SAIDI and SAIFI reliability reporting with every outage record proved against the meters and every major event day justified
Clean reliability indices, ready for the regulator.
See one case, screen by screen ↓SAIDI and SAIFI reliability reporting turns a period's outage records into the numbers the regulator sees: how long the average customer was without power (SAIDI), how often (SAIFI) and how long each interruption lasted (CAIDI), with and without major event days. Those records come from the outage management system, typed by dispatchers in the middle of storms. Duplicates, 23:59 placeholder restore times, start times taken from the first customer call, customer counts from the wrong transformer and three-minute interruptions logged as sustained all land in the index.
Before each filing a reliability engineer reconciles thousands of records in spreadsheets, works out the major event day threshold in a workbook and writes the justification by hand. Major event day exclusions are the biggest single lever on the reported number, so the regulator scrutinises each one, and an exclusion without its calculation and evidence is a finding. The smart meters already record the minute each customer lost and regained power, but no one has time to match them to the record.
Estimated split for one quarterly reliability report at a distribution utility, in working days, by hand and with the solution.
Six agents check every outage record against the meters, set the major event day threshold and draft the report; a reliability engineer decides.
Lumina Utilities serves 1,284,600 electricity customers on 1,146 feeders across five regions. Its Q3 2026 reliability report is due to the Regional Energy and Water Commission (REWC) on 31 October. On 19 August a line of thunderstorms crossed Harbour and Eastern Hills. Here is what happens next, screen by screen, in the working solution.
Nine months sit on one reliability calendar, each day coloured by customer minutes lost. Year-to-date SAIDI excluding major event days is 62.2 minutes, 11.2 over the 51.0-minute target to 30 September. Last night's run put one day at the top: Wed 19 Aug, 14.2 minutes, 3.5 times the major event day threshold, with 9 record exceptions that change its minutes.
“Overnight run 02:10 · 6 agents · ready 02:40”
The checker has matched 2,671 records this year to 41.6 million meter events. On the storm day it found three duplicates, each one interruption logged twice, and three restore times the meters disprove. OUT-26-48233 on feeder breaker CB-31 shows a 23:59 placeholder left open when the crew moved on; 1,846 of 1,902 meters report power back between 17:41 and 17:47, so it proposes 17:44.
Not trying to lower the number: EX-0818 moves a start time from the first customer call at 15:40 to the last-gasp messages at 15:12 — it adds 28 minutes.
Nadia accepts the 9 fixes. The customer count checker had found transformer T-55810 serves 148 meters, not the 412 customers taken from T-55801, and a 2 min 55 s operation on recloser R-2251-3 comes out of SAIDI and SAIFI as momentary. The day goes from 218 to 214 interruptions, 104,374 to 98,900 customers and 14.2 to 13.0 minutes; year-to-date SAIDI falls from 62.2 to 61.0.
“Run by the record checker and index calculator · accepted by Nadia Rahman”
The 2026 threshold comes from 1,811 days with interruptions in 2021–2025, with 15 days with none left out. The mean of ln(daily SAIDI) is α = −1.622 and the standard deviation β = 1.211, so T_MED = e^(α + 2.5β) = 4.08 minutes, fixed for the year. Only 10 of the 1,811 days were above it; earlier years flagged 2 to 4 major event days each.
“A day in 2026 whose daily SAIDI is above 4.08 minutes is a major event day.”
At 13.0 minutes the day is 3.19 times the threshold, and Nadia marks it. The assessor counts 37 interruptions that ran past midnight to 19 August, the day they began, and gathers the weather and crew evidence. SAIDI excluding major event days falls from 61.0 to 48.0 minutes, SAIFI from 0.742 to 0.665 and CAIDI from 82.2 to 72.2 — now 3.0 minutes inside the target.
“Major event day · excluded from the indices excluding major event days; kept in the all-days figures.”
The day file shows the storm hour by hour, from 13:00 on 19 August to the last restore at 06:10 the next morning, the largest interruptions with the crew's cause, and the threshold test. The evidence is cited: the Cape Lumen weather station with gusts of 96 km/h at 14:20, 1,240 lightning strikes, 31 Lumina crews and 12 from Tasman Line Services, and the 9 accepted record corrections.
“Before the test, each outage record for the day was checked against smart meter power-off and power-on events, and 9 corrections were accepted, which moved the day from 14.2 to 13.0 minutes.”
On Sat 2 May a heatwave overloaded transformers in Central and Riverside: 61 interruptions, 31,240 customers, 3.91 minutes of daily SAIDI — 0.96 times the threshold. Asked to exclude it, the assessor refuses and says why. The day counts in every index, and can be described in the report narrative instead.
“2 May cannot be excluded. Its daily SAIDI of 3.91 min is below the 2026 threshold of 4.08 min.”
Ranked by customer minutes excluding major event days, F-1102 Harbour Road 3 leads with 1,290,486, a feeder SAIDI of 253 minutes, 3,164 customers with three or more interruptions and equipment failure as the top cause. F-3108 Cape Lumen, first on all days, drops to fifteenth once the storm is set aside. Nadia sends F-1102 to the November capital planning review with its year.
Feeder F-1102: “214,000 customer minutes on the 19 Aug storm (major event day candidate).”
The report writer drafts the Q3 2026 report in the REWC template: summary, method, major event days, data quality, worst-performing feeders, notable events and Appendix A with the justification note. It shows 61.0 minutes across all days and 48.0 excluding major event days, against 51.0. The data-quality section records 2,667 sustained interruptions and 50 corrections accepted this year; source records are unchanged.
“Drafted 13 Oct 02:40 · every number links to its source · re-drafts when a record changes”
Graham opens the signature with SAIDI excluding major event days at 48.0 minutes and 1 major event day, 19 August, in front of him. He signs as Head of Network Performance on 21 Oct at 09:14. Leah Fontaine in regulatory affairs files with the REWC and gets receipt REWC-RR-2026-Q3-0417.
Meaning of signature: “I have reviewed the indices, the major event day exclusions and their justification, and approve this report for filing.”
The same Q3 report, seen by the four people who own it — what their quarter looked like, and what it looks like now.
Six agents check every outage record against the meters, set the major event day threshold and draft the report; a reliability engineer decides.
Matches every outage record to the smart meters' power-off (last gasp) and power-on events and to breaker operations, within ±10 minutes; finds duplicates, wrong restore and start times, and momentary interruptions logged as sustained, and proposes each fix with its meter evidence.
Counts the active meters below each device in the meter-to-transformer connectivity model at the time of the outage and checks them against the customers interrupted on the record.
Reads the crew close-out notes and, where a note names a different cause, proposes the code from the cause list and quotes the note — “tree limb on phase B, span 14–15, cleared” becomes Vegetation.
Computes SAIDI, SAIFI, CAIDI, MAIFI and CEMI-3 to IEEE 1366 from accepted records, all days and excluding major event days, by region, feeder and cause, and recomputes on every accepted change.
Builds five years of daily SAIDI, computes α, β and the threshold by the 2.5 beta method, flags the days above it, gathers weather station, lightning and crew log evidence and drafts each justification note with citations.
Drafts the quarterly and annual reliability reports and the major event day appendix in the REWC template — summary, method, major event days, data quality, worst feeders, notable events — with every number cited.
Accepts every fix and every exclusion. The agents propose; a named person decides.
Ask about any day, feeder, index or exclusion — or say what to change. Every answer points to its records.
Is 19 August a major event day?
Yes, it qualifies. After the 9 record fixes its daily SAIDI is 13.0 min, against the 2026 threshold of 4.08 min — 3.2 times over. 214 interruptions, 98,900 customers, thunderstorms with 96 km/h gusts. Marking it takes SAIDI excluding major event days from 61.0 to 48.0 min, inside the 51.0 target; the all-days 61.0 stays on the report beside it.
Can we exclude 2 May as well?
No. 2 May had a daily SAIDI of 3.91 min, below the 2026 threshold of 4.08 min. IEEE 1366 treats a day as a major event day only above the threshold, so the heatwave counts in every index. You can describe it in the report narrative instead.
Which feeders lose the most minutes outside storms?
Excluding major event days: F-1102 Harbour Road 3, 1,290,486 customer minutes, mostly equipment failure; F-5071 Weir Lane, 1,283,210; F-4430 Millbrook North, 914,008. Together that is 2.7 min of the system's SAIDI.
Add the 2 May heatwave to the report narrative
Added to section 6 of the Q3 draft: “On 2 May a heatwave overloaded distribution transformers in Central and Riverside: 61 interruptions, 31,240 customers, 3.91 minutes of system SAIDI.” It is counted in every index; section 3 says so.
13 screens from the working solution, on its sample data. Pick one to see it large.
The method, the day, the meter matching, the targets and who decides are settings; every change is saved as a new version and logged.
| Setting | Default | Choose from |
|---|---|---|
| Major event day method | 2.5 beta | 2.5 beta (IEEE 1366, threshold = e^(α + 2.5β) from daily SAIDI) · Regulator-defined |
| Years of daily data behind the threshold | 5 years | 3 years · 5 years |
| Threshold fixed for the reporting year | 2 January | Fixed for the reporting year · approved by the Head of Network Performance |
| Scope | Unplanned | Unplanned · All — planned interruptions are reported separately |
| Meter time tolerance | ±10 min | ±5 min · ±10 min · ±15 min — meter events this close to a record's times are a match |
| Meters that must agree to propose a fix | 80% | 70% · 80% · 90% of the meters on the device |
| Targets, excluding major event days | SAIDI 68 min · SAIFI 0.95 a year | From the regulator's performance scheme, shown pro rata — 51.0 min and 0.713 to 30 Sep |
| Who decides | Nadia Rahman, Owen Achterberg | Reliability engineers accept fixes and mark major event days · Graham Oyelaran signs · Leah Fontaine files |
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 network data · “target” = the design goal, measured in the live solution · “estimated” = our estimate (the working-days split) · “typical” = published figures (EIA, Today in Energy: US customers averaged about 11 hours of interruptions in 2024, nearly 9 from major events, and about 2 hours a year without them; LBNL, evaluation of the IEEE 1366 major event day method: the 2.5 beta method is designed to flag about 2.3 major event days a year). Method: IEEE Std 1366, 2.5 beta. Lumina Utilities, the Regional Energy and Water Commission (REWC), Tasman Line Services and the people, feeders and records named on this page are fictional — characters and sample data in the working solution.
It is the periodic report a distribution utility files on how long (SAIDI) and how often (SAIFI) its customers were without power, and how long each interruption lasted (CAIDI), with and without major event days. SAIDI and SAIFI Reliability Reporting checks every outage record against the smart meters, computes the indices to IEEE 1366, justifies each major event day and drafts the regulator report.
SAIDI is total customer minutes interrupted divided by customers served; SAIFI is total customers interrupted divided by customers served; CAIDI is SAIDI divided by SAIFI. Only sustained interruptions — longer than 5 minutes — count, and each counts to the day it began. The index calculator uses accepted records only and recomputes on every change.
It takes five years of daily SAIDI, leaves out days with none, takes the natural log of each and computes the mean α and standard deviation β. The threshold is e^(α + 2.5β); any day in the reporting year above it is a major event day. In the sample, 1,811 days of 2021–2025 data give 4.08 minutes for 2026.
The outage record checker finds the meters behind each device in the connectivity model and their power-off and power-on events within ±10 minutes. When at least 80% agree, it proposes a fix with the evidence: a duplicate merged, a placeholder restore time replaced by the meters' time, a late start moved, a momentary interruption taken out. Some fixes add minutes.
No. The major event day assessor refuses any day at or below the threshold and says why. In the sample, the 2 May heatwave at 3.91 minutes, against a threshold of 4.08, stays in every index and can be described in the report narrative instead.
Yes. Feeders are ranked by customer minutes, all days and excluding major event days, with feeder SAIDI, customers with three or more interruptions (CEMI-3), top cause and the year month by month. A feeder can be sent to the capital planning review, and the top five go into the report.
Yes. Agents propose and draft; they never edit a source record, accept a fix, exclude a day or file. A reliability engineer accepts each fix and marks each major event day, the Head of Network Performance signs with the meaning of the signature recorded, and regulatory affairs files.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your outage records, meter event feed, connectivity model, regulator template and past filings — and it goes live once every quality gate has passed. We will walk you through it on a past quarter of your own outage data first.
We'll run SAIDI and SAIFI Reliability Reporting on a past quarter of your own outage records and meter events.