GPT-based data-driven urban building energy modeling (GPT-UBEM): Concept, methodology, and case studies

Abstract

Achieving carbon neutrality is a critical global goal, with urban building energy modeling (UBEM) playing a pivotal role by providing data-driven insights to optimize energy consumption and reduce emissions. This paper introduces GPT-based urban building energy modeling (GPT-UBEM), a novel approach utilizing GPT’s advanced capabilities to address key UBEM challenges using GPT-4o. The study aimed to demonstrate the effectiveness of GPT-UBEM in performing UBEM tasks and to explore its potential in overcoming traditional limitations. Specifically, (1) basic analytics of urban data, (2) data analysis and energy prediction, (3) building feature engineering and optimization, and (4) energy signature analysis were conducted in four case studies. These analyses were applied to 2,000 buildings in Seoul and 31 buildings in Gangwon-do, South Korea. Through case study, the findings highlighted the ability of GPT-UBEM to integrate diverse data sources, optimize building features for high accuracy in prediction models, and provide valuable insights for urban planners and policymakers through the use of expert domain knowledge and intervention. Additionally, based on the results derived from GPT-UBEM in this study, the current limitations of GPT-UBEM (L1 to L3) and future research directions (F1 to F4) have been outlined.