Identification of zonal pollutant diffusion characteristics using dynamic mode decomposition: Towards the deployment of sensors

Sensor monitoring is a common manner for indoor environment control. However, the distribution of indoor air pollutant shows large variation both spatially and temporally in building environments, which poses great challenges to sensor deployment. Accurately describing the spatiotemporal distribution of indoor physical parameters through limited monitoring points has always been the core problem of intelligent monitoring. Thus, a reduced order method based on dynamic mode decomposition (DMD) was applied for model reduction to capture the characteristics of spatiotemporal distribution of contaminant transportation. The deployment of sensors should cover all the regions with similar diffusion characteristics to obtain valid data for fast prediction. This work focused on transient results from a step change in pollutant transportation. We first obtained the steady state airflow field from one stationary simulation and another three transient simulations were performed. Instead of solving coupled Partial Differential Equations (PDEs), the concentration field under the action of fluid dynamic can be represented by 12 DMD based eigenmodes that are dominant over the reduced space, contributing 98.2% to the concentration field. The transportation of contaminant among zones with similar diffusion characteristics can be represented by the dominant eigenmodes. The results of the fast prediction match well with the RANS simulation with a maximum error of 3.26% along two selected monitoring lines. We also discussed the spatiotemporal characteristic of typical zones for sensor deployment. The results are consistent to our previous research and will be a significant finding for the development of intelligent zonal monitoring and control.