How to read an online partial discharge (OLPD) survey report

An online partial discharge (OLPD) survey usually ends in a table of panels, cables or motors with a green, amber or red cell next to each one. A red cell often comes with a recommendation to de-energize and inspect, and an outage on a main switchboard is expensive to schedule. Before acting on the colour, check how the survey reached it. Six questions cover most of what can go wrong in a survey report, and all of them can be answered from the report itself.

What partial discharge is

Partial discharge is a small electrical spark that bridges only part of the insulation between conductors: a void inside cast resin, a gap at a cable termination, a contaminated surface, a sharp edge in air. Each discharge erodes a little insulation, and over months or years that can grow into a failure. Online surveys detect the side effects while the equipment stays energized: a voltage pulse on the metal enclosure, a current pulse in a ground strap, a burst of radio frequency, or ultrasonic sound.

1. Which sensor was used, and where?

Each sensor type sees some kinds of discharge well and others poorly or not at all.

• TEV (transient earth voltage) sensors sit on the outside of metal-clad switchgear and pick up pulses from discharge inside the enclosure. They are the standard tool for internal discharge in medium-voltage switchgear.
• Airborne ultrasonic sensors hear surface discharge and corona, but only where there is an air path out of the enclosure, such as a vent or a panel gap.
• HFCT (high-frequency current transformer) sensors clamp around a cable screen or ground strap and pick up discharge in cables, joints and terminations, and in motor and generator windings.
• UHF and acoustic contact sensors are used on transformers and gas-insulated switchgear, through drain valves, dielectric windows or on the tank wall.

A report should say which sensor was used on each asset and where it was coupled. If a cable run was only checked with a TEV probe on the switchgear it connects to, a clean result tells you little about the cable.

2. What units, and what is the red line based on?

TEV readings are usually in dBmV, ultrasonic in dBµV, and HFCT in millivolts or estimated picocoulombs. A dB value is relative to that instrument's own reference, so readings from different makers are not on the same scale. The green, amber and red bands are normally the instrument maker's or the survey company's own guidance, not a value from a standard.

Look for the report to state which bands it used and where they come from. Watch for units that change between assets or between this year and last, and for a conversion from one unit to another with no stated basis.

3. How was noise separated from discharge?

Substations are electrically noisy. Variable-frequency drives, switching power supplies, radio transmitters, lighting ballasts and corona on outdoor lines can all produce signals a sensor reads as discharge. A competent survey takes background readings (TEV on metalwork that is not connected to the high-voltage system, ultrasonic in open air) and explains how each reported indication was told apart from noise. That can be by pulse timing between two sensors, by correlation with the power-frequency phase, or by locating the source physically.

If a report flags a panel as red but shows no background reading and gives no discrimination method, ask for both before scheduling anything.

4. What was not measured?

Compare the report against the single-line diagram. Panels without access, cables with no exposed ground strap, and equipment that was de-energized on the day are common gaps. The report should list them. It should also avoid conclusions about equipment no sensor covered, and its equipment names should match the single-line, so you know which red cell belongs to which breaker.

5. What were the conditions on the day?

Discharge activity changes with voltage, load, temperature and above all humidity. Surface discharge and tracking on contaminated insulation can be quiet on a dry day and active on a humid one. A survey should record load, voltage, ambient temperature and humidity, and how long each reading was taken for. Without those, a rise from last year's survey to this year's may only mean this year's was a wetter day.

6. Do the PRPD plots support the stated cause?

A phase-resolved partial discharge (PRPD) plot stacks thousands of pulses against the point in the 60 Hz voltage cycle where each one occurred. Different defects tend to produce different shapes, which is how a survey names a likely cause. Commonly described patterns include:

• Corona from a sharp point in air, typically a single cluster around the negative peak of the voltage, with pulses of fairly even size.
• Internal (void) discharge, typically two similar clusters, one in each half-cycle, on the rising part of the voltage wave.
• Surface discharge, typically two clusters of unequal size, often broader and more sensitive to humidity than void discharge.
• Floating metal, such as a loose shield or bolt, often regular pulses of nearly constant size.

These shapes overlap, and real plots can show two sources at once. Pattern reading also depends on the phase reference: if the instrument was synchronized to the wrong phase, every cluster is shifted and the reading changes. A report that names a cause should include a legible plot, state the phase reference, and say what in the pattern led to its conclusion. A cause stated with no plot cannot be checked by anyone.

Trend beats a single survey

One survey is a snapshot. The more useful question is whether activity at a point is rising, and how fast. That requires the same sensor at the same location on the same equipment, under similar conditions, across several surveys. Comparison with sister equipment helps too: if every panel in a lineup reads about the same and one reads far higher, that one stands out even without history.

A report that calls an asset critical from a single reading, with no trend and no sister-unit comparison, is making a strong claim from thin data. It may still be right. Ask for a confirming measurement, ideally by a second method, before you commit to an outage.

What a survey report cannot tell you

Even a well-run survey cannot give a remaining life, and it only sees defects that discharge on the day. It is one input alongside thermography, inspection, maintenance history and the consequence of a failure on that bus. The decision on an outage belongs to the owner and their engineer. The six questions above help make sure that decision rests on a survey that was done properly.