Detecting Partial Discharge in Transformers Using DGA

Partial discharge (PD) is among the most significant insulation degradation mechanisms in high-voltage transformers, and among the most difficult to detect reliably through periodic DGA. Unlike thermal faults, which produce gas at rates that scale directly with fault intensity, partial discharge can cause substantial cumulative insulation damage while producing dissolved gas at rates that remain below standard alert thresholds for extended periods.

Understanding how DGA detects PD, where its detection sensitivity is limited, and how to interpret PD-consistent gas signatures within a broader diagnostic framework is essential for asset managers responsible for high-voltage transmission equipment.

What Partial Discharge Is and Why It Matters

Partial discharge occurs when the local electric field in a small region of the insulation system, such as a gas-filled void in solid insulation, an oil-gap near a high-field electrode, or a contaminated surface, exceeds the local breakdown field strength. The result is a brief, localised electrical discharge that does not bridge the full insulation gap (which would be a flashover), but deposits energy in the discharge zone.

Individual PD events deposit small amounts of energy, typically picocoulombs to nanocoulombs per discharge pulse, but can occur at rates of thousands to millions of pulses per second under energised conditions. The cumulative effect of sustained PD activity is:

Erosion and carbonisation of insulation surfaces. Discharge energy dissociates polymer bonds in cellulose and pressboard, creating carbonised conduction tracks that progressively reduce the local insulation resistance.

Production of X-wax. In oil-immersed transformers, PD in oil produces hydrogen-saturated degradation products (X-wax) that deposit on insulation surfaces, further contaminating and degrading their dielectric properties.

Progressive pit formation in oil. Sustained PD in oil creates micro-voids and chemical degradation products that reduce the oil's dielectric strength in the discharge zone.

Over extended periods, PD can progress from a localised incipient condition to a developed fault that bridges the insulation gap. The transition from incipient PD to advanced fault can be sudden and may not provide DGA warning at a rate that allows intervention under quarterly sampling.

Gas Signature of Partial Discharge

Partial discharge in mineral oil produces a characteristic dissolved gas signature dominated by hydrogen (H₂) and methane (CH₄), with minimal or no ethylene (C₂H₄) or acetylene (C₂H₂) [1]. The physical reason for this signature is the energy regime of PD: discharge events deposit energy through ionisation and bond dissociation at electron temperatures far above what thermal decomposition produces, but the total energy per event is small. The high-energy bond dissociation produces H₂ readily; CH₄ formation requires lower activation energy than the longer-chain hydrocarbons that require sustained high temperature for significant generation.

In the Duval Triangle [1], PD activity plots in the PD zone, characterised by a high proportion of CH₄ relative to C₂H₄, with minimal C₂H₂. The Duval Pentagon [1], which includes H₂ directly, provides additional discrimination between the PD zone and the low-energy discharge (D1) zone, which can overlap in the Triangle.

IEEE C57.104 [2] and IEC 60599 [3] both identify elevated H₂ with modest CH₄ and minimal heavier hydrocarbons as the signature pattern for PD activity. However, neither standard provides the sensitivity needed for early detection of incipient PD, because the absolute concentrations that PD generates in early stages are frequently below the threshold values that trigger follow-up actions.

The Detection Sensitivity Problem

The fundamental challenge of using DGA for early PD detection is that PD can be present and active at energy levels and repetition rates that deposit gas at rates of a few ppm per month, rates that are difficult to distinguish from normal background gas generation variation [4], particularly when annual or quarterly sampling provides only infrequent data points.

Consider a transformer with a stable hydrogen baseline of 40 ppm over several years. PD activity initiates; hydrogen begins rising at 3 ppm per month. After one quarter (90 days), hydrogen is at 49 ppm, a change of 9 ppm that is well within laboratory measurement uncertainty [4] and produces no threshold exceedance against IEEE C57.104 [2]. After two more quarters, hydrogen is at 67 ppm, still within the normal range but now showing a clear trend over three consecutive samples.

Against threshold-based interpretation, this transformer generates no alerts until concentrations cross a condition boundary, which may be months or years after PD activity became detectable as a trend. Against rate-of-change analysis, the trend is visible in the second quarter: three consecutive rising samples at an accelerating rate is statistically distinguishable from noise.

This is precisely the application where the Hazard Factor (HF) metric in R-DGA methodology [4] adds detection capability that threshold methods cannot provide. HF responds to changes in the gas generation rate relative to the transformer's historical baseline, flagging a transformer whose hydrogen generation rate has increased significantly over the past several samples, independently of whether any absolute threshold has been crossed.

Distinguishing PD From Other Low-Gas-Level Conditions

Several conditions can produce hydrogen and methane without active PD, and distinguishing them correctly avoids unnecessary investigations and interventions.

Stray gassing. Some new insulating oils produce small amounts of hydrogen and hydrocarbons through a process called stray gassing, a non-fault-related oil degradation that occurs in the presence of the cellulose insulation at normal operating temperatures in early service life. Stray gassing typically produces low, stable H₂ and CH₄ without clear trend acceleration. It is benign, but it must be distinguished from PD-related gas generation.

Residual assembly gases. Hydrogen can be trapped in transformer insulation from manufacturing and installation processes. Newly commissioned transformers may show elevated hydrogen that gradually decreases over the first months of service as residual gas dissipates.

Normal ageing background. Background gas generation from slow thermal ageing of oil and insulation produces modest H₂ and CH₄ that increases gradually over the transformer's service life. This background must be accounted for when assessing whether an observed trend represents fault activity.

The discriminating features between PD-related gas generation and these benign conditions are: (1) trend acceleration: PD-related generation should show clear acceleration from a previous stable baseline; (2) Duval Triangle position: PD plots in the PD zone, while stray gassing and background aging typically do not produce the same ratio patterns; (3) CIGRE TB 771 [5] provides updated guidance on distinguishing fault-generated gas from background generation for the most common scenarios.

Follow-Up Methods After DGA Suggests PD

DGA results consistent with PD activity indicate that discharge has occurred, but they do not identify where in the transformer or at what intensity. Follow-up diagnostic testing guided by IEEE C57.152-2013 [6] is typically warranted before any maintenance decision.

Acoustic PD detection. Acoustic emission sensors applied to the transformer tank wall can detect the acoustic signature of PD discharge events and, through triangulation using multiple sensors, localise the PD source within the tank. Acoustic PD is sensitive to active discharge, providing real-time information about whether PD is occurring and, approximately, where.

Ultra-high frequency (UHF) PD detection. UHF sensors placed at oil valves or in the gas space detect the electromagnetic emissions from PD discharge in the GHz frequency range. UHF is effective for detecting PD in the winding insulation and provides directional sensitivity for source localisation.

Frequency domain spectroscopy. Dielectric spectroscopy techniques can detect moisture and insulation degradation products in the insulation system that are associated with PD activity, providing a different dimension of information from the acoustic and UHF methods.

These confirmatory tests answer the question DGA cannot: where is the PD occurring and how severe is the discharge? The combination of DGA (detecting that PD gas has been generated) with UHF or acoustic detection (identifying the source location) provides the information needed for an informed maintenance decision.

For transformers with confirmed active PD activity and high operational consequence, continuous online DGA monitoring through Monitor Watch provides the rapid trend detection between sampling intervals that matters when PD can progress quickly.

For technical background on fault detection methodology, visit the Science page. For product information, visit the TOA page or contact us.