MW
4.0 MW->2.8 MW
Binding source: UPS_RESERVE_FLOOR
AI Data Center Virtual Capacity Control Plane
How much virtual capacity is safe, for how long, and under which evidence?
Explore a synthetic timed-dispatch record: requested MW, a model-screened envelope, binding constraints, and allow / repair / reject / no-proof outcomes. This separate teaching dataset does not establish site feasibility or operational authority.
Synthetic timed-dispatch teaching example. Separate dataset from DEMO-01; no operational acceptance or equipment authority.
Dispatch Envelope
The runtime-assured capacity aperture, not a dashboard claim
The intended dispatch envelope would constrain proposed actions against declared electrical, storage, cooling, workload, and policy limits. This synthetic example illustrates that record structure; it does not establish safe sellable capacity.
Authority boundary
Read-only proof before autonomy
This synthetic example illustrates a possible decision record. Displayed hashes and signatures are teaching data, not verified operational proof. The model-screened envelope does not establish physical feasibility, customer acceptance, or permission to operate.
Requested envelope
4.0 MWSynthetic illustrative scenario—not a customer or production result.
The operator or workload plan asks for a timed capacity maneuver.
Model-screened envelope
2.8 MWSynthetic illustrative scenario—not a customer or production result.
The synthetic model proposes a smaller envelope; no operating permission follows.
Binding source
UPS / BESSSynthetic illustrative scenario—not a customer or production result.
The evidence trail identifies the limiting domain instead of hiding it.
Dispatch Envelope
Constraint Aperture
REPAIR
4.0 MW->2.8 MW-1.2 MW repaired
Hold
18 min
Ramp
0.70 MW/min
Recovery
12 min
UPS reserve floor clipped the request from 4.0 MW to 2.8 MW and 18 minutes; cooling margin delayed start by one minute.
Proof eligiblepolicy operator-policy-v14topology sha256:746f706f6c6f67793a383363342e2e2e39316600000000000000000000000000proof sha256:70726f6f663a38663463326431362e2e2e393161370000000000000000000000
- Requested
- 4.0 MW
- Model-screened
- 2.8 MW
- Repair delta
- −1.2 MW
- Proof
- Eligible
Scenario
Illustrative data
MW over time
Requested vs. model-screened envelope
RequestedModel-screenedRepair delta
T+0.0; requested 0.0 MW; model-screened 0.0 MW; binding domain None; proof eligible; not pinned.
illustrative scenario
Envelope dimension audit
What changed beyond MW?
requested -> model-screened -> binding source
Hold
20 min->18 min
Binding source: UPS_RESERVE_FLOOR
Ramp-up
1.00 MW/min->0.70 MW/min
Binding source: UPS_RESERVE_FLOOR
Start
T+0->T+1
Binding source: COOLING_SAFE_START
Recovery
6 min->12 min
Binding source: UPS_RESERVE_FLOOR
Rebound
1.0 MW->0.8 MW
Binding source: UPS_RESERVE_FLOOR
Decision-to-proof trace
Decision evidence writes once, then settles
Illustrative sample · evidence chain complete
The trace links the RTA result to ledger, dispatch envelope, and proof root artifacts for read-only inspection.
Proof Artifacts
Every modeled result needs its evidence boundary
The visual exposes proof roots, policy versions, topology hashes, freshness, confidence, and artifact names so operators can inspect why a capacity claim was allowed, repaired, rejected, or withheld as no-proof.
Static fallback
Scenario summary
| Scenario | Decision | Requested | Model-screened | Binding evidence |
|---|---|---|---|---|
Grid stress Reserve-limited | REPAIR | 4.0 MW | 2.8 MW | Storage: reserve_floor_report.csv; Cooling: cooling_water_envelope.json |
Normal operation All limits clear | ALLOW | 3.0 MW | 3.0 MW | No binding constraint in the illustrative request. |
Cooling contingency Hard thermal block | REJECT | 4.5 MW | withheld | Cooling: cooling_contingency_report.json |
Telemetry loss BESS evidence stale | NO-PROOF | 2.2 MW | withheld | Trust: telemetry_manifest.json; Storage: no_proof_gap_register.json |
Bridge power Fuel-window limited | REPAIR | 5.0 MW | 3.4 MW | Bridge: bridge_power_fuel_window.json |
Conditional development path
From Shadow Mode to bounded autonomy, only after evidence
The page is a public-facing preview of the operator discipline: use Shadow Mode to watch dispatch proposals, publish proof artifacts, identify proof gaps, and only then expand authority inside strict dispatch envelopes.
01
telemetry + topology + policy
02
deterministic solver / verification
03
runtime assurance gate
04
versioned decision record
05
read-only visual
06
audit-ready proof pack
Related operator resources
Continue the proof path
- A control-plane direction, with evidence before authority.Explore the intended virtual capacity control-plane architecture, its evidence and authority boundaries, and the current bounded assessment offer.
- Evidence informs a decision. Authority stays explicit.See how Shadow Mode, replay, allow / repair / reject decisions, audit logs, and proof packs establish evidence before bounded autonomy.
- Claims must never outrun evidencePublication of this GridNinja technical resource awaits named ownership, evidence review, and permission. Inspect the synthetic decision brief meanwhile.
Capacity assessment
Scope a historical-data capacity assessment
Agree the capacity question, evidence inputs, and review responsibilities before paid assessment work begins.