Failures Are Rare, but They Are Also Expensive
Power transformers are among the most reliable assets in the grid — but when one fails, the consequences are disproportionate. A catastrophic failure takes a unit out of service for months, requires a replacement that can cost millions, and in the worst case becomes a fire or explosion. Published failure analyses attribute roughly a third of failures to design and manufacturing defects and about a fifth to aging and overheating — the two categories that dissolved gas analysis (DGA) monitoring is specifically able to see coming.
The question for an asset owner is not whether transformers fail, but whether the failure is detected while it is still preventable. That is precisely what online DGA monitoring is designed to do.
The Fault Timeline: From Incipient to Catastrophic
Internal transformer faults do not appear suddenly. They develop over weeks or months through identifiable stages, and each stage releases characteristic gases into the oil:
| Stage | Typical temperature | Key gas released | Online signal |
|---|---|---|---|
| Partial discharge | Low energy | Hydrogen | H2 rises |
| Low-temperature overheating | ~150–300 °C | Hydrogen, methane | H2 / CH4 |
| High-temperature overheating | ~300–700 °C | Ethylene, ethane | C2H4 rises |
| Arcing / discharge | > 700 °C | Acetylene | C2H2 appears |
Industry guidance is that DGA can detect an incipient fault 30 to 120 days before visible symptoms. That window is the entire value of monitoring: it is the difference between a planned repair and a catastrophic failure.
The Blind Zone Between Laboratory Samples
The reason this early window is so often missed is sampling frequency. A transformer sampled once a year — or even quarterly — is invisible in between. Industry reporting estimates that about 37% of sudden transformer faults develop during the “blind zone” between periodic offline tests: a fault starts the day after a clean lab result and is not caught until the next sample, by which time it has progressed.
A laboratory sample is a snapshot; it cannot show how fast gases are accumulating. Online monitoring replaces the snapshot with a continuous time series, which is what makes rate-of-rise analysis possible in the first place.
Hydrogen Is the First Warning Gas
Among the fault gases, hydrogen deserves a dedicated layer because it appears first. It is generated in partial discharge, low-temperature overheating and moisture ingress, it is the smallest molecule so it diffuses fastest, and it is a component of the gas signature for most developing faults.
- Earliest signal: H2 precedes ethylene, methane and acetylene in most fault chains.
- Fast transport: small molecular size means a short lag between the fault site and a sensor at the tank wall.
- Low-cost screening: a zero-consumable hydrogen sensor can sit on every transformer, while full gas chromatography is reserved for the units that need it.
For the compliance context that is pushing this monitoring onto more fleets, see the 2026 dissolved hydrogen compliance guide.
From Warning to Response
Detection is only half the equation; the alarm must trigger the right action. A practical staged response mirrors the fault timeline:
- Advisory: H2 above the unit’s own baseline but no sustained rise — schedule the next laboratory DGA and review load history.
- Investigate: sustained hydrogen rate of rise, or H2 combined with any C2H2 — escalate to a 9-gas system or lab sample for fault typing per IEC 60599 / IEEE C57.104.
- Urgent: rapid rise or rising acetylene near a critical threshold — prepare inspection and load reduction while confirming.
This is the same decision logic used in real incidents; for worked examples, see DGA fault detection in real-world cases and the fire-prevention case study.
Monitoring That Prevents the Preventable
PAS DGA supplies the monitoring layers that turn early gas signals into scheduled action: the DGA-500 hydrogen monitor for fleet-wide screening and the DGA-900 for full fault typing on critical assets. The objective is not to predict every failure — it is to remove the failures that were detectable.
Contact PAS DGA to review how online DGA monitoring would change the risk profile of your transformer fleet.