Multi-source data fusion-driven urban building energy modeling

Abstract

The energy efficiency of buildings is important for creating sustainable cities. In this context, urban building energy modeling (UBEM) is essential for comprehending and predicting building energy consumption and the patterns from a city-wide perspective. This study proposes a multi-source data-fusion-driven UBEM and its real-world applications in Seoul, South Korea. Unlike existing models that utilize energy data as input variables, models based solely on building features play a crucial role in various projects in which energy consumption data is unavailable, such as new building projects, retrofitting projects, and new city developments. In this modeling scenario, data fusion using multi-source data is crucial for developing highly accurate and robust models because of the limited availability of input features. In this study, the utilization of multi-source open data, along with simulated data such as shading factor, enables the fusion of these datasets to derive building features. These features are subsequently utilized to train the prediction model, leading to accurate and robust modeling outcomes. Thus, five case studies were conducted for 47,391 buildings, located in Seoul, South Korea, to demonstrate how and how much data fusion-driven features enhance the performance of UBEM. Discussions in each case provide insights into feature selection and effective utilization tailored to each modeling context. Finally, the developed model was tested in two additional regions, and it was observed that adding features to the model training improved the model performance in both regions. These findings emphasize the importance of data fusion for robust UBEM in the context of sustainable cities.