Research on a dynamic simulation method of atrium thermal environment based on neural network

As a widely used interior building space, atriums usually have a large volume and a large-scale glass curtain wall which make designing and organizing the indoor thermal environment much more complex than a normal building space. Methods of traditional air conditioning design can hardly meet all the requirements. The goal of this research is to find a new simulation method based on an artificial neural network. The artificial neural network is used to describe the annual dynamic process with detailed parameter data and a reasonable amount of calculation time. Computational Fluid Dynamics (CFD) enables the prediction of indoor thermal data for buildings, and energy simulation model provides a whole building energy analysis, the artificial neural network (ANN) is used as an integrating tool to couple the energy simulation model and CFD model. The ANN model is iterative with the energy simulation model (DeST software), and can reasonably predict reasonable dynamic energy consumption and thermal environment parameters.

► We develop and validate a dynamic simulation method of atrium thermal environment.
► The method can reflect the annual change of atrium thermal environment.
► It can describe the flow and temperature field of the atrium at different time.
► And it can represent the annual change feature of flow and temperature field.
► The annual dynamic simulation is achieved under reasonable amount of calculation.