Numerical modeling of non-uniform indoor temperature distribution for coordinated air flow control

The indoor thermal environment is usually non-uniformly distributed in parameters such as temperature and velocity. For precise control of indoor environment, it is necessary to figure out the indoor non-uniform temperature distribution (INUTD) that empowers independently zone control according to the different requirements of occupants. Computational Fluid Dynamics (CFD) tool is usually used to obtain the INUTD, but it consumes huge computational resources and time, which may not be suitable for coupling with control system. To address this research question, this study proposes a numerical modeling method for predicting the INUTD and air flowrate response of multiple diffusers for coordinated air flow control. Firstly, we establish a dataset with 5000 cases, along with the results of INUTD for each case. Then two models, namely room thermal response model (RTRModel) and air flowrate prediction model (AFPModel), are developed by machine learning algorithms to predict indoor temperature and supply air flowrate, respectively. The results show that proposed models are both fast in prediction, less than 1 s for each case. The standard deviation of error in RTRModel developed with support vector machine algorithm is 0.0041 while that in AFPModel developed with Convolutional Neural Network algorithm is 0.0198. Further, a comparative analysis has been conducted between the AFPModel with and without optimization algorithm. The results reveal that using optimization algorithm is more accurate, but it takes more time. While numerical modeling can instantaneous response with qualified accuracy. The proposed method can contribute to independently zonal environment control and occupant-centered micro-environment control.