Meet IDA: An interview with Megger’s Dr. Peter Werelius

Jill Duplessis - Global technical marketing manager and Editor

On September 27, 2018, IEEE-SA Standards Board approved IEEE C57.161-2018 – Guide for Dielectric Frequency Response test. This work follows that of other standards organizations, such as CIGRE’s TB 254, TB 414 and TB 445 – all technical brochures that address aspects of dielectric frequency response (DFR) testing.

A Dielectric Frequency Response (DFR) test is an insulation health assessment tool that is more discriminating than traditional dielectric screening tests such as power factor/dissipation factor. DFR testing can be used, for example, to assess the insulation health of transformers, bushings, instrument transformers and cables. In a complex insulation system such as that of a transformer, DFR testing provides a moisture assessment of the solid insulation, conductivity of the liquid insulation and information about the thermal behavior of the insulation system.

With increasing attention being paid to the DFR method, it seemed a very appropriate time for Megger’s Jill Duplessis to talk to Dr Peter Werelius who developed the first commercially available DFR test set for field use. Here’s what he had  to say.

JD: Thanks Peter. I’d like to begin with congratulations to you and your peers for your collaborative work on the newly published IEEE guide! Back in 1996, the first commercially available DFR instrument designed for field use, which is known today as the Megger IDAX, was introduced. As its developer, what advice would your “scientist self” of today give to your “scientist self” 22 years ago?

PW:  I am fortunate that my personal guiding principles have served the development of the DFR technology well. So I would offer to others my firsthand experience that being generous to my first and subsequent business partners (with whom I founded WaBtech and later Pax Diagnostics) and being honest and helpful to my former employer along the way as well as early customers has come back to me in unimaginably, positive ways. My partners, my employees at that time, my former employer GE Programma, Megger, early customers, and KTH (Royal Institute of Technology) have been very committed and supportive.

JD: I was really moved when I learned of the origin of the name “IDAX”. I understand that it is quite personal to you. Can you explain?

PW: Björn Bengtsson, who is now Björn Jernström, and I were the co-founders of WaBtech AB. When we discussed a name for our first product, we talked about a few requirements: preferably it should be a name as well as an acronym and ideally the acronym should work in Swedish and in English. My now grown up baby girl, born in 1995, is named Ida and besides being very personal to me, her name met our other requirements. In Swedish, it was IsolationsDiagnostikApparat, which was later changed to Isolations Diagnostik Analysator, and in English it was Insulation Diagnostic Apparatus and later Insulation Diagnostic Analyzer. When Pax Diagnostics took over the IDA from GE-Programma in 2006, the instrument’s name was modified slightly to IDAX (Insulation Diagnostic Analyzer by Pax). Both IDA and IDAX are pronounced as names, not acronyms.

JD: Can you tell us a little about the history of the IDAX instruments and how they have developed over the years since the first model  was introduced?

PW: The initial development of the DFR instrument was part of my Ph.D. project that began in early 1992 at KTH (Royal institute of Technology, Stockholm). In 1994, Björn Jernström joined me and helped me to turn my student set-up into a practical instrument for field use. Late in 1996, we sold the first commercial unit, IDA#1, to Vattenfall, Sweden. The second unit, IDA#2, which we delivered in spring 1997 to ABB Inc. in the USA, was already significantly changed from IDA#1, so that it would fit into a case about half the size of the previous one.  After Programma Electric acquired WaBtech in late 1998, the IDA 200 instrument was developed and released in 2000. IDAX 206 followed (2006) and IDAX 300 entered the market in 2008. At about this time, competitors started to appear. Since then, we have improved the hardware, software and digital signal processing algorithms. An important milestone was the introduction of the 2000V peak amplifier accessory in 2011.

JD: There has been some talk about the need to compensate DFR test results for ageing byproducts; otherwise, in some cases, moisture results may be misleading. For perceptive users, this “red herring” of sorts puts focus squarely back on the database.

Fundamentally, DFR analysis and the results it’s intended to provide, such as moisture content in solid insulation and oil conductivity, is based on fitting the measured DFR curve to its best match in a database of curves. The test software does this through mathematical modeling – adjusting parameters that affect the shape of the model curve until the software finds the best match. The parameters include: 

• Relation of solid (cellulose) vs. liquid (oil) insulation
• Moisture in solid insulation;
• Oil conductivity

I think that it is important for users to recognize that there is not one universal database shared by all instrument manufacturers. When you acquire a DFR instrument, you are selecting a database. The database determines the accuracy of moisture assessment results. If differences that are observed in moisture assessments using different DFR instruments (or in results from a DFR instrument as compared to other moisture assessment tools) are explained away as being due to some undefined compensation algorithm, this may rightly cause discerning users to grow uneasy.

How can a user go about validating an instrument manufacturer’s database? Also, is there a more pragmatic approach to take when a user has legitimate concerns that aging of the asset may be skewing moisture assessment results?

PW: The fact that aging byproducts, or more specifically low molecular weight acids, affect DFR results has been known by the industry for more than 10 years ( See: Dielectric response of mineral oil impregnated cellulose and the impact of aging, D.Linhjell et. al. IEEE Transaction of Dielectrics and Electrical Insulation, 2007). Our database and others are typically applied without any attempt to directly compensate the results.  One database attempts to compensate for the effect of aging products by using measured oil conductivity but in my opinion this approach is not successful. It was reported in the paper referenced above that low molecular weight acids influence DFR response, but high molecular weight acids, which also can affect oil conductivity, do not affect the DFR response.

In my opinion, the influence of low molecular weight acids is not a big problem. For dry insulation, the DFR interpretation will give dry insulation. If the insulation is moderately moist, say with moisture content of 2%, the level of low molecular weight acids is usually very low. The only case where low molecular weight acids may influence the moisture assessment is that of a very aged and wet transformer. However, acids are also very bad for the insulation, so a high content of acid with low molecular weight combined with what turns out to be an actual moisture content that is slightly less than estimated still requires actions to be planned.

Before investing, decide how to treat the transformer. Complementing analyses, including insulation liquid analyses, are usually performed  to get a more complete picture about the insulation condition.

JD: Some years ago, a DFR test on a healthy transformer may have taken well over one hour. Naturally, this was the source of some concern in the industry since adoption of a test requires that its value outweigh its cost – and the cost, of course, is greatly affected by the time the test takes to perform.

An early attempt to reduce DFR test time combined the preferred frequency domain spectroscopy (FDS) method, which has better immunity to interference, with the polarization and depolarization current (PDC) method by arranging for the instrument to switch to the use of DC voltage for measurement of polarization current (PC) at low frequencies. These are the frequencies where the FDS measurement is slow because it has to wait for the completion of several very time-consuming cycles of the applied test voltage. However, reliability suffered with this combined approach, especially in electrically noisy field environments, because the PDC method is particularly susceptible to error due to interference. Is there a way to perform DFR tests quickly without reverting to an approach that combines FDS with PDC?

PW: Any time that you can perform an entire measurement in only FDS mode, you have a better, more reliable measurement. The DC method has two major disadvantages. The first is fundamental and is that a DC current interference will appear as increased losses (power factor/dissipation factor), especially at lower frequencies. This may lead to under- or, more commonly, over-estimation of moisture content. In a true frequency domain measurement, DC current interference will not affect the measured losses. Secondly, you can look at a DC (PDC) measurement as applying all frequencies at the same time. This may shorten the measurement time but, by spreading the energy across all frequencies, it makes the method more sensitive to AC interference.

A true frequency domain method that speeds up measurement time is accomplished simply by applying several frequencies simultaneously, and extracting measurement data from the applied frequencies simultaneously. Of course, the voltage level of each frequency needs to be reduced, which makes the measurements a bit more sensitive to AC interference, but DC current interference will still not affect the measurements. This multi-frequency approach is an advance over the older approach of combining FDS with PDC, which is more sensitive to AC interference and also quite sensitive to DC interference.

The time savings of the multi-frequency method are significant. For example, using three frequencies simultaneously reduces measurement time by about 40%.

JD: Speaking of time, if I’m testing a three-winding transformer and I want to measure two insulation components concurrently (e.g., CHL and CLT), this requires an instrument with dual measuring channels. If an instrument claims to deliver DFR test results for two components in the same amount time it takes to test a single component, am I truly getting the same number of measured values per insulation component as I would if I had measured each insulation component separately?

PW: In some dual channel instruments, both channels share a single ammeter and, with these, either half as many measurements are made or no timesaving is realized. An instrument must have two ammeters if it is to perform two truly concurrent measurements.

JD: Most DFR test instruments provide a maximum 140 V RMS signal (200 V peak), primarily because this voltage is sufficient to provide an accurate moisture assessment of the solid insulation. In addition, the voltage source lends itself to a lightweight, easily transported instrument. However, in some instances, a 1400 V RMS is required to improve accuracy of the measurement – or even to acquire intelligible test results. Please share more about this.

PW: The most commonly used voltage level is about 140 V RMS and this is enough for measurement of the CHL-insulation in most conditions. However, in situations where there is a high level of interference or when measuring CH, CL, reactors, bushings and current transformers (CT’s), a 140V RMS signal does not provide a high enough measurement signal-to-noise ratio to get meaningful results. Using a higher test voltage (1,400 V RMS/2,000 V peak) in these instances will improve the measurement accuracy and is recommended. And, referring back to your earlier question, using a higher test voltage also makes it possible to successfully use multi-frequency measurements to shorten measurement times in medium interference conditions.

JD: No discussion about insulation is complete without addressing the important topic of temperature.  It is widely understood that the electrical characteristics of insulating materials change with temperature. Since diagnostic tests on insulation rely on detecting changes in the material(s)’s electrical characteristics, it is important that only data that has been obtained at the same temperature be compared. This way, any change in the measured electrical parameters can be attributed to a change in the state of the material and not a change in temperature. Since it is not practical to wait for same conditions every time data is needed, an alternate means to determine an insulation’s equivalent behavior at a baseline temperature of 20°C is extremely valuable. A lot of information about an insulation’s thermal behavior can be accessed from DFR test results. Can you explain a little more about this?

PW: The DFR interpretation used in IDAX –and in other instruments – is based on the so called master curve technique, where the DFR response of the impregnated cellulose material can be shifted to another temperature by simply changing the frequency axis based on the temperature difference while keeping the curve shape unchanged. The analysis is then performed at actual measurement temperature, e.g. 32°C.

However, network frequency (50/60 Hz) power factor measurements have traditionally been corrected to 20°C by using an approximate correction factor. It is well-known that this correction factor depends on insulation status, and in some cases, this introduces quite large errors. Fortunately, by using our knowledge of DFR interpretation and the fact that modern power factor instruments can sweep over a range of frequencies, we can now determine a much more accurate temperature correction factor by carrying out a limited frequency sweep on the actual test object.

JD: From what you’ve told us, Peter, it’s apparent that you were a true pioneer of DFR testing and that you’re still a very active innovator in the field. Is this a valid assessment?

PW: It’s a good summary, but it’s only fair to acknowledge that my early work on DFR testing built on work that had already been carried out by others, notably by A K Jonscher in the UK and by Professor Uno Gäfvert, Dr. Bo Nettelblad and Kenneth Johansson in Sweden. The big advance our research group at KTH, headed by Professor Roland Eriksson, and later my colleagues and I made was to take DFR testing out of the laboratory into the field. As you say, I have remained – and still remain – very closely involved with DFR testing and I am delighted to see how widely the techniques that my colleagues and I devised and developed are now being used. Of course, we no longer have the field to ourselves, but they do say that imitation is the sincerest form of flattery. With this in mind, I’m sincerely flattered by how many of our ideas other instrument manufacturers have now adopted!

JD: Thank you for your time, Peter.  I know you’re very busy but I’m sure your time has been well used and readers will be fascinated by the insights you’ve given us into DFR testing past and present.

PW: It’s a pleasure!