Partial Discharge in Power Transformers: What It Is, Why It Matters, and How to Detect It Early

Partial discharge is one of those failure mechanisms that maintenance engineers either take very seriously or ignore almost entirely. The ones who ignore it tend to do so because PD activity is invisible, silent, and easy to dismiss as background noise until the transformer fails in a way that is neither invisible nor silent.

This article explains what partial discharge actually is at a physical level, why it is a reliable early indicator of insulation failure, and what detection methods are practical for industrial maintenance teams without specialist high-voltage laboratory equipment.

The Physics, Without the Textbook Version

Transformer insulation — oil-impregnated paper, pressboard, cast resin in dry-type units — is not perfectly uniform. Manufacturing variations, aging, moisture ingress, and mechanical stress from thermal cycling all create microscopic voids, delamination zones, and contamination points within the insulation structure.

Under normal operating voltage, these defects are tolerable. The electric field distributes itself across the insulation and the defect regions carry slightly higher field stress than the surrounding material, but not enough to cause breakdown.

When the local field stress at a defect exceeds the dielectric strength of the gas or degraded material filling that void — and this threshold is lower than you might expect for a void containing air or moisture — a micro-arc occurs. The discharge lasts nanoseconds, dissipates a small amount of energy, and leaves behind chemical byproducts: ozone in gas-filled voids, hydrogen and low-molecular-weight hydrocarbons in oil.

The discharge also produces a high-frequency current pulse that propagates through the transformer's capacitive structure, and electromagnetic radiation across a broad frequency spectrum.

Each individual discharge event is trivial. The problem is cumulative. Each discharge chemically attacks the surrounding insulation material, enlarges the void slightly, and makes the next discharge more likely and more energetic. Over months to years, this progressive erosion creates conductive pathways through the insulation. When those pathways connect the high-voltage conductor to ground or to an adjacent phase, the result is a full dielectric breakdown — a failure mode that typically involves significant damage to the transformer and is not field-repairable.

Why PD Detection Is Complicated

If PD produces measurable electrical and electromagnetic signals, why is it not detected routinely?

The primary difficulty is signal-to-noise ratio. The current pulses produced by partial discharge are real but small — typically in the picocoulomb to nanocoulomb range. Industrial electrical environments generate electromagnetic noise orders of magnitude larger from variable frequency drives, switching power supplies, corona discharge on overhead lines, and radio frequency sources.

Distinguishing genuine PD signals from this background requires either careful sensor placement, frequency-domain filtering, or both. It also requires some interpretation skill — a high PD reading is not automatically a crisis, and a low reading does not guarantee a healthy transformer.

The second difficulty is accessibility. The most sensitive PD measurement methods require direct electrical connection to the transformer's capacitive tap points on bushings, or physical access to install sensors during a planned outage. Retrofitting a monitoring system to a transformer that has been in service for twenty years, in a substation where taking outages is operationally difficult, is a real constraint.

Detection Methods in Practice

*Capacitive coupling via bushing tap* is the most sensitive method for online PD monitoring. Most HV bushings have a capacitive tap — a small terminal on the bushing body — originally intended for power factor testing. A coupling capacitor connected to this tap, feeding into a high-frequency current transformer and signal conditioning electronics, gives a direct measurement of PD-induced current pulses on the HV conductor. Sensitivity is excellent and the installation is non-intrusive once the coupling capacitor is in place.

The limitation is that not all bushings have accessible taps, and the measurement is local to that bushing. A PD source deep in the main tank may have attenuated significantly by the time its signal reaches the bushing tap.

*High-frequency current transformers (HFCTs) on grounding conductors* are the most practical retrofit option for most installations. Clamp-on HFCTs installed on the transformer tank grounding strap, neutral grounding connection, or cable screen grounds capture PD-induced current pulses without requiring any modification to the transformer itself. Sensitivity is lower than bushing tap coupling but sufficient for detecting significant PD activity, and installation can be done live.

*UHF sensors* pick up the electromagnetic radiation from PD events in the GHz frequency range. They are the standard monitoring method for gas-insulated switchgear (GIS) because they can be installed through drain valves without taking the equipment out of service. For conventional oil-filled transformers, UHF sensors are less commonly used because the metal tank attenuates the signal, but they are viable for dry-type transformers and for transformers with dielectric windows or access points.

*Acoustic emission sensors* detect the ultrasonic pressure waves generated by PD events, typically in the 100–300kHz acoustic frequency range. Piezoelectric sensors attached to the transformer tank wall are non-invasive and require no electrical connection to the HV system. Their main advantage is that acoustic triangulation — using multiple sensors at different positions on the tank — can locate the PD source within the transformer, which is diagnostically very useful. The limitation is that signal attenuation through steel and oil means sensitivity drops off significantly for PD sources more than a meter or two from the sensor.

*Dissolved gas analysis* is not a direct PD measurement but it detects PD indirectly through the chemical signature. Hydrogen is the primary dissolved gas produced by partial discharge in oil. Sustained PD activity produces a characteristic DGA profile with elevated hydrogen, sometimes accompanied by methane, that is distinguishable from thermal fault signatures. Online DGA sensors provide continuous trending; laboratory DGA provides more accurate quantification. For facilities without installed PD sensors, a sharp increase in dissolved hydrogen on a routine DGA sample is often the first indication of developing PD activity.

Interpreting What You Find

A PD measurement gives you a magnitude — typically expressed in picocoulombs (pC) for electrical methods — and a pattern. Magnitude alone is not a reliable alarm criterion. A transformer with 500pC of PD that has been stable for two years is less concerning than one showing 100pC that doubled in the last month.

Trend rate matters more than absolute level. A PD monitoring system that records continuously and alerts on rate-of-change is more useful than one with fixed magnitude thresholds.

Phase-resolved PD analysis — plotting discharge magnitude and repetition rate against the phase angle of the supply voltage — produces patterns that are characteristic of specific defect types. Void discharge in solid insulation produces a symmetric pattern centered on the voltage peaks. Surface discharge produces an asymmetric pattern. Corona discharge from sharp electrode edges produces a pattern concentrated in one polarity. These signatures help maintenance engineers prioritize the urgency of follow-up action.

For most industrial maintenance teams, the practical workflow is: install HFCT sensors on grounding conductors, log continuously, set rate-of-change alerts, and use DGA to cross-check. When PD activity increases sharply or DGA shows correlated gas generation, that is the trigger for a more detailed investigation — ideally with portable high-sensitivity equipment during the next available planned outage.

The Cost Argument

A complete HFCT-based PD monitoring installation for a single transformer — sensors, signal conditioning, edge gateway, connectivity — costs in the $3,000–8,000 range depending on specification. A transformer rewinding or replacement following a failed insulation event costs $50,000 to $400,000 plus the operational cost of unplanned downtime.

The arithmetic is straightforward. The reason PD monitoring is not universal is not economic — it is organizational. Maintenance budgets are easier to defend than monitoring investments whose value is measured in failures that did not happen.