An energy-saving design method for residential building group based on convolutional neural network

Abstract

Current building energy saving methods mainly focus on the individual building, seldom focus on the building group. Different from the energy saving design method which takes the individual building as the object, the energy saving design of building group also needs to consider the influence of microclimate on building energy consumption, because microclimate is the boundary condition for building energy consumption calculation. The purpose of this study is to investigate the influence of the building group form and geometric factors on the energy consumption of residential buildings in Beijing. First, this study developed a complex plane generator based on convolutional neural network (CNN), and used the generator to extract and classify the plane of residential buildings in Beijing, and generate three standard residential building models. Secondly, the microclimate conditions of the three standard models in summer and winter were calculated by ENVI-met, and the energy consumption of individual building with different geometric form was calculated by DesignBuilder. Based on the above simulation results, the influence of group forms and geometric factors on microclimate and the influence of geometric form on energy consumption of individual building were quantified, and the influence law of group forms and geometric factors on energy consumption of buildings was revealed. The obtained law can be used to quickly estimate the change of total building energy consumption when the group form or geometric factor changes at the initial stage of the design of residential buildings.