Data-Driven Approach Enhances Atrium Design for Improved Energy Efficiency and Daylighting

Strategic Insights -

Researchers from Shandong Jianzhu University have developed a data-driven approach to optimize atrium design in public buildings, utilizing a multimodal neural network model to minimize carbon emissions and energy return on investment (EROI) while improving daylighting. The study, funded by the Science and Technology Program of the Ministry of Housing and Urban-Rural Development of the People's Republic of China, involves a case study of 7 cities in China. According to the research, the approach has achieved significant results, including a carbon emission reduction rate exceeding 89% and an energy return on investment of 9.37.

Key Takeaways:

  • The data-driven approach integrates simulation, prediction, and optimization to optimize atrium design, skylight, PV system, and fenestration factors.
  • The approach utilizes a multimodal neural network model, combining ResNet-50, ViT-32, and Inception-V4 for image features and MLP and TabTransformer for numerical features.
  • The study was conducted in 7 cities in China, and the results show that the approach can achieve significant reductions in carbon emissions and improvements in daylighting.
  • The key factors in optimizing atrium design are identified as atrium width, skylight-to-roof ratio, PV coverage, and horizontal panel.
  • The ViT-32+MLP model significantly improves prediction accuracy and efficiency compared to other models.
  • The research offers a practical method for SIPV design, supporting regional adaptation and scalable application in public building atriums.

Statistics:

  • Carbon emission reduction rate (CER): 89% or more
  • Daylight factor (DF): 1.2% - 3.4% increase
  • Daylight glare probability (DGP): 0.28 or less
  • Energy return on investment (EROI): 9.37

Sources:

  • Data-driven Approach of Atrium Skylights-integrated Photovoltaic Systems Design Based On Multimodal Deep Learning: Considering Different Regions In China. Energy, 2025;335.
  • Pergamon-elsevier Science Ltd, The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, England. (Elsevier - www.elsevier.com; Energy - www.journals.elsevier.com/energy/)
  • Yanqiu Cui, Shandong Jianzhu University, School of Architecture and Urban Planning, Jinan 250100, People's Republic of China.