In short: Energy modeling programs inherit their occupancy assumptions from standard databases and those databases disagree so badly that the same building can return results varying by up to 4,000–6,000% . The FREDS occupancy database is the fix: a harmonized set reconciling six major sources into one coherent default, published in Energy & Buildings (Rugani, Picco et al., 2024).
Every energy simulation contains a hidden co-author: the database that decides how the building is occupied. When you have not measured how many people will occupy a space, when the lights will run, or what equipment will be plugged in (and at design stage, nobody has) ; the simulation has to borrow those numbers from a standard library. That choice is normally made by the modeller, not the software.In some tools a default database is applied without being explicitly flagged to the user.
Our previous article documented what that silence costs: the same building, simulated with six different industry-standard occupancy databases, returned results spanning up to 4000–6000% , averaging 500–1000% per city within a single building-use category (Rugani, Picco, Salvadori, Fantozzi and Marengo, 2024). This article is about the constructive half of that research: how the team built a replacement and why it shares a name with this platform.
The harmonization method
The research team’s diagnosis separated the problem into two layers. Group A: absolute load values the watts per square meter of people, lighting and equipment a profile assigns. Group B: schedules the hourly shape describing when those loads occur across days and weeks. Both varied wildly between the six sources analyzed.
The fix proceeded in disciplined steps rather than one heroic average:
Step one: Harmonize the loads. The six sources’ absolute values were reconciled, category by category across nine building-use types, into a single coherent set (the harmonized “Group A*” ). Each source’s schedules were retained. Re-simulating across Reykjavik, London and Madrid measured exactly how much of the chaos the load values were responsible for: the worst discrepancies collapsed from thousands of percent down towards roughly 90% .
Step two: Compare the residual. A ~90% spread is vastly better than 4000%, but it is not done. The remaining variance sat in Group B the hourly schedules, where sources disagree about when buildings are actually used. So the work continued: retrospective analysis of the schedule structures, identification of outliers, and the definition of new harmonized hourly profiles (Group B*) to complete the database.
Step three: Validate the whole. The completed database was re-simulated against the originals to confirm it behaves as a robust central estimate rather than simply adding a seventh opinion to the six existing ones.
The published outcome is a unified occupancy database covering the nine use categories, designed specifically as a defensible default for dynamic simulation at early design stage, the moment when project-specific occupancy data does not yet exist and the database is your occupancy model. In the paper, the team gave it a name: FREDS.
Because the platform and the database share a name
That naming is not branding garnish; it describes the architecture. The FREDS platform’s promise is validated simulation from limited inputs: you describe the building in concept-stage terms, and the platform supplies research-grade defaults for everything you cannot yet know. The credibility of that promise lives or dies on the quality of those defaults, and occupancy is the most consequential default of all.
So the harmonized database is not an optional library you can import into the platform. It is the platform’s native assumption set. Select “building type” in FREDS and the loads and schedules underneath are the peer-reviewed harmonized profiles not whichever single national standard a developer happened to wire in years ago. The 4000% input lottery is resolved before your first simulation runs, and the resolution is documented in Energy & Buildings rather than in a changelog.
This is, we would argue, the right division of labor between tool and user. Engine fidelity (EnergyPlus) handles the physics. The harmonized database governs the most volatile inputs. The user’s attention is reserved for what they genuinely know better than any database: the building itself.
What this does not claim
Honesty about limits is part of the method, so three caveats belong in this article rather than in small print.
Harmonization is not measured. A harmonized default is the best available assumption when no project data exists. If you have real occupancy information about a tenant’s actual hours, metered equipment loads that data beats any database, and the platform is built to accept it.
Schedules still carry irreducible uncertainty. The research itself shows hourly usage patterns remain a genuine source of variance even after harmonization; buildings are used by people, and people are not standards-compliant. The database narrows the uncertainty band dramatically; it cannot abolish it.
A default is a starting point, not a verdict. For decisions that hinge specifically on occupancy densification studies, ventilation strategy and sensitivity testing across occupancy assumptions remains good practice, and running those variants is precisely what a minutes-fast platform makes affordable.
The takeaway for anyone choosing energy modeling programs
When evaluating energy modeling programs, the interface is what demonstrates well but the assumption layer is what determines whether two simulations of the same building agree. The question this research arms you with is simple to ask and revealing to hear answered: where does this tool’s occupancy data come from, and has that choice ever been tested?
For FREDS the answer is unusually short. It comes from a harmonization of NREL, EN 16798-1, ISO 17772-1, ASHRAE 90.1 and the academic literature; it was tested across nine building types and three climates; and the test is published, peer-reviewed and open access under the same name as the platform you would be running.
Every FREDS simulation starts from the published database. Run one and inspect the assumptions
Frequently Asked Questions
What is the FREDS occupancy database?
Why do energy simulations of the same building return different results?
How was the harmonised database validated?
Can I use my own occupancy data instead of the default?
How does the FREDS database compare with UK NCM or CIBSE occupancy profiles?
References
Written by Marco Picco, PhD co-founder of FREDS4Buildings and a building-physics lecturer at the University of Lancashire, whose work on building energy simulation has been published in Energy & Buildings , Building and Environment and the IBPSA Building Simulation conference proceedings. About Marco Picco
Reference: Rugani, R., Picco, M., Salvadori, G., Fantozzi, F. and Marengo, M. (2024). A numerical analysis of occupancy profile databases impact on dynamic energy simulation of buildings. Energy & Buildings, 310, 114114. View paper
