Estimation of instantaneous airflow distribution in cubic building group model using multi-time-delay LSE-POD

Abstract

The airflow in a city has strong turbulence characteristics, and strong winds may have a significant influence on pedestrians; hence the rapid calculation of urban airflow distribution can protect pedestrians from danger. Hu et al. (Building and Environment, 2022, Vol.221) successfully applied single-time linear stochastic estimation-proper orthogonal decomposition (single-time LSE-POD, referred to as SLSE) to estimate the instantaneous airflow distribution in an urban model using limited sensors. However, some problems such as low-precision estimation in the wake region have not been solved. Therefore, this study used a method for airflow distribution estimation based on multi-time-delay linear stochastic estimation and proper orthogonal decomposition (multi-time-delay LSE-POD, referred to as MLSE). This study used a validated dataset simulated by large-eddy simulation (LES) in an urban canopy model. The optimal delay time was calculated for the MLSE, and the estimated airflow distribution with the MLSE was compared to the LES data and SLSE. The following conclusions were made: 1) For the velocity field error, the spatially averaged values of MLSE are lower than those of SLSE in both reconstruction and prediction; 2) The Reynolds stresses of the velocity in the MLSE are more accurate than those in the SLSE; 3) The shapes of the probability density functions of velocity for MLSE at representative points are closer to the LES results than to the SLSE results; 4) The time interval to the optimal delay time does not vary with the sampling time step of the LES. This is approximately equal to one eddy cycle time.