation in cables or any other device.

AC hipot source frequency plays the largest role in the amount of power required to charge the capacitance of a test specimen. The use of low frequency — such as 0.1Hz that is normally used — reduces the power requirement by 600 times as compared to 60 Hz tests. This means the lower the frequency, the larger the Xc, resulting in a smaller, more portable power supply that can drive larger, longer cables.

The measurement and diagnostic side is tan delta or dissipation factor. Most technicians routinely perform power factor (PF) testing. If they understand the theory of PF testing, they also already understand dissipation factor and just didn’t realize the correlation between the two.

In Figure 4, power factor is the cosine of the angle theta formed by the ratio of total current and resistive current, whereas dissipation factor is the tangent of the angle delta formed by the ratio of total current and capacitive current. The cosine of theta and the tangent of delta are comparable for results up to 10% PF.

VLF Tan Delta

In a perfect capacitor, voltage and current are shifted by 90 degrees with the current leading the voltage. In a perfect cable, current is purely capacitive with zero resistive current. In reality, there is always a small amount of impurities in the dielectric; ageing and damage to the cable creates additional resistive paths and passes current to ground that can be identified and measured.

Since the cable is not perfect and there is resistive current, the phase angle is not 90 degrees, but something less. The amount of phase angle shift is proportional to the level of contamination and damage. Test equipment and associated software measures the shift in phase angle, displays the applied voltage, creates the graphical waveforms, and calculates the tan delta number, which provides excellent and easy to interpret results.

Tan Delta Performance Measurement

The cable to be tested would be de-energized, disconnected, isolated, and tested for absence of voltage. Anything that is part of the test circuit will affect the results, including items such as terminations and equipment that cannot be disconnected. The ends of the cable must be prepped, cleaned, and bagged to prevent corona and partial discharge issues that would adversely affect the test results. The voltage is then raised in steps (Figure 5), typically one-half the normal operating voltage starting at 0.5Uo, where the letter “U” represents a unit of voltage based on dividing the operating voltage into quarters. The test steps would equal 0.5Uo, 1.0Uo, 1.5Uo, and 2.0Uo. At 2.0Uo, the test voltage is held for a predetermined amount of time, typically 30–60 minutes, called withstand. This test is Go/No-Go. At the end of the time, the voltage is reduced to zero.

The tan delta numbers at the lower voltage are then compared to those at the higher voltage (Figure 6). If the insulation could be perfect, the line would be horizontal when the T/D numbers for each voltage are graphed. In less than perfect insulation, the tan delta numbers will increase as the voltage is increased.

IEEE Std. 400-2.2013, IEEE Guide for Field Testing of Shielded Power Cable Using Very Low Frequency (VLF)(less than 1 Hz) provides recommended test voltages, test times, and withstand hold times for different types of cables.

Interpreting Results

Understanding the results requires some experience. Results are divided into good, marginal, and deteriorated categories:

• Good indicates the cables test well and can be returned or placed into service for the first time.

• Marginal means tan delta values indicate areas of poor or damaged insulation in the cable system and further information on the cable is required to determine if the cable is safe to return or place in service.
• Deteriorated indicates the cable system has extremely high tan delta values indicating a significant problem and should not be returned to service.

In all cases, a previous benchmark test provides the easiest means to compare the latest test results. However, a benchmark is not necessary to obtain accurate assessment as can be observed in the actual test data sets seen in Figure 7, Figure 8, and Figure 9.

Just as in power factor, dissipation factor/tan delta is very linear for any applied test voltage. If the cable is damaged, the change in the capacitance and/or resistive watts loss will cause a change in the graphical analysis and will no longer be flat or linear; therefore, the sharper the climb (tip-up) with increased voltage, the greater the existing damage. A change in either the resistive element or the capacitance directly shifts the results, although it is most often a resistive shift rather than a capacitive shift that causes the changes in T/D results and curves. This may be due to contamination, water treeing, partial discharge, moisture, or poorly installed terminations to name a few. It is a test technician’s responsibility to recognize that a problem exists and determine whether the cables are safe to energize. With practice, most techs will be comfortable performing this much more reliable test technique.

Conclusion

This brief description of hipot versus tan/delta demonstrates that tan/delta incorporates the “good” of DC and AC hipot as well as eliminating many of the “bad” of both by incorporating low-frequency AC to minimize equipment power supply needs and providing test results that are easy to obtain and evaluate. The graphical analysis provided by the software provides uncomplicated identification of problems for both the technician and the end user or owner of the cables. Depending on the equipment manufacturer, the results can be sent by Bluetooth to a laptop; this makes the test much safer for the test technician and practically fully automated.

Rick Youngblood is responsible for utility contracts as well as in-house training for all technicians at Electrical Maintenance and Testing in Carmel, Indiana. After leaving active duty Air Force, Rick finished his engineering degree at Purdue University and worked for PSI Energy as a Project Engineer responsible for substation maintenance and testing. He finished his 25-year career with Duke Energy as Manager of Substation Services responsible for all aspects of substation maintenance and test. After retiring from Duke, Rick and his partner opened up a branch of American Electrical Testing in Indianapolis concentrating on utility business, where he earned his NETA Level 2 and Level 3 Technician. Rick retired a second time after seven years and closed the Indianapolis division of AET. Finding retirement unsatisfying, Rick became a Client Service Engineer for Doble Engineering, where he dealt with all utilities from a training perspective for the Great Lakes region for seven years.

For more information, please visit dissipation factor tester supplier.