Building energy consumption optimization method based on convolutional neural network and BIM

Abstract

The increasing tension of energy supply and demand makes the optimization of building energy consumption more and more concerned by researchers. Based on the theory of convolutional neural network and BIM (Building Information Modeling), a building energy consumption optimization model is constructed. The optimization parameter solving problem of convolutional neural network is solved. In the simulation process, a calculation model of the same size as Revit's three-dimensional model is established in eQUEST software, and the basic analysis parameters of the model, such as geographical location, meteorological data and other information, component materials, and running time table are set as unified standards. The energy consumption simulation analysis was carried out for the self-built model and the model automatically generated by the improved DOE-2 file in eQUEST software. The body coefficient of the building is 0.370, and the window-wall ratios in the east, west, south and north directions are 0.07, 0.21, 0.30 and 0.16, respectively, which all meet the requirements of relevant specifications. Compared with the scheme before optimization, it is found that building energy consumption is reduced by 24.53%, natural lighting is increased by 18.98%, and natural pressure par hours are increased by 10.57%.

Keywords

• Convolutional neural network (CNN)
• BIM
• Building energy consumption
• Consumption optimization
• Low carbon
• Emission reduction