Data-driven modelling of pressurized corridor ventilation system performance in a multi-unit residential building

Area of Science:

• Building Science
• HVAC Systems Engineering
• Environmental Engineering

Background:

• Pressurized corridor (PC) ventilation is common in multi-unit residential buildings (MURBs) for air supply and contaminant control.
• Ventilation in PC systems relies on pressure differentials, which are influenced by occupant behavior, weather, and building operation.
• Understanding these pressure dynamics is crucial for ensuring adequate ventilation and preventing contaminant transfer.

Purpose of the Study:

• To develop and validate an XGBoost regression model for predicting inter-zonal pressure differentials in MURBs with PC systems.
• To assess the model's utility as a diagnostic tool for evaluating ventilation performance.
• To investigate factors influencing ventilation and contaminant transfer in PC systems.

Main Methods:

• Collected 6-month field data from a 17-storey MURB in Toronto, Canada.
• Measured corridor-to-unit and exterior-to-unit pressures, airflow, temperature, and humidity.
• Developed an XGBoost regression model, including feature selection and hyperparameter tuning.

Main Results:

• The XGBoost model accurately predicts inter-zonal pressure differentials in PC systems.
• The study identified key variables impacting building pressure differentials.
• The model demonstrated potential as a diagnostic tool for ventilation system performance.

Conclusions:

• The developed XGBoost model provides a valuable tool for understanding and improving ventilation in MURBs with PC systems.
• Accurate prediction of pressure differentials is key to optimizing ventilation and mitigating contaminant risks.
• The model can serve as a virtual testing environment for ventilation strategies.