This historical case concerns a 100 MVA, 100/110 kV generator step-up (GSU) transformer, reported as built in 1980, with elevated hydrogen, ethylene and acetylene in its oil. The utility used filtration and degassing, but oil treatment alone could not establish whether the source of gas generation had been removed. The case record does not reconcile the originally reported failure date with the later samples; it therefore supports a retrospective discussion of the data, rather than a verified failure timeline.
Transformer condition through data
The utility collected periodic DGA results and other oil-test data for assessment against IEEE and IEC guidance. Occasional moisture and paper-degradation information was also available. Periodic laboratory samples provide snapshots and trends, rather than continuous surveillance. The account contains no inspection record sufficient to rule out leakage or establish every internal defect.
Gas trends and fault interpretations
The supplied history describes unusual main-tank gassing following an outage in 2016. Hydrogen, ethylene and acetylene were the main concerns. Their concentrations and changes need to be assessed together: hydrogen can arise from several mechanisms, and its presence alone is not a thermometer or a fault diagnosis.
DGA trends can help identify possible fault mechanisms and active gas generation. They do not, on their own, locate the defect or determine when failure will occur. Seetalabs reviewed the available condition data with RONIN to help organize the evidence and identify priorities for further investigation.
The condition data was uploaded to the RONIN dashboard to obtain a transformer health-index score and its trend. This provides a relative screening view. As CIGRE TB 761, section 2.2, explains, a numerical ranking is not automatically a calibrated probability of failure or an estimate of remaining service life.
Ronin AI discoveries
The supplied dataset is described as covering 2016 to 2021, with periodic DGA and most oil-quality parameters, but no interfacial tension (IFT). That missing input and the unresolved event chronology limit the conclusions. We retain the reported score history below as an illustration of how the assessment changed with the available measurements.
The reported health-index score dropped from 71.6% to 63.8% over 20 days in 2016, moving from the historical dashboard’s good band to its fair band. The account associates this change with rising gas concentrations and reports degassing afterward. These labels belong to that implementation; they are not universal CIGRE condition thresholds.
The account reports a further score of about 50% at the beginning of 2017 and about 38% in 2019. The downward trend warranted a closer look at the input measurements and maintenance history. It does not, by itself, demonstrate a particular failure probability or establish the cause of deterioration.
The original review also raised concerns about carbon oxides, 2-FAL and moisture. The detailed values and applicable reference conditions are not reproduced here, so we cannot independently classify those results. They should be reviewed alongside gas trends, oil treatments and the available paper-condition evidence.
The reported oil treatments were followed by further low index scores. That observation is a reason to investigate persistent gas generation and other condition indicators. It cannot show, on its own, that insulation degradation was rapid, that a specific internal hot spot existed, or that a repair had failed.
Conclusions
The latest score quoted in the historical account is 61.6%, in the dashboard’s fair band. A higher score after oil treatment is not proof that an internal defect has been repaired: treatment changes dissolved-gas concentrations and the subsequent trend must be reviewed. The available record does not support a six-month intervention deadline or a claim that replacement was avoided. The useful lesson is to retain the measurements, treatment dates and inspection findings together, then let an engineer determine the next action from that complete record.
