Dissolved gas analysis measures the gases that thermal and electrical faults produce inside an oil-filled transformer, and dissolve into the oil long before anything trips. One sample a year is the normal baseline for a distribution unit; every three to six months is realistic for a heavily loaded or critical transformer. Taking the sample is straightforward field work. Interpreting the result is not, and belongs with a specialist working to IEC 60599.
Which gases matter and what they point to
Each fault mechanism cracks the oil or the paper in a characteristic way, so the mix of gases tells you what kind of problem you have, not just that you have one.
- Hydrogen: partial discharge, and the earliest indicator of almost every developing fault.
- Methane and ethane: a low-temperature thermal fault, roughly below 300 °C.
- Ethylene: a hotter thermal fault, above roughly 300 °C, such as a poor joint or a stray-flux hot spot.
- Acetylene: arcing. A few ppm on a unit that was previously clean is enough to justify immediate investigation.
- Carbon monoxide and carbon dioxide: degradation of the cellulose paper insulation; the ratio between them matters more than either value alone.
- Oxygen and nitrogen: not fault gases, but a check on seal integrity and on whether the sample was taken cleanly.
Absolute concentrations matter less than the trend. A stable 150 ppm of hydrogen is a different situation from 60 ppm rising by 20 ppm a month. IEC 60599 sets out the ratio methods used to classify the fault, including the Duval triangle and the Rogers ratios; these are not a lookup you do by eye, and a wrong call in either direction is expensive.
How to take a sample that is worth analysing
Sample from the bottom drain valve with the transformer at normal operating temperature and under load, so the oil is mixed. Flush one to two litres through the valve first to clear the dead volume, then fill a glass syringe or a stainless steel cylinder without letting a single air bubble in. Record load, oil temperature, ambient temperature, tap position and date on the label, and keep the sample out of sunlight until it reaches the laboratory.
Poor sampling is a common cause of false alarms. An air bubble adds oxygen and nitrogen and strips out the light gases, which is exactly the part of the result you care about most.
What else the same sample tells you
Ask the laboratory for the physical and chemical oil properties alongside the gas analysis, because they come from the same bottle. Breakdown voltage to IEC 60156 shows the dielectric margin, water content in ppm shows moisture ingress and drives paper ageing, and the neutralisation number in mg KOH/g shows oxidation of the oil itself.
For an older unit, add a furan analysis. The 2-FAL content gives an indirect estimate of the degree of polymerisation of the paper, which is the honest measure of how much service life is left in the insulation.
How often to sample
Take a baseline at commissioning, then sample annually as a default. Move to six-monthly for units above roughly 80 % continuous load, for anything the site cannot run without, and for transformers older than 25 years. Sample additionally after any through-fault, overload event or protection operation, and before and after major work.
The practical rule is that a single DGA is a data point and three are a trend, so budget for the sequence rather than for one analysis. For the units you genuinely cannot afford to lose, continuous hydrogen and temperature monitoring through Protrafo Insights fills the gap between samples, but it does not replace the laboratory result or the specialist who reads it.
