Full Library
Developing a deep neural network model for predicting ventilation rates in public housing buildings in Hong Kong
Resource type
Journal Article
Authors/contributors
- Dai, Ho Kam (Author)
- Shi, Yifu (Author)
- Chen, Chun (Author)
Title
Developing a deep neural network model for predicting ventilation rates in public housing buildings in Hong Kong
Abstract
In Hong Kong, 45 % of the population live in public housing buildings. This study aimed to develop deep neural network (DNN) models for efficiently predicting the ventilation rates in public housing buildings in Hong Kong in the era of modular flat design. First, a database of ventilation rates in Cheung Tai House, a representative public housing building, under 23 wind conditions and 32 different ventilation settings, was obtained by means of computational fluid dynamic (CFD) and multi-zone airflow network models. On-site measurements were conducted in two residential units to validate the numerical models. The database was then used to train the DNN models for the 32 ventilation settings, respectively, for predicting the ventilation rate. The trained DNN models can accurately predict the ventilation rates in Cheung Tai House with a mean absolute percentage error (MAPE) of 4.3 %. The trained DNN models were further validated by numerical simulation results for Hung Fuk Estate and field measurement results for Chun Yeung Estate. The results show that the trained DNN models based on the Cheung Tai House database could predict the ventilation rates of other public housing buildings reasonably well, with a MAPE of 10 % regardless of the building’s structural shape.
Publication
Energy and Buildings
Pages
113993
Date
2024-02-10
Journal Abbr
Energy and Buildings
ISSN
0378-7788
Accessed
12/02/2024, 21:28
Library Catalogue
ScienceDirect
Call Number
openalex:W4391906572
Extra
openalex: W4391906572
Citation
Dai, H. K., Shi, Y., & Chen, C. (2024). Developing a deep neural network model for predicting ventilation rates in public housing buildings in Hong Kong. Energy and Buildings, 113993. https://doi.org/10.1016/j.enbuild.2024.113993
Link to this record