Oil Sampling Points and Frequency by Transformer Rating
Send your test requirement — get a tailored solution and a quote within one working day.
A single sampling interval applied to every transformer in a fleet is either too frequent for the units that never change or too infrequent for the units that do. Transformer oil sampling frequency should be set per unit class from the rating and the criticality, and then adjusted by the trend the sample itself produces.
The cost argument supports that approach rather than a uniform one. The sampling visit costs more than the analysis, so a programme that samples everything at the same interval spends most of its budget on mobilisation. Sampling where the information is worth having, at an interval the trend justifies, spends it on decisions.
Sampling interval by transformer rating
The starting interval should reflect the consequence of failure and the rate at which the unit’s condition can change. Higher ratings generally carry higher consequence, and larger units generally have more oil, more insulation and slower thermal response but more stored energy in a fault.
| Unit class | Starting interval logic | What adjusts it |
|---|---|---|
| Large transmission unit | Short interval from the operator’s standard | Trend, fault history, loading profile |
| Medium industrial unit | Moderate interval, aligned with other maintenance | Duty cycle and process criticality |
| Distribution transformer | Longer interval, or sampling on a fleet rotation | Loading, environment and age |
| Unit supplying a critical load | Shorter interval regardless of rating | Consequence of failure rather than size |
| Newly commissioned unit | Short initial interval to establish a baseline | Results of the first two or three samples |
The last two rows are the ones that override the rating. A small transformer whose failure removes a critical process justifies more frequent sampling than a large one with redundancy behind it, and a new unit needs early samples whatever its class.
Load and thermal history as triggers
Loading drives the thermal ageing of the paper insulation, and dissolved gas and furan concentrations respond to thermal stress. A unit that has been operated near its rating through a hot period is a unit whose oil has experienced more ageing than the calendar suggests.
The practical trigger is a period of abnormal loading rather than a single peak. Where monitoring data or operating records show sustained high loading, a sample at the end of the period captures the effect. Where no such record exists, the sampling interval itself has to be short enough to catch it.
Temperature is the second thermal variable. A unit in a hot climate or a poorly ventilated installation accumulates thermal ageing faster than an identical unit in a temperate well-ventilated one, and the interval should reflect that difference.
After a fault or a trip
A fault that the protection cleared is an event that justifies a sample regardless of where the unit sits in its interval. The gas produced by an internal fault appears in the oil quickly, and a sample taken soon after the event establishes whether anything is still happening.
The follow-up sample matters as much as the first. A concentration that rises and then plateaus indicates an event that has finished; one that continues to rise indicates active degradation and determines how quickly the next action is needed.
A through-fault that the transformer carried without protection operating is a subtler case. It may have produced no gas at all, or it may have produced a thermal signature that only appears in the subsequent samples. The interval should be shortened for a defined period afterwards rather than relying on the routine schedule.
New versus aged units
A new transformer is establishing its baseline. Manufacturing residues, first thermal cycles and the oil filling process can all release products into the oil, and the first samples at a shorter interval are what separates those transient effects from a developing condition.
An aged unit has a history and the interval is set by the rate of change in that history rather than by the age. A unit that has been stable for fifteen years can be sampled less often than one that has shown movement over the last three, even though both are the same age.
Where the history does not exist, as after a change of owner or a loss of records, the practical approach is to sample at a shorter interval for a period to establish a baseline, then extend it. The first few samples cost more than the eventual interval will, and they are what makes the later comparison possible.
Connecting intervals to DGA trends
The interval should be a function of the trend. A unit whose dissolved gas concentrations are flat can move to a longer interval; one whose concentrations are moving should be brought forward. That rule is what converts a fixed schedule into a managed one.
The rule should be written explicitly. A statement such as shortening the interval to a defined period when a concentration rises by a defined proportion between two samples is auditable and produces consistent decisions across a team. A statement about increasing frequency where results are of concern is not.
The interval should also have a floor. Continuing to shorten a sampling interval produces more data but not necessarily more information, and where a condition is progressing the response is an intervention rather than a faster measurement cycle.
Cost and logistics of routine sampling
The sampling visit dominates the cost. Transport, access and the technician’s time generally exceed the laboratory analysis many times over, which has two implications for the programme.
The first is that a sample should be worth taking. Running the full screening set while the technician is at the unit costs little more than running one determination, and it produces a complete picture rather than a partial one.
The second is that visits should be consolidated. Sampling several units in one substation on the same visit, or aligning oil sampling with other maintenance activity, reduces the cost per sample without reducing the information. The schedule should be built around access opportunities rather than around an interval that ignores them.
Adjusting intervals after an abnormal result
An abnormal result should trigger a defined response rather than an ad hoc one. The response has two parts: a shortened interval until the condition is understood, and a diagnostic measurement that addresses the specific question the abnormal result raised.
The shortened interval should have a stated end point. A rule that returns the unit to the normal schedule once a defined number of consecutive samples have been stable prevents an escalated programme from becoming permanent.
Where the abnormal result is a step change, the first response is a repeat sample rather than an investigation. A single result that departs from a stable history may reflect sampling, laboratory variation or a genuine change, and the repeat distinguishes between them at low cost. The interpretation framework for the gas results is published as IEC 60599, and the supervision guidance for the oil in service is published as IEC 60422, with the breakdown voltage method defined in IEC 60156 and the corresponding ASTM dielectric breakdown method published as ASTM D1816. Research and utility practice are coordinated through CIGRE study committees and published by EPRI, and the oil testing range is grouped on the oil testing hub.
Writing the sampling schedule
The schedule should state the interval per unit class, the adjustment rule with its inputs, the sampling point for each unit, the determinations to be run, the laboratory or on-site method, and the trigger conditions that override the interval.
It should also name who reviews the results and within what timescale, because a schedule whose results are filed rather than read is a cost without a decision. Where several units share a substation, the schedule should group them so that the visit is efficient.
Finally, the schedule should be reviewed annually against the results it produced. An interval that has never led to a finding on any unit in the class is either well chosen or too long, and the review is where that judgement is made on evidence rather than on habit.
An interval that is the same for every transformer is set by the calendar rather than by the fleet.
Send your transformer list with ratings, duties and current intervals to our engineering team and we will help you set the interval per class and the adjustment rule. Oil sampling and analysis equipment is grouped on the oil testing hub.
FAQ
How often should transformer oil be sampled?
The starting interval should follow the unit rating, its criticality and its duty, and then be adjusted by the observed trend. A large transmission transformer carrying a critical load justifies a shorter interval than a lightly loaded distribution unit. After two or three results establish a baseline, the interval can be extended where the values are flat and shortened where they are moving.
What events should trigger a sample outside the interval?
A fault that the protection cleared, a through-fault that the transformer carried, an abnormal thermal excursion, an oil addition or treatment, and any change in the dissolved gas trend that is not yet conclusive. Each of those is a condition under which the interval between routine samples is too long to capture what is happening.
Do new and aged transformers need different intervals?
Yes. A new transformer is establishing its baseline and may still be releasing products from manufacture and from the first thermal cycles, so early sampling at a shorter interval is justified. An aged transformer has a history, and the interval is set by the rate at which that history is changing rather than by the age itself.
Where should the sample be taken from?
From a point that gives access to oil that has circulated, ideally a dedicated sampling valve on the tank. Sampling from a dead leg, from an isolated cooler circuit or from a point that has been standing gives oil that may not represent the bulk. The sampling point should be recorded so that the next sample is comparable.
How does cost influence the interval?
The sampling visit dominates the cost rather than the laboratory analysis, which is the strongest argument for drawing a good sample and running the full screening set while the opportunity exists. It is also the argument against frequent visits for a single determination, which cost more in mobilisation than in analysis.
