Partial Discharge in Transformers and GIS

Learn more about Partial Discharge in Transformers and GIS

Partial discharge (PD) is a localized electrical breakdown in insulation that does not completely bridge the conductors. PD appears as tiny sparks or pulses, typically occurring at voids, cracks, or sharp interfaces where the local electric field exceeds the dielectric strength. In power transformers and gas‑insulated switchgear (GIS), PD is a critical indicator of insulation health: it often initiates in microscopic defects and can erode insulation over time. Indeed, high-voltage transformers—being key to system reliability—are known to experience most of their insulation failures from PD. Similarly, in GIS the presence of internal PD signals distortion of the field in the SF₆‑insulated system; untreated PD leads to insulation deterioration and risk of flashover. Because PD pulses can only occur where small gaps or impurities exist, their detection and analysis provide early warning of defects before catastrophic failure.

Common PD Defect Types

In both transformers and GIS, PD arises from characteristic defect types:

• Voids and Bubbles (Internal PD): Small cavities in solid insulation (e.g. air or gas gaps in oil‐impregnated paper, epoxy, or PCB) concentrate the field and initiate PD. In transformers this includes voids between pressed paper layers, imperfect impregnation, or air bubbles in oil. In GIS, void-like “tip” or floating electrode defects (e.g. loose metal particles or conductor tips) act similarly, causing internal discharges within the SF₆ gas.

• Delamination/Cracks: Layers of insulating paper or pressboard can delaminate under stress, creating extended void channels. Sharp cracks or delaminated layers often originate in transformer windings under thermal/mechanical stress and lead to internal PD.

• Corona Discharge: PD also occurs as corona on sharp edges or protrusions at air/gas interfaces. In transformers, corona often appears at HV bushings, terminations, or on oil–air boundaries. In GIS, corona can form on conductor edges or at internal guard rings before breakdown. Corona is characterized by small, continuous glow-discharge and typically produces UV, acoustic and RF emissions (though pure corona is less destructive than internal void discharges).

• Surface Tracking/Discharge: When PD propagates along the surface of solid insulation under contamination or moisture, it is called surface discharge or tracking. In transformers, this may occur on tank wall coatings, porcelain bushings or surge arresters. In GIS, the inner surfaces of solid insulation (insulators, spacers) can undergo surface PD if exposed or contaminated. Surface PD has long, branching paths and is highly erosive.

In practice, mixed defects are common. For example, a transformer winding may have internal bubbles and surface creepage on its top layers. In GIS, a single event may combine corona and floating metal effects. Recognizing these defect types is crucial, as each produces distinct PD signatures in measurement data.

Offline vs. Online PD Testing Techniques

Offline (Type/Acceptance) PD Testing: Offline PD tests are conducted with the equipment de‑energized, typically during factory acceptance, commissioning, or major maintenance. The standard method follows IEC 60270 (“conventional PD measurement”). A high-voltage source applies a controlled AC or DC stress (often up to ~1.5–2× rated voltage) to each winding or phase, while sensitive PD detectors (via capacitive coupling) record discharges. A precision coupling capacitor (CC) and measuring impedance are inserted between the DUT (device under test) and ground to pick up the high-frequency PD pulses. The applied voltage is usually pre-stressed above the intended test level, then held at the test level while PD activity is recorded. Offline tests precisely measure PD inception/extinction voltages and pulse magnitude (in picocoulombs) for each phase, under low-noise conditions. This method is highly repeatable and standardized, as defined by IEC 60270, but it requires taking the equipment out of service.

Online (In-service) PD Testing: Online testing monitors PD under normal operating or light-loaded conditions, using non-invasive sensors. It is widely used for asset condition monitoring without service interruption. Online PD detection relies on high-frequency coupling of PD signals into measurement devices. For transformers, this typically means clamping high-frequency current transformers (HFCTs) on the tank ground or neutral conductors, and placing acoustic/ultrasonic sensors on the tank or bushings. For GIS, UHF antennas or VHF capacitive couplers are installed inside gas compartments or on voltage transformer (VT) ports, capturing electromagnetic pulses. Online techniques often cover broad frequency ranges (e.g. 30 MHz–3 GHz) and use digital filtering to differentiate PD from noise. Common methods include VHF/UHF radio-frequency sensing and acoustic emission (AE) detection. Online monitoring is less standardized (IEC 60270 does not cover it) and typically relies on trend analysis and pattern recognition to identify developing PD.

PD Sensor Types and Installation

Various sensors capture PD signals in transformers and GIS:

• Coupling Capacitors (CC): Used in offline tests per IEC 60270, CCs provide a capacitive path for the PD pulse current from the DUT to the PD detector. In practice, voltage transformers or test capacitors are connected at each high-voltage terminal or bushing. During factory tests, each phase’s coupling cap is connected to the PD instrument, which measures displacement current (in pC) through a 50 Ω impedance. CCs also supply a reference for phase-resolved PD measurements (allowing φ–q correlation).

• HFCTs (High-Frequency Current Transformers): HFCT clamps are placed around available ground or neutral connections to capture PD current pulses magnetically. In transformers, an HFCT is typically installed around the transformer grounding lead or neutral bushing tap cable, allowing PD measurement without high-voltage isolation. Similarly, an HFCT may be placed on the metallic enclosure or gas-grounding conductor of GIS. HFCTs are sensitive to pulses up to tens of MHz and are ideal when a ground path is accessible.

• UHF Antennas (RF Sensors): UHF/VHF sensors (patch antennas or monopoles) detect electromagnetic emissions from PD discharges. In GIS, UHF sensors are usually installed inside the metal enclosure at dielectric joints or accessible ports, since SF₆ is an excellent RF waveguide. In transformers, UHF probes can be inserted via oil ports or fitted on the tank (often near windings). UHF detection covers roughly 200 MHz–3 GHz, a band where PD pulses are strong and many lower-frequency interferences are attenuated. UHF sensors have been standard in GIS for decades and are now increasingly used on transformers as well.

• Acoustic (Ultrasonic) Sensors: Piezoelectric AE sensors mounted on the equipment structure (transformer tank, GIS chamber walls or bushings) pick up sound waves from PD. They respond to PD-induced ultrasonic transients (typically 20–300 kHz) propagating through metal. Acoustic sensors are useful for locating PD sources (time-of-flight) and separating noise, especially in GIS where sound attenuates across compartments.

• Bushing-Coupled (Voltage Tap) Sensors: Many transformers have built-in capacitive taps on bushings (PD couplers) that can feed PD signals to monitors. These provide low-noise PD access without external coupling capacitors. Similarly, some GIS designs include capacitive couplers in VTs or dedicated PD taps on conduits.

Sensor Installation: In transformers, typical installations include HFCT clamps on the ground lead(s), UHF probes through oil outlets, and acoustic sensors on the tank walls. For GIS, standard practice is to equip each bay with at least one UHF sensor inside the housing and several HFCTs or capacitive couplers on the grounding bus. Sensor placement should cover all high-field regions. For example, gas compartments are often interconnected by RF absorbers, so antennas may be placed at tank ends or on busbars. If multiple sensors are used, data correlation (time-stamping to the AC cycle) helps discriminate internal PD from external noise.

PD Testing Procedures

Transformer PD Testing (Offline)

1. Preparation: De-energize and ground the transformer. Verify oil level/quality. Attach high-voltage leads to one winding at a time. Connect coupling capacitors or test capacitors to the high-voltage bushings, referencing the transformer grounding. Install HFCTs on the ground/neutral conductor(s). Attach acoustic sensors on the tank. Use shielded connections to the PD instrument.

2. Calibration: Using an internal calibrator (IEC 60270), inject a known charge pulse to calibrate the measurement “k-factor” for each channel. Ensure the instrument’s bandwidth and filters match IEC 60270 (conventionally <1 MHz).

3. Voltage Ramp: Apply AC voltage gradually. Often a pre-stress step is used: ramp to ~1.3–1.5× the intended test voltage (to charge the dielectric) and hold for a few seconds. Then reduce to the PD test voltage (e.g. 1.0–1.2× rated) and maintain for a specified duration (usually 30–60 seconds per phase).

4. PD Measurement: Record PD pulses over many AC cycles, capturing phase (0–360°), magnitude (pC), and time. Construct Phase-Resolved PD (PRPD) patterns, plotting pulse magnitude versus AC phase. Measure PD Inception Voltage (PDIV) (on the rising ramp) and Extinction Voltage (the fall). Note the average PD magnitude, pulse count, and apparent charge (Σq) per cycle.

5. Repeat for All Windings: Test HV, LV (if applicable), and tertiary windings/bushings in sequence. Swap coupling caps as needed for each phase.

Offline PD test results yield the baseline PD level of a transformer. A passing test typically requires PD to stay below a manufacturer-defined threshold (often ~10 pC maximum) and show stable PRPD patterns. Any abnormal PD (e.g. >10–50 pC or continuous discharges) usually warrants further investigation.

Transformer PD Testing (Online / In-Service)

1. Sensor Deployment: Mount PD sensors under operating conditions. Clamp HFCTs on the transformer’s neutral or ground lead (or on the tank’s ground cable). Attach acoustic sensors on the tank; install UHF antennas in any available oil/flange ports. Ensure the sensors and cables can withstand full system voltage.

2. Data Acquisition: Use a portable or permanent PD monitor synchronized to the power frequency. Measure continuously or during special on-line tests (e.g. stepped voltage up to 1.0–1.2× nominal while energized). Monitor PD pulses in real time, recording PRPD patterns and pulse magnitudes for each phase.

3. Noise Management: In-service tests contend with noise (arcing contacts, corona elsewhere). Use filtering, time gating (correlation with power cycle), and sensor diversity (e.g. both HFCT and UHF) to filter out external sources.

4. Interpretation: Compare measured PD activity to historical baselines. Look for increases in pulse count or amplitude, emergence of new patterns, or changes in PDIV under nominal voltage. Trending PD data over days to years is key for condition assessment.

GIS PD Testing (Offline/Commissioning)

1. Preparation: After GIS installation (assembly or repair), evacuate to specified vacuum and pressurize with SF₆ to operating pressure. Ground all non-tested compartments and ensure one phase/bay is isolated for testing while others are grounded (per design).

2. Sensor Connection: Insert UHF antenna probes into coupling flanges or use built-in capacitive PD couplers in the enclosure. Clamp HFCTs on the common ground bus or connect an external PD detection unit to the grounding conductor. Acoustic sensors may be attached to the enclosure.

3. High-Voltage Application: Apply AC voltage to the test phase following IEC 60270 technique (usually 0.8–1.2× rated at 50/60 Hz). Ramp voltage slowly and maintain for the standard duration while measuring PD. Typically, the test is repeated at nominal and peak line frequencies or at ±5 kV increments around rating.

4. Data Recording: Capture PD pulses and PRPD patterns for each phase. Since GIS PD is usually measured with UHF, record pulse envelopes (e.g. envelope detector data) and correlate pulses to phase. Online-style monitors may also log PD over multiple cycles.

5. Evaluation: Compare PD levels with acceptance criteria (from type test specs). GIS type tests often require no PD above a small threshold (e.g. <10 pC in IEC test mode) or PD only at the initiation step and cessation at the test voltage. Any persistent PD under test voltage indicates a defect (metal particle, void, etc.) that must be corrected (e.g. repair or filter cleaning).

GIS PD Monitoring (In-Service)

For in-service GIS, permanent PD monitoring is common. UHF antennas are installed at strategic locations (e.g. at gas compartment access points or on busbars). These feed into a continuous PD monitoring unit. HFCTs on the SF₆ ground line and acoustic sensors on the enclosure complement UHF data. Periodically (or continuously), the system logs PD magnitudes and PRPD patterns. Alarms are set if PD exceeds preset levels or if patterns change, prompting inspection or maintenance (see Use Cases).

Analysis and Interpretation of PD Data

Quantitative Metrics: The raw PD data consist of millions of pulses per test. Analysis reduces these to metrics: pulse amplitude (charge in pC), count rate (pulses per cycle), cumulative charge, PDIV/PDEV, etc. These are often plotted as PRPD histograms (phase angle vs. charge) or time-series. Key indicators include: a sudden rise in PD count or charge, lower PDIV, or any continuous discharge (high repetition). For diagnostic trend monitoring, one typically watches for incremental increases in PD magnitude/count that could signal developing defects.

PRPD Pattern Recognition: Plotting PD pulse amplitude vs. phase angle of the AC cycle (PRPD plot) is the principal analysis tool. Different defect types produce distinct PRPD signatures. For example:

Figure: Typical phase-resolved PD (PRPD) maps for four GIS defect types. Each subplot shows PD pulse magnitude (y-axis) vs. AC phase (x-axis) for thousands of discharges in one cycle, with the AC voltage waveform overlaid (grey curve).

• Internal (Corona-type) PD: Discharges concentrate near the voltage peaks (around 90° and 270°). The pattern is relatively symmetrical and pulses occur on both positive and negative peaks.

• Free-Metal Discharge: This yields scattered, low-density pulses spread irregularly across the cycle (especially near mid-cycle), since floating particles discharge only sporadically. The pulse count is low and amplitude moderate.

• Floating-Electrode Discharge: When a small conductor floats near a main conductor (as in some GIS faults), the PRPD shows twin lobes around each voltage peak (symmetric clusters around 90° and 270°), reflecting discharges in both half-cycles.

• Surface (Tracking) Discharge: Surface PD tends to occur on the rising or falling edges of the waveform rather than at the peak. In PRPD, pulses cluster near 0°–90° (rising positive edge) and/or 180°–270° (falling negative edge).

These pattern signatures allow experienced engineers to infer PD source type. For example, an internal void in transformer oil usually shows the “double-lobe” peak pattern of (a) or (d) above, whereas a surface discharge on a bushing might resemble (d). Tools like PRPD imaging and pattern-classification algorithms (e.g. ANN, SVM) can aid classification, but visual inspection of PRPD scatter plots is often enough.

Time-Domain Signatures: In addition to PRPD, the raw PD pulse waveforms (nanosecond transients) carry information. For instance, internal PD pulses are often shorter and more damped than corona pulses. High-speed oscilloscopes or spectrum analyzers can show these differences. Correlating PD pulses across multiple sensors can also localize the source (e.g. time-of-flight between acoustic sensors, or comparing HFCT vs UHF arrival times). Modern PD analyzers therefore record both phase and time signatures, enabling advanced diagnosis.

Phase-Amplitude Correlation: Engineers often compute statistical features from the PRPD, such as the mean pulse magnitude in each phase bin, kurtosis/skewness of the pattern, or the correlation between positive- and negative‐half cycles. For example, strong asymmetry between + and – half-cycles might indicate surface discharge (which tends to occur more on one polarity), whereas internal voids produce more balanced distributions. These analyses are guided by standards and IEEE guides (e.g. IEEE Std C57.113 on transformer PD) but ultimately require expert interpretation.

Standards and PD Thresholds

There is no single universal PD limit, as acceptable PD depends on equipment type, voltage class, and test conditions. The IEC 60270 standard governs how to measure PD (coupling, bandwidth, calibration) but explicitly provides no pass/fail criteria. Equipment-specific standards and manufacturer specifications set the thresholds:

• Transformers: Type tests (IEC 60076‑11 for oil-filled, IEC 60076‑15 for dry, or IEEE C57.12.90) usually require measuring PD at 1.5–2× rated voltage. A common acceptance criterion is that PD must stay below about 10 pC (quantified charge) per phase during the routine test (at 2× voltage). Indeed, many factory tests specify “no PD > 10 pC.” Field-test guides (IEEE C57.152) similarly suggest that new transformers exhibit very low PD, though as [58] notes, operational assessments may relax the test voltage (e.g. 0.8–1.0× rated) and focus on defect type rather than absolute level. In on-line monitoring, alarm levels are often set relative to background (e.g. a sudden doubling of baseline PD), or absolute values (some utilities use ~20–50 pC as high alarms for oil transformers).

• GIS: For gas-insulated switchgear type tests (IEC 62271‑203 for metal-enclosed, or 62271‑100 for general IEC 66kV+), a PD test is required at 1.2× rated voltage. Standards often state that no PD above a defined level shall occur under this test. In practice, similar thresholds (~5–10 pC) are used, reflecting the high insulation strength of SF₆ and the expectation of a nearly PD‑free result. (Note: IEC actually limits the detection sensitivity to ~5 pC in test gear.) Like transformers, GIS in-service PD is trended rather than compared to a fixed “pass/fail” number.

The IEEE community agrees that interpretation is key: as one survey noted, IEEE/IEC standards provide test procedures but leave acceptance criteria to the user. Thus, engineers must set alarm thresholds based on experience, often guided by literature. For example, many experts cite ~10 pC as a practical upper limit for factory PD in new HV equipment. Ambient noise must also be considered: on-site PD monitoring typically has noise floors of a few pC, so effective alarms usually exceed that by a safety margin.

Use Case Examples: Power Transformers

1. Factory Acceptance Testing: Before shipping, each transformer undergoes a routine induced-voltage test with PD measurement (typically 1.5–2× rated voltage). The goal is to verify insulation integrity. For a new transformer, acceptance criteria often demand PD <~10 pC on each phase. Any measured PD is logged and analyzed (PRPD pattern, PDEV). A passed test means the transformer is free of major voids or defects.

2. Commissioning (On-site) PD Test: After installation, a transformer may be re‑tested before initial energization. This catches damage during transport. On-site testing usually uses portable PD analyzers with coupling capacitors to the bushing taps and HFCT on ground. The procedure mimics factory tests (though sometimes at slightly lower voltage if oil refill is pending). Results are compared to factory values; new PD indicates handling damage or moisture ingress.

3. Scheduled Maintenance / In-Service Inspection: During a planned outage, PD testing is performed to assess aging. This can be done offline by repeated tan-δ or PD tests, or online via monitoring (see below). For example, an annual check might involve briefly de-energizing and injecting an AC ramp to look for new PD activity, especially if the transformer is >10–15 years old. Another approach is applying an impulse (lightning impulse) while measuring PD with HFCTs. The data help decision-makers determine if oil treatment or winding repair is needed.

4. Failure Investigation: If a transformer partially discharges (arcs) or fails, forensic PD analysis is conducted. Technicians will examine any recorded online PD trends preceding the event and perform detailed offline PD testing on the wreckage. PRPD patterns can help pinpoint the failure site (e.g. failed coil interturn insulation) by matching old patterns or distinguishing internal vs. surface discharge signatures. Such post-mortems often combine PD analysis with dissolved gas-in-oil (DGA) and physical inspection to identify root cause.

5. Retrofit / Rewind Commissioning: When a transformer is rewinded or retrofitted, PD testing is used to verify the new insulation. After the work, an induced-voltage PD test is mandatory. If the rewind involved changes (e.g. using new materials or spacers), the PD test ensures no voids were introduced. Any detected PD (especially >5–10 pC) would require correction (re-impregnation or re-coiling) before the unit can return to service. These acceptance tests follow essentially the same steps as new transformer tests, and are often conducted under the guidance of IEEE C57.152.

Use Case Examples: Gas-Insulated Switchgear (GIS)

1. Commissioning Tests: Upon assembly (or after any major outage), GIS bays are tested with PD measurement. A single phase or section is energized while neighboring sections are grounded. UHF sensors and HFCTs (or coupling taps) capture PD during a stepped voltage ramp. The acceptance test typically requires that PD above ~5–10 pC only appears at the ramp initiation and then extinguishes. This verifies, for example, that there are no floating metal particles. Any continuous PD would indicate a manufacturing or assembly defect requiring rework.

2. Periodic In-Service Monitoring: Modern substations often have permanent PD monitors on GIS. For example, an SF₆‑insulated 220 kV GIS may have UHF antennas in each bay feeding a central unit. These run continuously or at scheduled intervals. Trending PD allows detection of slow degradation (e.g. particle movement) before faults occur. Operators analyze PRPD patterns from these monitors to detect changes in PD activity (e.g. shifts in discharge phase or amplitude) that could signal an emerging problem.

3. SF₆ Leak/Degradation Studies: Since humidity and contaminants in leaking SF₆ can promote PD, PD tests are done when a GIS is suspected of gas loss. After a suspected leak, engineers measure PD (often online with UHF) to assess any deterioration. An increase in surface-discharge signatures might confirm moisture ingress. These tests guide maintenance: if PD rises past alarm levels, the GIS may be taken out of service for drying and gas refill.

4. Circuit Breaker Contact Checks: Many GIS include integrated circuit breakers. PD tests can check these elements as well. For example, before performing maintenance on a GIS breaker, technicians may apply a test voltage and measure PD specifically in the breaker section. A high PD level could indicate contact wear or contamination. Since PD in breakers often presents as short bursts during closing/opening transitions, time-domain analysis of PD during breaker operation can also be performed.

5. Predictive Maintenance / Troubleshooting: When a particular GIS bay shows unusual behavior (e.g. partial flashovers, or after lightning strikes), targeted PD surveys are done. Teams might clamp HFCTs on the common ground bus and use portable UHF detectors to scan different compartments and cables. By comparing to reference patterns, they can locate the fault (e.g. a defect in a feeder cable or disconnect switch). This use-case also includes correlating PD data with other diagnostics (e.g. SF₆ gas analysis) to plan maintenance only where needed.

References: In all of the above, best practices and acceptance criteria are guided by IEEE/IEC standards and manufacturer guidelines. For instance, IEEE Std. C57.152 provides a framework for transformer field PD tests (though it defers acceptance criteria to the manufacturer). Likewise, IEC 60270 defines the conventional PD measurement method but explicitly offers no fixed acceptance levels. Technical literature and industry surveys are therefore used to set practical PD thresholds (e.g. ≈10 pC for new transformer tests) and to understand defect signatures. The methods above integrate such guidance to ensure comprehensive, reliable PD diagnosis in both transformers and GIS.