Power transformers are designed to pass large amounts of energy through with very high efficiency, so in normal operation and in the absence of unusual stress there should be no degradation of the oil and paper insulation inside the transformer. Insulation damage or degradation resulting from an abnormal condition (“fault”) is almost always detectable by the formation of characteristic gaseous byproducts in the transformer oil. The main concern of transformer dissolved-gas analysis (DGA) is to detect, quantify, and interpret the active production of fault-related gas when it occurs.
The first step in the interpretation of DGA data is to check for problems with the integrity of the data. Issues with the data must be recognized and dealt with before any conclusions can be reached about the status of the transformer. Interpretation of the gas analysis results for an oil sample requires comparison with results for earlier samples. If successive sample results agree well, it is usually safe to conclude that they represent the state of the transformer fairly. If successive sample results do not agree, or if they don’t seem to make sense, then the possibility of bad data must be considered before concluding that a transformer fault is the best explanation.
With that in mind, here are six data issues commonly encountered in transformer DGA.
1. Gas concentrations vary greatly from sample to sample
Poor repeatability of gas concentration measurements can mask small changes indicating a problem in its early stages and make it very difficult to identify and assess any fault that is detected. Good gas concentration measurement quality depends on the collection of oil samples that are representative of the bulk of the transformer oil. With good sampling and good lab quality control, measurement uncertainty is typically about 10-15%.
2. One oil sample’s gas concentrations are very different from the other samples
A human error may have occurred such as: mislabeling a sample from a different oil compartment or a different transformer; interchanging gas values, such as N2 and O2; or duplicating, reversing, or dropping digits in a number. It is also possible that the exceptional sample results are explained by an isolated bad oil sample due to errors in sampling, handling, or lab processing.
3. Contradictory information
If hydrogen concentration decreases sharply while other gases remain roughly the same as before or even increase, the oil sample might have been exposed to air or a gas bubble might have been expelled from the sample syringe.
If a gas concentration other than hydrogen decreases significantly while the other gases are unchanged or increasing, the decrease is likely due to a transcription error as mentioned in (2). A gas increase while the other gases are unchanged or decreasing might also be the result of a transcription error or might be caused by stray gassing.
4. Hydrogen concentration is extremely low in all samples
Hydrogen gas escapes quickly if there is any chance, and a transformer will gradually lose other gases if it “breathes” by design or there is a leak such as a ruptured conservator diaphragm. Persistent extremely low hydrogen concentrations may indicate that gas loss is occurring, which could result in faults to be missed or their severity to be underestimated.
It is also possible that consistently low hydrogen readings can be due to a sensor problem or misconfiguration of the lab’s gas chromatograph, in which case the reported concentrations of the other gases are not necessarily inaccurate.
5. After oil is degassed or replaced, gas concentrations trend upward again.
Oil confined within the layers of paper winding insulation is largely unaffected by degassing or replacement of the main tank oil. After the degassing or replacement, gas from the winding insulation oil diffused over weeks or months into the clean oil until the gas concentrations inside and outside the paper insulation equalize. Usually (assuming that no new gas is produced meanwhile) the final effect is to restore about 10% to 15% of the gas level reduction. The rise in gas concentrations following degassing or oil replacement must not be mistaken for active gas production.
6. Recurrent gas loss.
What you need to know: For a sealed or gas-blanketed transformer, recurrent and simultaneous decreases of gas concentrations can mask fault-related gas production and turn a steady upward trend into a “saw tooth” pattern. That kind of behaviour is sometimes seen when operational thermal cycling of a transformer causes periodic expulsion of headspace gas. There are also cases where transformer operators degas a transformer whenever gas concentrations exceed a limit, in the mistaken belief that the gases themselves are harmful to the transformer. Recurrent gas loss, like the gas leaks mentioned in (4), can conceal or badly misrepresent evidence of fault-related gas production.
Recommendations
1. When abnormal DGA results are discovered, get a verification sample to confirm that there is a transformer problem rather than a data problem before taking drastic action.
2. Use commercially prepared gas-in-oil standards for checking the quality of lab DGA results. Also occasionally collect a transformer oil sample in several syringes at the same time and distribute them to different labs to compare results.
3. Data problems are not always the lab’s fault. Be as careful as possible with sampling, labelling, and shipping oil samples. Keep the lab informed of all changes to equipment inventory and identification.
