Dissolved Gas Analysis (DGA) – Interpreting Transformer Oil Test Results

Every fault inside an oil-filled transformer leaves a chemical signature. Localised heating, partial discharge and arcing each break down mineral oil and cellulose insulation in a different way, releasing a characteristic mixture of gases that dissolves in the oil. Dissolved gas analysis reads that mixture and identifies what is happening inside a sealed tank without opening it.

It is the single most informative diagnostic test available for a transformer in service, and the earliest warning of a developing fault.

The Key Gases

Typical Concentration Limits

IEC 60599 gives ranges of gas concentrations observed in transformers without known faults. Values above the 90th percentile are considered abnormal and warrant closer monitoring.

Rate of Change Matters More Than Absolute Value

A transformer that has operated for twenty years with 120 ppm of hydrogen is behaving normally. The same transformer showing 120 ppm after a sample taken three months earlier showed 30 ppm is telling you something has changed.

The rate of gas generation is the more useful indicator:

• Below 10 % increase per month — normal ageing, continue routine sampling
• 10 to 30 % per month — increase sampling frequency, begin trending
• Above 30 % per month — active fault, investigate without delay

This is why a single DGA result has limited value. The test becomes diagnostic when there is a history to compare against.

The Duval Triangle

The Duval Triangle plots three gases — methane, ethylene and acetylene — as relative percentages of their sum, and reads the fault type from which zone the point falls into. It is the most widely used graphical interpretation method and is referenced in IEC 60599.

How to Plot a Point

Take the concentrations of the three gases in ppm and express each as a percentage of their sum:

Worked Example

A DGA report returns CH₄ = 90 ppm, C₂H₄ = 250 ppm, C₂H₂ = 5 ppm.

The high ethylene with very little acetylene is the signature of severe overheating without arcing. Likely causes include a bad joint in the winding circuit, a circulating current in the core or tank, or a blocked oil duct producing a local hotspot.

Fault Types and What They Mean

Limitations of the Method

The Duval Triangle always returns a diagnosis, whether or not a fault exists. Because there is no "normal" zone, a healthy transformer with trace levels of all three gases will still plot somewhere on the triangle.

This produces the most common misuse of the method: applying it to a unit whose absolute gas concentrations are well within normal limits, and then acting on a diagnosis that has no physical basis.

Other limitations worth knowing:

• The triangle identifies fault type, not location. A T3 result does not distinguish between a winding hotspot and a circulating current in the tank.
• Two simultaneous faults produce a point somewhere between their individual signatures, which may fall in a zone corresponding to neither.
• Zone boundaries differ for alternative fluids. Natural and synthetic esters require the adjusted boundaries of Duval Triangle 3.
• Units where on-load tap changer oil communicates with the main tank show elevated acetylene from normal switching operations, not from a fault.

Sampling Practice

Poor sampling invalidates the analysis. The gases of interest are dissolved in the oil and will escape if the sample is exposed to air.

• Use a gas-tight syringe or an approved sealed container, never an open bottle
• Flush the sampling valve before drawing the sample to clear stagnant oil from the pipework
• Take the sample with the transformer in normal service and at stable temperature, not immediately after energisation
• Exclude all air bubbles from the syringe — a visible bubble can significantly alter the reading
• Record the oil and ambient temperature, and the transformer loading at the time of sampling
• Deliver to the laboratory promptly and keep the sample out of direct sunlight

Recommended Sampling Intervals

Practical Notes

• Establish a baseline. The first DGA on a new transformer is the reference against which every later sample is judged.
• Use the same laboratory throughout. Inter-laboratory variation on the same sample can exceed the changes you are trying to detect.
• A sudden appearance of acetylene after a system fault may reflect that fault rather than an internal defect. Sample again after the unit has run normally for a period.
• Read DGA alongside the other oil tests — moisture content, dielectric strength, acidity and interfacial tension. Gases indicate what is happening; the physical tests indicate how much life the insulation has left.
• Thermal faults identified as T1 or T2 can often be correlated with the insulation thermal class and observed temperature rise. See Transformer Insulation Thermal Classes and Temperature Limits.