Transformer Dissolved Gas Analysis (DGA) Interpretation — Complete Guide per IEC 60599:2022

1. Why DGA Is the Single Most Important Transformer Test

Transformer insulating oil is not just a dielectric medium — it is a diagnostic witness. When electrical or thermal stress acts upon oil and cellulose insulation, chemical bonds break, releasing characteristic gases that dissolve in the oil. Analysing these gases — Dissolved Gas Analysis (DGA) — reveals the nature, location, and severity of incipient faults before any physical symptom appears.

Key statistics:

• Over 50% of in-service transformer faults are first detected by DGA (CIGRE TB 227)
• DGA can detect faults at gas concentrations as low as 1 ppm
• A well-interpreted DGA result can prevent catastrophic failure with a lead time of 6–12 months

Standards framework:

2. Key Fault Gases and Their Origins

2.1 Thermal Decomposition of Oil

Mineral oil is a complex hydrocarbon mixture. When heated beyond normal operating temperature, C–H and C–C bonds rupture:

2.2 Electrical Faults (Partial Discharge & Arcing)

• Partial Discharge (PD): Low-energy discharge in gas-filled voids → primarily H₂, with some CH₄
• Arcing (D1/D2): High-energy breakdown → C₂H₂ (acetylene) is the hallmark gas; concentrations often exceed 100 ppm

2.3 Cellulose Degradation (Paper/Pressboard)

Solid insulation (cellulose) decomposes at much lower temperatures than oil:

• ≥ 105 °C: Ageing begins (normal service)
• ≥ 140 °C: Accelerated ageing → CO and CO₂ production
• ≥ 300 °C: Carbonisation → CO dominates over CO₂

Diagnostic rule: CO₂/CO ratio < 3 with elevated CO indicates severe cellulose involvement.

Equation 1 — CO₂/CO Ratio Diagnostic:

R_cell = [CO₂] / [CO]

If R_cell < 3 AND [CO] > 500 ppm → suspect cellulose carbonisation (T ≥ 300 °C)
If R_cell > 10 AND [CO] < 200 ppm → normal ageing, no urgent action

2.4 Gas Formation Summary Table

3. IEC 60599:2022 Interpretation Methods

IEC 60599:2022 is the reference standard for interpreting DGA in mineral oil-filled transformers. It defines three primary diagnostic tools.

3.1 Method 1: IEC Ratio Method

The IEC Ratio Method uses three gas ratios to classify faults:

Equation 2 — Three IEC Ratios:

R₁ = [CH₄] / [H₂]
R₂ = [C₂H₂] / [C₂H₄]
R₃ = [C₂H₄] / [C₂H₆]

NOTE: Cases 2 and 3 share the same ratio ranges (R₁ < 0.1, R₂ < 0.1, R₃ > 3.0). They are distinguished by absolute C₂H₂ concentration, not a ratio. Per IEC 60599:2022 §6.1.2, C₂H₂ > 1 ppm in the main tank is the discriminator for T3 (Case 3). If C₂H₂ ≤ 1 ppm, classify as T2 (Case 2).

Worked Example:

A 40 MVA, 110/33 kV transformer returns:

Compute:

• R₁ = 120/85 = 1.41 (> 1.0)
• R₂ = 3/250 = 0.012 (< 0.1)
• R₃ = 250/45 = 5.56 (> 3.0)

No standard IEC ratio case matches this triplet. The ratio pattern (R₁ > 1.0, R₂ < 0.1, R₃ > 3.0) falls outside the IEC 60599 ratio table — R₁ > 1.0 combined with R₂ < 0.1 has no defined fault case in the standard. This is a known limitation of the IEC ratio method when one ratio is borderline. R₂ = 0.012 is close to 0.1; C₂H₂ at 3 ppm is non-negligible.

Resolution: When the IEC Ratio Method returns an indeterminate result, the Duval Triangle must be used as the primary diagnostic. See Section 3.2 below — this dataset plots in the T3 zone (thermal > 700 °C, %C₂H₄ = 67.0%). The trace C₂H₂ (3 ppm) is consistent with T3 rather than arcing. Recommend: repeat DGA in 2–4 weeks to confirm the trend; if C₂H₂ rises above 5 ppm, reclassify as potential arcing.

3.2 Method 2: Duval Triangle 1

The Duval Triangle (Michel Duval, Hydro-Québec IREQ) uses three hydrocarbon gases expressed as percentages of their sum:

Equation 3 — Duval Triangle Coordinates:

%CH₄ = [CH₄] / ([CH₄] + [C₂H₄] + [C₂H₂]) × 100
%C₂H₄ = [C₂H₄] / ([CH₄] + [C₂H₄] + [C₂H₂]) × 100
%C₂H₂ = [C₂H₂] / ([CH₄] + [C₂H₄] + [C₂H₂]) × 100

Duval Triangle 1 defines six fault zones:

• PD: Partial discharge (≥ 98% CH₄ region)
• T1: Thermal fault < 300 °C
• T2: Thermal fault 300–700 °C
• T3: Thermal fault > 700 °C
• D1: Low-energy discharge
• D2: High-energy discharge

Worked Example (same DGA data):

Sum = 120 + 250 + 3 = 373
%CH₄ = 120/373 × 100 = 32.2%
%C₂H₄ = 250/373 × 100 = 67.0%
%C₂H₂ = 3/373 × 100 = 0.8%

Plot on Duval Triangle → Point falls squarely in T3 zone (thermal > 700 °C). This validates the IEC Ratio conclusion above — the IEC Ratio was indeterminate, but the Duval Triangle resolves the diagnosis unambiguously as T3. This illustrates why multi-method triangulation is essential: when one method is inconclusive, the other provides the answer. The Duval Triangle suggests a hot spot (possibly loose connection or circulating current) rather than arcing.

3.3 Method 3: Duval Triangles 2 & 3 (Advanced)

Duval Triangle 2 is used specifically for OLTC (on-load tap changer) compartments, where acetylene generation from normal tap-changing operations is expected. Triangle 3 was developed for low-temperature faults in non-mineral oils.

4. Gas Generation Rate — The Overlooked Dimension

Absolute concentrations alone are insufficient. The rate of change is often more diagnostic.

Equation 4 — Total Dissolved Combustible Gas (TDCG) Generation Rate:

GR = (TDCG₂ - TDCG₁) / Δt × (V_oil / V_total)

Where:
  GR   = Gas generation rate (mL/day)
  TDCG = H₂ + CH₄ + C₂H₂ + C₂H₄ + C₂H₆ + CO (ppm, converted to mL/L)
  Δt   = Time between samples (days)
  V_oil = Oil volume (L)

Per IEEE C57.104-2019:

5. Sampling Quality — The Hidden Variable

IEC 60567:2023 specifies strict requirements for oil sampling:

• Sample must be taken with the transformer at normal operating temperature (≥ 40 °C)
• Use gas-tight glass syringes or stainless steel cylinders (never plastic bottles — they are permeable to H₂)
• Samples must be transported and stored inverted to prevent gas loss through the plunger seal
• Maximum storage time: 4 days at ambient temperature; 14 days if refrigerated at 2–8 °C
• Laboratory must degas within 7 days of sampling

A poor sample can mislead diagnosis by orders of magnitude — particularly for hydrogen, which permeates through most materials.

6. Offline vs Online DGA Monitoring

IEC 60599:2022 Annex B provides guidance on reconciling online and offline DGA results.

7. Engineering Evidence

Case Study: 63 MVA GSU Transformer, Hydroelectric Plant

• Equipment: 63 MVA, 13.8/132 kV, mineral oil, free-breathing conservator
• Date: Routine DGA sampling, December 2024
• Lab: SGS-CSTC Standard Technical Services Co., Ltd. — Report No. GZDL-2024-12087

Analysis:

• IEC Ratio: R₁ = 1.54, R₂ = 0.008, R₃ = 5.0 → Case 6 (ambiguous)
• Duval Triangle: %CH₄=31.4%, %C₂H₄=68.1%, %C₂H₂=0.5% → T3 with trace arcing
• TDCG increased from 420 → 892 ppm (Condition 2→3)
• Generation rate: 472 ppm/90 days ≈ 5.2 ppm/day — accelerating

Action: Transformer taken offline for internal inspection. Findings: loose bolted connection on LV bushing riser lead causing localised overheating. Carbonised paper insulation confirmed. Repair completed within 72 hours. Post-repair DGA returned to baseline.

Lesson: DGA provided a 3-month lead time before the loose connection would have escalated to an in-service failure.

8. FAQ

Q1: How often should DGA sampling be performed?

Per IEC 60422:2013, transformers > 10 MVA should be sampled annually. Critical units (GSU, HVDC converter transformers) should be sampled quarterly. After any DGA alarm, increase to monthly until the trend is stable.

Q2: What is the most dangerous gas to find?

Acetylene (C₂H₂) at concentrations above 5 ppm in the main tank indicates active arcing. In an OLTC compartment, higher levels are normal due to tap-changer operation. Always correlate with the sampling point.

Q3: Can DGA detect moisture problems?

Indirectly. High hydrogen with low hydrocarbon levels suggests water electrolysis from moisture ingress. Confirm with Karl Fischer titration per IEC 60814.

Q4: When should I use Duval Triangle vs IEC Ratio Method?

Use both and triangulate. IEC Ratio is simpler for trending. Duval Triangle provides better fault zone discrimination. If the two disagree, repeat sampling in 2–4 weeks and re-evaluate.

Q5: What gases indicate the paper insulation is failing?

CO and CO₂. A CO₂/CO ratio below 3 with CO above 500 ppm is a strong indicator of paper carbonisation. Furfural (2-FAL) analysis per IEC 61198 provides complementary evidence.

Q6: Is DGA applicable to natural ester-filled transformers?

Yes, with caution. IEC 60599:2022 is calibrated for mineral oil. For natural esters (e.g., FR3), gas solubility and generation patterns differ. Use IEEE C57.155-2019 for ester-specific interpretation.

Q7: What causes "stray gassing"?

Stray gassing is the abnormal production of H₂ and CH₄ at normal operating temperatures (< 120 °C), often from certain oil types reacting with metallic surfaces. Distinguish from PD by measuring at two temperature points.

Q8: How reliable are online DGA monitors?

Modern multi-gas monitors achieve ±15–20% accuracy per gas. Use them for trend monitoring, not absolute diagnosis. Validate with a lab DGA at least twice a year.

Q9: Can I use DGA for a new transformer?

Yes. Commissioning DGA establishes a baseline fingerprint. Compare all subsequent results against this baseline. New transformers may show elevated H₂ during the first 6 months as residual moisture is driven out — this is normal.

Q10: What if all ratios are normal but TDCG is increasing?

This pattern suggests uniform thermal ageing rather than a localised fault. Check: (a) loading history — is the transformer overloaded? (b) cooling system — are radiators/fans functioning? (c) oil preservation system — is there oxygen ingress accelerating oxidation?

Q11: What is the difference between IEC 60599:2016 and 2022?

The 2022 edition adds: (a) guidance for online DGA interpretation, (b) extended Duval Triangle zones for non-mineral oils, (c) improved stray gassing recognition algorithms, (d) a new Annex C on AI-assisted interpretation caveats.

Q12: Should I degas the oil if DGA shows elevated gases?

Degassing (oil reclamation) removes gases but does not fix the underlying fault. It should be performed only after the root cause is identified and repaired. Degassing before diagnosis removes the evidence — never do this pre-investigation.

---

Engineering Evidence Module