Co-zyBench is a benchmark for evaluating thermal comfort control strategies with an EnergyPlus-based co-simulation loop.

The repository entry point for the full benchmark is main.py. The core runtime path uses:

• main.py
• occupants.py
• comfort_collector.py
• strategies.py
• sim_ep.py
• result_collector.py

• Windows

• EnergyPlus 22.2.0

The code uses EnergyPlus IDF models together with the EnergyPlus Python API helper.

• Python 3.10 or 3.11

Install the Python packages imported by the runtime call graph:

• colorama
• tqdm
• eppy
• pandas
• numpy
• pytz
• scikit-learn
• energyplus_api_helpers

Example setup:

python -m venv .venv
.venv\Scripts\activate
python -m pip install --upgrade pip
python -m pip install colorama tqdm eppy pandas numpy pytz scikit-learn energyplus_api_helpers

Before the first run, update these paths to match your local EnergyPlus 22.2.0 installation:

• main.py: idd_file
• sim_ep.py: CoSimulation.ep_path

functional_test.py also accepts these paths as command line arguments:

• --energyplus-dir
• --idd-file

Example installation path:

C:\EnergyPlusV22-2-0

Example IDD path:

C:\EnergyPlusV22-2-0\Energy+.idd

Run the short functional test first:

python functional_test.py

This test runs a real EnergyPlus simulation on the small office model with:

• models/office/CZ4/in.idf
• models/office/trajectories
• models/office/occ_config.txt
• city paris
• aggregation strategy majority
• HVAC controller fixed_rule
• a one-day temporary RunPeriod

The script creates a temporary model copy in .functional_test_runtime/functional_test.idf, runs EnergyPlus, and writes a result file under results/.

Optional arguments:

python functional_test.py --city paris --start-date 2010-01-01 --days 1

Supported --start-date months are 1, 2, 6, 7, 8, and 12, because the control loop in sim_ep.py is exercised in those months.

Expected output:

• results/result_<timestamp>/result-functional_test_majority_fixed_rule_1.txt

The full benchmark is configured directly in main.py.

Main configuration fields:

• path_ep_model
• path_trajectories
• path_profile
• hvac_controller
• algorithms
• cities

In the committed default setup:

• path_ep_model = models/synthetic_large/10floors_V2.idf
• path_trajectories = models/synthetic_large/trajectories
• path_profile = 1200
• hvac_controller = ["preference_estimation"]
• algorithms = ["majority", "fair", "drift"]
• cities = ["paris"]

path_profile can be either:

• a path to an occupant profile file
• an integer occupant count for synthetic occupant generation

After installing the Python packages and setting the EnergyPlus paths:

python main.py

The committed default benchmark run uses:

• models/synthetic_large/10floors_V2.idf
• models/synthetic_large/trajectories
• 1200 synthetic occupants
• city paris
• algorithms majority, fair, drift
• HVAC controller preference_estimation
• annual simulation mode launched from sim_ep.py
• control dates 2010-01-01 to 2010-12-30

During the full benchmark run, the repository generates:

• models/temp/temp.idf
• results/result_<timestamp>/result-<city>_<algorithm>_<hvac_controller>_<repeat>.txt

Each result file contains:

• cooling energy in kWh
• heating energy in kWh
• fan energy in kWh
• total energy in kWh
• CO2 value
• ITC
• average ITC
• per-occupant TCE-related values
• standard deviation of TCE-related values

Distributed under the MIT License. See LICENSE for details.