PYTHON · NUMPY

Replay the shift

Use a deterministic twin to check a filed claim without visiting the site.

01Challenge

An unattended plant files a production claim for a shift. You cannot visit it. You have a twin that will re-run the same commanded sequence. Find the discrepancy.

Predict firstA plant files a shift with a six-hour undeclared derate. Against a twin at ±2% fidelity, how many hours are flagged?

02Model

The regulator's problem with an unattended plant is not that it might be unsafe. Passive design handles a great deal of that. The problem is evidentiary: something happened for eight hours with nobody present, and a document arrived afterwards describing it. Trust is the only available response unless there is something to check the document against. A deterministic twin is that something. It is not a monitoring tool and not a simulation for design; its defining property is reproducibility. Given the same commanded sequence it produces the same output, every time, so a filed claim can be re-run rather than believed. An undeclared derate that would pass any single reading shows up immediately as a run of hours where the filed number and the replayed number part company.

03Guided practice

Replay the filed shift against the twin, count the departing hours, and total the discrepancy.

04Feedback

PASS WHEN You checked a claim without visiting the plant.

If your run fails, check:

Six hours depart from the replay and the filed total is 1.32 MWh optimistic.
Widen the fidelity band past the depth of the derate and the same shortfall passes in silence. A fidelity claim is exactly what a passing check is buying.
Note what the twin is for: not accuracy, reproducibility. A twin that cannot be re-run identically cannot check anything.

05Retrieve & space

What property of the twin makes the check possible — accuracy or reproducibility?
What happens to the check as the fidelity band widens, and why does that matter commercially?
Why is an undeclared derate harder to catch than an unreported trip?

06Mastery & project

Check a filed claim against a reproducible replay, and state the fidelity band alongside the verdict.

Take any self-reported operational figure you rely on. Identify what could be re-run to check it. If nothing could, say what that means.

GO FURTHER · THE LIVE PLAYGROUND

Open the replay bench ↗ · Re-run the shift, then change where the log came from.

← Capacity or timing Who wrote the log →

PYTHON · NUMPY

Replay the shift

A filed shift, a deterministic twin, and a fidelity band.

You read

the arrays and values already in scope

You change

the code you write in each cell

Fixed

the dataset and the checks that grade you