An Innovative Method for Detecting Transformer Air Leaks

Presented By:
Toni Mellin
Senja Leivo
Vaisala Oyj
TechCon 2023


Oxygen is one of the main components among heat, moisture, and acids, to cause degradation of the insulation paper, which leads to shortening the lifetime of the transformer. The source of oxygen in transformers is mostly ambient air and thus, it is common to use a membrane or rubber bag-sealed transformer design to protect it. Despite this, air leaks still do occur for various reasons, and it is important to detect them and to perform corrective actions to minimize paper degradation, extending the transformer’s lifetime. This paper presents an innovative total gas pressure (TGP) method, for detecting air ingress in its early stages. This is based on measuring the total gas pressure of all gases dissolved in oil, the main components of which are oxygen and nitrogen, which will show rising levels of TGP in case of an air leak. The paper includes a few case studies of transformers in different conditions and their total gas pressure behavior.


Transformers lifetime is greatly dependent on the condition of its solid insulation. Oxygen is one main contributor together with heat and moisture to cause premature aging of the solid insulation in the transformer windings, therefore shortening the residual life of the transformer by years [1]. This has led the industry to prefer sealed designs, which has increased their share of the power transformer market.

Increased use of sealed transformers has made the issue of air leaks more important and a primary source increasing oxygen and moisture content in new transformers. Air leaks into the transformer are often caused by the embrittlement of gaskets or the rubber bag of the conservator. Thus, prolonging a transformers operational life is dependent on confirming proper sealing against ambient air ingress and monitoring this becomes crucial to maximize the lifetime. Global power demand is rising and electrification is growing, which keeps growing the importance of transformers, as well as the need for new innovative solutions to monitor and manage these critical assets.

Prior art

The diagnostic value of gases in their gaseous phase or dissolved in the transformer oil and their behavior inside the transformer has been recognized long time ago. The invention of the Buchholz relay around the 1920s by M. Buchholz, has been the practical method of using the gas contents inside the transformer to indicate issues for a very long time and is still in active use [2]. The Buchholz relay can catch active, quickly developing faults from the fast gas rate or cumulating gases trapped in it, but unless air ingress happens extremely rapidly, the relay cannot indicate leaks in the transformer. Air ingress usually occurs slowly over time and the gases should have time to dissolve properly in the oil, thus the air ingress will not be detected by the Buchholz relay. Not that it would necessarily even be the intended and wanted manner of operation of the relay.

Methods to detect air leaks during transformer maintenance or commissioning exist, but that usually requires the transformer to be de-energized for the duration of the testing. For some time, Freon gases were used for leak testing [3], but since the unnecessary usage of Freons has been widely banned in many countries, this is not common practice anymore. The current practical way of leak testing is using pressure measurement either by over-pressurizing or under-pressurizing the tank to near-vacuum, either of which would have to be done while commissioning or during a maintenance operation while the oil is removed [4]. This means that there haven’t really been good practical methods to check for leaks while the transformer is energized until oxygen began to be used as an indicator of leaks in sealed transformers [5].

Transformer owners want to avoid oxygen and moisture getting into the transformer, as oxygen accelerates the oxidization of the oil and both of them accelerate the aging of the paper insulation and thus aging of the whole transformer [1]. The initial oxygen content of a transformer should be very low, and since there are no natural sources of oxygen inside a transformer all the oxygen comes from ambient air through a free-breathing conservator or leaks in through the sealing of the tank, through other connections, or even possible manufacturing and assembly errors. Oxygen concentration does not always increase inside the transformer and dissolve into the transformer oil. Oil oxidation and insulation aging can also consume oxygen, thus decreasing the concentration [6]. Traditional method of measuring oxygen has been in a laboratory from the oil samples together with the fault gases as part of standard dissolved gas analysis (DGA) to confirm that the transformers maintain their sealing against ambient air. Oxygen can alternatively be measured directly with multi-gas online DGA monitors, which can tell about the possible leaks, but as oxygen can also decrease due to aging mechanisms, the interpretation of the oxygen readings is not always clear or straightforward. When performing a laboratory analysis of oxygen concentration, from the point an oil sample is taken and shipped to a laboratory, there are multiple occasions for the sample to be contaminated with air, which makes the interpretation even more difficult. On the online monitoring side, the biggest drawback is that oxygen sensors have lifetime limitations, which most often means that they need to be replaced during the lifetime of the transformer and the monitor, increasing the overall maintenance need.

Total Gas Pressure

A new method for continuous online monitoring

A new method, called total gas pressure (TGP), has been introduced for real-time detection of any air leaks in sealed transformers. In this method, all gases dissolved in the oil are extracted under vacuum, and the pressure – i.e. the sum of partial pressures of all gases ± is measured in controlled, stable conditions. TGP measurement has been implemented as an additional complementary measurement parameter to a multi-gas DGA monitor. In practice, measuring TGP online must be done in tandem with a hermetically sealed oil sampling mechanism, that uses vacuum extraction to extract all gases, especially the dominating nitrogen and oxygen. This does mean that the method is not directly implementable to all multi-gas DGA monitoring systems, as not all of them utilize vacuum extraction.

Pressure sensors are robust and reliable, with long lifetimes, which gives them an edge in maintainability over oxygen sensors for the same purpose. As implementing the described method requires relatively intricate mechanical operation to create a vacuum for the oil and gases, it wouldn’t be economically feasible as a standalone measurement device without the other measurement parameters a multi-gas DGA monitor can provide.

In a well-sealed transformer with degassed oil, the pressure of dissolved gases is and will remain low, well below atmospheric pressure around 1000 hPa (~14.5 psia). If air leaks into the tank, it dissolves in the oil, and the total gas pressure starts to increase. The dominating gases in TGP are nitrogen and oxygen, the portion of fault gases is negligible. For example, 1000 ppm of carbon dioxide (CO2) corresponds to only 1 hPa (~0.0145 psia) in pressure. Even if all the oxygen is consumed in aging reactions inside the transformer, nitrogen would increase the TGP value in case of continuous air ingress. Real-time monitoring with a hermetic system eliminates the risk of oil contamination with air during sampling. Interpreting the rise of a pressure value in a sealed system is intuitive and understandable without further expertise. If TGP indicates a sealing issue, the problem can be addressed during the transformer’s next service break to minimize further degradation of the transformer insulation due to oxygen. Since it is a continuous online condition monitoring method, the potential for detecting leaks is the highest possible as you continuously receive new condition data every hour, and setting reasonable alarm limits can enable the asset manager to be notified quickly if there are changes.

Uncertainty compared to lab work

Taking an oil sample intended for dissolved gas analysis (DGA) needs to be done with extreme care, as contamination with ambient air and possible loss of fault gases during sampling is always a significant element to the total uncertainty. This applies specifically to oxygen which is abundant in ambient air. When taking an oil sample from a sealed transformer, the sample’s oxygen levels are low, and it is nearly impossible to keep oxygen in the air from contaminating the sample. There are many steps in the process where samples may be contaminated with air, including at the site during oil sampling, during transportation due to temperature or pressure variations, or at the laboratory while the oil sample is transferred from its initial vessel to a gas chromatography vial. The end result may even be that the sample is slightly contaminated or changed during all of the mentioned parts. A contaminated sample may lead to a false interpretation – of air ingress in the transformer tank or to such high variance between readings that it becomes a difficult task to diagnose anything based on the data.

It is also typical that O2 and N2 concentration laboratory results vary significantly between the periodically taken samples, making it hard for trending inspection or interpretation whether air ingress has occurred or not. The variation is mainly due to pressure changes, which are due to volume changes that derive from the normal temperature changes inside the transformer, causing thermal expansion in the insulating oil [7]. Temperature changes themselves are consequently caused by ambient temperature changes and loading variation. All of which makes it hard to have consistent laboratory measurements for oxygen levels, which again makes it difficult to use the results for meaningful analysis. These are the reasons why a sealed continuously operating automatic condition monitoring always has great advantages over manual sampling and laboratory work, as error sources are minimized while the sampling is consistent and frequent. In laboratory reports, the parameter equivalent to TGP may be called ‘total gas content’ or ‘total partial pressure’. It is not commonly expressed in all laboratory reports, and because of the reasons previously mentioned, the laboratory-indicated value may not be the most effective in indicating leaks. To further illustrate this point of view, laboratory-provided measurement results of oxygen and nitrogen are presented in Figure 1. We can see the amount of variance in oxygen and nitrogen measurements between samples from the data. It is difficult to see trend development reliably from the data, which is not uncommon with oxygen and nitrogen results.

Figure 1 Oxygen and nitrogen concentration in oil measured from DGA samples of a power transformer during 1991-2020

In the context of laboratory work and DGA diagnosis, the IEEE standard C57.104 2019 revision categorizes transformers typical values into several subcategories based on transformer age and oxygen-to-nitrogen ratio (O2/N2). [8] The reason for splitting transformer DGA typical values to different categories based on oxygen-to-nitrogen ratio, is to distinguish sealed and free-breathing transformers from each other, which was determined to have an effect on the gas concentrations generated and lost through conservator. TGP method can easily distinguish between sealed and free-breathing transformers. In a free-breathing transformer, TGP should be close to atmospheric pressure around 950 « 1070 hPa (13.49 « 15.52 psia), but of course, this depends on geographical location and altitude, so one should look up natural barometric pressure variation in the area to confirm the correct levels. TGP values well below atmospheric pressure is most certainly an indication of a sealed transformer. If a sealed transformer is at atmospheric pressure level, it simply has leaked so much that it has saturated to ambient air. TGP values well above atmospheric pressure is an indication of a nitrogen-blanketed transformer, which is also sealed, but instead of ambient air leaking in, it will more likely leak nitrogen out. At some point, a leak will cause the nitrogen gas tank to run out of gas and the internal pressure along with TGP will reduce. Nitrogen blanketed transformer leaks can also cause oxygen and air to leak in, but the nitrogen blanket should at least slow down the effect.


Sealed transformer

TGP readings measured in real-time from the oil of a new sealed power transformer are presented in Figure 2. We can see that the TGP levels stay very low during its first two years. Right after commissioning, some gases are released into the oil, likely from the oil-impregnated paper raising the TGP slightly, but it stabilizes to a low level. The data in the figure was averaged to show the trend better, but we can still see that slight variations are happening. The average increase in TGP after stabilization is only 7 hPa/year (~0.10 psi/year). At that rate, it would take 130 years to reach equilibrium with the atmospheric pressure of 1000 hPa (~14.5 psia). In the real world, nothing usually is completely gas-tight and without any minimal leaks or gas diffusion, so we have to look at the rates of change critically. Even NASA has allowed a maximum leak rate of 3 psi/minute (~20.7 hPa/minute) in their pre-checks for space suits to be used for EVA (Extra Vehicular Activity) i.e. spacewalks [9]. All data indicates that the transformer in question is well sealed and not leaking to a degree requiring actions.

Figure 2 Total gas pressure data of a membrane sealed power transformer during its first two years after commissioning

Another sealed GSU transformer’s TGP data is presented in Figure 3 without any additional averaging. We can see a slight change in TGP during June 2021, but it stabilizes later. Unfortunately, we don’t have further details on this transformer or its history, but it looks to be well-sealed based on the data. It would still be good to follow if there would be further developments in TGP later during the lifetime of this transformer.

Figure 3 Total gas pressure data from a membrane sealed GSU October 2020-January 2022

Sealed transformer leaking

Averaged TGP data of a sealed but obviously leaking GSU after its degassing is shown in Figure 4. Right after degassing, there is a roughly 30 hPa (~0.44 psi/year) quick increase, probably due to the gas in the oil-impregnated paper. After that, there is a steady increase with a significantly higher rate than the non-leaking examples presented previously in Figure 2 and Figure 3. In roughly two years, the TGP has changed about 120 hPa (~1.74 psi) ± a rate of change of 60 hPa/year (~0.87 psi/year). Keeping the same rate of increase, it would take roughly 15 years for the TGP to reach atmospheric pressure. It is rather obvious that there is more oxygen and possibly also moisture coming into the transformer all the time, which may reduce the lifetime of the transformer. At this point, it would be important to weigh the lifetime expectancy of the transformer and the costs of investigating and repairing the leakage or having other mitigation actions. Because details matter in these decisions, it is not always reasonable or economically relevant to try and fix the leakage, but the decision should be thoroughly discussed with transformer experts.

Figure 4 Total gas pressure data of a sealed GSU right after it was degassed and the following two years

Free-breathing transformer

Averaged TGP measurement data from a free-breathing transformer during and after degassing of its oil are presented in Figure 5. The degassing drops the TGP from ambient pressure level to a very low level after the vacuum treatment of the oil. We can see that the pressure level keeps slowly but steadily rising towards the ambient pressure level as air dissolves slowly into the oil. This transformer is running at a stable and relatively low load; thus, there is hardly any actual breathing, but the increase in TGP is probably only through the diffusion of air in the conservator through the connecting pipe into the tank. We can see that it takes a rather long time to fully reach ambient air level. The TGP changes around 500 hPa (~7.25 psi) in half a year, of which a calculated rate of change would be around 1000 hPa/year (~14.5 psi/year). Thus, extremely severe leaks and free breathing transformers are easy to distinguish.

Figure 5 Total gas pressure data from a free-breathing 100 MVA power transformer before and after repair and degassing.


In the paper, a new method for detecting air leaks into sealed transformers was described. The total gas pressure (TGP) is measured with a pressure sensor and hermetic gas extraction system included as a part of a multi-gas DGA monitor. Real-world use and measurement of TGP was presented with measurement results from the continuous online monitoring system. Compared to oxygen concentration, which is the traditional leak indicator, TGP is a straightforward and intuitive parameter; the pressure measurement will immediately show decision-makers if there’s an air leak – no specialist opinion nor interpretation is needed. The authors presented that a good method to analyze the leaks current severity is to calculate the most representative current leak rate with TGP. This leak rate then can be converted to time until reaching saturation to atmospheric pressure, which is an intuitive value to evaluate against the remaining service lifetime of the transformer. If the leak is so severe that it will likely reach close to ambient or ambient level before the transformer is intended to be retired, the owner should consider acting and trying to locate the leak and prevent the acceleration of aging. The comparison between a leaking transformer and non-leaking presented in Figure 6, shows that leaks in a transformer can be observed using TGP even without using the discussed leak rate per year calculation, just by comparing the TGP values over time. All of which should make it an efficient, yet simple to use new tool for transformer owners to detect issues during operation and give early indication of leak issues, so that actions can be taken to extend the life expectancy of a transformer.

Figure 6 Orange line TGP comparison between a leaking sealed transformer and dark blue line a well-sealed transformer with a leak


[1] W. A. Cigré, Thermal performance of transformers, vol. Technical Brochure vol. 393, 2009.
[2] E. Gross, “Simplicity in Transformer Protection,” Electrical Engineering 66(6), pp. 564-569, 1947.
[3] H. Wagner, “Five Years’ Experience with Transformer Total Combustible Gas Fault Detector Tests.,” IEEE Transactions on Power Apparatus and Systems, vol. 84, no. 8, pp. 700-706, 1965.
[4] W. A. Cigré, On-Site Assembly, On-Site Rebuild, and On-Site High Voltage Testing of Power Transformers, vol. Technical Brochure vol. 857, 2021.
[5] E. G. X. Zhang, “Asset-management of transformers based on condition monitoring and standard diagnosis,” IEEE Electrical Insulation Magazine, vol. 24, no. 4, pp. 26-40, 2008.
[6] D. Cigré WG, Advances in DGA interpretation, vol. Technical Brochure vol. 771, 2019.
[7] T. Oomnen, “Gas pressure calculations for sealed transformers under varying load conditions,” IEEE Transactions on Power Apparatus and Systems, no. 5, pp. 1278-1284, 1983.
[8] IEEE, “Guide for the Interpretation of Gases Generated in Mineral Oil-Immersed Transformers,” IEEE Std C57.104-2019, 2019.
[9] National Aeronautics and Space Administration, Mission Operations Directorate EVA, Robotics, and Crew Systems Operations Division, “EVA Checklist,” March 2005. [Online]. Available: https://www.nasa.gov/centers/johnson/pdf/492872main_EVA_G_H_20.pdf. [Accessed 1 12 2022].

Join our email list

We use cookies to give you the best online experience. By using this website you agree with our cookie policy.