Eight technical webinars that do not belong to any single asset: what online monitoring is for, what makes the data worth having and which equipment everybody forgets until it fails. A condition assessment is only as good as the data underneath it.
Key points
- Three tests for any monitoring programme: quality, breadth and timeliness of the data. An old report is not necessarily an out-of-date one.
- Numbers alone do not always tell the story: metadata matters as much as the measurement.
- Two assets are monitored far less than they should be: instrument transformers and the station battery, the one piece of equipment that has to work when everything else has stopped.
- Programmes stall at getting the data out; measuring is the easy part.
- There is value, too, in monitors already installed that nobody uses.
Three tests before buying anything
The introductory session sets out three tests for any monitoring programme, worth applying before any purchase: quality (is the measurement consistent and repeatable?), breadth (how many parameters, across how many assets?) and timeliness (how often, and how long until someone sees it?). Then, and only then, what it costs. A programme that fails any one of the three produces data nobody acts on.
An old result is not the same as a wrong one
Timeliness is the test the sessions treat most carefully, and the argument is subtler than it sounds. A test report being old does not mean it is out of date; it becomes out of date only if the circumstances underneath it have changed, and it is generally impossible to know that without repeating the test. So over time the result becomes less and less likely to be current, which is the whole case for measuring continuously instead. The more failure modes are monitored, the higher the confidence in the assessment that results.
The dimensions of data quality, and they all bite
Quality is broken down into the dimensions used in data management, among them accuracy (is the number right?), completeness (is anything missing?), consistency (does it agree with itself?) and validity (does it mean what you think?).
A case makes the point better than any definition. An operator sampling every six months, with the same people drawing the samples, had records showing suspiciously low fault-gas readings over a long period. The unit then tripped off line on a gas relay operation. Samples were sent to three different laboratories at once: two found significant gas, the third did not. The record had been poor data quality, not a healthy transformer.
Numbers alone do not tell the story
The session on metadata has the best single line of the set. An arc-furnace transformer showed rising moisture and furan results, and the recommendation from the numbers was simply to retest in six months. The real story was in a comment attached to the record: the secondary bus seal had gone, and the unit was running through a tank of nitrogen a week. It was no longer a sealed transformer. Metadata is what lets a record be found, understood and trusted later; without it, a number is just a number.
Identical units are not identical
Two units of the same make, rating, vintage and years in service, in the same substation and operating in parallel, always exhibit different patterns of behaviour. They have to be treated as individuals: a fleet ranking is a starting point, not an answer.
Two assets that are rarely watched
Instrument transformers. Oil-immersed CTs and VTs fail in the same insulation-driven ways as larger equipment, but because they are small they are rarely on any monitoring programme — even though they sit in the middle of the switchyard when they fail. The same capacitance and power factor approach applies, and so does the same argument: the condition changes between offline tests.
The station battery. Safety comes first: battery rooms hold acid and generate hydrogen, and monitoring reduces the need to enter the room at all. Building codes increasingly shape what has to be installed. The real requirement is simple — confidence that it will work — and that cannot be obtained by looking at it once a quarter.
Getting the data out: where programmes stall
One session is devoted entirely to communications, because that is where programmes stall: fibre, Ethernet, twisted pair, and cellular where nothing else reaches; DNP3, Modbus and IEC 61850 into the substation RTU; a single point of communication rather than one per device; and, where the network is difficult, start hosted and move in-house later. The measurement is rarely the obstacle; the path to the control room is.
The data is already there. Nobody is reading it
Many transformers arrive with a monitor fitted by the manufacturer. It is commissioned, and then nobody is assigned to watch it. Alarms need owners, thresholds need reviewing and trends need a habit; geospatial views and fleet dashboards exist to make that habit possible. An unread monitor is an expense, not an asset.
What the eight sessions agree on
- Judge a programme on quality, breadth and timeliness before cost.
- An old test is not wrong, but it is progressively less likely to be current.
- Bad data looks exactly like a healthy asset. Check the data first.
- Numbers need their context: the comment is often the real finding.
- Identical units behave differently. Assess them one at a time.
Related equipment
For station batteries, see the BVS battery monitoring system and the hydrogen area monitor. For transformers, the E3 transformer monitor and the rest of transformer testing and monitoring. To review a monitoring programme with us, contact our engineers.