Discharge activity is often intermittent. It can appear under particular load conditions, at particular temperatures, or after a switching operation, and be absent when a survey is performed. That behaviour is the argument for monitoring, and it is also the reason a clean survey result is not proof that an asset is free of discharge.

What continuous monitoring captures that surveys miss

A survey produces a measurement at one moment under one set of conditions. Its value depends on the test voltage or operating voltage, the load, the temperature and the noise environment at that time.

A monitoring system produces a continuous or frequent record, so the conditions in which activity appeared are captured along with the activity itself. That correlation is the additional information: it establishes whether the discharge is related to load, to temperature or to switching.

The third thing monitoring captures is the trend within a period rather than between periods. A slow increase over weeks is visible; the same increase measured at two annual surveys appears as a single step of unknown shape.

Sensor types and their practical limits

What matters for the selection is the detection threshold the chosen method achieves in the actual environment. A sensor with excellent theoretical sensitivity installed in a position where it couples poorly performs worse than a less sensitive one placed well.

Placement and data quality

Placement determines what the sensor can see. A discharge site inside a winding couples to a bushing tap through a path that attenuates the signal, and the measured amplitude depends on that path rather than on the discharge alone. A sensor placed at the far side of a gas-insulated section may see nothing from a discharge at the other end.

Data quality also depends on what is recorded alongside the measurement. A monitoring system that records the discharge quantity without recording the load, the voltage or the temperature produces data that cannot be interpreted when an alarm occurs, because the conditions that produced it are unknown.

The practical requirement is that the system records the quantity with its conditions and its own status, so that an alarm can be examined rather than simply acted on. A system whose status is not recorded cannot distinguish a genuine quiet period from a period when the sensor was disconnected.

Alarm thresholds and nuisance alerts

An alarm threshold is a trade-off. Set low and the system produces alerts for events that are not discharge; set high and it misses small activity that may be the earliest indication of a developing defect.

The useful approach is to set the threshold from the measured noise floor of the installation rather than from a generic figure, and to use more than one criterion. A level threshold combined with a rate-of-change criterion catches both a sudden event and a gradual development, and it reduces the number of alerts that are merely a level drifting near a limit.

The response to an alarm should also be defined in advance. An alarm that leads to an immediate shutdown decision is a different instrument from one that leads to a planned investigation at the next access, and the threshold should be set with that response in mind.

Cost structure of monitoring versus periodic testing

Monitoring has a capital cost per asset and a continuing cost for data handling, alarm review and system maintenance. The capital cost includes the sensor, the installation and, where the equipment was not prepared for it, the work of fitting it.

Periodic measurement has a lower capital cost and a recurring cost per visit that is dominated by access and mobilisation rather than by the measurement itself. On a site where access is difficult, that recurring cost is high; on a site where the equipment is readily accessible, it is low.

The comparison should be made over the life of the asset and over the population, not on one unit. Monitoring becomes more attractive as the asset becomes more critical, as the population of monitored assets grows, and as the interval between surveys would otherwise have to be short.

Where monitoring is justified by criticality

Monitoring is justified where an unplanned failure has a consequence beyond the asset: a transformer whose failure removes a critical transfer, a gas-insulated installation whose repair requires a long outage, or a cable circuit whose failure affects a large number of customers.

It is also justified where the equipment type has a known susceptibility, or where the specific unit has shown activity in a survey. In the second case, monitoring converts a single observation into a record of how the condition develops.

Where neither applies, a periodic survey at a defined interval is sufficient, and the money spent on monitoring would produce less risk reduction than the same money spent on surveying more of the fleet.

Integrating monitoring data with test records

Monitoring data and test records answer the same question at different times, and they are most useful when they are held together. A survey result provides an absolute measurement under controlled conditions; the monitoring record provides the context in which that measurement sits.

The integration requires that both use the same asset identification and that both record their conditions. Where they do not, the two datasets accumulate separately and the comparison that would make either of them more useful cannot be made.

The measurement method that the survey results are interpreted against is defined in IEC 60270, and the requirements for the gas-insulated equipment where UHF monitoring is commonly applied are published as IEC 62271-203.

Deciding for a specific asset fleet

The decision should be made per asset class rather than for the whole fleet. A transmission transformer fleet may justify monitoring on the units whose failure has the largest consequence, while the same operator may survey the remaining units periodically at a defined interval.

The criteria for inclusion should be written down: criticality, consequence of failure, known susceptibility of the type, and any history of activity on the specific unit. Applying the criteria consistently across the fleet is what makes the decision defensible and what allows it to be reviewed as the fleet changes.

Where monitoring is installed, the arrangement should include the baseline survey that establishes the starting condition. A monitoring system with no baseline produces alarms whose significance cannot be judged. Method documentation for the measurement arrangements is published by suppliers such as OMICRON, asset management practice is coordinated through CIGRE study committees and published by EPRI, and the discharge measurement instrument range is grouped on the partial discharge testing hub.