A study on the advancement of building energy simulations: Utilizing ChatGPT to enhance occupancy rate predictions and sustainability outcomes

Abstract

Various countries around the world are implementing a system to evaluate and certify the energy efficiency of buildings. Energy simulation plays an essential role in the process of evaluating and optimizing the energy performance of buildings, contributing to operation of sustainable buildings and reduction of greenhouse gas emissions. In an energy simulation, the occupancy rate, or the percentage of people using the building, serves as a critical parameter. This is essential for accurately predicting energy consumption, ventilation, and thermal comfort in a building. Various methods and tools have been developed to collect accurate occupant data, but they often face significant challenges in terms of time and cost. Existing occupancy schedules often fail to adequately reflect building types, user characteristics, and regional variations. To address this problem, research studies are being conducted to obtain more precise occupancy schedules through various methods such as surveys, physical sensors, Wi-Fi data, and deep learning models. ChatGPT has the capability to answer questions in a wide range of fields due to its extensive knowledge base. It finds applications in various domains, including building energy management, education, medicine, military affairs, construction projects, human resources, and resource movement predictions. This study introduces a methodology for estimating a building’s occupancy rate based on its use and user characteristics, utilizing ChatGPT. The reliability of the occupancy schedules generated by ChatGPT was confirmed through a survey, and this methodology exhibited high consistency compared to existing published occupancy rate references. Furthermore, after conducting energy simulations using a modeling approach similar to a real residential space, the results closely matched the actual energy consumption when employing the occupancy rates generated by ChatGPT as input values. These findings confirm the effectiveness of occupancy schedules that take into account the building’s use and characteristics in predicting energy consumption patterns for the building.