Asymmetric Sizing and 780 Unmet Hours in an Appendix G Comparison

No. The number is not defensible as produced, for two independent reasons.

1. Asymmetric sizing. The baseline carries a 15% cushion on coils and boiler; the proposed carries whatever the autosize algorithm produced, with no cushion. In most engines, fan power, auxiliary power, and some part-load behaviour scale with capacity, so a smaller proposed plant books energy savings that have nothing to do with the envelope, lighting, or DCV measures. Part of the 38% is a nameplate artifact.

The 15% is also undocumented and unverified against the actual design intent - hard-sizing from 2019 drawings imports whatever the engineer drew, plus an arbitrary adder.

2. The two designs were sized under different rules. A comparison is only meaningful if both sides are sized consistently. Either autosize both, or hard-size both from drawings with the same adder. Never one of each.

A note on the 30% smaller AHUs. Some of that reduction is real and expected: DCV cuts outdoor air, which cuts the coil load, which reduces the autosized coil and fan. That is a genuine measure effect and should appear as reduced coil and fan energy. But it must not appear as a smaller nameplate relative to a baseline that was never allowed to shrink by the same rule - the baseline's DCV is not required to be modeled, but its sizing basis must be the same as the proposed's.

Fix: re-establish a single sizing rule for both models, re-run both, and re-report. Expect the savings to fall materially, and be prepared for the DCV benefit to shrink once the baseline is autosized.

Treat the comparison as invalid until resolved. A model that cannot meet loads is not delivering the same service as one that can; unmet loads are shed, not served, so the proposed model systematically under-reports energy and over-reports savings. The 735-hour delta is large enough to explain a good part of the 38%.

Program criteria (verify the exact edition and rule set your program cites - these are commonly applied as): no more than 300 unmet load hours in either model, and the proposed model not exceeding the baseline by more than 50 hours. The run fails both tests.

Diagnostic sequence - find the zone before touching capacity:

1. Determine whether the unmet hours are heating or cooling, and in which zones. Hundreds of hours almost always trace to one or two zones, not to a global capacity shortfall.
2. Check thermostat setpoints and setpoint schedules against the availability schedule, and confirm the availability schedule during the sizing run matches the occupied period. A zone that is 'occupied' during the sizing run but conditioned on a different schedule in the annual run is a classic source.
3. Check zone minimum air fractions and whether the DCV schedule starves a zone of ventilation or of reheat-capable airflow.
4. Confirm the design days used for autosizing actually bound the annual run - a design day at a lower altitude, or with economizer lockout at a different temperature, will undersize the plant.
5. Check coil capacity at the actual entering conditions (altitude correction, entering water temperature, refrigerant characteristics).
6. Confirm terminal reheat capacity was autosized and not defaulted to zero.

Re-run after each fix and track the hour count; do not hard-size your way out of the problem before you know the cause, or you will mask a control error as a capacity error and carry it into every parametric run.

It is a defensible choice, provided it is applied symmetrically.

If you hard-size the proposed from the drawings, the same 15% adder must be applied to both models - or to neither. Otherwise you have simply reversed the asymmetry: now the baseline is oversized and the proposed is sized exactly, and the savings are again inflated by a sizing mismatch rather than by the measures.

What you give up. Sizing the proposed to the actual design means the model no longer credits the project for HVAC downsizing. For a code or program comparison this is usually the right trade: equitability of the sizing basis matters more than capturing a benefit the design team may never install. For an owner-facing 'what will this building actually use' study, the as-designed sizes are also more realistic than an autosize output.

What to document:

• The sizing rule chosen, and the reasoning.
• The capacity of every autosized or hard-sized component in both models, in the assumptions log.
• A note that 90.1 editions differ on whether proposed equipment is autosized or modeled as designed - cite the edition in force for this project.
• Confirmation that the rule was held identical across every parametric run, not just the headline case.

Also reconsider the 15%. If it was the modeler's own invention, remove it from both models or replace it with a documented design margin from the mechanical drawings. An invented cushion on both sides is less harmful than on one, but it still distorts part-load behaviour and should be justified.