Generation of a large synthetic database of office tower’s energy demand using simulation and machine learning

Alammar, Ammar and Jabi, Wassim ORCID: https://orcid.org/0000-0002-2594-9568 2023. Generation of a large synthetic database of office tower’s energy demand using simulation and machine learning. Presented at: 6th International Symposium on Formal Methods in Architecture (6FMA), 24-28 May 2022. Published in: Mora, Plácido Lizancos, Viana, David Leite, Morais, Franklim and Vaz, Jorge Vieira eds. Formal Methods in Architecture. FMA 2022. Digital Innovations in Architecture, Engineering and Construction. Singapore: Springer, pp. 479-500. 10.1007/978-981-99-2217-8_27

Preview

PDF - Accepted Post-Print Version Download (1MB) | Preview

Abstract

Machine learning (ML) has proven to be an effective technique serving as a predictive surrogate model for evaluating the performance of buildings. This approach provides considerable benefits such as reduced processing time, simplified predictions and computational efficiency. This study presents an alternative approach using a decision tree (DT) model to predict the hourly cooling loads of adaptive façade (AF) in significantly less time than when applying building performance simulation (BPS). Due to the absence of real-world data, generative parametric modelling of a prototypical office tower with an adaptive façade shading system situated in an urban setting was carried out along with simulation of its energy demand using the Honeybee add-on for Rhino/Grasshopper software. The generated large synthetic datasets were fed in so as to train and test the decision tree model. The prediction results revealed an extremely accurate model capable of estimating cooling loads in a matter of seconds. The paper concludes by arguing that decision tree surrogate models can be effectively used by researchers and designers to assess their future adaptive façade design.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Published Online
Status:	Published
Schools:	Architecture
Publisher:	Springer
ISBN:	9789819922161
Date of First Compliant Deposit:	14 August 2023
Last Modified:	02 Aug 2024 01:30
URI:	https://orca.cardiff.ac.uk/id/eprint/161526

Actions (repository staff only)

Edit Item