Integrating Gradient Boosting and Parametric Architecture for Optimizing Energy Use Intensity in Net-Zero Energy Buildings

Achieving net-zero energy building (NZEB) status requires accurate energy use intensity (EUI) calculations, as conventional methods often fail to capture the complexity of design and climatic conditions. In this research, a parametric energy modeling approach was used to conduct 1,350 simulations and analyze the impact of design parameters on building EUI. These simulations covered six building types—an existing building and I-, L-, T-, U-, and H-shaped buildings—across eight locations in different climate zones. A case study was conducted in Busan, Korea, where on-site measurements were obtained using portable devices to validate the simulation results. I-shaped buildings exhibited the lowest EUI, reaching 109 kWh/m²/yr at 0° and 180° orientations. The simulation results indicated that building orientations of 140°, 90°, 135°, and 270° tended to produce higher EUI values, whereas 0° and 180° showed lower EUI values of 122 and 123 kWh/m²/yr, respectively. The use of triple-pane insulated glass effectively reduced the I-shaped building's EUI to 103 kWh/m²/yr. Implementing photovoltaic (PV) systems further reduced the EUI significantly, with the I-shaped building achieving an EUI of −14 kWh/m²/yr at a 20% PV efficiency. Analysis using an extreme gradient boosting (XGBoost) model revealed that the climate zone, PV area, and type of heating, ventilation, and air-conditioning system significantly affected the EUI. This model, processed using Colab, was highly effective, with an R-squared value of 0.99, a root mean square error of 4.57, and a mean absolute error of 1.99. These findings demonstrate that the XGBoost model can effectively capture complex data patterns and can be used for high-accuracy EUI estimation. Doi: 10.28991/CEJ-2025-011-03-06 Full Text: PDF