IEEE C57.104 is the principal North American guide for interpreting dissolved gas analysis (DGA) results in mineral oil-immersed transformers. Where IEC 60599 asks "what kind of fault is it?", IEEE C57.104 leads with "how concerned should I be?" — assigning each result a DGA status that drives sampling frequency and intervention decisions. The current edition is a substantial rebuild grounded in a very large statistical study of in-service transformer data.
What it covers
The guide applies to mineral oil-immersed transformers and reactors of essentially all sizes, voltage classes and constructions. Its central tool is a status classification built from statistical gas-concentration norms, sample-to-sample change, and multi-point rate-of-change analysis. Rather than a single reference table, it stratifies its norms by the transformer's oxygen-to-nitrogen balance — a practical proxy for sealed versus free-breathing preservation — and by equipment age, recognising that gas accumulation behaves differently across these populations.
For identifying the underlying fault once a result is flagged, the guide carries Rogers Ratios and the Duval Triangle as first-tier methods, with further graphical methods, the Key Gas approach, and an energy-based severity metric set out in its annexes. It also devotes unusually careful attention to data quality — transcription errors, swapped samples, air exposure and contamination — on the principle that a confirmation sample should always precede emergency action. The guide deliberately excludes tap-changer DGA, ester and silicone fluids, and factory-test sampling, deferring each to its dedicated companion document.
Why it matters in practice
A status framework answers the question an asset manager actually asks: given this result, what do I do next, and how soon? By separating gas level from rate of change, the guide distinguishes a transformer carrying old, stable gas from one that is actively gassing — a distinction that absolute concentrations alone obscure. Its caution that DGA status is not the same as transformer condition is well taken: equipment can fail with little prior gassing, and units can run for years with elevated gas and no fault. The guide is explicit that wind turbine and network transformers were excluded from its statistical base, so applying its norms there tends to over-flag — a limitation worth respecting rather than ignoring.
How we use it
We use IEEE C57.104 as a severity-framing layer alongside our primary fault-type diagnosis, particularly for clients operating under North American practice or where a structured status classification is contractually required. Reading it in parallel with IEC 60599 gives useful cross-validation: when ratios point to a fault but every gas sits below the norms, the fault is likely historic or very early stage. We treat any confirmed rise in acetylene as significant regardless of absolute level, and we record preservation type independently, because the oxygen-to-nitrogen split that the norms rely on can be ambiguous in real fleets.