| name | wattlab-twin-ops-reheat-dial |
| description | As-operated Twin dial for dual-fuel G14: fan/OA runtime, DAT/reheat coupling, daytime HW (not plant kill), adaptive elec-first vs gas-first. Use when monthly ±% shows shape fail after envelope, gas spikes after cool DAT/long fans, or elec CV fails in peak cool months while gas is near pass. Triggers on: ops dial, reheat, HW availability, fan hours, 0% OA, daytime HW, elec-first G14, dual-fuel chess, tradeoff trap. |
WattLab Twin ops + reheat dial (Phase 2c)
Start here: ../../docs/AGENT_CONTEXT.md.
Method SoT: ../../docs/BUG_REPORT_TWIN_DIAL_AI_CONTEXT.md.
Envelope/SAT phases: ../wattlab-twin-calibrate-dial/SKILL.md +
../../docs/TWIN_DIAL_PLAYBOOK.md §2c.
EnergyPlus stays autosize. This skill is schedules / setpoints / HW
availability by hour — not more glass when the residual is ops-shaped.
Gate (always)
Both fuels: |NMBE| ≤ 5% and CVRMSE ≤ 15%. Annual % alone ≠ calibrated.
Barely-≤15% CV is provisional — freeze after margin or stamp “edge pass.”
Never promote a run that flips a previously-passing fuel.
Adaptive fuel order
Default envelope playbook is gas-then-elec. Flip when monthly ±% says so:
| Signal | Order |
|---|
| Elec long in shoulders / short in peak cool months | Elec-first ops (fans, OA, DAT), then gas recovery |
| Gas CV fail with elec already pass | Gas-only knobs (HW hours, winter fans, soft HW) — do not undo elec knobs |
| Both fail | One hypothesis per run; score both fuels every time |
Elec-shape (as-operated)
| Symptom | Try |
|---|
| Shoulder elec long | Shorten AHU fan hours those months |
| Peak cool month elec short | Longer fans and/or cooler DAT and/or lower OA |
| Peak short + dump says recirc | Fans on, OA ≈ 0 (even 24/7) |
Gas recovery without killing elec
| Symptom | Try | Trap |
|---|
| Gas spike after cooler DAT / longer fans | Daytime-only HW in spike months | Full HW off → that month gas ≈ −100% vs bills |
| Still high gas | Soften HW OA-reset high SP; warm DAT slightly; shorten HW hours | Cooler DAT alone without HW control |
| Winter gas high | Shorter winter fan hours; softer HW | Cutting OA/fans can cut useful shoulder reheat |
| Shoulder gas short | Cooler DAT and/or higher VAV min those months | Raises chiller elec — watch elec CV |
| Near gas CV gate (~15–17) | Stack small knobs | One big plant-off usually overshoots |
Reheat coupling (always state this)
more mechanical cooling (cool DAT, long fans, low OA)
→ more reheat opportunity
→ gas up unless HW is scheduled/softened
Dial both sides of the coil story in the same campaign narrative.
FDD / dump → knob (no IDF surgery in openfdd-mcp)
- Read monthly ±% (Studio dial charts) + dump DAT/fan/OA when present.
- Form one schedule/HW hypothesis (not three knobs at once).
- Patch + simulate via EnergyPlus-MCP / vibe20 tools — never invent savings.
- Score both fuels; keep or revert if the other fuel flips.
Open-FDD agents: see docs/mcp-agents/fdd-ops-to-twin-knobs.md (pointer only).