CONTEXT(CLONE)
Sign in

Check as-installed protection settings against the approved study

For: Protection engineer at a utility, a DC microgrid or data centre operator, or a commissioning contractor

Pattern: Map, verify, reduceNeeds scaleDesigned for 30 to 500 agents
Get early accessSign in to use

The pain today

The coordination study says one thing, the issued setting sheet another, the commissioning record a third. Across hundreds of relays and DC breakers, a stale pickup or time delay stays hidden until a fault is cleared by the wrong device.

The ask

“I attached the approved coordination study tables, the issued setting sheets and the as-left commissioning records for our relays and breakers, all as text exports. For every device, compare pickup, time delay, curve type and enabled functions across the three, and list each difference with the revision it came from.”

Plain words, as you would say it to a colleague. Edit it to fit your case before you send it.

What you attach or connect

  • Coordination study setting tables as text
  • Issued setting sheets per device
  • As-left commissioning and test records
  • Device list with feeder, bus and revision history

The unit of work

One worker task per one protective device.

Why a swarm fits

Each device is checked against its own three records and nothing else. Devices are independent, the pile is wide, and one missed difference matters more than any summary.

Not for

It compares stated values. It does not compute fault currents, plot time-current curves or judge whether the study itself gives selectivity.

The decision tree

6 typed decisions, each with an action for every answer

At fixed moments in a run, the engine puts one narrow question to a decision model. The decision model never writes text: it answers yes or no with a probability, picks from listed options, or gives a score, about a small slice of the material. The engine then does exactly what this tree says, which is what makes the run auditable. The thresholds are the template's design values, not measured results.

  1. Planner, while planning

    Scope checkA choice among options

    Before work starts on a unit

    Does this device identifier appear in the study table, the issued setting sheet and the as-left record?

    Sees only: One device identifier and the index of each of the three sources

    Why: Lists missing records as findings instead of silently skipping the device.

    • In all threethenAccept
    • Missing from one sourcethenMark unresolved
    • Identifier variants need matchingthenEscalate to a strong model

  2. Before workers, before a task runs

    Small worker or strong modelA choice among options

    Before a task runs

    Does the setting sheet for this device list a single overcurrent element, or several enabled functions and setting groups?

    Sees only: The function summary block of one setting sheet

    Why: Keeps simple breakers and fuses on cheap workers and sends multifunction relays up.

    • Single elementthenAccept
    • Several functions or setting groupsthenEscalate to a strong model

  3. After workers, the judge checks

    Evidence checkA choice among options

    After a worker answers

    Do the two quoted values give the pickup for the same element of this device in the same unit and on the same base, primary or secondary?

    Sees only: Two quoted pickup entries with their units and ratio notes

    Why: Separates true setting differences from primary and secondary scaling differences.

    • Same element and base, values equalthenAccept
    • Same element and base, values differthenMark unresolved
    • Different base or unitthenEscalate to a strong model
    • Different elementsthenReject and retry

  4. Evidence checkYes or no, with a probability

    After a worker answers

    Is the quoted setting sheet revision the latest one listed for this device in the revision history?

    Sees only: The sheet's revision block and the device's revision history row

    Why: Avoids reporting differences against a sheet that has already been superseded.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenReject and retry

  5. Evidence checkYes or no, with a probability

    After a worker answers

    Does the as-left record show the same functions enabled and the same curve type as the quoted study row?

    Sees only: One study row and the matching as-left record block

    Why: Catches an element left disabled or on another curve, which a value comparison alone misses.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenReject and retry

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does the finding report any difference in pickup, time delay, curve type or enabled function, or a device with no as-left record?

    Sees only: One device's difference list with quotes

    Why: Every setting difference goes to the licensed protection engineer; the swarm never says which value is right.

    Accountable: The licensed protection engineer owns every setting change and the sign-off that a scheme is coordinated.

    • Yes: 0.40 or higherthenAsk a person
    • Unsure: 0.15 up to 0.40thenAsk a person
    • No: below 0.15thenAccept

The fleet: who does what

Model tiers by role, not brands: you choose the models. Strong reasoning models plan and reconcile, small fast models do the wide work, and the judge is a decision model from a different family, so it does not share the workers' blind spots.

  1. Planner

    A strong reasoning model matches device identifiers across the study, sheets and records and fixes the fields to compare.

    Decisions here:1. Scope check

  2. Workers

    Small fast workers from an open-weight family compare one device's values across its three records with quotes.

    Designed for 30 to 500 agents, one worker task per one protective device. Each worker receives only its own unit.

    Decisions here:2. Small worker or strong model

  3. Judge, from a different model family

    A decision model from a different family confirms each difference is real, including units, ranges and naming variants.

    Decisions here:3. Evidence check4. Evidence check5. Evidence check

  4. Reconciler

    A strong reasoning model groups differences by feeder and revision and lists devices that could not be matched.

  5. Accountable person

    The licensed protection engineer owns every setting change and the sign-off that a scheme is coordinated.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Each difference quotes all records involved, with revision and date
  • Unit and scaling differences are separated from true setting differences
  • Devices missing from any of the three sources are listed, not skipped

What comes back

  • Difference list per device with quotes and revisions
  • Devices with no study entry or no as-left record
  • Feeders where several devices differ from the study
  • Clean devices, listed for sign-off sampling

What to measure

  • Differences confirmed by the protection engineer
  • Differences found later in the field that were missed
  • Engineer hours per device checked
  • Cost per device

Names of measures only. No result is claimed for this template.

Templates open in the workspace chat with the ask filled in. Nothing runs until you send it.

Get early accessSign in to use

More in Energy and utilities

Reconcile study assumptions across an interconnection queue

For: Interconnection manager at a developer, or a transmission planner reviewing clustered studies

Each project's study rests on a base case, on earlier-queued projects assumed in service, on short-circuit levels and on shared network upgrades.

Pattern: Map, verify, reduceNeeds scale6 decisionsDesigned for 30 to 400 agents

Check a flexibility contract portfolio against dispatch obligations

For: Portfolio manager at a flexibility aggregator, or a utility demand-response programme lead

Every site contract has its own notice time, duration limit, activation cap, availability window, baseline method and penalty.

Pattern: Specialist panelNeeds live models6 decisionsDesigned for 8 to 300 agents

Review maintenance and test records across a substation fleet

For: Asset manager or maintenance engineer at a transmission or distribution network operator

Oil analysis reports, breaker timing tests, relay test sheets and battery discharge tests sit inside work orders.

Pattern: Hierarchical decompositionNeeds a connector6 decisionsDesigned for 40 to 600 agents