Meter-to-transformer connectivity validation that corrects the GIS from AMI voltage, and sends a technician only where voltage cannot decide
Meters relinked to the right transformer from voltage alone, with 87% fewer field checks.
See one case, screen by screen ↓The GIS says which transformer each meter is on, and often it is wrong. The link was copied from the old meter's record at an exchange, taken from a design drawing for a new connection, or left behind after a rebuild that was never posted. Long rear-lane services run to the pole you would not expect. The usual check is a distance and orphan screen in the GIS, then a crew to trace the service, so the error stays until someone pays for a truck roll.
Every system downstream inherits it. The outage model rolls customer calls up under the wrong device and sends crews to the wrong place. Planned-work notices are built from GIS links, so homes that will lose power are not warned. Transformer loading studies count the wrong houses, missing real overloads and planning upgrades nobody needs. Energy-balance checks flag honest transformers as losses, and solar and EV hosting is assessed on the wrong secondary.
Estimated split for one feeder of about 1,200 meters with about 30 suspected links, in working days, by hand and with the solution.
Six specialist agents read the voltage, outages and locations and propose each correction; the GIS data steward moves the link.
Brookmere 12 is an 11 kV feeder out of Brookmere substation at Lumina Utilities: 1,184 meters and 96 transformers in 8 sections. At 02:10 on Tue 13 Oct 2026 the nightly check reads 28 days of 15-minute voltage. Here is what happens next, screen by screen, in the working solution.
The voltage data checker reads 3,182,592 reads for 1,184 meters — 99.6 % complete, 2 meters re-aligned for clock drift. The phase identifier matches the 96 transformer groups to the SCADA phase voltages at Brookmere and finds 4 recorded on the wrong phase. The correlation analyst compares every meter with every transformer within 380 m: 31 follow a different transformer from the one the GIS records. Outage events support 10 of the 31 and contradict none.
Correction packager: “Ranked 31 proposals and 4 phase changes · 22 high · 6 medium · 3 need a field check · Nadia Okafor decides”
Nadia owns the connectivity model for 214 feeders, and Brookmere 12 is this week's review feeder. Section 3, Wren Close and Alder Row, has 148 meters, 12 transformers, 6 suspected links and 1 transformer on the wrong phase. 27 Wren Close, meter 20418836 with a 7.4 kW EV charger, is drawn with a dashed red link to T-1207 on Alder Row, behind the rear gardens.
“Voltage follows T-1209 at 0.91 and its recorded T-1207 at 0.34”
Each dot is one 15-minute read, the meter's swing against the group's. Against T-1207 the dots are a round cloud at 0.34; against T-1209, a 25 kVA unit on Wren Close, they line up on a tight diagonal at 0.91. No other transformer within 380 m fits better than 0.36. Install GPS from the meter exchange on 14 Mar 2025 puts the meter 22 m from T-1209 and 214 m from T-1207: the old meter's link was copied.
Location checker: location alone never moves a link — here it agrees with the voltage.
On 28 Sep at 14:12 the fuse on T-1207 operated (OE-58113): 8 of its 9 recorded meters sent a last-gasp, and 27 Wren Close kept reporting. On 3 Oct the cut-out on T-1209 was replaced (WO-774120): the notice went to 8 homes, and this meter lost power with the T-1209 group, along with 26 Wren Close, recorded on T-1212. A recloser trip on 21 Sep blinked every meter in sections 3 to 8 and is counted as neutral.
“Every meter in section 3 blinked together. It says nothing about which transformer — the agent counts it as neutral.”
Nadia presses Move link to T-1209. The service link swings across the map to T-1209, the recorded T-1207 is struck through, and the change joins the GIS update with its evidence. Undo is one click away.
“27 Wren Close linked to T-1209 · added to the GIS update”
Corrections ranks all 31 proposals on Brookmere 12 by how strongly the evidence agrees: address, meter, section, recorded and proposed transformer, the voltage match, what else supports it, and a confidence. Nadia accepts the high-confidence links in bulk. The 6 medium links, such as 26 Wren Close at 0.12 → 0.78, are reviewed one by one, and 3 go to the field — 25 Mill Lane fits T-1204 at 0.77 and T-1202 at 0.69.
“Confidence joins voltage, outage and location evidence”
Owen's week has 18 stops across the network, three of them on Brookmere 12. At 25 Mill Lane the secondaries of T-1204 and T-1202 are tied through link box LB-77, so voltage cannot separate them. The job sheet says what to check: open LB-77, trace the service to its pole, read the tag. He finds LB-77 open and the service running to pole MP-4471 under T-1202, and records it with two photographs.
“25 Mill Lane: fed from T-1202 · added to the GIS update”
Edit version CONN-BRK12-1013 gathers 23 meter links, one of them from Owen's field check, each with its evidence and who decided it; 8 are still open on Brookmere 12. Seven checks run: one transformer per meter, every target in service, same feeder, phase against the primary tap, loading after the move, outage-model references and no open edits. The loading check flags T-1209 and sends it to planning.
“T-1209 at 25 kVA goes to 112% of its rating — sent to Tom Hale in planning”
Nadia approves the 23 changes with a note for the record: voltage correlation, 28 days to 12 Oct; outage evidence on 2; field check on 25 Mill Lane. They post to the GIS edit version. Grace Lindqvist, the outage model administrator, publishes it at the 22:00 outage-model build, so a second person signs every publish.
“Published. Grace Lindqvist published version CONN-BRK12-1013 to the outage model at 22:00. 23 changes are live in the GIS.”
With the meters really on it, T-1209 runs at 112 % of its 25 kVA rating, not the 85 % the record showed, driven by the EV charger at 27 Wren Close. T-1120 goes the other way, from 103 % to 94 %, so its upgrade is not needed. Replaying the 3 Oct notice shows 8 homes warned and 10 that lost power; with the corrected map the list is the 10.
“With the corrected map: the notice list is the 10 homes that lost power — nobody missed, nobody warned for nothing”
The same feeder, seen by the five people who keep the network model right — what their week looked like, and what it looks like now.
Six specialist agents read the voltage, outages and locations and propose each correction; the GIS data steward moves the link.
Reads 28 days of 15-minute voltage for every meter from the head-end and checks it is fit to compare: drops meters with more than 20 % of reads missing, flat-lined or stuck values, and re-aligns meters whose clocks drift.
Correlates each meter's read-to-read voltage changes with the median of every transformer group within 380 m, leaving the meter out of its own group, and proposes a move only when another group fits by the margin the steward sets.
Correlates each transformer group's voltage with the SCADA phase voltages A, B and C at the feeder head, and flags a transformer whose recorded phase is not its best match by a clear margin.
Lists the last-gasp and power-restore messages for every device operation and planned outage in the window. A meter that goes dark with a group supports a link to it; one that keeps reporting while its recorded transformer is off contradicts the record.
Compares install GPS and the address with transformer locations, the side of the street and rear-lane services.
Scores each proposal from voltage, outage and location evidence, books field checks where voltage cannot decide, builds the GIS edit version with the evidence for each change and runs the seven checks.
Decides each change; the outage model administrator publishes. The agents propose; a named person decides.
Ask in plain words about any meter, transformer or feeder — or say what to accept. Every answer points to its records.
Why does 27 Wren Close belong on T-1209?
Three things agree. Its voltage follows the T-1209 group at 0.91 and its recorded T-1207 group at only 0.34. On 28 Sep the fuse on T-1207 operated and 8 of its 9 meters went dark, but this one kept reporting; on 3 Oct it lost power with T-1209 during planned work. Its install GPS is 22 m from T-1209 and 214 m from T-1207. The GIS link was copied from the old meter's record.
Which links on Brookmere 12 still need a field check?
Three, where voltage cannot separate two transformers: 25 Mill Lane, 25 Samphire Lane and 84 Harrow Road. At 25 Mill Lane the two secondaries are tied through link box LB-77, so Owen Pryce will open it and trace the service. The other 28 links are decided from data.
What changes for planning if we accept these?
On Brookmere 12, T-1209 (25 kVA) goes from 85 % to 112 % of its rating: it really carries 10 meters, not 8, and one has a 7.4 kW EV charger. T-1120 goes the other way, from 103 % to 94 %, so its upgrade is not needed. Across the network the corrected map finds 38 transformers over their rating and drops 52 planned upgrades. T-1211 also moves from phase B to C.
Accept the 22 high-confidence links on Brookmere 12
Done — the links moved to the transformer their voltage follows, each with voltage above 0.85. They are in the GIS update with the evidence for each. The 6 medium ones need your review one by one, and 3 wait for Owen Pryce. Nothing reaches the GIS until you approve and Grace Lindqvist publishes.
13 screens from the working solution, on its sample data. Pick one to see it large.
How sure the agents must be, who decides, and where the data comes from are settings, chosen by Network Records. The preview shows what each would do on Brookmere 12 before it applies from tonight's check.
| Setting | Default | Choose from |
|---|---|---|
| High confidence — voltage match with the proposed transformer at least | 0.85 | Stepped from 0.75 to 0.95 |
| Medium confidence — reviewed one by one | 0.70 | Stepped from 0.60 to 0.84 |
| Margin over the recorded transformer | 0.30 | Stepped from 0.10 to 0.60 — how much better the proposed transformer must fit |
| Send for a field check when two transformers fit within | 0.15 | Stepped from 0.05 to 0.30 |
| Search radius and window | 380 m · 28 days | Transformers within 380 m, the last 28 days of 15-minute reads |
| Accept high-confidence links in bulk | On | On · Off — medium links are always reviewed one by one |
| Second person publishes to the outage model | On | On · Off — the outage model administrator, after the steward approves |
| Do not propose again for 90 days after Keep as recorded | On | On · Off — unless the evidence changes |
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 utility data · “target” = the design goal, measured in the live solution · “estimated” = our arithmetic, inputs shown: $3.7M = 14,990 links settled from data × $250 a visit, on the demo network of 412,600 meters; 87 % fewer field checks = 2,250 of 17,240 suspected links still needing a visit; 52 and 38 = the demo network after one nightly check, loading recomputed with corrected links; the feeder's day split is our estimate · “typical” = published figures: T&D World (2016) — a first estimate of about 193,000 meters, nearly 5 %, at a large US utility pointing to a wrong or non-existent transformer; S&C Electric via Ventavid (truck roll $250–$500). Method: T. A. Short, IEEE Transactions on Smart Grid 4 (2013); EPRI, Meter Connectivity (Phase ID). Lumina Utilities and the people, streets, substations, feeders, meters and records named on this page are fictional — characters and sample data in the working solution.
It checks that the GIS links each meter to the transformer it is really on, and corrects the links that are wrong. Meter-to-Transformer Connectivity Validation does it from AMI voltage: six agents compare each meter's 15-minute voltage with every nearby transformer group each night, confirm it with outage events and install GPS, and propose the correction. The GIS data steward decides.
Meters on the same transformer see the same voltage swing whenever that transformer's load changes. The voltage correlation analyst correlates each meter's read-to-read changes with the median of every transformer group within 380 m, leaving the meter out of its own group, and proposes a move when another group fits by the margin you set. In the sample case, 27 Wren Close follows T-1209 at 0.91 and its recorded T-1207 at 0.34.
Yes. The phase identifier correlates each transformer group's voltage with the SCADA phase voltages A, B and C at the feeder head. On Brookmere 12 it found 4 transformers recorded on the wrong phase, such as T-1211 on Fenn Street, recorded on phase B but following phase C. It never changes the phase of a three-phase transformer.
A technician goes only where voltage cannot decide, such as two secondaries tied through a link box. On Lumina's 412,600 meters, 14,990 of 17,240 suspected links are settled from data, so 87 % fewer field checks than visiting every suspect — an estimated $3.7M of visits at $250 each.
No. It reads the meter head-end, the GIS, outage management, the SCADA historian and field install records, and writes corrections to a GIS edit version. Your GIS stays the system of record, and the outage model administrator publishes the approved update to the outage model.
Yes. The agents propose and never edit the GIS. The GIS data steward accepts, keeps or sends each change for a field check, bulk accept is allowed only for high confidence, seven checks run before approval, and a second person publishes every update to the outage model.
Once each meter is on the right transformer, loading is counted on the meters really on it. In the sample, T-1209 runs at 112 % of its rating, not 85 %, and T-1120 drops from 103 % to 94 %, so its upgrade is not needed. Planned-work notices built from corrected links reach the 10 homes that lose power, not the 8 the GIS recorded.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your connectivity rules, GIS data standard, thresholds and a month of meter voltage — and it goes live once every quality gate has passed. We will walk you through it on your own feeders first.
Every held bill released inside its cycle, each estimate with its method and source.
1-min demoEnergy Theft DetectionAn estimated $8.4M a year recovered, with 3 in 5 site visits finding theft.
1-min demoMeter Event Management18,214 alarms become 23 actions, and every safety alarm reaches a person in 5 minutes.
1-min demoWe'll run Meter-to-Transformer Connectivity Validation on a month of your own meter voltage and GIS links.