CFD- and BPNN- based investigation and prediction of air pollutant dispersion in urban environment

Abstract

<p>This study employed a Computational Fluid Dynamics (CFD)-based back propagation <a href="https://www.sciencedirect.com/topics/social-sciences/neural-network" title="Learn more about neural network from ScienceDirect's AI-generated Topic Pages">neural network</a> (BPNN) to investigate and predict the pollutant dispersion in an ideal urban environment. The training and test datasets were generated using the Reynolds-averaged Navier–Stokes (RANS) model, which involved 60 cases by varying reference inflow speeds, NO & NO₂, and ambient O₃ concentrations. The results indicated that the ambient flow was gradually decoupled from the reversed flow at each building's leeward zone further downstream, leading to heightened difficulty in pollutant elimination. Upstream buildings located in front of the emission source were rarely affected by pollutants, regardless of changes in <a href="https://www.sciencedirect.com/topics/engineering/damkohler-number" title="Learn more about Damköhler numbers from ScienceDirect's AI-generated Topic Pages">Damköhler numbers</a> (<em>Da</em>). While varying <em>Da</em>_NO provides a limited influence, confining <em>Da</em>_O3 can effectively reduce hazardous pollutants' impact around the emission source vicinity, on the urban streets, and in the further downstream area of the building array. A budget analysis showed that increasing <em>Da</em>_O3 primarily enlarged the chemical reactions’ effects on NO increase within the urban residential area. Finally, the BPNN model demonstrated a high accuracy in predicting wind speed and NO concentration within a few seconds. These findings provide valuable insights for designing effective strategies to mitigate hazardous pollutants’ impacts in urban environments.</p>