What distribution transformer monitoring actually detects

A distribution transformer rarely announces that it is dying. It runs hot for months, carries an unbalanced load nobody measured, loses oil slowly through a seal, and then fails on a Friday evening — taking a feeder, a few thousand customers and a weekend with it. Continuous monitoring exists to convert that sequence from a surprise into a work order.

The four failure modes worth instrumenting

1. Sustained overload

Distribution transformers are sized against a load forecast that is usually years old. Load grows, nobody re-rates the unit, and the transformer spends its life above nameplate. Insulation ageing is roughly exponential with temperature — the classic rule of thumb is that every 6–8°C above rated hot-spot temperature halves insulation life. A unit running 15% over rating is not failing today; it is quietly converting a 25-year asset into a 10-year one.

What monitoring sees: load current per phase against nameplate, with duration. The useful output is not an instantaneous reading but a duty profile — how many hours per week this unit spends above rating.

2. Phase imbalance

Single-phase loads distributed unevenly across a three-phase transformer produce neutral current, additional losses and localised heating. On a distribution network where connections are added opportunistically, imbalance is close to universal and almost never measured.

What monitoring sees: per-phase current and voltage, and therefore imbalance percentage. This is one of the cheapest problems on the network to fix — rebalancing connections costs a crew and an afternoon — and it is invisible without instrumentation.

3. Oil level, temperature and seal integrity

Oil is both insulation and cooling. A slow leak through a gasket or a valve lowers the level until windings lose cooling and then lose insulation. It is a failure mode with a long, silent warning period and a sudden end.

What monitoring sees: oil temperature trend and oil level. A rising temperature at constant load is the signal that matters, because it means cooling has degraded, not that demand has grown.

4. Tampering and unauthorised access

On Nigerian distribution networks, physical interference — illegal connections at the low-voltage terminals, oil theft, removal of components — is a real and expensive failure mode, not an edge case.

What monitoring sees: enclosure access events, and the electrical signature of a connection that appeared without a corresponding billed customer.

What monitoring does not do

It is worth being blunt about the limits, because oversold monitoring projects are how utilities learn to distrust the category.

• It does not replace dissolved gas analysis. Incipient internal faults — partial discharge, arcing, cellulose degradation — are detected in the oil chemistry. Online DGA exists but belongs on power transformers, not on a distribution unit where the whole asset costs less than the instrument.
• It does not predict a random failure. A lightning strike or a mechanical impact will not be foreseen by a temperature trend. Monitoring catches the failures that develop, which is most of them, not all of them.
• It does not fix anything by itself. Telemetry without a maintenance process that acts on it produces a dashboard nobody opens. The organisational half of the project is the harder half.

The economics, roughly

The case for monitoring a distribution transformer is not usually made on the replacement cost of the transformer. It is made on three other numbers:

• Unserved energy. A failed distribution transformer takes its feeder out for as long as replacement takes — which, if the failure was unplanned and the spare is not local, is days rather than hours. That is billed energy that never gets sold.
• Emergency versus planned intervention. A planned oil top-up or load rebalance is a scheduled crew visit. The same problem discovered after failure is an emergency mobilisation, a crane, and a replacement unit pulled from wherever one exists.
• Asset life extension. Keeping a fleet within rating rather than chronically above it changes the replacement cycle across the whole network, not one unit.

Where to deploy first

Nobody instruments an entire distribution fleet at once, and nobody should. A sensible first tranche covers:

1. Units serving high-value or high-visibility load — where an outage is commercially or politically expensive.
2. Units with a failure history — the network already knows which ones these are.
3. Units in locations that are slow to reach — where the difference between planned and emergency intervention is largest.
4. A control group of healthy units — so that six months in you can say what normal looks like, and quantify what the programme found.

The connectivity problem, and why solar matters

A distribution transformer sits where the power is, which sounds like it solves the monitoring unit's own supply problem until you consider that the moments you most want telemetry are the moments the transformer is not supplying anything. A monitoring node powered from the asset it monitors goes dark exactly when it matters.

That is the reasoning behind Tranxwatch-Solar: a solar-powered sensing and telemetry unit, independent of the transformer's own supply, engineered for Nigerian network coverage with local buffering when connectivity drops. The hardware, cloud infrastructure, interface software and Android application were designed and developed in-house.

Starting small

The most useful pilot is small, boring and instrumented on both sides: twenty units, six months, a control group, and an agreed definition of what counts as a catch. That produces a number you can take to a board. A fleet-wide rollout without it produces a dashboard and an argument.