Things to Consider when Retro-filling Transformers with Natural Esters

Presented By:
Kerry Williams
K-BIK Power Pty Ltd
& Brian Sparling, SMIEEE
Dynamic Ratings
TechCon 2018


The question of whether a mineral oil immersed transformer can be retro-filled with a natural ester fluid has been answered with numerous units in service operating relatively normally. There are however several questions that have not been answered and this paper provides a few more answers and advice on what needs to be considered when changing fluids. It is not just a case of deciding to change fluids and then being happy that you have resolved a potential environmental or fire risk issue. There are many additional things to consider which can have a profound impact on the operation, performance and monitoring of a retro-filled transformer.

The University of Queensland in Brisbane QLD, has invested in a transformer research centre where the transformer has undergone a refurbishment and changing of the insulating fluid from mineral oil to a natural ester. During the refurbishment, the transformer was re-tested in the factory and a temperature rise test was performed. The results are discussed in this paper and the outcomes may surprise some readers who believe that simply changing fluids will give an increased transformer rating due to the cooling and moisture absorption capabilities of natural esters.

The paper also discusses the peripheral issues associated with data management, monitoring devices, field equipment and identification of a unit that has been retro-filled. These peripherals can quickly add costs to a project or, if forgotten, create serious issues for asset managers if not correctly recorded and identified. This is a holistic view of what is involved in changing the fluids and how an organization needs to be very aware that such a change is not limited to only the cost of the new fluid.

Australasian Transformer Innovation Centre


The Australasian Transformer Innovation Centre (TIC) is located within the University of Queensland and has devoted a facility for transformer research and development. The University has installed a 486 kVA, 22/4.5/0.415 kV, three-phase 50 Hz, transformer, built with windings in the classical core form and disc and layer winding configuration. The original transformer was built in 2006 as an ONAN/ ONAF/OFAF unit. In the latter part of 2016 it was refurbished, re-processed and retro-filled with natural ester (FR3) by the original OEM, Wilson Transformers, Glen Waverly AU, into a unit that is now KNAN/KNAF/KFAF.

The objective of the TIC is to carry out numerous research projects to understand the performance characteristics of this unit initially from a thermal rating point of view, and any other observable difference with the change in dielectric fluid.

The transformer had a series of upgrades made to the monitoring systems, and new moisture sensors installed, for the purposes of not only monitoring, but includes on board acquisition and archiving of the data measured and information gleaned from the data, into useful transformer information.

This paper will outline these additions, and analyze the heat run outcome with the natural ester in respect to the original heat run carried out.

History of the Unit

The original transformer was purposely built for the Monash University Transformer Life Management Centre in Victoria. During its time at Monash the transformer was subjected to numerous load tests that included overloading to 1.5 p.u. for continuous periods of several days. The unit underwent many high voltage electrical tests and was primarily used for research and student education. When Monash University changed strategic focus, the unit was acquired by the University of Queensland (UQ).

Location of fiber optic temperature probes

The unit was originally built with 16 fiber optic temperature probes inserted into the windings. Table 1 provides a quick overview of the location of each probe within the windings and Figure 1 shows the fiber probes being inserted into the windings during original manufacture. The unit was fitted with 5 Vaisala moisture probes to allow detection of the moisture content at varying points throughout the transformer. These temperature and moisture probes along with magnetic oil flow meters allow researchers to assess the changes in moisture throughout the transformer under different loading conditions. Additionally, there some PT100 temperature sensors fitted in the top lid of the transformer to allow a correlation between the winding probes and the oil probes.

fiber optic temperature probes inserted in MV Winding
HV Windings Fiber Optic Temperature Probes

The transformer is not currently fitted with an on-line DGA monitor as there appears to be no suitable device available that is correctly calibrated for ester fluids. Manufacturers are working toward a suitable device which would allow units of this type to be continuously monitored. Having said that, it is noted that a great deal of work has been done by many researchers and industry “innovators” in fitting on-line DGA monitors to larger power transformers and recording the results. An example of this is in New South Wales in Australia, where AusGrid has fitted two, ester fluid filled 50MVA 132kV transformers with on-line monitoring devices and the gas results were recorded over a few years and published1.

Following a survey of likely industry sponsors of research, a decision was made to not just refurbish the unit but to retro-fill it with natural ester fluid. The survey had indicated that industries and power utilities in Australia were interested in a greater understanding of the application of ester fluids in transformers. A few utilities were interested in how a retro-filled transformer would perform. This was an opportunity for UQ to investigate this area and provide a better understanding of the changes in performance and operational requirements with such a unit.

The TIC was set up as a semi-independent research centre within the UQ. The centre has had initial financial support from the University and as it was developed industry participants have now committed to financially support the centre into the future. There is an Advisory Steering Committee for the centre governance and a Research Project and Education Committee for assessing and overseeing projects and delivering Continuous Professional Development courses for the power industry. These Committees are made up of university and industry members and thereby provide a balance between academic needs and power industry needs.

Much of the data obtained for this paper has been obtained from projects undertaken with this transformer for the specific purpose of informing the industry on the performance of this unit. One of the primary criteria of all projects undertaken in the centre is the benefit to industry and how the outcomes can be applied in service.

Upgrades to the unit in addition to FR3 retro-filling

The original Dynamic Ratings T3 monitoring system, was completely upgraded to the latest technology including latest CPU, and firmware, as well as extended communications features, such as a remote User Interface module (Figures 3, and 4), installed in the control room. With this feature, no one needs to go into the room with the transformer energized to download data, or observe the parameters monitored via a WEB browser interface.

Having all sensors able to be monitored from outside the room (remotely) is the way it is generally set up in the field with the data coming back to the office and in this instance, where students are involved, the facility meets stringent health and safety requirements for access to live HV equipment.

Following are some of the specific monitoring equipment features:

  • An additional ETM (Electronic Temperature Monitor) was installed.
  • All 5 original moisture in oil sensors were replaced with the latest technology from Vaisala to enable lower detection limits than what was available 10 years ago. All 5 sensors are connected to one communication port of the DR-E3 system via a multi-drop RS 485 serial link.
  • A bushing monitor system capable of electrical PD detection, as well as leakage current measurement, is installed on each of the HV bushings, with an RFCT connected around the Neutral earthing terminal. The HV bushings are 66kV rated bushings. In addition, the MV bushings do not have DLA taps at the base.
  • The unit was originally built with a total of 16 fibre-optic temperature probes embedded into the three windings, in HV and MV sections. There are 4 separate control devices for these 16 FO probes, all are connected via individual serial links to the E3 system. Unfortunately, one of the probes has failed, but 15 remain active and working.
  • A serial port remains open for the addition of a suitable (with natural ester fluid) on line DGA monitor.
  • At this point in time, data is not archived by any central historian style server. The UQ uses the transformer for specific periods of time and so they have a few data loggers that they can connect via an ethernet, USB or RS port as required. The reason being that their individual projects do not rely on a main system and they do not need to filter out data which is not relevant to their project.
Remote User Interface and B100 ETM
Overall View of control room

Comparison of the Routine and Heat Run Test Results

After the unit was refurbished and retro filled in the factory, it was subjected to routine testing as if it were a new unit. The transformer was originally tested as new in December 2006 and after a refurbishment and retro filling it was tested in December 2016 almost exactly 10 years later.

The following tables give an indication of the changes in the test results across a number of parameters and a summation of the key points is given for each.

Table 1 provides the details of the No-load Loss (iron loss) and excitation currents. In this test, the same voltage was applied as in the original test. And the observations made here are:

  • No-Load loss is very close to original but a few watts lower
  • The No-load currents again were as near as the original as would be expected.
no load loss and no load currents

Table 2 provides a summary of the test results for the Load Loss and Impedance testing and fundamentally nothing has changed since the original tests were performed. This confirms that the windings have not been changed in any way to adjust for the changes in fluid types. Therefore, when energized and bought up to full load temperature the inputs remain the same and the only influencing factor on the transformer performance will be the fluid.

load loss and impedance test results

Table 3 provides the values of the Dielectric Dissipation Factor and Capacitances. These results were expected to vary not just because of the aging of the unit and its previous service life but there was an expectation as well that the new fluid may have an impact on these results. Normally changes in DDF and Capacitance results of greater than 15% would indicate the unit is aging. There may well be the start of a breakdown somewhere in the dielectric system but based on the way this unit has been tested after refurbishment, this is highly unlikely.

Therefore, it can be concluded that the change in fluid will trigger a substantial change in the DDF and Capacitance results. It would also be of interest to record changes in these results if several transformers were to undergo refurbishment and a fluid change. From a correlation of that data it would then be likely that there could be a consistent change in the data and that may be of use to utilities for future reference.

results of dielectric dissipation factor and capacitance tests

Table 4 shows the results of the temperature rise tests and it is clear for this that there are some substantial variations across the ratings between the mineral oil and ester fluid. The yellow column shows the variation in degrees Celsius and the light blue columns show the variation in percentage of the original result.

We have analyzed these results closer in the coming section on Observations

Results of temperature rise tests

Table 5 and Figure 5 show the temperature variations at each probe. With the number of active probes and the ability to extract the data easily the variation in the cooling of the transformer can be quickly seen as shown in Figure 5 below. This data is analyzed with that of Table 4 above in the Observations section below.

Fiber optic probes measurements
Plot of fiber optic probe


With the ability to have the two sets of heat run data and with both tests having followed the same protocol as demanded by the IEC standards there is an opportunity to understand the clear differences of the cooling properties of each fluid. It is important to note that the winding types are LV: Layer wound enameled copper, MV: Layer wound enameled copper and HV: Disc wound paper insulated aluminum.

From Table 4 in most instances the ester fluid has cooled the transformer slightly better than the mineral oil. With that said it is important to look closer at some of the critical areas.

HV Winding Temperatures: There appears to be a direct impact on the HV winding hotspot temperature when fans or pumps are operating. That is, when there are no fans or pumps in operation the HV winding hotspot appears to be some 2°C(K) higher for the ester fluid than mineral oil. This is reflected in the winding gradients, However, when the data from the fiber optic probes is reviewed in conjunction with this it does not reflect the same change. The probes that are closest to the HV winding hotspot in each phase are Probes 4 (B), 15 (A), 16(C) as these were positioned for the hotspot. It should be noted that with fans and pumps operating either independently or together, the ester fluid clearly has a greater cooling effect on the HV winding than would have been expected, with all gradients and hotspots more than 10°C(K) cooler than the mineral oil.

MV Winding Temperatures: There is a considerable change in the cooling of the MV winding because of the change to the ester fluid. The results for all cooling modes show the temperature differential was in favor of the mineral oil by around 3 to 5°C. There may be some consideration of the location of the winding and the ability of the oil to pass easily through that winding given it has a higher viscosity than mineral oil. The data extracted from the fiber probes 11, 13 and 14 show these increases in temperature in the winding, however probe 12 has no data as the probe does not work.

LV Winding Temperature: When comparing the HV, MV and LV for the ONAN vs KNAN the LV winding performed better with the ester fluid by around 5°C(K). As there are no fiber probes in the LV winding, it is not possible to compare the probe results and the only data available is that shown in Table 3.

A possible theory based on the above results is that the type of winding construction and conductor insulation may have an impact on how the ester fluids cool the transformer compared to the traditional mineral oils. Where paper insulation is used, it appears the ester fluids are not as effective as mineral oils when only cooling by natural convection. Conversely, when forced cooling is involved, a paper covered conductor may be a better option as it appears to be better suited for cooling by ester fluids.

When an enamel coated conductor is involved, this appears to be completely the opposite with the most effective cooling taking place when there is not forced cooling. There seems to be a correlation between the ester fluid’s higher viscosity and the type of conductor insulation used. That is, where a transformer will generally operate with natural convection cooling then the insulation preference would be enamel coated conductors. However, if the transformer will be run at loads where the fans or pumps will operate on a very regular basis then a paper covered conductor would be more suitable.

There is also some need to look at the winding construction as there are clear differences between the layer and disc type windings. This difference may also play a large part in the natural cooling of the unit. If the layer winding is designed for the higher viscosity ester fluid, then the cooling would be more in line with the mineral oil. As we are discussing retro filled units there is no opportunity to make that change unless the transformer is being rewound.

It is common knowledge that the cellulose paper used to insulate windings in mineral oil filled transformers degrades faster than when using ester fluids. 2This has been confirmed by the tensile strength and degree of polymerization derived from accelerated aging tests using the methods from IEEE Standard C57.100. From these tests paper aged at 170°C in ester fluids takes approximately 5 to 8 times longer to reach its end of life than paper aged in standard mineral oil.3 It is estimated that paper submersed in ester oil at 110°C will last at least 2.5 times longer than paper submersed in mineral oil.4 When this type of research is applied to the above test data from the transformer, it can be proposed that the impact of the additional elevated temperatures in some windings may not have a detrimental effect on the paper aging.

Whilst specific details of moisture have not been factored into this document, it is worthy of comment as the relative moisture absorption of mineral oil is lower than that of ester fluids. In work done at Monash University in Australia5 it was found that the ester fluids are highly water absorbent and appear to remove moisture from the paper insulation used in transformers. As the moisture is generated in transformers as the cellulose ages at elevated temperatures and since ester fluids can absorb moisture to a greater extent than mineral oil, moisture will migrate from the paper into the ester fluid.

Therefore, retro filling an un-refurbished and factory dried aged mineral oil transformer with ester fluid could potentially help remove water from the insulation. In a refurbished transformer it then follows that the ability of the unit to stay in a relatively drier state is greater. Applying the increase in some areas of temperatures within the winding could in fact help this process and so the life of the transformer would not theoretically be reduced due to the slower paper aging rate.

It will also pay to have a review of the oil flow pump on the transformer if one is fitted as the oil flow rate will be affected by the higher viscosity oil. This is an area that can easily be overlooked but plays a vital part in the cooling of the transformer and even though the pump should be suitable, a check on the effect of electrostatic charging and general suitability of the pump for the esters is a must.

Other observations

There seems to be an issue with the thermal expansion of the ester fluid in a retro filled transformer which may mean the conservator may not be adequate for the expansion space required. This would mean a larger conservator must be fitted and when that happens then a new conservator bag can be fitted to prevent direct contact with oxygen. The alternative is to fill the existing conservator with nitrogen gas or allow a slightly larger expansion space in the sealed tank. Our US colleagues suggest that the latter may cause higher gas pressures in sealed and/or nitrogen pressurized units, leading to venting, and therefore to a higher consumption rate of nitrogen, meaning, more requests for nitrogen bottle replacement.


Fluid Handling

There are many articles and instructions written by the suppliers of the ester fluids and these should be followed. What they do not seem to advise is how to deal with the equipment needed to handle the ester fluids. Based on research work done at Monash University (1) between 2006 and 2010 it was found that a 1mm (1/32”) deep solution of ester fluid when left exposed to the atmosphere took almost 2 years to solidify.

Therefore, when using pumps or filters the oil will not immediately congeal in the pipes when left. It is advisable to clean them as the fluid does collect water and it is not desirable for use on the next transformer. Interestingly, this is the same process for when mineral oil is used yet we seem to be more concerned about esters than the mineral oil.

There has been mention of a situation in the UK where a small weep occurred on the sunny side of an ester filled transformer. The weep was only slight and so the fluid congealed over a few weeks and sealed the weep. There is no need to change types of gasket materials as there are now thousands of units in service with standard neoprene cork gaskets that work effectively.

As mentioned previously, the ester fluids absorb higher levels of moisture and operate at higher temperatures and therefore extracting the moisture from the fluid requires heating the ester fluid to around 110°C. Therefore, the oil seals in the pumps and vacuum units need to be designed to withstand the higher temperatures or else they will fail after only a few runs. It is also advisable to use a pump with suitable higher rating as the higher viscosity of the ester fluid will work the pumps harder.

Identification and Mixing

Two issues that have been around for some time now are related to mixing the oils and how to identify each.

The easier way to go now is to identify the ester fluid units as they are progressively installed. Figure 6 shows one example where the pole top transformer has a clear indication (decal), on the side identifying the fluid. Other solutions include changing the color of the paint however this may introduce other issues associated with the thermal performance. The paint color does have an impact on how a transformer radiates or absorbs heat and a color change would need careful consideration. The nameplates are change from O for oil to K for non-mineral fluids. This is not enough as there are many different fluids available and all have slightly different characteristics. To date there does not seem to be a driver in the industry to set a uniform standard for identifying the units in the near future.

Identification is important as it leads into the second issue of mixing oils. If a service technician had a call to a site for a low oil level alarm, which oil would they take? Once they get there they may be under pressure to get the unit topped up to get supply back on. He may not look at the nameplate and add mineral oil to an ester unit or vice versa. The good news is that it is not that much of a problem when dealing with small quantities. The suppliers of ester fluids suggest up to 10% of mineral can be mixed without a significant impact on the ester fluid properties. Where an ester fluid unit has been installed in a high fire risk or an environmentally sensitive location, then any cross-contamination risk needs to be considered and mitigation strategies put in place.

identifying ester fluid filled transformers

If the unit is refurbished, retro filled or has a top up of incorrect oil then the key issue is that the mixed fluids cannot be considered as biodegradable and if the levels of contamination are higher than 10% it may have a fire risk impact. This is important when disposing of the fluid later in life. What is paramount in any network to prevent inadvertent cross contamination is that the transformers are clearly identified in the field and in any asset management data base.

On-Site Work6

First step should be a condition assessment of the unit. A visual inspection to confirm integrity of all seals/bolted connection, and proper operation of all gauges and ancillary devices (OTI, WTI etc.) should be performed. This inspection and assessment may indicate whether additional maintenance operation should be performed while the unit is out of service.

Some important reminders (regardless of the fluid to be replaced);

  • Source and have available a complete set of gaskets on site, just in case there is some seeping and/or weeping fluid.
  • Elastomers including NBR types with higher nitrile content, silicone, or fluoropolymer are recommended. Gaskets with higher temperature demands warrant the use of silicone or fluoropolymer (Viton), compositions.
  • Be sure to make, have available, and install (close to the sampling valve as well as the rating plate) a Retro-Fill label, indicating the type of fluid now in the tank. • After waiting 24 hours after the fill, take samples of the fluid, for the requested ASTM tests, as well as DGA test. This will provide the baseline measurement from which to look at future trends/changes.

Bushings and Internal clearances

It is vital that the user consult with the original manufacturer of the transformer, including the OLTC and bushings, to inform them of the planned change, and what, if any modifications and/or recommended actions need to take place because of changing the dielectric fluid. Some of the dielectric properties of a natural ester are not the same as mineral oil. Care for dielectric clearances between phases and phase to ground paths may require additional work or components to ensure proper dielectric performance. A change in the dielectric properties of the oil may mean a change in the operating characteristics of the bushing. In particular, that part of the bushing that is immersed in the new fluid. The bushing base may see different stress levels which may need addressing and so the manufacturer should be consulted.

Operation of the Retro filled transformer

As stated earlier in the temperature rise analysis there are considerations that need to be dealt with when operating a retrofilled ester fluid transformer. This includes understanding the change in cooling characteristics associated with the fluids. The transformer manufacturers change the design requirements slightly when they are aware the transformer will be ester filled at the start of its life.

This is not possible with an existing transformer and so understanding the impact and the change in operating conditions is vital to getting the maximum performance from the transformer. It is strongly recommended that a reasonable level of design review be done which takes into account any concerns about pumps, bushings, tapchangers and other internal components. Then after retro filling, a full temperature rise test should be conducted and these results compared with the original. This will help understand what the changes to the operating characteristics will be.

This type of information needs to be passed to the system operators and those doing load forecasts as they need to know what constraints have been imposed on the transformer after a retro fill. It may well be that in situations as shown in the research transformer that under some cooling modes the unit can be pushed well beyond its original rating simply by knowing there is adequate cooling in hand. In other types of transformers there may well be an opposite affect when different cooling modes are activated. This knowledge must be made available or all the good work and original purpose for retro-filling will be lost and the cost benefits will not realized.


  1. D. Martin and N. Lelekakis, “Monitoring and Diagnostics of AusGrid’s Two New 50MVA Envirotemp FR3 Fluid Filled Transformers: Final Report”, Monash University Centre for Power Transformer Monitoring, Diagnostics and Life Management, April 2012.
  2. K.J. Rapp, C.P. McShane, and J. Luksich,. Interaction mechanisms of natural ester dielectric fluid and kraft paper. In Dielectric Liquids, 2005. ICDL 2005. 2005 IEEE International Conference on, pages 393–396, June 2005.
  3. IEEE standard test procedure for thermal evaluation of insulation systems for liquid-immersed distribution and power transformers. IEEE Std C57.100-2011 (Revision of IEEE Std C57.100-1999), pages 1–37, Jan 2012.
  4. AA Abdelmalik, J.C. Fothergill, and S.J. Dodd. Aging of kraft paper insulation in natural ester dielectric fluid. In Solid Dielectrics (ICSD), 2013 IEEE International Conference on, pages 541–544, June 2013.
  5. D. Martin, N. Lelekakis, J. Wijaya, and K. Williams. Water uptake rates of transformer paper insulation impregnated with vegetable oil. Electrical Insulation Magazine, IEEE, 29(5):56–61, September 2013.
  6. Cargill Service Information Bulletin R2040, Guide for Retrofilling Power Class Transformers > 7500 kVA.
  7. Cargill Service Information Bulletin S10, Dielectric Fluids, EnvirotempTM FR3TM Fluid Storage and Handling Guide

Join our email list

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