heating-setback-optimizer

Use when designing or questioning a thermostat schedule: computes real fuel savings from night/away temperature setbacks (degree-hour physics with thermal-mass and heat-loss-recovery effects), compares constant vs setback strategies, finds optimal setback depth, and warns when setback BACKFIRES (heat pumps with strips, radiant floors, short setbacks).

Overview

"Turn the heat down at night" is folk wisdom with failure modes. The savings exist — a heated house loses energy at a rate proportional to (inside − outside), so every degree-hour you don't maintain is fuel not burned — but the size of the saving depends on your building, your climate, and critically your heating system. A deep setback on a heat pump that triggers resistive backup strips on recovery can cost MORE than it saves. This skill computes, for your house:

• Degree-hour savings of a candidate setback schedule vs constant holding, using your balance point and thermal mass
• Recovery penalty: the extra burner/strip capacity needed to climb back (and the comfort lag in the morning)
• System-specific verdicts: furnace/boiler (setback wins), heat pump with strips (setback depth capped, or use smart recovery), radiant slab (setbacks must be shallow and slow — slab time constants are 6–12 h)
• Optimal depth search: scans 0–6 °C setback × schedule shapes and ranks by net saving; flags comfort tradeoff (morning warm-up time)
• Dollars: converts saved energy to fuel cost for gas/oil/ electric- resistance/heat-pump COP

When to Use

• "Is it worth setting back at night, or is that a myth?" (answer: measured, not folklore)
• Designing a weekly program: workday away 9–17, night 22–7, weekend shape
• Heat-pump owners told to "never touch the thermostat" — get the real cap
• Comparing smart-thermostat claims (Nest-style "10–12% savings") for YOUR house and climate
• Don't use for: cooling setbacks (different physics: latent load — humidity), multi-zone commercial, or exact billing prediction (model accuracy is ±20%, good for decisions, not invoices).

Steps

1. Estimate your house: heat-loss coefficient (or last winter's fuel use — script converts), indoor setpoint, outdoor design/average temps, system type and efficiency/COP, thermal mass class (light = timber frame + air, medium = drywall+contents, heavy = slab/stone).

   bash

   python3 scripts/setback.py --k 220 --tin 21 --tout-avg -3 --hours-night 8 \
       --system furnace --fuel gas --price 1.30
   python3 scripts/setback.py --scan   # optimal depth search with defaults
   

2. Read the comparison table: constant vs your setback schedule, % and $ saved, recovery ramp minutes, morning comfort temp at wake time.
3. Check the verdict block for your system: allowed depth, cycle-length floor (setbacks shorter than ~3 h rarely pay), smart-recovery advice.
4. Iterate: try deeper/shallower, weekend shapes, away-day — pick the schedule on the ranked list.

The Model (summary — full in references/setback-physics.md)

• Steady loss: Q_loss = K·(T_in − T_out) W; setback saves K·ΔT·t_setback minus recovery overrun.
• Recovery needs the system's excess capacity: ramp time ≈ C·ΔT / (Q_max − K·(T_in − T_out)) with mass C (light≈3, medium≈6, heavy≈12 kWh/K effective).
• Heat pump with auxiliary strips: if recovery load > compressor capacity, strips (COP≈1) burn the savings; effective cap ≈ 1.5–2 °C unless the thermostat does staged/smart recovery at compressor-only speed.
• Furnace/boiler: recovery is cheap (oversized burners); savings ≈ 0.8–1.2% per °C-setback per 8 h — the classic field-study band.

Output

Savings table (energy, %, $ per heating season), recovery ramps, ranked schedule candidates from the scan, system verdict. --json for machine use.

Common Pitfalls

1. Deep setback on a strip-equipped heat pump. The morning recovery calls strips (resistance, COP 1) and erases the night's savings. Cap at 1.5–2 °C or use gradual/smart recovery; the script prices this in.
2. Setback shorter than the house's time constant. A 1 h dip in a heavy slab house barely moves the temperature; savings ≈ 0, cycling wear ≈ up. Keep setback windows ≥3 h (light) / ≥6 h (heavy).
3. Counting recovery as free. Raising 3 °C costs an overshoot of mass heating; the model subtracts it — folk math usually doesn't.
4. Ignoring outdoor-reset/radiant systems. Radiant slab with outdoor reset: use 1 °C max setback or none; ask the model (system=radiant).
5. Vented/uninsulated setback arithmetic on old houses overestimates savings if K is taken from design-day instead of seasonal average — use fuel-bill-derived K when possible.
6. Humidity swing: deep setbacks in tight winter houses raise morning RH at windows → condensation; check glazing class if you see frosting.

Verification Checklist

• System type honored: heat-pump caps / radiant floors / furnace freedom
• Recovery energy subtracted, not ignored
• Setback window ≥ system-appropriate minimum flagged if violated
• Savings quoted as % of heating energy, not total bill
• Scan ranked list shows the depth curve, including where it flattens