Tan delta testing: variable frequency method

Author: Jill Duplessis, Global technical marketing manager

Southern California Edison (SCE) also makes extensive use of narrowband dielectric frequency response (NB DR) testing, which it describes as “preferred … over conventional line-frequency testing in that anomalies such as moderately wet insulation may not be detectable using conventional tests.” This article looks in more detail at variable frequency test techniques, such as NB DFR tests, and the benefits they offer.

Variable frequency tan delta (narrow-band dielectric frequency response, NB DFR)

A variable frequency tan delta 1 test is an expansion of a traditional tan delta test, whereby tan δ tests are performed on every insulation component (e.g., CH, CHL, and CL) at multiple frequencies (e.g., between 1 – 500 Hz) including line frequency. Ultimately, the test is a (narrowband) dielectric frequency response measurement.

The advantages of a narrowband DFR test are noteworthy, including:

„„♦  Provides earlier indication of a problem in the dielectric.

„„♦  Distinguishes between the “Tan delta look alikes” [e.g., separates the case whereby a 0.3% line frequency tan delta is truly acceptable (0.5% water content) from the case whereby a 0.3% line frequency tan delta is hiding increasing moisture (2.0% water content)].

„„♦  Allows the possibility of determining the temperature-corrected tan delta at 20 °C for the insulation system based on its actual condition and not from standard tables (ITC, individual temperature correction). 

One benefit of performing Tan delta tests at multiple frequencies is that it provides context for the line frequency tan delta measurement. Knowledge of the behaviour of the dielectric’s response at points on either side of the line frequency measurement provides a better sense of whether or not the 50/60 Hz tan delta test result is representative of a healthy state for that particular insulation system, and also whether an observed 50/60 Hz tan delta change from previous needs to be investigated immediately or not. Ideally, the slope of the dielectric response (i.e., a line fitted to the multiple power factor measurement points) is positive through the line frequency. This indicates low losses of the system.

Individual Temperature Correction (ITC)

While a line frequency power factor test is not acutely sensitive to an emerging dielectric problem, it is sensitive to temperature. For example, it is generally expected that a power factor measurement at a top oil temperature of 30 °C will be higher than a power factor measurement on the same insulation component at 25 °C simply because of the influence of temperature. Therefore, it is important to compensate for any variances in temperature between tests if one is to trend test data and trust that a change in tan delta is truly due to a change in the insulation system’s condition. This temperature dependency variable is removed by correcting all tan delta test results to their equivalent 20 °C value.

For years, the industry has relied upon a couple of curves to correct for the temperature dependency of all transformers: whether new, service-aged, lightly loaded, over-loaded, clean, or contaminated, etc. But, while generic correction factors were available in IEEE standard C57.12.90-2006, section 10.10.5, they were subsequently removed in C57.12.90-2010 with the following note: “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.”

A NB DFR test allows for the determination of an insulation system’s unique, or individual temperature correction (ITC). This is significant as testing has revealed that not only does every transformer exhibit unique sensitivity to temperature and require individual temperature compensation, but, over its life, the temperature dependency of a transformer can change. Generally, as insulation deteriorates an increase in temperature causes tan delta to increase dramatically.

The ITC method is based on the fact that a tan delta measurement, at a certain temperature and frequency, corresponds to a measurement made at a different temperature and frequency. By measuring power factor at different frequencies [1-500 Hz] and at any given insulation temperature, one can determine power factor at any temperature [5 - 50 °C] and at nominal frequency.

Dielectric Frequency Response (DFR)

A “classic” dielectric frequency response (DFR) test is similar in concept to the NB DFR test but includes more power factor measurements over a wider frequency range (typically 1 mHz – 1000 Hz). A DFR test measurement at a given temperature is compared to a database of curves through an analysis tool, which ultimately delivers the additional benefits of:

„„♦  Providing an accurate assessment of moisture content in the solid insulation of the transformer based on international standards and guides

„„♦  Providing an estimation of oil conductivity

„„♦  Going the furthest of the three methods towards identifying the contaminant by assessing the two major sources of losses above (moisture and oil conductivity, respectively).

DFR testing is typically performed solely on the interwinding insulation system(s) of a transformer (CHL) since this is where the configuration is most well defined and least influenced by other factors, e.g., bushings, creep, LTC and more. This is in contrast to the narrowband version of the test in which all three insulation components of a two-winding transformer (CH, CHL and CL) are routinely measured in conjunction with a traditional tan delta measurement.

Fig.1: Typical dielectric response curve

A typical tan delta plotted versus frequency is given in Fig. 1. Moisture influences the low and high frequency areas. The pseudo-linear, middle section of the curve reflects oil conductivity. Insulation geometry conditions determine the “knee points”, which are located to the left and right side of the steep gradient. As temperature increases, the dielectric response curve shifts to the right. Therefore, the influence of temperature needs to be accounted for so the tester simply must input this value.

Moisture determination is based on a comparison of the transformer’s measured response to a modelled dielectric response. The insulation model, which represents volume fractions of the insulation system components, is the internationally recognized X-Y model described in guides such as CIGRE TB 254 and 414.

The insulation structure of the power transformer is represented by the relative number of spacers (sticks) and barriers in the cooling duct, as shown in Fig. 2. Parameter X is defined as the ratio of the sum of all barriers in the duct, lumped together, and divided by the duct width. The spacer coverage Y is defined as the total width of all the spacers divided by the total length of the periphery of the duct.

Fig. 2: XY model of a transformer

The permittivity of oil (ɛoil), spacers (ɛspacers), and barriers (ɛbarriers), are complex functions of both frequency and temperature. The equivalent permittivity of the XY model is given by:

Using DFR for moisture determination is based on the comparison of the transformer’s measured dielectric response to a modelled dielectric response. An analysing algorithm rearranges the modelled dielectric response and delivers a new modelled curve (for which the moisture content and oil conductivity are known) that reflects the measured transformer. Results are presented as moisture content and oil conductivity for the transformer.

Fig. 3: DFR moisture analysis of transformer insulation

An example is shown in Fig. 3. The transformer is a 20 MVA distribution transformer. Moisture content of cellulose and oil conductivity are obtained by varying the parameters in the model curve to the measured dielectric frequency response at a certain insulation temperature.

Temperature is an important test parameter that must be provided by the user as an input in the software. The winding temperature or top oil temperature may be used as a close approximation to the insulation temperature. DFR measurements on a distribution transformer at various temperatures are shown in Fig. 4. While the influence of temperature on the dielectric response is clear (shifting the curve to the right as the temperature increases), the moisture analysis tool is still able to identify the correct value of moisture in solid insulation independent of insulation temperature.

Fig. 4: DFR measurement and analysis at various temperatures

An additional note is DFR’s ability to distinguish between moisture and oil conductivity as sources of increasing losses. Fig. 5 provides DFR test results for three different transformers, all of which have identical tan delta test results at 50 Hz (approximately 0.5%) but each with very different problems.

Fig. 5: DFR analysis for three transformers

DFR is an indirect method for assessing moisture in solid insulation. There are no practical ways to directly measure moisture in transformer paper insulation so most available tools utilize indirect measuring methods, whereby properties of insulation that can be related to moisture content are measured.

The indirect methods that have been traditionally applied in the industry to assess water contamination of the paper insulation, such as moisture in oil measurements and use of equilibrium charts, only provide accurate assessments if moisture equilibrium has been achieved. However, during the normal operation of a transformer, the temperature inside the transformer varies throughout the day, so moisture equilibrium between paper and oil will rarely be attained since the time constants of thermal and moisture dynamic processes are very different. Due to the inaccuracies associated with most other methods, the dielectric frequency response (DFR) method has emerged as an attractive alternative. This electrical test method (based on models) is a non-intrusive, very reliable test with high repeatability. There is no need to wait for equilibrium, there are no inaccuracies due to the sampling and handling of oil, and DFR testing can be performed as part of the suite of electrical tests planned during a maintenance outage.

For decades, the industry has relied on a simple,
line frequency, tan delta measurement to determine dielectric health. Unfortunately, this test method has a lot of shortcomings. Experiences abound throughout the industry demonstrating that a traditional tan delta test is not sufficiently sensitive to changing levels of contamination. This forces the tester to take any changes in power factor seriously and trending has emerged as the best approach for analysing the results. The problem here is filtering all the test variables that can influence the power factor measurement so that the results are truly representative of the dielectric under test. A significant variable is temperature and the conventional methods to normalise every test result to a common base for comparison  are inaccurate. Consequently, problems are often missed, but an equal worry is that hyper-vigilance can lead to too many expensive false alarms.

Dielectric response methods have greatly advanced the understanding of dielectric behaviour and revealed more efficient means by which to uncover and identify dielectric problems. The industry is now incorporating dielectric frequency response methods [narrow-band (NB DFR) and “classic” (DFR)] into its test programs; the reward is heightened awareness about asset condition and better planning opportunities. These methods enable the determination of a dielectric’s unique and at times changing sensitivity to temperature and provide the tester with the means to accurately correct power factor results to a 20° C base.

Narrow-band DFR measurements, confirm when seemingly good power factor values actually are good, and reveal when they are not, thereby allowing earlier detection of problems. DFR tests provide highly accurate moisture in solid insulation and oil conductivity assessments.

For more information on ND DFR test equipment click here 

For more information on "classic" DFR test equipment click here 

1 or power factor . For simplicity of text, "tan delta" substitutes for power factor/tan delta throughout the article.