From Canopy Flow to Cooling: Can 1D Ventilation Models Predict Natural Cooling from LES?

Natural cooling offers a sustainable alternative to energy-intensive mechanical cooling by utilizing cooler outdoor air, typically at night, to maintain comfortable indoor temperatures during the day. This process relies on buoyant and wind forces to drive ventilation. In building energy models, natural ventilation is often estimated using one-dimensional (1D) flow models driven by pressure differences. These pressures are typically derived from empirical models based on wind tunnel experiments. However, for accurate predictions, both the pressure estimates and flow models must be sufficiently precise.

A key limitation is that these models often fail to capture the complexity of urban environments, where surrounding buildings significantly influence airflow. Large Eddy Simulations (LES) provide a powerful alternative, offering detailed and accurate representations of wind flow through urban areas. Moreover, LES can explicitly simulate building interiors, enabling a fully coupled analysis of wind-driven natural ventilation. However, simulating building interiors presents challenges. First, interior modeling requires a fine mesh, adding computational expense. Second, resolving building interiors depends on detailed knowledge or assumptions about the indoor layout.

To address these challenges, we compare LES simulations with and without building interiors. For simulations without interiors, we instead predict ventilation rates using 1D flow models. Preliminary results indicate reasonable agreement between the 1D models and simulated ventilation rates. Predicting ventilation rates from LES simulations without building interiors opens up exciting possibilities. First, we can estimate ventilation rates for various interior layouts using a single simulation of the exterior flow. Second, we can combine 1D interior flow models with exterior flow fields generated from machine learning models trained on LES data—without the need to train these models on interior flow fields.