...Right size your transformer DGA device

More Monitored DGA Gases do not Necessarily Equal Better Transformer Protection ... The Case for the simpler, proven, pragmatic ‘Calisto 2’

Very worryingly, in the past few years, however, our industry would now appear to have become more attracted to the enticingly impressive data quality and quantity now offered by a plethora of nine gas ‘diagnostic’-style monitors.

That said, for the vast majority of transformers, a well‑designed three‑parameter DGA monitor, such as the ‘Calisto 2’ [which tracks hydrogen (H₂), moisture (H₂O) and carbon monoxide (CO)], delivers every piece of information needed for timely, confident and practical condition management of all relevant transformer main tank issues that may arise either from various concerning events which are manifested in oil molecule breakdown, or paper aging/heating.

Risk of Overcomplicating DGA Monitoring:

Indeed, a strong argument now exists that there is an emerging danger of overcomplicating the task of main tank DGA monitoring by way of specifying nine gas devices over the seemingly humbler ‘Calisto 2’ simplified DGA devices.

Such complications may (and do) risk resulting in:

• Significantly elevated (typically 3 times higher)monitoring costs per transformer, making business cases more challenging and protracted
• Higher real associated maintenance and device management costs; an inability or possible stress, for our field practitioners to set up the required (at least) 18 separate alarms and to assess the monitors competently on an on-going basis
• The associated potential to incur the cost of specialist consulting inputs simply to set up the monitors at the outset or to conduct basic assessment of generated data streams
• Or for our operators either to be unable, or seriously confused, in determining at-a-glance the fundamental nature of an issue (or to assess its severity) were an alarm to be raised.

All of the above are most concerning matters and we must confront them carefully.
This application note considers when a multi-gas DGA monitor is genuinely required, and when it is not. A practical decision framework is provided to help owners select the right DGA architecture for each transformer, rather than reflexively choosing a more expensive option.

---

Both DGA monitor architectures have their place, but currently the simplied DGA monitors, such as the ‘Calisto 2’, remain by far the most commonly deployed monitor architecture in our region.

The ‘Calisto 2’ device is the longer trusted of these detection-style technologies, it being lauded by the insurance companies for their already 25 years of proven operational history in Australasia. Their appeal lies in their delivering a dependable single main output trendline that is simple to interpret at a glance by any operator with basic training.

That said, nine gas DFA monitors are being rightly taken up on an increasing basis for certain very specific reasons: watching a major transformer with a known DGA problem; strategic rationales (such as protecting transformers of State or National significance) or managing sites with immense cost of risk; or where access to the monitored site in under 1-2 days may be impractical. Multi-gas monitors will continue to play a strong role in the monitoring field in the future where such monitor architecture is so merited.

In all other situations where a multi-gas configuration is harder to justify, it is logical that the balance of monitor architectures will progressively shift back in favour of the more proven and simpler 2-gas monitors such as the ‘Calisto 2’. This is particularly the case as formal DGA skill levels amongst field practitioners continue to decline, and the use of more complex devices thus becomes increasingly problematic.

Common Misconception: “More Monitored DGA Gases does not Equal Better Protection”:

• Simplified 2-gas DGA monitors of optimal design, advanced specification, simplified data trending and interpretation, and appropriate alarm set-up, of which the ‘Calisto 2’ excels, are capable of warning of a developing main tank problem just as quickly as a complex 9-gas DGA monitor (see below).
• Adding the measurement of up to 4 to 8 more gases does not inherently allow an adverse main tank situation to be signaled any earlier. Rather, with a least 18 alarms to set, adjust, and interpret, it still only refines the diagnostic process once an adverse condition has arisen and risks even missing an alarm state or diagnosis due to operator confusion.
• In 25 years of Australasian field experience, transformers equipped with ‘Calisto 2’ in the generation, transmission, heavy industrial, and distribution sectors have frequently illustrated their capability to determine an adverse main tank condition in a timely manner and then to guide the asset owner and field operator alike toward an appropriate response and actions.

Multi Gas DGA Monitors Offer Just One Key Contribution:

What this application note wishes to illustrate, is that detection-style monitors are not simply “alarm units”, as is often wrongly assumed. If suitably designed and configured, they are as capable of early detection of abnormalities as multi-gas DGA monitors. This is so both in terms of quantifying an issue [by trending and tracking gassing levels, and rates of change] and in qualifying the issue as to its probable cause [correlated by observing both the patterns of gassing and the effect of interventions such as transformer load reduction]. In other words, if both monitor architectures are commissioned and managed suitably, the only notable additional contribution offered by a nine gas DGA monitor is the ability to perform a formal diagnosis of the fault remotely.

Hydrogen is Present and Observable in all Fault Types:

Having said that, and in elaboration of the above comments, a lot more can be achieved with the ‘Calisto 2’ monitors to initially determine the likely issue being seen by the monitor than is generally realized. This may seem a surprising comment, given that Hydrogen is not a diagnostic gas per se (unless for PD issues only). Referring to Figure 2, in reality, Hydrogen is produced in an almost linear fashion over the fault temperature range in the main tank (say 120 Deg C for PD, to circa 300-800 Deg C for thermal fault ranges, to 900-1200 Deg C for arcing). This chemical reality, coupled with the ‘Calisto 2’s’ outstanding performance specification (below), allows it to work so effectively at reliably signaling adverse main tank events in a timely fashion that can easily be interpreted.

The design and specification of the ‘Calisto 2’ monitor has clearly been engineered for an unrivalled ability to perform the duties expected of it. Its list of credentials includes:

• unmatched reading accuracy, stability and repeatability over life
• very high resolution
• the lowest detection limit (‘LDL’) of any such monitorpumped oil circulation
• fast gas extraction and analysis with no oil contamination
• a unique per-reading calibration (nominally 3 hourly)
• an active temperature conditioning of the incoming oil (pre-measurement) to both normalise solubility coefficients of dissolved gas and moisture in oil, plus cushions the internal electronics to seeing just a 0.1 Deg. C variation over life
• an ability to measure incoming oil temperature to present moisture levels in ppm, or %RS at T Deg. C, or % RS at 25 Deg. C
• very good reliability over its >15-year EOL.

In other words, and by design, what the ‘Calisto 2’ reads can be taken as ‘fact’, then acted upon without hesitation or delay.

With only brief training, as depicted in the illustrative case studies of this Application Note, field operators can interpret the single-line trend plots from the ‘Calisto 2’ [watching hydrogen growth, rate of change, and patterns; moisture drift; and load-driven CO and H2 gassing shifts] to gain an instant picture of what’s unfolding inside the main tank. Such information permits the operations team to judge whether:

• the transformer in question has no notable issue and can stay on load
• has an emerging issue and should be observed regularly for adverse change
• should be run derated to continue operation whilst a response is planned
• or must come offline immediately.

In many routine events, this rapid triage alone is enough to safeguard the asset, remain calmly in control, and minimise operator stress through the process.

A full dissolved-gas analysis (DGA) at an external oil laboratory remains the gold-standard confirmation of the assessed and initially corroborated matter. To that end, the ‘Calisto 2’ streamlines the oil sampling process: its integrated DGA sampling valve supplies oil directly into the DGA syringe at a known oil temperature (35 Deg C +/- 0.1 Deg C), so a follow-up lab sample can verify the diagnosis with minimal effort and highest confidence.

‘Calisto 2’ Monitors Deliver ‘Actionable’ Intelligence:

In reality, site logistics, courier schedules, and laboratory queues often mean a validated DGA report arrives between one to five days after the alarm. During that critical time window, ‘Calisto 2’s’ live gas trends become the operator’s main decision tool. By tracking how quickly key gases climb and whether they plateau when the transformer’s load is eased, for example, operational staff can form a sound preliminary diagnosis, gauge severity, and choose between continued monitoring, load reduction, or an urgent outage. In short, Calisto 2 delivers actionable intelligence that protects the transformer long before the laboratory results are to hand.

In other words, a device of the specification of the ‘Calisto 2’ might well see its key role being competently, and safely, discharged by the site operator interacting with it prior to a final DGA sample coming back. Of course, devices such as the ‘Calisto 2’ are equipped with outstanding communications options (including IEC61850) so that all data recorded, along with real-time self-check alarms, may be viewed remotely to the site by third parties who can also guide the hand of the field operator managing the matter on site. That blend of on-site visibility and remote expertise delivers high-performance protection at a fraction of the cost and operational complexity of a multi-gas system—making ‘Calisto 2’ an economical yet powerful transformer DGA monitoring choice.

The following illustrations, all sourced from Morgan Schaffer ‘Calisto 2’ DGA monitors in real-world adverse main tank situations, are intended to offer a quick overview of the above points, and particularly to demonstrate the unique capability of the ‘Calisto 2’ to faithfully qualify and quantify main tank issues from PD to the full thermal fault range to the arcing range:

---

Illustration One…Random Arcing:

Here we see a classic random arcing situation. Under this scenario we observe very sudden and sharp rises in H2 that then decay over the next week and would otherwise be missed. A DGA taken at that very time would show very high levels of H2 and Acetylene but a week later just some acetylene that would otherwise puzzle the observer. The ‘Calisto 2’ has caught this rapid event and it is unmistakeable, without needing a DGA to confirm the situation. Just to prove this point, the DGA Duval analysis for one such event is shown below.

Fig 3: Illustration of a series of 3 major main tank arcing events.

Fig 4: Duval triangle corroboration of one of the major arcing events of Fig 1.

---

Illustration Two....PD events in the Main Tank

Here we see two illustrations of PD events. These are typically modestly high levels of H2 (nominally a few hundred ppm) and reasonably slow evolving. Wind farm transformer PD issues look like this but are unusual in that they can reach huge levels (5,000-15,000 ppm not being uncommon). Main tank PD issues tend to be continuous and fairly stable over time but may in some cases rise in level at a steady rate of change (unlike thermal faults, described below). PD faults seldom have a load-related correlation. In some cases they may pertain to a design issue (e.g.: a core earthing matter) and simply remain very stable with time, effectively acting as an H2 ‘baseline offset’.

The first illustration below is a tap change diverter resistor issue and the second was a spanner left in the main tank causing PD off its edges.

Fig 5: Tap change diverter resistor PD issue.

CFE

---

Illustration Three...T2 thermal fault trending toward runaway:

The illustration below shows an unmistakeable thermal runaway (T2 zone area) which is characterised by a number of key features. The first is a steadily increasing H2 trend with time, characterised by a generally increasing rate of change. Whilst this trend may appear for short times to stabilise, it almost certainly continues the said trend pattern. Thermal faults will almost certainly reduce in rate of change, and even perhaps gradually in level, by the act of dropping back on transformer loading to force the corroboration and thus the diagnosis. Indeed, it may be possible to manage interim continued, but qualified, transformer use via such an intervention. A typical thermal issue as described might run from normal to severe manifestation over of period of just 1-2 months.

Again, such a characteristic is clearly evident without recourse to an initial DGA corroboration being needed to reach a conclusion as to fault type.

Table 1: H₂ generating rates from May 30th to Sept. 26th

Time section (refer to Fig. 4)	Time period	H2 generating rate (ppm/day)
A	May 3 to May 20	<1.0
B1	May 30 to June 15	3.8
B2	June 16 to June 28	5.0
B3	June 29 to Aug 12	2.8
B4	Aug 13 to Aug 25	4.1
B5	Aug 26 to Sept 12	1.4
C1	Sept 13 to Sept 26	12.1

Fig 7: T2 Thermal fault developing in a GSU.

---

Illustration Four...Thermal to Arcing fault transition:

The next two illustrations show thermal issues transitioning to T3 thermal, then to Arcing faults. Characteristics of this are very sudden and typically exhibit rapid positive-going changes to both H2 gas levels and rates of change. This behaviour may ‘play out’ over mere hours. Again, this trend is unmistakeable to the observer. A DGA Duval analysis is included with the first of the 2 illustrations to confirm the observation.

It is worthy of note that even in such extreme events of short duration, the Calisto’s three-hourly verified reporting rates are perfectly pitched to the scenario, these being likely to give multiple alarms in that short time frame to ensure a timely response from the asset owner. This scenario does make the point clearly that any alarm generated by a Calisto (as long as it is set suitably, as previously stated), should be acted upon promptly as there may be a limited time to respond once the adverse scenario is revealed.

Table 2: Sudden increase in Hydrogen generating rate

Time section (refer to Fig. 6)	Time period H2 concentration (ppm)	Maximum H2 generating rate (ppm/hour)
A	27/09/2002 5:52 to 28/09/2002 5:53, 1,367 ppm	100
B	28/09/2002 8:53 to 28/09/2002 23:42, 5006 ppm	300
C	29/09/2002 11:43 to 29/09/2002 23:42, 9494 ppm	392

Fig 8: GSU transitioning to T3 then Arcing event over c.2 days!

ComEd

• 765 KV, 333 MVA Transformer, single phase, placed in service in 8/1971. Large shell form with 16,580 gal. main tank;
• Loading had been low for several months prior to the failure, but gassing was creeping up very steadily to high ppm levels of H2;
• On Oct. 28, and for the next three days, gassing rose dramatically and quickly, the ‘Calisto 2’ monitor alarmed, and the transformer was then quickly taken out of service before any further damage resulted;
• A DGA sample was taken on three occasions, as the gassing increased, and showed a significant increase in both ethylene and acetylene. The transformer was confirmed to be no longer usable and was wisely taken out of service in a controlled fashion. Rewinding was then considered, as the timely removal from service preserved such options being possible.

Fig 10: Another illustration of a rapid transition from T2 to a T3 then Arcing event.

---

Illustration Five: CO gas indicating a thermal fault issue:

The illustration below focusses on the Calisto 2’s CO channel (shown in orange at right) catching a
fast-developing thermal issue in a main tank (potentially a localised heating matter, blockage, design issue etc). The H2 in the main tank is stable and low-level (2 ppm), indicating no electrical issues. The observer can thus make an immediate deduction from the pattern and makeup of the data, and thence plan an appropriate ‘next steps’ response.

Fig 11: Illustration of a concerning change in paper aging rates. Note: H2 remains very low, & stable.

---

Summary and Concluding Remarks:

We now serve in a world where many projects routinely demand harder-working transformers built to a minimised specification under a tight budget rigour, are installed with minimal or zero redundancy, and where asset owners and stakeholders alike must strive to manage the associated increased risks and associated cost of risk. Insurers have rightly noted the confluence of these issues and are now seeking a good level of demonstrated effort in the transformer DGA monitoring front particularly.

In parallel with the above unfolding constraints there are two relevant, but complicating, factors noted to be in play recently. One is a widespread rationalisation of Capex budgets and project spends. The other is a progressive loss since Covid of much of the required technical skillsets to assess and respond to an unplanned transformer issue, including a decreased awareness amongst field operators and practitioners of the science of DGA analysis.

Perhaps as an unexpected corollary, and one made even more perplexing yet perhaps explainable in the light of the above observations, our industry would now appear to be abandoning long-standing and well-proven transformer on-line DGA solutions of simpler and more cost-effective nature. Instead, perhaps driven by a failure to understand the quality of the contribution possible by simpler devices, a current trend is embracing by preference a bewildering selection of high-priced and complex multi-gas DGA devices which, for the reasons outlined above, may fail to deliver their expected outcomes effectively.

Whilst this Application Note has observed and respected that multi-gas DGA monitoring devices do indeed have a tightly prescribed range of clear and rightful scenarios where their deployment is clearly merited, it has lamented an observed recent over-use of these devices where their level of both implementation and operational complexity could be argued to be more a risk than an advantage.

The Note has made the case, well-illustrated with real-world events and convincingly advocated via technical reasoning, that for many applications where multi-gas DGA devices are currently being rolled out or planned, a far more rational, pragmatic, economic, long-proven, long-respected, widely deployed, and practicable transformer on-line DGA monitoring solution lies in the guise of the uniquely capable ‘Calisto 2’ device. Currently serving in the Australasian region as the numerically largest-deployed DGA monitor, it has reliably delivered simply interpreted precision DGA information to its asset owners in the generation, large industrial, transmission, and distribution markets for the past 25 years and contributed well on many occasions where adverse events occurred.

The point has been well-made that the ‘Calisto 2’ is capable of alarming on the evolution of an adverse main tank condition in an identical time frame to a multi-gas device, may readily interpreted by field practitioners with modest training only, and has the fullest support of the insurance industry in managing such risk.

To conclude, multi-gas DGA monitors have a clear and rightful place in only a few specific situations. In most deployment scenarios, the pragmatic simplicity, long-proven acceptance by our industry and insurers alike, and its significant cost savings over life commend the humbler but technically sophisticated ‘Calisto 2’ DGA monitoring device for the majority of online transformer DGA applications.