October 20, 2026 · Industry Standards

An auditor who reviews partial discharge work almost never repeats the measurement. They read what was written down and ask whether the number could have been produced any other way. That is why PD testing audit requirements are mostly about the record, not the instrument. In August 2026 State Grid reported completing a six-category field verification chain covering ultrasonic, high-frequency, core grounding current, vibration, UHF and bushing end-screen monitoring devices. The framework makes the point in operational terms: a reading is evidence only when the conditions behind it can be reproduced. A pC value without its bandwidth, calibration and background noise is an anecdote.

What the auditor is actually testing

PD testing audit requirements come down to three questions. Was the measurement made with a defined method, so another engineer could repeat it? Was the instrument known to be working at the time, rather than assumed to be? And is the pass or fail criterion stated before the result was known?

Partial discharge is not a single quantity. A reading in picocoulombs depends on the coupling path, detector bandwidth, gain and calibration pulse. Two engineers measuring the same defect with different set-ups can produce numbers a factor of two apart and both be right. The record has to carry the set-up, or the numbers cannot be compared with anything.

The record fields that make a value defensible

These fields make a record reusable in a later dispute or fleet comparison.

Record field Why the auditor asks for it What a weak record looks like
Method and standard basis Defines the measured quantity “PD tested” with no method named
Detector bandwidth and gain Sets what the reading can see Amplitude only, no instrument settings
Calibration pulse and date Converts mV to pC Calibration claimed but not dated
Background noise before energising Separates defect from site noise No pre-test background level
Sensor position and coupling Explains sensitivity and path loss Channel count without locations
Test voltage and duration Fixes the stress at which PD appeared Single PDIV figure, no ramp record
Acceptance criterion Proves the threshold predates the result Limit quoted only after the test

A record built from these fields can be trended against earlier tests on the same asset, which is the only comparison that carries diagnostic weight. The reasoning behind trend-first thresholds is in PD alarm threshold and trending.

Calibration and background: the two numbers that matter

Calibration is not a formality. In conventional measurement the injected charge defines the scale, and if the calibration pulse is applied at the terminals while the defect sits at the far end of a winding, the reading underestimates the discharge at source. The record should say where the pulse was injected, not merely that one was used.

Background noise is the second number. A site with 200 pC of ambient noise cannot demonstrate 50 pC sensitivity, whatever the detector specification claims. Recording the background before the test voltage is applied turns that limitation into a documented condition rather than an omission.

Proving an online monitor still works

Permanently installed monitors raise the same question continuously: is the sensor still sensitive? Two approaches are common. The first is periodic injection of a known calibration pulse at the sensor or its coupler, comparing the response with the expected value. The second is relative-amplitude verification, where the response to a defined pulse is compared with a modelled threshold to detect a sensor that has degraded or detached. A published patent application for UHF sensor verification on bushings uses this principle.

For an online system the record should therefore include the verification event, its date and the response obtained. Without that entry a flat trend is ambiguous: it can mean the insulation is quiet, or that the channel has gone deaf. A monitor designed in line with IEC TS 62478 / GB/T 42287 and DL/T 1498.1 provides the measurement chain; the verification record proves it was alive. The conventional and non-conventional measurement routes are compared in online partial discharge monitoring, and the laboratory method it complements in IEC 60270 compliance.

What do PD testing audit requirements usually ask for first?

The method and the instrument settings, because they define what the number means. Calibration, background and the pre-stated acceptance criterion follow immediately after.

Is a screenshot of the PRPD pattern enough as evidence?

It is supporting evidence, not the record. A pattern is meaningful only with the settings, gain, calibration and background level that produced it, together with the date and channel location.

Building the record set is easier when the platform writes it as the test runs. The TPD-400 four-input unit records UHF, HFCT, acoustic and RF channels on a common time base (vendor data), which makes channel-by-channel comparison possible in review; sensitivity and sampling figures should be quoted with the same label. Pairing it with a dissolved gas channel through the DGM-900 gives the auditor electrical and chemical evidence for the same event, and the reasoning for combining them is in PD versus DGA monitoring. The applicable standards map is collected under technology and standards. If you are preparing a record template for a new substation or a fleet retrofit, send the current form through the contact form and we will mark it against the fields above.