Enhancing Energy Intelligence in Taiwanese Office Buildings: Utilizing a Novel BIM-Derived Dataset for AI-Driven Energy Consumption Prediction

Abstract

The construction sector's substantial contribution to global greenhouse gas emissions necessitates urgent action to address the environmental impact of building activities. This research centers on the critical issue of energy consumption in office buildings, constituting a significant portion of global emissions. With a specific focus on Taiwanese office buildings, characterized by unique structural features, this study introduces an innovative framework that combines Building Information Modeling (BIM) and Machine Learning (ML) to enhance energy intelligence. The primary objective is to develop the TaiBIM_EnergyOffice dataset, a comprehensive and meticulously curated repository of energy consumption patterns specific to typical office structures in Taiwan, using BIM techniques. This compilation will encompass diverse facets such as architectural configurations, material specifications, and environmental conditions, ensuring a holistic representation of the factors influencing energy consumption in the targeted context. The dataset is envisaged as a valuable foundation for subsequent ML-driven energy consumption predictions. Envisioned as a dual-fold approach, the first facet involves creating single-objective ML models tailored to accurately predict energy consumption patterns in Taiwanese office buildings. Anticipated to exhibit high accuracy by assimilating insights derived from the BIM-derived dataset. Simultaneously, the second facet involves implementing multi-objective ML models, which may include estimating energy consumption and optimizing building features to mitigate energy usage. This offers a more holistic and sustainable approach to construction. The expected outcome is a suite of ML models that not only predict energy consumption with precision but also provide insights into the trade-offs inherent in the optimization process. This research bridges academic inquiry and practical utility, with the results envisioned to contribute significantly to developing a robust and versatile framework for energy consumption prediction and optimization in the context of office buildings in Taiwan.