Dynamic analysis and diagnosis of building life cycle carbon emissions based on regrouping-adversarial transfer learning

In the context of global low carbon development, the measurement and diagnosis of lifetime building carbon emissions has become a hot and basic problem to be urgently solved. However, due to the limitations of technology, models, and real-world factors, only a small number of buildings can currently measure their carbon emissions. The study introduces the concept of building carbon emission boundaries and proposes a method for measuring building carbon emissions based on antagonistic migration algorithm fusion. The improved machine learning is used to enhance the measurement of building carbon emissions, and feature extraction is utilized to identify diagnostic factors such as outdoor temperature, and outdoor humidity among others, which contain valuable information about specific constructions or regions. Taking the building in Shaanxi, China as an example, the results show that the performance can significantly improve after optimization, accompanied by a decrease of 19.41% in the mean square error value compared to the linear model. And the material production phase contribute 63.39% of the carbon emissions. The study presents a feasible approach for building dynamic analysis and diagnosis, laying the foundation for future research in reducing carbon emissions, developing diagnostic analyses, and improving diagnostic techniques. The study therefore has theoretical research value and practical application prospect.