The 72-Hour Dispatch
The third rung — run the system hour by hour through the event, spend the storage, start the reserve, shed what you must, and record the cost.
Stage 3 of 4 · Weeks 7–10
- Due
- 9 October 2026
- Weight
- 25% of the course
- Takes as input
- a portfolio with a correlation profile
- Produces
- a dispatch log and a counterfactual
What you are producing#
Two things. A dispatch log — what you did in each hour of the event and why — and a counterfactual: the same event run again with one decision changed.
What earlier work this uses#
The portfolio from A2, and your A1 definition to bound the event. You are operating your own design, which is the point. A design nobody has operated is an assertion.
How the canonical definition constrains this#
Duration is the constraint that bites. Storage sized in megawatts answers a different question from storage sized in megawatt-hours, and a multi-day event is where confusing the two becomes expensive — that is week 7. The recovery allowance matters here too: a brief return above the threshold is not a reset, and dispatching as though it were is a decision your log has to own.
What this feeds#
Design for the Worst Week is where the dispatch log becomes evidence. Your final system has to answer what the dispatch exposed.
What counts as a defensible answer#
A log that shows judgement under the information actually available at that hour, and a counterfactual that isolates one variable cleanly. Losing the event and understanding exactly where it was lost scores better than claiming a survival the log does not support.
The 72-hour dispatch console
Commit the reserve, choose a storage policy, then run it.
Dispatch stack against demand — % of installed capacity
Generation, storage, reserve and unserved demand across a 72-hour window, hour by hour. On the opening decisions the run loses supply from hour 38 onwards.
Energy left in storage — capacity-percent-hours
Stored energy remaining on the same time axis, plotted on its own scale because stored energy and hourly output are different measurements.
- Outcome
- lost
- Shortfall from
- h38
- Unrecoverable at
- h21
- Unserved energy
- 359 %·h
- Hours short
- 30
Synthetic scenario Synthetic trace: flat demand at 40% of capacity, storage rated 40% / 1050 %·h, reserve 24% with a 6-hour lead time, 6% demand reduction available. Generation is the shared teaching trace. No megawatt figure is claimed. This scenario is survivable — but not on the opening decisions.
The spec
The brief poses the problem and leaves room for a response. The spec is the fixed part: what a marker checks. Some lines can be checked mechanically; the rest call for judgement.
- every dispatch decision is recorded with the information available at that hour
- your counterfactual changes exactly one thing, and you say what it shows
- the log states what was shed, when, and to whom
How it is marked
| Criterion | Weight |
|---|---|
| Quality of the hour-by-hour decisions | 40% |
| The counterfactual, and what it isolates | 35% |
| Honesty of the log about what was sacrificed | 25% |