Whole-building energy simulation of natural and hybrid ventilation couples a multizone airflow network (each zone a well-mixed node, openings modeled as orifice-type links) with a transient thermal solver. A 2011 assessment of the EnergyPlus Airflow Network model against three monitored European office buildings found good agreement for a single-story building but systematic over-prediction of buoyancy-driven (stack) flows, overheating hours, and atrium exhaust flows in multi-story, atrium-connected buildings, together with strong sensitivity to effective vent areas, discharge coefficients, local wind and solar data, and thermal mass, and functional tool limitations in vent and fan control and in roof-opening and atrium-connection modeling. The study left open an attribution problem: whether such over-prediction reflects network model-form deficiencies (opening and connection formulations, sequential thermal-airflow coupling) or input uncertainty, and it called for archetypal field datasets with measured flow rates to settle the question. Later work closed many tool-capability gaps (a horizontal-opening link component, sensor-based venting control, co-simulation couplings) and quantified input sensitivity, but no study or dataset set conclusively separates model-form from input error for multi-story, atrium-connected buildings, and current EnergyPlus still restricts stratified-atrium network modeling and retains sequential coupling. The open problem is to deliver validated, attributable prediction: improved formulations and coupling guidance, plus multi-building field datasets, sufficient to quantify an error budget separating model-form from input-uncertainty contributions for archetypal naturally ventilated geometries.

Background

Natural ventilation (NV) supplies outdoor air and removes heat using wind pressure and thermal buoyancy alone; hybrid (mixed-mode) ventilation supplements it with mechanical components such as exhaust fans. Both are central strategies for low-energy cooling, and their seasonal performance is evaluated with whole-building energy simulation, in practice by coupling a multizone airflow network (AFN) model to a transient thermal solver such as EnergyPlus. In an AFN each building zone is one well-mixed node with a single temperature and pressure, and airflow through each opening or crack follows an orifice-type or power-law relation in the pressure difference ΔP\Delta P, scaled by an effective opening area and a discharge coefficient CdC_d that accounts for vena-contracta and geometry losses; wind driving enters through wind pressure coefficients CpC_p on each facade, and buoyancy (stack) driving through air-density differences between connected zones at different heights. Fully resolving the flow with computational fluid dynamics (CFD) is too costly for annual, whole-building use, so the network approximation — no air recirculation or temperature stratification within a zone — is the accepted price of long-term simulation. Two structural issues follow. First, coupling: the network needs zone temperatures to compute buoyancy while the thermal solver needs airflow to compute zone temperatures; a ping-pong (sequential) scheme computes each from the other's previous values, whereas a simultaneous scheme solves both together, and the two can diverge at large time steps. Second, attribution: predicted airflow rates and indoor temperatures inherit error both from this model form (a tall atrium is not one well-mixed node, and a large unobstructed internal connection does not behave like a reduced-area orifice) and from inputs (effective vent areas, discharge coefficients, wind pressure data, thermal mass) that are rarely measured in real buildings.

A companion review by the same group (Johnson, Zhai and Krarti 2012) compared four network airflow models against measured natural-ventilation data from the literature and found ventilation rates generally predicted within roughly 35%, except for wind-driven single-sided ventilation. The source study (Zhai, Johnson and Krarti 2011) then assessed the EnergyPlus Airflow Network model against three monitored European office buildings. For the single-story Renson building the coupled model reproduced measured zone temperatures and the summer comfort metric of temperature-excess hours within about 15%. For the multi-story Bang & Olufsen headquarters it predicted 39 hours above 27 °C against 3 measured hours, with ventilation rates near 8 air changes per hour (ach, hourly airflow divided by zone volume) against a 3 ach design exhaust; for the three-storey atrium building Houghton Hall it under-predicted temperatures in moderate weather, over-predicted them in extreme heat, and predicted atrium stack flows larger than expected, with agreement improving when modeled thermal mass was increased. Predictions were highly sensitive to the net effective vent area, the use of local versus distant solar data, and thermal mass, and ping-pong coupling gave indistinguishable results at coupling time steps of 1 to 15 minutes but degraded at 1 hour. The tested tool version also lacked native environmental-sensor control of operable vents, native modulating control of exhaust fans, and adequate treatments of horizontal roof openings and open atrium connections.

Subsequent tool development closed several of these capability gaps: current EnergyPlus (v26.1) provides a dedicated horizontal-opening link component with bidirectional-flow treatment and a user-supplied discharge coefficient, zone- and surface-level venting control by indoor-outdoor temperature or enthalpy difference (including adaptive-comfort modes) with opening modulation, occupant-based ventilation control, and Energy Management System routes for arbitrary sensor-based control. The same current documentation, however, still lists as restricted the representation of air circulation and temperature stratification within a zone — explicitly warning against dividing an atrium into network subzones with the expectation of realistic subzone temperatures or inter-subzone flows — treats Airflow Network exhaust fans as constant volume, and its engineering reference states that zone temperatures from the previous system time step are prescribed inside the airflow solution, i.e., the internal thermal-airflow coupling remains sequential. In parallel, field and sensitivity studies have established that input choices (wind pressure data sources, discharge coefficients, thermal mass) significantly change agreement with measurements, and full-scale validation experiments with uncertainty quantification now exist for individual buildings, but no multi-archetype dataset set with measured airflow rates is available. The unresolved target is stated below.

Problem Statement

The problem, as left open by Zhai, Johnson and Krarti (2011) and not resolved since, is to make whole-building network simulation of natural and hybrid ventilation a validated, attributable predictor of buoyancy-driven airflow and thermal response for multi-story, atrium-connected buildings. Concretely, determine with evidence:

(1) whether the systematic over-prediction of buoyancy-driven (stack) ventilation and of the resulting zone temperatures in multi-story, atrium-connected buildings arises primarily from network model-form deficiencies - the treatment of large internal connections and stratified tall zones, horizontal roof-opening and thermal-chimney link formulations, and the sequential (ping-pong) thermal-airflow coupling scheme - or from uncertainties in input parameters such as effective opening areas, discharge coefficients, wind pressure data, and thermal mass;

(2) validated discharge relationships and bidirectional-flow parametrizations for horizontal roof openings and thermal-chimney geometries usable in network models;

(3) coupling-strategy guidance - the maximal acceptable coupling time step, and whether simultaneous solution is required for certain configurations - established across building typologies rather than single case studies; and

(4) the measured-data coverage required to attribute simulation error conclusively between model form and input uncertainty - measured volume flow rates at inlets, exhaust stacks and fans; on-site wind speed and direction and solar irradiance; documented effective opening areas and discharge coefficients; vent-position and fan-status logs; thermal-mass inventories - for multiple archetypal naturally ventilated building geometries, assembled from openly available field measurements, with any gaps in this coverage stated explicitly.

The completion boundary is a demonstrated, reproducible error budget: predictions of ventilation rates and zone temperatures for archetypal multi-story naturally or hybrid-ventilated buildings - covering at least the monitored single-story naturally ventilated, multi-story stack-ventilated, and multi-story atrium-connected geometries represented in the cited literature - that agree with measurements within bounds the answer states and justifies, benchmarked against the measured ventilation rates and zone temperatures reported for those cited buildings, with the residual error decomposed into model-form and input-uncertainty contributions to the extent openly available data and models support. The four demands above define what this error budget must cover: documenting the data-coverage gaps that limit the decomposition is a required part of the budget rather than a substitute for it, and the budget is to be established from an assembly of openly available measured building studies and documented models rather than from a new field-measurement campaign.

References

  1. Zhai, Z. J., Johnson, M.-H., & Krarti, M. (2011). Assessment of natural and hybrid ventilation models in whole-building energy simulations. Energy and Buildings, 43(9), 2251-2261. DOI: 10.1016/j.enbuild.2011.06.026. https://doi.org/10.1016/j.enbuild.2011.06.026
  2. Johnson, M.-H., Zhai, Z. J., & Krarti, M. (2012). Performance evaluation of network airflow models for natural ventilation. HVAC&R Research, 18(3), 349-365. DOI: 10.1080/10789669.2011.617291. https://doi.org/10.1080/10789669.2011.617291
  3. Ji, Z., Su, Y., & Khan, N. (2012). Performance evaluation and energy saving potential of windcatcher natural ventilation systems in China. International Journal of Architecture, Engineering and Construction, 84-95. DOI: 10.7492/IJAEC.2012.010. https://doi.org/10.7492/IJAEC.2012.010
  4. Carrilho da Graça, G., & Linden, P. (2016). Ten questions about natural ventilation of non-domestic buildings. Building and Environment, 107, 263-273. DOI: 10.1016/j.buildenv.2016.08.007. https://doi.org/10.1016/j.buildenv.2016.08.007
  5. Arendt, K., Krzaczek, M., & Tejchman, J. (2017). Influence of input data on airflow network accuracy in residential buildings with natural wind- and stack-driven ventilation. Building Simulation, 10(2), 229-238. DOI: 10.1007/s12273-016-0320-5. https://doi.org/10.1007/s12273-016-0320-5
  6. Monari, F., & Strachan, P. (2017). Characterization of an airflow network model by sensitivity analysis: parameter screening, fixing, prioritizing and mapping. Journal of Building Performance Simulation, 10(1), 17-36. DOI: 10.1080/19401493.2015.1110621. https://doi.org/10.1080/19401493.2015.1110621
  7. Dols, W. S., Emmerich, S. J., & Polidoro, B. J. (2016). Coupling the multizone airflow and contaminant transport software CONTAM with EnergyPlus using co-simulation. Building Simulation, 9(4), 469-479. DOI: 10.1007/s12273-016-0279-2. https://doi.org/10.1007/s12273-016-0279-2
  8. Saber, E. M., Chaer, I., Gillich, A., & Ekpeti, B. G. (2021). Review of intelligent control systems for natural ventilation as passive cooling strategy for UK buildings and similar climatic conditions. Energies, 14(15), 4388. DOI: 10.3390/en14154388. https://doi.org/10.3390/en14154388
  9. Chen, C., & Gorlé, C. (2021). Full-scale validation of natural ventilation models in Stanford's Y2E2 building. Proceedings of Building Simulation 2021: 17th Conference of IBPSA. DOI: 10.26868/25222708.2021.30679. https://doi.org/10.26868/25222708.2021.30679
  10. Chew, L. W., Chen, C., & Gorlé, C. (2021). CFD-based analysis of the discharge coefficient for buoyancy-driven ventilation in a full-scale operational building. Proceedings of Building Simulation 2021: 17th Conference of IBPSA. DOI: 10.26868/25222708.2021.30482. https://doi.org/10.26868/25222708.2021.30482
  11. He, G., Wu, Q., Li, Z., Ge, W., Lv, D., & Cong, L. (2021). Ventilation performance of solar chimney in a test house: Field measurement and validation of plume model. Building and Environment, 193, 107648. DOI: 10.1016/j.buildenv.2021.107648. https://doi.org/10.1016/j.buildenv.2021.107648
  12. Michalak, P. (2022). Thermal-airflow coupling in hourly energy simulation of a building with natural stack ventilation. Energies, 15(11), 4175. DOI: 10.3390/en15114175. https://doi.org/10.3390/en15114175
  13. Hu, H., Kikumoto, H., Ooka, R., Lin, C., & Zhang, B. (2022). Comprehensive validation of experimental and numerical natural ventilation predictions based on field measurement with experimental house. Building and Environment, 207, 108433. DOI: 10.1016/j.buildenv.2021.108433. https://doi.org/10.1016/j.buildenv.2021.108433
  14. Simões, J. C., & Carrilho da Graça, G. (2025). Experimental validation of neural network-based prediction of natural ventilation bulk airflow rate. Energy and Buildings, 342, 115871. DOI: 10.1016/j.enbuild.2025.115871. https://doi.org/10.1016/j.enbuild.2025.115871
  15. Schreck, C., Rouchier, S., Foucquier, A., & Wurtz, E. (2026). In situ air change rate estimation from metabolic CO2 measurements and state-space modelling. Journal of Building Performance Simulation, 19(4), 694-713. DOI: 10.1080/19401493.2026.2660965. https://doi.org/10.1080/19401493.2026.2660965
  16. EnergyPlus Version 26.1 Documentation: Input Output Reference, Group - Airflow Network (U.S. Department of Energy). https://bigladdersoftware.com/epx/docs/26-1/input-output-reference/group-airflow-network.html
  17. EnergyPlus Version 26.1 Documentation: Engineering Reference, AirflowNetwork Model (U.S. Department of Energy). https://bigladdersoftware.com/epx/docs/26-1/engineering-reference/airflownetwork-model.html
  18. EnergyPlus Version 8.0 Documentation: Input Output Reference, Group - Airflow Network (U.S. Department of Energy). https://bigladdersoftware.com/epx/docs/8-0/input-output-reference/page-029.html