Multi-objective optimization of indoor comfort and carbon emissions using a feature-fusion CNN-MLP model for morphological analysis of mixed residential areas

Abstract

As China pursues sustainable urbanization, the transformation of old mixed use communities provides a key opportunity to improve living conditions and reduce environmental impact. With the help of artificial intelligence (AI) and multi-source big data, this study builds typical multi-functional residential types in provincial capitals in different climate zones. A feature fusion convolutional neural network multi - layer perceptron (CNN - MLP) model is developed. According to the key design parameters such as building height, window characteristics, function composition ratio, six representative residential form types are automatically extracted. In engineering application, the real shape model derived from artificial intelligence is combined with traditional simulation. In addition, the performance prediction model based on artificial neural network (ANN) not only ensures the high accuracy of the performance parameters calculation of multifunctional buildings, but also significantly shortens the calculation time. Sensitivity analysis focuses on solar thermal gain coefficient and window U value, and determines the key factors affecting carbon emissions and indoor thermal comfort. Using NSGA - II algorithm to carry out multi-objective optimization, the Pareto optimal design scheme of balancing carbon emission reduction and thermal comfort is determined, and these strategies are analyzed under different building forms and climate conditions. The results show that the SHGC and U value of windows are the most important factors affecting carbon emissions and thermal comfort. Through multi-objective optimization, the efficiency and residential comfort of low-carbon reconstruction of old communities can be significantly improved, which provides a quantitative basis for differentiated climate responsive design. Through the comprehensive control of window wall ratio, function ratio and other parameters by multi-objective algorithm, the thermal environment adaptability can be significantly improved while reducing building energy consumption, and the collaborative optimization of ecological benefits and living quality in low-carbon transformation can be realized. This study provides data support for the low-carbon reconstruction of old urban areas by constructing the optimization strategy of building form and the parameter analysis framework driven by artificial intelligence in the mixed functional community of multi climate areas, and deeply combines the real data and performance simulation to guide the sustainable architectural design and urban renewal. The proposed artificial intelligence integration framework significantly improves the analytical efficiency of complex building parameters, and its multi-objective optimization mechanism provides an innovative methodology for balancing carbon emissions and thermal comfort, promoting the development of green building theory and helping the practice of low-carbon city regeneration.

Topics & Concepts