Building Energy and HVAC Digital Twin

Vendor-neutral, reproducible thermal modeling, parameter calibration, held-out validation, uncertainty, sensitivity, and comfort-aware energy optimization.

View on GitHubDownload one-page PDFBuilt with Vicena

Transparent physics

Inspect the single-zone 1R1C balance, gains, ideal loads, COP, fan power, interval energy, cost, and comfort.

Calibratable

Fit bounded parameters on a declared chronological training period and report approximate local uncertainty.

Validation first

Run held-out metrics, residual diagnostics, analytical convergence, physical response, and invalid-input gates.

Research extensible

Upload documented weather, occupancy, temperature, meter, equipment-status, price, and setpoint series.

Completed synthetic reference answer

32.18 kWh for the deterministic 24-hour baseline. Increasing envelope UA by 25 percent increases electricity use by 14.13%. The fine-step analytical free-decay maximum error is 0.093 degC. These are Level 0 synthetic results, not measured building validation.

Quickstart

git clone https://github.com/vicena-labs/building-energy-hvac-digital-twin.git
cd building-energy-hvac-digital-twin
python -m pip install -e .
python scripts/run_reference_cases.py
hvac-twin run datasets/example/building_timeseries.csv --output outputs/reference_run
Building Energy and HVAC Digital Twin one-page overview

Scientific status and limitations

This is a Level 0 executable synthetic reference and a framework for a calibratable R&D twin. It is not production-ready. Major limits include one well-mixed zone, idealized control, constant COP, simplified occupancy and solar gains, no humidity or ventilation physics, and no measured multi-season validation.

Citation, license, and agent use

Use CITATION.cff. Repository code and documentation are MIT licensed. External data retain their own licenses.

Clone https://github.com/vicena-labs/building-energy-hvac-digital-twin.git. Read AGENTS.md, AGENT_PLAYBOOK.md, RUNS.md, and the repository skill. Run tests and the reference cases before scientific changes. Report what is synthetic and what measured data are required. Do not infer missing units or metadata.