Transformer DGA interpretation that trends every oil result against the unit’s own history, with advice set by how fast the gas is rising
Every oil result trended the day it lands, with advice set by how fast the gas is rising.
See one case, screen by screen ↓When something inside a transformer overheats or arcs, the oil breaks down into a fingerprint of dissolved gases: hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide and carbon dioxide. A lab measures them once or twice a year, or an online monitor hourly. The report comes back as a PDF, with the gases in µL/L, and it gets filed. The value that matters is the change since the last sample, and nobody works it out until an engineer copies five years of results into a spreadsheet. A unit can sit below every screening value and still be heating fast; IEEE C57.104-2019 added the rate of gas increase between samples for this reason.
Then the methods disagree. On the same sample the Duval triangle may say a thermal fault of 300–700 °C, the IEC 60599 ratios below 300 °C, and the Rogers ratios no code at all. Settling it takes an expert, and the experts are few and retiring. The context sits in other systems too: the loading, the sister unit’s outage, the tap-changer design, the last oil treatment. So a gas rise caused by extra load looks the same as one caused by a failing joint.
Estimated split for one abnormal lab result, in calendar days, by hand and with the solution.
Six agents read, interpret and trend every oil result; the engineer signs and the maintenance manager approves load caps.
Saltmarsh Road T2 is a 132/33 kV, 60/90 MVA transformer built in 1998. It supplies 38,400 customers and the Eastwick Pumping Station. Since 2 July its sister unit T1 has been out for a bushing replacement, and T2 carries both halves of the load. On Tuesday 13 October a new lab report arrives. Here is what happens next, screen by screen, in the working solution.
Report BOL-26-14882 from Brightwater Oil Laboratory arrived at 07:48, sampled on 7 October. Reports are read every 30 minutes; Nadia does not wait and presses Read it now. The six agents work in turn beside the Duval triangle, each step with its time and what it found.
Load and history checker: “Serial HT-98-4471 → Saltmarsh Road T2 · 5 earlier samples since 2024 · carrying T1’s load since 2 Jul, peak 104% on 14 Aug”
The reader takes 21 values off the report with 0 missing and checks the units and the sampling: bottom drain valve, glass syringe, oil at 54 °C. Methane is 128 ppm, above Lumina’s 90 ppm screening value for the first time. Ethylene is 41 ppm, still under its 50 ppm screening value, but up 18 since 11 July. Acetylene is flat at 0.6 ppm, moisture is 14 mg/kg, and 2-FAL at 0.21 mg/kg is unchanged since 2024.
Lab comment: “The lab asks for ‘review by the asset owner’. It cannot see the history; the speed of the change is in Lumina’s record, not on the report.”
For two years Saltmarsh Road T2’s samples crept along the T1 zone of Duval Triangle 1. The 7 October sample puts it at 75.5% methane, 24.2% ethylene and 0.4% acetylene: over the 20% ethylene line, into T2, a thermal fault of 300–700 °C. Ethylene is rising at 74.7 ppm a year, 3.7 times Lumina’s rate limit of 20, and four gases are above their rate limits. A dashed projection shows where the path goes at this speed.
Advice: “Resample in 7 days — by Tue 20 Oct · cap the load at 85% (76.5 MVA) · infrared scan this week”
Daily peaks ran at about 60% of the 90 MVA rating before T1 went out, then above 80% through the summer, with a peak of 104% on 14 August. The advice history shows the same story in the rates: 12.2 ppm a year in March, 15.1 in July, 74.7 now. Each sample’s advice is kept with its reason, so the record shows why the advice changed.
“T2 carries both halves of Saltmarsh Road until T1 returns on 30 Oct. The gas speeded up when the load did.”
Duval Triangle 1 says T2, 300–700 °C. The IEC 60599 ratios say T1, below 300 °C, because C₂H₄/C₂H₆ is 0.55 and still under 1. Rogers returns no code; Doernenburg does not apply, because no gas is above twice its validity limit; key gas says overheated oil. The interpreter does not force them to agree. The trend settles it: C₂H₄/C₂H₆ has risen every sample, from 0.20 to 0.55, as a hot spot heats.
“3 of 5 methods give an answer, and all 3 say thermal.” Confidence: medium–high. Would change it: acetylene rising (arcing), or CO rising fast (paper).
Nadia sends the case to Dr. Vos with the triangle, the five methods and the load history attached. The interpretation already names the likely place: a connection or lead heated by the extra load. His reply lands under it and on the record.
Dr. Vos: “Agree. On the 1998 Halvorsen units the 33 kV leads have bolted joints that loosen under heavy cycling; Saltmarsh Road T1 had the same in 2011. Infrared scan first, then winding resistance if the resample is still fast.”
Under the triangle, five speed bands run from routine to outage review. Nadia drags the what-if slider to 30 ppm ethylene: the point moves back into T1, the rate falls to 29.0 ppm a year, 1.5× the limit, and the advice becomes “Resample in 90 days — by Mon 11 Jan”. At the measured 41 ppm it is 7 days and a load cap.
“A what-if only — the record keeps the measured 41 ppm.”
Resample in 7 days, with Callum Price taking the syringe from the bottom valve. Cap T2 at 85% of nameplate, 76.5 MVA, until the resample: the forecast peak this week is 79.8 MVA, so 6 MVA of 33 kV load moves to Ashby Cross T1, which runs at 71% after. Eastwick Pumping Station, 2.4 MVA, stays on Saltmarsh Road. An infrared scan of the bushing connections, leads and coolers this week, and an outage review if the resample is still above 60 ppm a year.
Not recommended now: “degassing the oil — it would erase the trend the resample needs; reconsider after the cause is found.”
Nadia signs the interpretation with the meaning of her signature: “I have reviewed the evidence and agree with the interpretation”. At 08:26 the resample request goes to Brightwater and the load cap to Omar Bekele. At 08:27 Omar approves the cap and books the infrared scan for Thu 15 Oct. At 08:31 Grace Liu applies the 85% cap at Grid Control.
“Grid Control applied the cap · 6 MVA moved to Ashby Cross”
The fleet view puts the latest sample of all 347 transformers, 286 on the grid and 61 at water sites, on one Duval triangle: 7 act and 39 watch. Samples due in the next 15 days are listed, overdue first, and Saltmarsh Road T2 now shows its resample booked for Tue 20 Oct.
By zone: 251 in T1, 12 in T2, 1 in T3, 70 partial discharge, 6 D1, 4 D2, 3 DT.
The same oil sample, seen by the five people who act on it — what their week looked like, and what it looks like now.
Six agents read, interpret and trend every oil result; the engineer signs and the maintenance manager approves load caps.
Reads each lab report — the seven dissolved gases with their units, the oxygen/nitrogen ratio, every oil condition test and the furan result, the sample date, point and oil temperature — and matches it to the transformer by serial, citing the line of the report for every value.
Watches the 38 online monitors’ daily averages and alarms, and separates a real step change from a calibration spike; a step change counts only when it persists over at least 6 readings.
Runs Duval Triangle 1, the IEC 60599 ratios, Rogers, Doernenburg and key gas on every sample, reports each method’s answer, its inputs and why, and uses the trend across samples to explain a split.
Works out each gas’s change and rate a year since the previous sample against Lumina’s rate limits, sets the speed band, and projects the Duval path 30, 60 and 90 days ahead at the current speed.
Joins loading, top-oil temperature, sister-unit outages, tap-changer design and the oil treatment history to each sample, from the control system and maintenance history.
Turns the speed band into actions with dates and owners — resample, load cap, tests, outage review — checks the forecast peak, the transfer options and any water site supplied, and drafts the lab request, the crew request and the control-room note.
Signs every interpretation. The agents propose; a named person decides.
Ask about any transformer, gas, method or action — or say what to change. Every answer points to its sample, report or record.
Why is Saltmarsh Road T2 now in the T2 zone?
The 7 Oct sample moved it. Methane 128 and ethylene 41 ppm put the point at 75.5% CH₄, 24.2% C₂H₄ and 0.4% C₂H₂ — over the 20% ethylene line into T2, a thermal fault of 300–700 °C. For two years the path crept along T1; it crossed when ethylene rose 18 ppm in 88 days, after T2 took T1’s load on 2 Jul. Acetylene is flat at 0.6 ppm, so there is no arcing.
Do the methods agree on Saltmarsh Road T2?
On the kind of fault, yes; on the temperature, no. Duval Triangle 1: T2, 300–700 °C. IEC 60599 ratios: T1, below 300 °C. Rogers: no code. Doernenburg: does not apply. Key gas: overheated oil. All three that answer say thermal. The trend settles the temperature: C₂H₄/C₂H₆ has risen every sample, 0.20 → 0.55.
Which transformers are moving fastest toward a fault zone?
Ravensholme WTW T3, acetylene +8.7 a year, 17× the limit — check the tap-changer seal first. Saltmarsh Road T2, ethylene +74.7 a year, 3.7× — resample in 7 days. Ferrymead T1, hydrogen +101 a year, 2.5× — resample in 30 days. Ashby Cross T1 is slow (3 ppm a year) but sits 1 point from the T2 line, so it is resampled within 90 days.
Cap Saltmarsh Road T2 at 85% until the resample
Drafted for Omar Bekele: 85% of nameplate, 76.5 MVA, until the resample on Tue 20 Oct is read. The forecast peak this week is 79.8 MVA, so Grid Control moves 6 MVA of 33 kV load to Ashby Cross; Eastwick Pumping Station stays on Saltmarsh Road. It goes to Omar when you sign the interpretation.
13 screens from the working solution, on its sample data. Pick one to see it large.
The gas limits, the speed bands that set the advice, and who signs and approves are settings, chosen by the utility.
| Setting | Default | Choose from |
|---|---|---|
| Rate limit for ethylene | 20 ppm a year | Set per gas with − / +, from IEEE C57.104-2019 Table 1 · each change saved as a new rule version with its reason |
| Speed band: 3–5× the rate limit | Resample within 7 days · cap the load at 85% · infrared scan | Five bands: Routine · Resample in 90 days · Resample in 30 days · Resample in 7 days · Outage review |
| Near a zone boundary, resample within 90 days | 3 points | 2 points · 3 points · 5 points from a Duval boundary |
| Signs every interpretation | Nadia Farouk | Nadia Farouk · Dr. Pieter Vos |
| Approves a load cap or an outage | Omar Bekele | Omar Bekele · Head of substations |
| Applies a load cap | Grace Liu · senior control engineer | Grace Liu (senior control engineer) · Duty control engineer — Lumina Grid Control |
| Read new lab reports | On | On · Off — every 30 minutes, or at once from the triangle |
| Include online monitor alarms in this week’s results | On | On · Off — alarms are confirmed with a lab check sample before any action |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample fleet data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures: replacement cost of AUD 23.1M and 45.5M in place (AusNet project planning reports to the Australian Energy Regulator, Oct 2025); a 3–4 year lead time (Powerlink spare transformer case to the Australian Energy Regulator, Jan 2026); more than 1 in 8 major failures ends in fire or explosion (CIGRE Working Group A2.37) and roughly one major failure per 200 transformers a year (CIGRE global estimate), both reported by INMR; 65% of respondents sample oil for the lab once or twice a year (T&D World survey). Methods: IEEE C57.104-2019 · IEC 60599:2022. Lumina Utilities, its people, transformers, sites, labs and records named on this page are fictional — characters and sample data in the working solution.
Software that reads dissolved gas analysis (DGA) results from transformer oil and says what kind of fault, if any, they point to, and what to do next. Transformer DGA Interpretation reads every lab report and online monitor alarm the day it arrives, runs five interpretation methods side by side, trends the sample against the transformer’s own history and drafts the advice. The condition engineer signs.
Duval Triangle 1 plots the shares of methane, ethylene and acetylene in a sample and reads the fault zone where the point lands: PD, T1, T2, T3, D1, D2 or DT. Here every sample of a transformer is drawn as a dated path, so the engineer sees a unit moving toward a fault zone before it gets there, with a projection of where it goes at the current speed.
The DGA interpreter shows each method’s answer and inputs and never forces them to agree; a method that gives no answer says so. In the sample case Duval said 300–700 °C and the IEC 60599 ratios below 300 °C. The trend settled it: the C₂H₄/C₂H₆ ratio had risen every sample, from 0.20 to 0.55, as a hot spot heats.
The trend and rate analyst works out each gas’s rate a year since the previous sample and compares it with the utility’s rate limit. The speed sets the advice: routine below the limit, then resample in 90, 30 or 7 days, and an outage review above 5× the limit or when acetylene appears. A sample within 3 points of a Duval boundary is resampled within 90 days, however slow.
Yes. The online monitor watcher reads daily averages and alarms, and separates a real step change from a calibration spike. A single reading is never treated as a trend, and an alarm is confirmed with a lab check sample before any action.
Yes. In the working solution the fleet holds 347 transformers, 286 on the grid and 61 at water sites such as treatment works and pumping stations. When a load cap is drafted, the action recommender checks any water site the transformer supplies — in the sample case the Eastwick Pumping Station stays on.
Yes. The agents read, interpret and recommend; they never send a load cap, book an outage or change a limit. The condition engineer signs every interpretation, the substation maintenance manager approves any load cap or outage, and the control room applies an approved cap.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your DGA practice and limits, lab report formats, transformer register and loading history — and it goes live once every quality gate has passed. We will walk you through it on your own oil results first.
Defects caught while they grow: every structure compared with last year's flight.
1-min demoAsset Health IndexThe same capital budget removes 2.5× the risk, every score traced to its evidence.
1-min demoMaintenance Knowledge SearchDecades of fixes, searchable before the expert retires.
1-min demoWe’ll run Transformer DGA Interpretation on a sample of your own lab reports and transformer history.