Approach and contribution
30-second case brief
A model-based investigation of loads, simulation settings and HVAC performance.
- Problem
- Identify which changes reduce a small office’s complete annual energy use, and how HVAC settings affect energy and delivered thermal service.
- My contribution
- Individual project. Reference-model audit, simulation setup, parametric comparisons, HVAC sensitivity and sizing, and technical reporting.
- Outcome
- The recorded envelope/load sequence reduced modeled annual site energy from 363.95 to 288.13 GJ (20.8%). Separate PTAC and VAV branches were also evaluated.
- Explore the evidence
ApproachReference inputs → simulation-method checks → staged load changes → orientation → HVAC sensitivity and sizing → recorded branch outcomes.
Project description and responsibility
- Context
- AE551 Building Energy Systems · Drexel University · Spring 2022
- Role
- Individual project · all modeling, analysis, results and reporting
- Reference
- DOE post-1980 Small Office · one conditioned floor · approximately 511 m²
- Climate
- Albuquerque, New Mexico · mixed dry (4B) · TMY3 station 723650
- Tools
- EnergyPlus · reference IDFs · HVAC templates · parametric runs and end-use reports
- Accounting
- Annual modeled site energy (GJ/year); HVAC intensity and unmet hours identified separately
Interactive tool · recorded simulation results
Compare the energy sequence.
Choose two stages. Inspect total demand and the changes inside each end use. Values are reconciled to the original EnergyPlus reports.
ReferenceComparison GJ/year · common bar scale
Totals use the EnergyPlus Total Site Energy rows; rounded end uses may sum to 0.01 GJ more. The infiltration-labeled transition also changes the glazing gap and attic infiltration. These are annual site-energy comparisons, not independent measure savings.
Inputs, settings, and full reportA common model and energy boundaryEstablish the reference office before comparing changes
Use a documented building and weather basis
This individual project investigated the DOE post-1980 Small Office reference model in Albuquerque, New Mexico. The building has one conditioned floor of approximately 511 m², a core zone and four perimeter zones beneath an unconditioned attic. Albuquerque’s mixed-dry climate and TMY3 weather file supplied the environmental basis for the annual simulations.
I audited the reference inputs and used the model to investigate three related questions: how simulation methods change calculated outputs, how envelope and internal-load measures affect end uses, and how two replacement HVAC configurations respond to sizing and operating inputs. Holding the case and accounting boundary explicit made the comparisons useful for engineering decisions.
The annual end-use comparison includes heating, cooling, interior and exterior lighting, equipment, fans and water systems. The separate heat-transfer and HVAC-intensity studies retain their own units and definitions. This provides a consistent way to connect a model input to the quantity it changes.
Original course presentation · slide 21 · DOE Small Office reference-model view
Make the model specification inspectable
The overall window-to-wall ratio is 21.20%, with 24.44% on the original south façade and 19.81% on the other orientations. The reference SimpleGlazingSystem inputs are U-factor 4.08816 W/(m²·K) and SHGC 0.36, with visible transmittance left unspecified. EnergyPlus reports derived glass properties of U=4.092 W/(m²·K), SHGC=0.364 and visible transmittance 0.225.
The zone arrangement separates interior conditions from façade exposure. Insulation, infiltration and lighting were examined with the reference HVAC configuration. The PTAC and VAV studies are separate branches using the reference envelope and internal loads, with their own thermostat and sizing assumptions.
| Input | Value |
|---|---|
| Building | DOE Small Office · post-1980 · one conditioned floor |
| Conditioned area | Approximately 511 m² |
| Thermal zones | Core + four perimeter zones; unconditioned attic |
| Weather | Albuquerque TMY3 · station 723650 |
| Climate / elevation | 4B mixed dry · 1,619 m |
| Overall / south WWR | 21.20% / 24.44% |
| Reference window inputs | U=4.08816 W/(m²·K) · SHGC=0.36 · VT unspecified |
Numerical methods and runtimeCheck the effect of calculation choices
Compare surface convection and zone-air methods
I used conduction transfer functions for the opaque construction calculation and compared inside and outside convection algorithms alongside the zone-air heat-balance solution. The small-office study combined Analytical or Euler zone-air solutions with TARP inside convection and either TARP or DOE-2 outside convection.
The reported envelope-convection plus zone-exfiltration heat emissions ranged from 64.44 to 65.94 GJ across the four combinations. These runs cover February 20–27 and July 20–27 inclusive: 384 simulated hours. Historical runtime was 1.50–1.66 seconds per run. These quantities are neither annual site energy nor a test of accuracy against measured data.
2026 source audit · archived EnergyPlus reports and model inputs
Test whether the same question matters at a larger scale
The large-office comparison made the influence of simulation choices more visible. Changing outside convection from TARP to DOE-2 increased the reported envelope-convection quantity from 8,831.72 to 10,201.21 GJ when inside convection remained TARP. The result is a change in the calculated exchange, so its interpretation depends on the selected physical model and boundary.
The historical zone-air method comparison took 266.68 seconds for Analytical, 245.85 for Euler and 235.14 for Third Order Backward Difference. These elapsed times describe the original comparison; hardware and repeated-run statistics were not documented.
2026 source audit · archived EnergyPlus reports and model inputs
2026 source audit · archived EnergyPlus reports and model inputs
Five measures and their end-use interactionsReduce loads through a sequence of model decisions
Select insulation from the diminishing-return sweep
The insulation study varied thickness from 2 to 50 cm while retaining a conductivity of 0.049 W/(m·K). Annual site energy fell from 438.21 GJ at 2 cm to 354.51 at 12 cm, with smaller further decreases as thickness increased. The response therefore supports a thickness decision through the shape of the curve, rather than through the largest tested thickness alone.
I retained 12 cm for the subsequent measures. In the archived sequence, total site energy decreases from 363.95 to 354.51 GJ/year. The model sets both wall and attic-floor insulation layers to 12 cm; their reference thicknesses are approximately 2.48 and 12.66 cm respectively. The independent thickness sweep has its own starting case and does not establish a cost optimum.
2026 source audit · archived EnergyPlus reports and model inputs
Compare glazing as a thermal and solar component
The glazing sweep compared single, double and triple configurations, clear and low-emissivity glass, and air or argon cavities. I selected the double low-E/15 mm argon/low-E configuration for the continuing sequence. The comparison shows why this was a multi-end-use decision: reducing heat transmission also changes solar gains and the cooling and fan response.
Within the recorded sequence, heating decreased from 30.74 to 22.31 GJ after the glazing change, while cooling increased from 30.23 to 34.16 and fans from 48.22 to 51.94. The net annual total decreased by 0.79 GJ to 353.72. Keeping those component changes visible explains the relatively small total improvement.
2026 source audit · archived EnergyPlus reports and model inputs
Test lower infiltration and its coupled loads
The next saved case uses scheduled 0.10 ACH in the conditioned zones. The reference core used 0.36 ACH, while perimeter zones used 0.001133 m³/(s·m²) of exterior surface area. Attic infiltration also changes from 1.0 to 0.5 ACH, and the selected glazing gap changes from 15 to 10 mm. Annual site energy decreases from 353.72 to 340.02 GJ, a 13.70 GJ change across this combined transition.
Across that combined change, heating falls from 22.31 to 7.88 GJ/year, cooling rises from 34.16 to 36.27 and fans fall from 51.94 to 50.56. The archived comparison does not isolate the contribution of each input; it records the net response of the saved models.
2026 source audit · archived EnergyPlus reports and model inputs
Reduce lighting power and read the thermal consequences
Lighting power density decreased from 19.48 to 12 W/m². Interior-lighting energy fell from 102.64 to 63.23 GJ. The lower internal heat gain also reduced cooling from 36.27 to 29.39 and fans from 50.56 to 42.42, while heating increased from 7.88 to 12.44 GJ.
The annual total decreased by 49.86 GJ, from 340.02 to 290.16. This was the largest single decrease in the recorded envelope/load sequence. The result connects an electrical-load decision to the building’s thermal balance and shows why the analysis tracked all end uses together.
2026 source audit · archived EnergyPlus reports and model inputs
Combine permanent solar protection with an adjustable blind
The shading sweep compared blinds, fabric shades, fins, overhangs and combined schemes. The cumulative final model has 1.2 m overhangs with 0.2 m side extensions on the original south and east openings, plus interior Venetian blinds on all four façades. Blind deployment responds to incident solar above 350 W/m²; the slat angle is fixed, not continuously optimized.
The cumulative shading stage decreases cooling from 29.39 to 26.52 GJ/year and fans from 42.42 to 41.37, while heating rises from 12.44 to 14.34. Total site energy is 288.13 GJ/year. The separate scheme sweep uses a clear inner pane instead of the cumulative model’s low-E inner pane; its totals must be read within that comparison set.
2026 source audit · archived EnergyPlus reports and model inputs
Four rotated casesCheck how orientation changes the selected envelope
Compare four directions with the same shading approach
The separate orientation cases rotate the model in 90° increments. Original case labels South, East, North and West correspond to Building North Axis values 0°, 90°, 180° and 270°. They retain the overhang and blind arrangement as the building rotates and use low-E/10 mm argon/clear glazing, distinct from the cumulative final model.
Annual site energy in this sweep was 294.87 GJ for east, 288.83 for north, 296.14 for west and 289.00 for south. North and south were lower than east and west in this comparison. Keeping this as its own sweep shows the orientation response without turning it into an additional stage of the recorded six-stage result.
2026 source audit · archived EnergyPlus reports and model inputs
Sensitivity and delivered thermal serviceInvestigate separate reference-based HVAC systems
Build PTAC and VAV branches from HVAC templates
I introduced two HVAC configurations into the reference building. The PTAC branch uses single-speed direct-expansion cooling and hot-water heating from a natural-gas boiler. The VAV branch uses a central water-cooled electric chiller with a cooling-tower/condenser-water loop, and hot-water heating from a gas boiler. These are independent reference-envelope branches, not systems applied after the envelope/load reductions.
The sizing study used 18°C heating and 24°C cooling thermostat inputs. Original interface captures show the PTAC zone fields and the VAV system fields before optimization, including autosized flows, fan properties, coil types and design supply temperatures. These inputs establish what the later COP, efficiency and setpoint sweeps actually changed.
Original course presentation · slide 28 · EnergyPlus IDF Editor
Original course presentation · slide 31 · EnergyPlus IDF Editor
Use sensitivity to identify inputs worth testing further
The sensitivity study increased each tested input by 10%, one at a time, within its own baseline. PTAC cooling COP 3→3.3 changes the reported HVAC utility intensity from 145.09 to 140.41 MJ/m²; boiler efficiency 0.80→0.88 gives 137.26 MJ/m². These values sum the HVAC-row fuel intensities in EnergyPlus, not whole-site energy.
For the separate VAV sensitivity baseline, HVAC utility intensity is 368.09 MJ/m². Chiller COP 4→4.4 gives 354.40; fan total efficiency 0.70→0.77 gives 360.53. Sensitivity cases use a scheduled cooling setpoint and have whole-site baselines of 307.99 GJ/year (PTAC) and 421.98 (VAV), distinct from the later constant-24°C optimization baselines of 311.53 and 418.67.
| Branch / input | Baseline → test | HVAC utility (MJ/m²) |
|---|---|---|
| PTAC reference | Default inputs | 145.09 |
| PTAC cooling COP | 3 → 3.3 | 140.41 |
| PTAC boiler efficiency | 0.80 → 0.88 | 137.26 |
| VAV reference | Default inputs | 368.09 |
| VAV chiller COP | 4 → 4.4 | 354.40 |
| VAV fan total efficiency | 0.70 → 0.77 | 360.53 |
| VAV boiler efficiency | 0.80 → 0.88 | 357.86 |
Final report · Tables 15–16, PDF pp. 21–22 · HVAC utility intensity
Read sizing against both energy and unmet cooling hours
I varied zone and plant sizing factors by ±20%. Matched PTAC cases changed both zone and boiler factors; matched VAV cases changed zone, boiler and chiller factors together. The sizing comparison pairs whole-site annual energy with the report’s occupied cooling-setpoint unmet hours, so the amount of thermal service remains part of the engineering decision.
For VAV, factors of 0.8 gave 383.02 GJ/year and 440.17 occupied unmet cooling hours. At 1.0, the outputs were 418.67 GJ and 67.67 hours; at 1.2, 461.68 GJ and 2.17 hours. The PTAC response was different: its annual totals stayed near 311 GJ while occupied unmet cooling fell from 72.67 hours at 0.8 to 0.50 at 1.0 and zero at 1.2. Comparing these outputs together reveals the capacity and operating consequences of the sizing choice.
| System | Matched factors | Site energy (GJ/year) | Unmet cooling (h) |
|---|---|---|---|
| PTAC | 0.8 zone / boiler | 312.09 | 72.67 |
| PTAC | 1.0 zone / boiler | 311.53 | 0.50 |
| PTAC | 1.2 zone / boiler | 310.95 | 0.00 |
| VAV | 0.8 zone / boiler / chiller | 383.02 | 440.17 |
| VAV | 1.0 zone / boiler / chiller | 418.67 | 67.67 |
| VAV | 1.2 zone / boiler / chiller | 461.68 | 2.17 |
Original sizing-run reports · Annual Building Utility Performance Summary and occupied setpoint unmet hours
Separate operating-input comparisonsSelect settings by their complete energy response
PTAC: cooling efficiency and supply-temperature decisions
The PTAC cooling-COP sweep tested 3, 4, 5 and 6. Annual total site energy decreased from 311.53 to 296.72 GJ across those cases, with cooling energy falling from 29.61 to 14.81. I retained COP 6 for the branch.
The cooling design supply-temperature comparison tests 14°C down to 12°C. Total site energy rises from 296.72 to 297.36 GJ/year, and occupied cooling-setpoint unmet time rises from 0.50 to 48.50 hours. I retained 14°C. This parameter is a zone sizing input, not the zone thermostat or a measured delivered-air temperature.
2026 source audit · archived EnergyPlus reports and model inputs
2026 source audit · archived EnergyPlus reports and model inputs
PTAC: heating supply and boiler efficiency
The saved heating-supply sweep actually tests 55, 60, 70 and 80°C, yielding 295.95, 295.27, 294.03 and 292.51 GJ/year. The report labels its first point 50°C, but the source IDF and expanded model both specify 55°C. The preceding retained 50°C case gives 296.72 GJ/year and is included separately in the corrected plot. This resolves the apparent discontinuity.
The boiler-efficiency sweep increases efficiency from 80% to 94%. Heating decreases from 37.80 to 32.17 GJ/year and total site energy from 292.51 to 286.89. Final PTAC inputs are COP 6, 14°C cooling design supply, 80°C heating design supply and 94% boiler efficiency. The original 292.52 value is a sum of rounded end uses.
2026 source audit · archived EnergyPlus reports and model inputs
2026 source audit · archived EnergyPlus reports and model inputs
VAV: chiller and fan efficiency
The VAV chiller study tested COP 3 through 6, with COP 4 as the branch baseline. Moving from COP 4 to 6 reduced the annual total from 418.67 to 393.36 GJ and cooling energy from 73.47 to 48.98. This established the retained chiller efficiency before the fan comparison.
Supply-fan total efficiency was varied from 70% to 90%. Annual fan energy falls from 32.26 to 25.09 GJ, cooling from 48.98 to 47.95 and total site energy to 384.76. Total efficiency includes the motor; 90% total with the saved 90% motor efficiency implies ideal 100% impeller efficiency. The endpoint is an idealized parameter case rather than a specified commercially achievable fan.
2026 source audit · archived EnergyPlus reports and model inputs
2026 source audit · archived EnergyPlus reports and model inputs
VAV: distinguish the coil setpoint from zone supply temperature
The cooling-coil design-setpoint sweep compared 13.5, 12.8, 12.1 and 11.4°C. Lowering the coil setpoint from 12.8 to 11.4°C changed heating from 56.69 to 59.09 GJ, cooling from 47.95 to 47.54, and fans from 25.09 to 22.30. The combined annual total decreased from 384.76 to 383.85 GJ.
The final VAV run retains an 11.4°C cooling-coil design setpoint; zone cooling design supply remains 12.8°C. Raising boiler efficiency to 94% gives 375.05 GJ/year. Occupied cooling unmet time is 70.17 hours, compared with 67.67 at the branch baseline. Both cases have zero occupied heating unmet hours.
2026 source audit · archived EnergyPlus reports and model inputs
Recorded stages and independent HVAC branchesResults: where the reductions came from
The six-stage envelope and load result
Using the original EnergyPlus Total Site Energy rows, the archived envelope/load sequence decreases from 363.95 to 288.13 GJ/year: 75.82 GJ or 20.8%. Successive changes are 9.44 GJ for insulation, 0.79 for selected glazing, 13.70 for infiltration plus the glazing-gap change, 49.86 for lighting and 2.03 for shading. The original report’s rounded-component sums give 353.73 and 288.14 GJ at the glazing and shading stages.
Lighting accounted for approximately 65.8% of the total decrease. Its effect included lower cooling and fan demand as well as lower lighting consumption. The stacked result retains heating increases at the lighting and shading stages, so the final total can be understood as the balance among end uses rather than as a list of independent percentage claims.
| Stage | Site energy (GJ/year) | Decrease from prior (GJ) |
|---|---|---|
| Baseline | 363.95 | Reference |
| Insulation | 354.51 | 9.44 |
| Windows | 353.72 | 0.79 |
| Infiltration + glazing gap | 340.02 | 13.70 |
| Lighting | 290.16 | 49.86 |
| Shading | 288.13 | 2.03 |
| Total decrease | 75.82 GJ | 20.8% of baseline |
2026 source audit · archived EnergyPlus reports and model inputs
The HVAC branch outcomes
PTAC decreases from 311.53 to 286.89 GJ/year, a 24.64 GJ reduction (7.9%), with occupied cooling unmet time unchanged at 0.50 hours. VAV decreases from 418.67 to 375.05 GJ/year, a 43.62 GJ reduction (10.4%), while occupied cooling unmet time increases from 67.67 to 70.17 hours (+2.50). All four cases have zero occupied heating unmet hours. Branch savings cannot be added to the envelope percentage.
The occupied-unmet metric uses a 0.556°C setpoint tolerance and counts facility time when at least one occupied zone is outside its applicable heating or cooling threshold. It is not summed zone-hours, an adaptive-comfort result or a general comfort certification. No acceptance criterion was stated, so the 2.50-hour VAV increase is reported without declaring either a material comfort failure or compliance.
2026 source audit · archived EnergyPlus reports and model inputs
2026 source audit · archived EnergyPlus reports and model inputs
2026 audit · original reports and independent EnergyPlus 9.5 reruns
Reproduction and limits
An October 2026 audit reran 13 original models with EnergyPlus 9.5. All checked site totals, occupied heating/cooling unmet hours and tabulated end uses matched the archived reports at their displayed precision. HVAC reruns used preserved expanded models. The recovered TMY3 file matches station 723650; the exact original weather-file bytes were unavailable.
Successful execution does not remove model limitations. Original runs retain coil operating-range and psychrometric warnings, and the final fan setting is idealized. These are deterministic course-study scenarios with no measured-building calibration, cost optimum, statistical significance test or specified comfort acceptance threshold. The original report remains available as a historical source alongside this correction record.
| Verification item | Result |
|---|---|
| Original cases rerun | 13 |
| Engine | EnergyPlus 9.5.0-de239b2e5f |
| Compared outputs | Total Site Energy; occupied heating/cooling unmet hours; full End Uses rows |
| Numerical agreement | All checked outputs match archived reports at displayed precision |
| Weather basis | Recovered Albuquerque station-723650 TMY3; original file bytes not established |
2026 scientific audit · verification and source reconciliation
Verified technical record
EnergyPlus scientific audit
October 2026 verification of the original models and reported outputs. The original 2022 report remains available separately.
The complete project record
Full book
Read the complete report, thesis, or presentation here. Use the page controls to explore the original figures and methods.
Building Energy Systems - Final Report
The original document, page by page.
The reader loads when you open the book.
