Author:

Maik Koch, Michael Krueger and Markus Puetter

### Abstract

This paper discusses and compares dielectric spectroscopy in time and frequency domain to traditional dielectric measurements such as dielectric dissipation factor at power frequency and polarization index.

Condition assessment of oil-paper-insulated power transformers, particularly of water content, is becoming increasingly important for aged transformers and also for quality control of new transformers in the factory. The demand for sophisticated and at the same time reliable and easy to use diagnostic techniques drove the development of dielectric response methods. The first approach, called Recovery Voltage Method, is now outdated. The newer methods, Polarization and Depolarization Currents and Frequency Domain Spectroscopy, have proven their suitability for transformer diagnostics and are now frequently used.

This paper describes in detail the interpretation of dielectric response measurements in frequency domain. Since especially the low frequencies reflect water concentration, their measurement is of outmost importance for reliable data analysis. Beside a frequency sweep, the response of a dielectric to a voltage sweep is experimentally investigated and discussed.

Special focus is given on a comparison of the currently available dielectric spectroscopy methods to traditional measurement techniques like dielectric dissipation factor tests at power frequency and 0.1 Hz, dielectric adsorption ratio and the polarization index. The traditional methods suffer from a limited time or frequency range which impedes the discrimination of specific dielectric properties. If for example increased losses appear, it is impossible to discriminate whether they are caused by the insulating oil or the cellulose insulation.

Examples of Dielectric Spectroscopy transformer measurements illustrate the advantages of analysis in a wide frequency range.

**Introduction to Electrical Insulation Tests**

Power transformers are the most expensive links in the chain of transmission network for electrical energy connecting generation to utilization. Nowadays three factors stress them: the increased demand for electrical energy, whereas the average age of transformers increases as well and maintenance strategies change forced by the cost pressure in liberalized energy markets. Electric utilities try to suspend the investment in new devices and shift maintenance from time based to condition based strategies. To realize this strategy the demand for new diagnostic methods arises, methods which reliably evaluate the actual condition of the equipment.

** Transformer Aging as a Worldwide Issue. **In all developed countries large proportions of the transformer fleets are approaching the end of their design life. For example in the United States the average age of large electrical transformers is 35 years; the design life is 40 to 50 years [1]. The authors conclude: “The data we see indicates that of the 110’000+ large electrical transformers installed in the United States, up to 2 % will fail this year; that is 2200 transformers. … Utilities need quality, reliable, maintenance free solutions that provide timely information on the transformers health.” Economical interests boost this statement and it mirrors a worldwide trend.

In Germany, a recent investigation found the average age of transformers for 110 kV rated voltage is 31 years, for 220 kV 34 years and for 380 kV 30 years, [2]. Figure 1 illustrates failures of power transformers classified by voltage level and inducing subsystem and the dependence on operating time. It is important to precisely know the error rate of the equipment and develop appropriate diagnostic procedures.

** Electrical insulation tests **are able to detect probable failures of the active component, mainly the insulation system and of the bushings. Therefore, researchers looked for decades for methods to find out the condition and the aging state of the insulation system. The dissipation or power factor has been used for decades to evaluate the losses in insulation materials. Figure 2 illustrates the equivalent circuit for losses in insulation materials and the corresponding vector diagram. Any solid or liquid insulation can be modeled by a capacitance C, representing the “ideal” behavior of insulation, and a resistor R, representing the electrical losses. The dissipation factor tan d indicates the quality of insulation materials by the tangent of the ratio of resistive current I

_{R}to capacitive current I

_{C}. The power factor is the cosine of the ratio of resistive current I

_{R}to total current I. For small angles, dissipation and power factor are identical.

While the conventional tests like power factor, dissipation factor and polarization index look on narrow frequency points or areas, the newer dielectric spectroscopy methods conduct information over a very wide frequency range and thus enable for the discrimination between different effects and more dependable information about the assed condition. This article describes the current state of knowledge about the frequency and voltage dependent characteristics of oil-paper-insulations; it then lists the available measurement methods with their specific advantages and compares them at practical study cases.

**Dielectric Behavior of Oil-Paper-Insulations**

**Response to a Frequency Sweep**

Applying a frequency sweep over a very wide range means to measure the dielectric response. For oil-paper-insulated power transformers, the dielectric response consists of three components: The response of the cellulose insulation (paper, pressboard), the response of the oil and the interfacial polarization effect. Moisture, temperature, insulation geometry, oil conductivity and conductive aging by-products influence the dielectric response.

** Superposition of Dielectric Properties. **Figure 3 (left) displays the dissipation factor of only pressboard with a moisture content of 1, 2 and 3 % measured at 20°C. Figure 3 (right) shows the dissipation factor of solely oil with a conductivity of 1 pS/m measured at 20°C. Note, that at low frequencies the losses are much higher compared to pressboard and that the dissipation factor is just a line with a slope of – 20 dB / decade, which is due to the fact that insulation oil shows conductive behavior nearly without polarization processes.

The dielectric properties of pressboard and oil are superimposed together with the interfacial polarization process. Interfacial polarization is typical for non-homogeneous dielectrics with different permittivity or conductivity. Here charge carriers such as ions accumulate at the interfaces, forming clouds with a dipole-like behavior. This kind of polarization is effective only below some ten Hertz.

Figure 4 displays the dissipation factor of pressboard having 1 % moisture content and oil together with the interfacial polarization effect (insulation geometry). The frequency range of 1000-10 Hz is dominated by the cellulose insulation, however also the measurement cables and the connection technique influence this region. Oil conductivity causes the steep slope at 1-0.01 Hz. Dissolved conductive aging by-products, soot and high molecular weight acids increase the oil conductivity and thus influence this area. The interfacial polarization (insulation geometry, ratio of oil to pressboard) determines the local maximum or “hump” at

0.003 Hz. The higher the ratio of oil to pressboard, the more dominating is this effect. Finally, the properties of the cellulose appear again at the frequencies below 0.0005 Hz, here reflecting moisture, the manufacturing process and low molecular weight acids. The frequency limits correspond to Figure 4, but will vary in a wide range with moisture, oil conductivity, insulation geometry, temperature and amount of conductive aging by-products.

** The Effect of Moisture. **Moisture particularly increases the losses in the low frequency range of the dielectric response of pressboard. Thus, the point of inflexion on the left hand side of the area dominated by insulation geometry is required for a reliable moisture determination.

Since pressboard also dominates the high frequency area above 10 Hz in Figure 4, it might appear that it is sufficient to measure this frequency range. However, moisture especially affects the low frequency branch of the dissipation factor curve. Figure 3 illustrates, that the high frequency part of the dissipation factor curve is very similar for various moisture contents, but the low frequency part differs. Consequently, if the measurement range is restricted to the high frequencies, the accuracy of water determination will be very low allowing only for a rough discrimination between wet and dry.

With increasing moisture content and oil conductivity, the curve shifts toward higher frequencies, but the shape remains similar. Figure 5 (left) depicts the dissipation factor over frequency for 3 % moisture content and 10 pS/m oil conductivity at 20°C insulation temperature.

** The Effect of Temperature. **Figure 5 (right) illustrates the influence of temperature on the same insulation system. At 50°C the losses of pressboard along with the oil conductivity increase while the shape of the curve remains similar.

Conclusively, a frequency sweep of dissipation factor of oil-paper-insulations as depicted in Figure 4 and Figure 5 provides information about the pressboard, the oil and the interfacial polarization effect.

**Response to a Voltage Sweep**

Applying a voltage sweep on oil-paper-insulations is an old diagnostic procedure and also known as “tip-up test”. This variation of the conventional capacitance and dissipation factor measurement makes use of the fact that the power or dissipation factor will change with applied AC voltage. Here the test voltage is increased from e.g. 6 kV to 36 kV and the dissipation factor recorded. The frequency is kept constant at e.g. 0.1 Hz. According to the common interpretation, the magnitude of change indicates aging of the insulation.

To investigate what insulation properties are reflected by a voltage sweep, a large insulation model called “Pancake Model” was used, [3]. The model consists of eight pancake shaped coils with oil ducts between them. The ratio of barriers and spacers to oil ranged from 15 to 90 %, simulating the main insulation of different transformer geometries. Service-aged transformer oil (conductivity 16.5 pS/m) filled the tank. The moisture content in cellulose was 1.1 %, measured at paper and pressboard samples.

Figure 6 (left) displays the dissipation factor at 0.1 Hz as a function of voltage for various ratios of oil to pressboard; that is 15-90 %. From the results it becomes obvious, that a high amount of oil increases the dissipation factor. A measurement of dissipation factor on a single frequency point will reflect not only the material condition (e.g. aging) but also the insulation geometry. The diagram further depicts the voltage dependence of the dissipation factor. It becomes obvious, that a high amount of oil increases the voltage dependence. Figure 6 (right) proves this assumption. Here the dissipation factor of oil alone is depicted as a function of the applied field strength; the higher the field strength, the lower the dissipation factor. In science, this effect is known as Garton effect [4]. The dissipation factor of insulation oil itself depends on conductive aging by-products such as soot, acids and moisture [5].

Conclusively, a voltage sweep of dissipation factor at a power transformer mainly reflects the condition and voltage dependence of the oil in conjunction with the insulation geometry.

**(To be continued)**