November 2, 2026 · DGA Fundamentals

At its 2026 session CIGRE carried a paper arguing that total gas pressure flags air leaks in sealed preservation systems more reliably than oxygen concentration does. Taken alone, an oxygen number does mislead easily. But dissolved oxygen in transformer oil is not a channel to drop: it is the one gas in the tank that is both a contaminant and a reactant, and read beside nitrogen it answers two questions at once.

Oxygen is not a fault gas — it is a reactant

The fault gases of dissolved gas analysis are products. IEC 60599:2022 sorts them into six fault types, and each appears because something inside the tank has gone wrong. Oxygen is different: it is there at commissioning because air is, and its level is set as much by the oil’s chemistry as by any fault.

What earns it a channel is that oxygen is consumed by the reactions it drives. Hydrocarbons oxidise, oxygen leaves the oil, acids form, and the acids attack the paper. The level moves both ways — down as insulation ages, up when the tank admits air — and the two look identical on a single sample, opposite on a trend.

What oxygen does to oil and paper

The chain is slow and well established. Oxygen attacks a hydrocarbon chain into a hydroperoxide, which breaks down into acids, aldehydes and ketones; the acids catalyse further oxidation, and the end products settle as sludge. Acid number rises, and sludge laid on winding surfaces insulates the conductor from the oil meant to cool it — hotter locally, so the reaction speeds up. The chemistry is set out in oil decomposition chemistry.

Cellulose fares worse and less reversibly. Acid hydrolysis cleaves glycosidic bonds, the degree of polymerisation falls, and paper loses mechanical strength long before dielectric strength — which is why a unit can pass a power-factor test and still fail under a through-fault. Inhibited oils buy time with DBPC, which scavenges peroxide radicals but is consumed rather than regenerated. Cut the oxygen supply and every rate in that chain drops — the whole logic of a sealed conservator or a nitrogen blanket.

Ingress or consumption: why oxygen needs nitrogen beside it

Two opposite stories end in the same reading. A gasket, bladder or breather admits air and oxygen rises; the oil oxidises and oxygen falls. Nitrogen separates them, because nitrogen is inert in service — no fault generates it, no ageing reaction consumes it. Air is 78.08 per cent nitrogen and 20.95 per cent oxygen, so nitrogen is the fingerprint that air has been there.

Sample shows Nitrogen says Oxygen says Reading
O₂ and N₂ rise, hydrocarbons flat Fresh air entered Not yet consumed Ingress
N₂ steady, O₂ falling, CO/CO₂ climbing No new air Being used up Oxidation or thermal fault
Both fall, fault gases flat Gas leaving Gas leaving Top-up or degassing
Both high on one sample High too High Air in the syringe

The ratio needs a caveat. Air in the gas phase is 3.7 parts nitrogen to one part oxygen — the figure quoted everywhere as the ingress signature. In oil it does not hold: oxygen is roughly twice as soluble, so air-equilibrated oil settles near O₂/N₂ ≈ 0.5, and a sustained slide below 0.3 is read as consumption. Act on the ratio against the unit’s own baseline.

That is also why CIGRE’s pressure argument works: nitrogen dominates the partial pressure of incoming air and is neither created nor destroyed in service, so total pressure tracks the leak while oxygen is consumed.

Is low dissolved oxygen in transformer oil a problem?

Not by itself. A sealed or nitrogen-blanketed unit that has always read low is behaving as designed. The problem is a falling number on a unit that used to read high, with nitrogen steady and CO, CO₂ or furans climbing: oxygen consumed faster than it is replaced, which points at oxidation or a thermal fault rather than the seal. Judge the slope against the unit’s history, not the level.

How is dissolved oxygen in transformer oil measured?

In the laboratory by gas chromatography after vacuum or headspace extraction, per IEC 60567:2023 and ASTM D3612. Online, a monitor circulates oil across a membrane or draws a vacuum-extracted gas sample and measures oxygen with the fault gases. Oxygen is the easiest channel to spoil: a bubble in the syringe lifts O₂ and N₂ together and reads exactly like a leak. Draw through dry, airtight tubing, discard the dead volume, and follow the oil sampling guide.

On the DGM-900 both channels are optional: the standard configuration covers the seven IEC 60599 fault gases plus moisture, and nitrogen is calculated for free-breathing transformers only.

Whether to add oxygen depends on the question. For fleet ageing it is the most direct measure of the reaction that limits insulation life; for leak hunting it is half a pair, and nitrogen does the pointing. A dissolved hydrogen sensor covers the fault-gas side cheaply, this comparison shows where single-gas monitoring stops being enough, and a recent lab report with nitrogen and oxygen columns can be read with you through our contact form.

Sources

  • IEC 60599:2022 — six fault types PD, D1, D2, T1, T2, T3.
  • IEC 60567:2023 and ASTM D3612 — sampling, extraction and gas chromatography.
  • CIGRE D1_11377_2026, Total gas pressure of dissolved gases in power transformers — air-ingress indicator argued to be more reliable than oxygen alone.