Efficient Prediction of Indoor Airflow in Naturally Ventilated Residential Buildings Using a CFD-DNN Model Approach

Abstract

Predicting indoor airflow in multi-storey residential buildings is crucial for energy-efficient natural ventilation systems. The indoor environment significantly affects human well-being due to extended indoor time and potential health risks. Precise and efficient airflow simulations are necessary to ensure thermal comfort and air quality. This study introduces a novel approach combining Computational Fluid Dynamics (CFD) simulations with machine learning techniques to predict indoor airflow. Specifically, we explore using a Deep Neural Network (DNN) model for accurate indoor airflow forecasting. The DNN effectively reproduces airflow patterns and temperature distributions. Integrating CFD simulations halves test scenario anticipation time, highlighting efficient indoor airflow prediction potential. Using a data-driven approach, this research demonstrates the feasibility of swiftly and accurately predicting indoor airflow in naturally ventilated residential buildings. Such models can optimize indoor air quality, thermal comfort, and energy efficiency, contributing to sustainable building design and operation.