Validated room by room against EnergyPlus®

The reference building

The RIME Loads Validation Building (RLVB-2026.01) is a three storey, 45,000 SF office in Seattle: 63 rooms per the published plan set, punched and ribbon glazing on four orientations, steel stud framed walls, slab on grade, and a full internal load schedule per space: conference rooms, open offices, a cafeteria, huddle rooms, restrooms, mechanical spaces. The complete definition (plans, constructions, loads) is in the reproduction package below.

The method

The same building was modeled twice: in Revit with RIME HVAC Loads (ASHRAE Heat Balance engine, all twelve monthly design days), and in EnergyPlus 25.1 (ideal loads, one zone per room, the same Seattle-Tacoma design days). Inputs were then audited room by room (people counts, people heat rates, lighting and equipment densities, setpoints, infiltration) until the two models were provably identical: 59 rooms, zero discrepancies. Where the engines differ by design (partition conventions, slab methods, film models), each difference is named, quantified, and published, not hidden in the residuals.

What the engines agree on

• Design heating within 2.1% building wide, with the reference set to RIME’s own stated conventions
• The median room lands at +0.2%; 38 of the 43 rooms carrying a real cooling load sit inside ±10%
• All 42 interior rooms within +0.3% in aggregate; interior design heating is zero in both engines
• Exterior wall design heating within 3.5% (29,232 vs 30,306 Btu/h); every Level 2 room within ±1.0% and every Level 3 room within ±1.6% on cooling
• Peak months and peak hours coincident, room by room
• Adding identical partition mass moves both engines by the same amount: +18,125 vs +18,063 Btu/h, 0.3% apart
• Building cooling +2.4%, with every residual a named convention, all in the conservative direction

Documented conventions: the honest fine print

Two independent engines never agree to the last Btu, and pretending otherwise is how load tools lose your trust. Every residual difference in this study is a named, deliberate modeling convention:

1. 1

   Inputs verified identical, room by room

   59 rooms audited: occupancy by seat count at 450 Btu/h·person (250 sensible / 200 latent), lighting and equipment W/sf per the published spec, 75/70°F setpoints, infiltration 0.05 cfm/sf of exterior wall in both engines. Zero discrepancies.

2. 2

   Interior surfaces adiabatic in both engines

   The standard single zone loads convention. The E+ reference sets all 408 interzone wall, floor and ceiling faces adiabatic (thermal mass retained). Against the realistic multizone run, RIME reads +3.0% cooling / +2.7% heating; that sensitivity is published too.

3. 3

   Interior partition mass in both engines, agreeing to 0.3%

   Both models carry the same gypsum partitions on every interior boundary as adiabatic mirror mass. The sensitivity experiment: adding those partitions to otherwise identical models moved EnergyPlus by +18,125 Btu/h and RIME by +18,063 Btu/h, agreement to 62 Btu/h on an 18,000 Btu/h effect. An earlier +5.3% figure compared a RIME run without partition mass against a reference with it; that inconsistency flattered the total by about two points and is corrected.

4. 4

   The reference adopts RIME’s own stated conventions

   So the comparison measures the engine, not the inputs. Interior convection is set to E+’s “Simple” algorithm, an exact match, since those are the ASHRAE constant film coefficients RIME uses (0.542 / 0.167 / 0.712 Btu/h·ft²·°F for wall / ceiling / floor); conduction is stated as CTF, already the E+ default; the timestep is 6, because RIME averages six subhourly samples per hour. Exterior absorptances are deliberately left alone: the drawings govern them (wall α 0.70, roof α 0.55) and both engines already carry those values.

5. 5

   Slab on grade heating by the ASHRAE F-factor method

   Q = F·P·(Tin − Tdesign air), the standard sizing convention. E+ drives the slab to winter soil temperature and reports about a third of the loss; RIME is conservative and faithful to the method.

6. 6

   Ground coupled slab cooling

   RIME’s monthly shallow ground model gives a smaller cooling side sink than E+’s soil and storage model, so Level 1 glazed rooms read higher. They are the only rooms outside ±15%, all in the conservative direction.

7. 7

   Exterior film: the same published convection model in both engines

   RIME’s Heat Balance computes the exterior film with the DOE-2 convection correlation (Walton natural convection by tilt and direction of heat flow, MoWiTT wind driven convection at the surface’s height, surface roughness), the same published correlation the EnergyPlus reference uses, at the station’s ASHRAE design wind, which the weather database now carries and the site syncs (4.1 m/s from 350° here, on every design day of the reference too). Both engines radiate longwave explicitly to a Clark and Allen sky temperature and both evaluate film and sky loss at the outside face, which on a massive wall sits far below the sol-air temperature because most of a solar pulse goes into storage. On identical inputs RIME’s outside face is within 1 °F of E+’s at every hour on the brick wall and the roof, and the anchor box’s walls agree on day conduction within 0.1 to 2% with E+ running its own film, nothing pinned. A Building option, “ASHRAE Sol-Air Longwave (Conservative)”, restores ASHRAE’s tabulated constants on the combined h₀ = 3.0 film and reads a few percent higher on sunlit surfaces. The margin is a visible choice, not something padded into the calc. It is off for the published figures.

8. 8

   Reference roof insulation carries its mass

   The E+ model had idealised its R-30 polyiso as a zero mass layer. A real board (6 in at 2.7 lb/ft³) has a small capacitance sitting in the middle of R-30, which lags the roof by roughly an hour; RIME resolves that mass, the zero mass reference could not, and the difference read as a RIME roof lag. All RLVB models now carry the board with its mass; the reference moved −0.7% on building cooling (Level 3 −1.5%), heating unchanged.

9. 9

   Roof basis correction, disclosed

   The RIME model’s roof had been entered at solar absorptance 0.45 (RIME’s saved construction default) over R-26.5 of board, where the drawings specify 0.55 over R-30, the values the E+ reference carries. Lining the Level 3 rooms up by roof share of load made it a straight line: rooms with only a roof −4 to −6%, rooms where the roof dominates −1 to −3%, and a same input twin reproduced the model’s roof number to 2 Btu/h from those two inputs. Corrected in the model; every Level 3 room now sits within ±1.6% and the floor as a whole at −0.3%.

10. 10

    Fenestration basis correction, disclosed

    The RIME model’s glazing had been entered as a custom low-e stack computing to U 0.341 / SHGC 0.46, while the E+ reference carried the plan’s W-1 at U 0.36 / SHGC 0.38. An input mismatch, not an engine difference: on identical inputs an envelope only twin of Conference B put the two engines within 1.8% at the same hour. The project glazing was reset to the plan values and the study rerun. The earlier published +7.6% cooling figure belongs to the mismatched basis; every Level 2 room now sits within ±1.0%.

11. 11

    Glass inward flowing fraction, corrected and disclosed

    RIME’s simple glazing optics (the LBNL single layer model both engines use to represent a rated U and SHGC) carried the rated inward flowing fraction of absorbed solar at the NFRC rating films (h₀ ≈ 29 W/m²K, winter) and held it fixed; the E+ reference recomputes that split every step at its simulation films. On a design afternoon N is about 13% higher, so the glass delivered a tenth less of its absorbed solar than it should and rooms whose peak sits on the falling side of the solar curve (the rooftop floor) read about 2% low. Found on the same input twin with the interior mass removed from both engines, where transmitted solar matched to 0% and the deficit was a flat level error. The split is now made at the films the glass actually sees, and the interior absorptance defaults were aligned (partitions and steel deck ceilings 0.50, concrete floors 0.60). Level 3 moved from −2.4% to −1.9%, Level 2 rooms from ±1.6% to ±0.8%, building cooling from +1.1% to +1.5%.

12. 12

    Sensible only infiltration heating

    Standard air density, no humidification credit; E+’s component includes cold air density and latent effects.

13. 13

    Window gain bookkeeping

    RIME reports full SHGC solar plus pure UA·ΔT conduction; E+ books the inward flowing absorbed fraction under conduction, so Solar + conduction is the like for like metric. One more step on the upper floors: E+’s pulse decomposition parks the beam solar storage held by the adiabatic floor and partitions as negative “Interzone” rows instead of netting it against the window, which RIME’s superposition attribution does. Adding those rows back, the Level 2 / 3 Conference B windows agree within 6% (+1.8% / +5.8%); the same input envelope only twin lands at +2.4% at the Level 2 peak and +0.4% at Level 3.