DGA Alarm Limits and Trending: Norms, Baselines, and False Positives

The number that trips most DGA alarms was measured on somebody else's fleet. IEEE C57.104-2019's concentration tables are 90th-percentile statistics from a large transformer population [1]; IEC 60599 publishes ranges of 90% typical values and typical rates of gas increase the same way [2]. Neither document claims those figures are limits for your transformer - and a scheme that treats them as limits produces the two failure modes this guide is about: alarms that cry wolf, and faults that hide under the line.

The transformer DGA fault gases guide covered what the gases mean and the DGA gas ratios guide how they are interpreted; this one covers the operational layer: where limits come from, why rate of change beats thresholds, and the false positives to rule out before anyone opens a tank. Camlin monitors are the instrument; the logic below is what they need to be set with.

Where the Limits Come From: 90th-Percentile Norms (IEEE C57.104)

A norm is a fleet statistic. C57.104-2019 rebuilt its tables against a far larger dataset than the 2008 edition, and stratifies them by factors such as transformer age and the oxygen-to-nitrogen ratio of the oil, because sealed and free-breathing units gas differently [1]. IEC 60599's Annex A carries the same idea as ranges of 90% typical concentration values observed in power transformers, with separate tables for instrument transformers and bushings [2]. The point of a 90th percentile is calibration: a reading above it puts the unit in the top tenth of the population, which justifies attention, not a verdict.

Trending Beats Thresholds: Rate of Change

A concentration is a position; a rate is a direction. IEC 60599 publishes typical rates of gas increase alongside its concentration ranges [2], and C57.104-2019 evaluates concentration and rate of change together rather than either alone [1]. The reason is physical: a 30-year-old unit with a stable, elevated hydrogen level is not failing, while a young unit whose ethylene doubles in a month is, even if every value sits under the table. IEC 60599 also ties the resampling interval to the suspected fault type [2] - a slow cellulose signature and a developing arc are watched on different clocks - and CIGRE TB 296 documents the limits of single-sample interpretation that trending exists to overcome [4].

Build the Baseline: Commissioning Values Are Your Real Zero

The defensible alarm compares the unit to itself. A commissioning DGA, repeated after any oil processing or degassing, sets the zero that every later reading is judged against; without it, the first alarm has no context and the fleet norm is the only reference left [1]. Online monitors make the baseline continuous: Camlin's TOTUS G5 measures the five headline gases plus moisture at fixed intervals using photo-acoustic spectroscopy, without carrier gas or the frequent recalibration of gas chromatography [5], so the unit's own signature accumulates from day one instead of at annual sample points.

The 5 False Positives to Rule Out First

Before a reading becomes a fault, eliminate the ways it can be wrong:

1. Sampling error. IEC 60567 governs how oil and free gas are sampled, transported, and extracted [3]; a syringe that saw air, a warm sample left in sunlight, or an unflushed valve changes the answer before the lab sees it.
2. Air ingress. Oxygen and nitrogen rising together indicate the oil is breathing, not faulting, and the O₂/N₂ ratio is exactly why C57.104-2019 stratifies its norms [1].
3. Tap-changer gas transfer. Where the OLTC compartment communicates with the main tank, its normal switching acetylene migrates; IEC 60599 treats tap-changers separately, with their own Duval Triangle [2]. Rule the path out before calling arcing.
4. Stray gassing. Some oils generate hydrogen and methane at moderate temperatures with no fault present - a behaviour the interpretation guidance flags as stray gassing [2][4]. A hydrogen rise with no ethylene, ethane, or acetylene movement deserves that check.
5. Method offset. IEC 60567 describes three extraction methods - vacuum, stripping, and headspace [3] - which do not return identical numbers. Compare online to online and lab to lab, and correlate the two deliberately rather than treating a difference as a trend.

Common Alarm-Setting Mistakes

The 5 recurring errors under IEEE C57.104 and IEC 60599:

• Fleet norms as unit limits. The tables are 90th-percentile statistics [1][2]; the unit's baseline and rate carry the alarm.
• Absolute thresholds with no rate. A rate-of-change alarm catches the young unit going bad; a threshold alone waits for it [1].
• One resampling interval for every fault. IEC 60599 ties the interval to the suspected fault type [2]; a D2 suspicion is not watched monthly.
• Alarming before ruling out false positives. Sampling, air, tap-changer gas, stray gassing, and method offset explain a large share of first alarms [3][2].
• A baseline that was never taken. Without commissioning values, the only reference is a stranger's fleet [1].

What to Specify When Enquiring: 5 Items

A monitoring enquiry that quotes cleanly states:

1. The fleet - unit ages, sealed or free-breathing, and any known tap-changer communication with the main tank
2. Baseline status - which units have commissioning DGA on record and which need one at installation [1]
3. Alarm philosophy - the norm set, rate-of-change window, and resampling logic the monitor must implement [1][2]
4. Lab correlation - the periodic laboratory schedule the online data will be reconciled with, per IEC 60567 [3]
5. Gas set - 5-gas G5 for DGA and moisture, or the 9-gas G9 where PD, bushing, and through-fault data belong in the same unit [5]

Sourcing and Support in Pakistan

Pacific Engineering & Automation is the authorized Camlin reseller in Pakistan, supplying the TOTUS range for transmission, K-Electric, DISCO, and IPP transformer fleets - and we help set the alarm logic from the unit's baseline and rate of change, not from a generic threshold pasted from a table.

To put defensible DGA alarms on a critical transformer, request a catalogue or quotation or contact our engineering team.

Field-derived case studies will be added to this post as Pacific Engineering & Automation accumulates engagement records. The current version is grounded in published specifications, regulatory documentation, and standards body references.

Sources

1. IEEE C57.104-2019 - IEEE Guide for the Interpretation of Gases Generated in Mineral Oil-Immersed Transformers
2. IEC 60599:2022 - Mineral oil-filled electrical equipment in service, guidance on the interpretation of dissolved and free gases analysis (4th ed.)
3. IEC 60567:2023 - Oil-filled electrical equipment, sampling of free gases and analysis of free and dissolved gases in mineral oils and other insulating liquids (5th ed.)
4. CIGRE Technical Brochure 296 - Recent developments on the interpretation of dissolved gas analysis in transformers
5. Camlin TOTUS G5 - 5-gas online DGA monitor