BIM-based carbon emission prediction model for prefabricated housing design

Abstract

Driven by the global dual carbon goals, the construction industry faces urgent tasks in carbon emission reduction. As one of the key approaches to achieving low-carbon development, prefabricated housing requires accurate quantification of carbon emissions during the design phase. However, traditional design methods lack effective tools for early-stage carbon emission estimation, making it difficult to meet emission reduction goals. Therefore, this study proposes a BIM-based carbon emission prediction model for prefabricated housing during the design phase. First, the model integrates comprehensive construction data through BIM and subsequently combines these data with life cycle assessment to quantify carbon emissions across different life cycle stages. Following dimensionality reduction via principal component analysis, a backpropagation neural network leverages its nonlinear mapping capability to mine data associations for rapid prediction. Experimental results indicate that, compared with benchmark models, the proposed model achieves a root mean square error of 8.32 kgCO 2 /m 2 and 7.85 kgCO 2 /m 2 on the two datasets, respectively. In a case study of prefabricated residential buildings in Beijing, the loss value drops to 0.031, while the mean absolute error of the carbon emission prediction reaches 5.23 kgCO 2 /m 2 with an accuracy rate of 92.3%, demonstrating superior predictive performance. This study delivers a practical quantitative approach to support low-carbon decision-making during the design stage of prefabricated housing, while also introducing an alternative pathway for achieving intelligent carbon emission forecasting.