CFD-based turbulent reactive flow simulations of power plant plumes

•Capabilities of CFD in modeling chemically reactive power plant plumes are examined.•Jet-dominated region (JDR) and ambient-dominated region (ADR) are defined.•RANS with velocity inlet (VI) and volume source (VS) and LES are compared.•RANS-VI is most appropriate for power plant plume simulations in the JDR.•RANS-VS is most appropriate for power plant plume simulations in the ADR.

This paper examined the capabilities of computational fluid dynamics (CFD) techniques in modeling the transport and chemical transformation of power plant plumes. Based on turbulence characteristics, we divided the plume evolution into two stages. The first stage is referred to as the jet-dominated region (JDR), characterized by a high momentum jet flow of flue gas. The second stage is referred to as the ambient-dominated region (ADR), driven by atmospheric boundary layer turbulence. Then, we compared the three methods in simulating plume transport in the JDR, i.e., Reynolds-averaged Navier-Stokes (RANS) model with velocity inlet (RANS-VI), RANS with volume source (RANS-VS) and Large-Eddy Simulation (LES). The VI method treats the stack exit as a surface inlet to the simulation domain, while the VS method defines a volume region containing the source with a specific emission rate. Our evaluation against a relevant wind tunnel experiment suggested that RANS-VI is most appropriate for power plant plume transport in the JDR. LES can achieve more accurate results, but the improvement in accuracy over RANS-VI may not justify its high computational costs. Nevertheless, LES is still preferable for JDR simulations if computational costs are not a constraint. The VS method requires refined mesh in the source region in order to achieve accurate results, making it no different from the VI method in the JDR. Next, for our ADR evaluation, we simulated plume chemical evolution in a well-characterized 1999 TVA Cumberland aircraft plume transect field study. RANS-VS was adopted, as proper RANS-VI and LES simulations would be exceedingly expensive in terms of computational costs. The overall model performance was satisfactory, evinced by the predicted concentrations of SO2, O3, NOx as well as NO2/NOx ratios fell within the variations in the observed values for large portions of the plume distributions. An indirect JDR evaluation by comparing the predicted plume centerline trajectory with that estimated using a semi-empirical equation and Cumberland-specific parameters indicated that RANS-VS can reasonably predict plume evolution in the JDR as well. Our study suggested that properly configured CFD simulations (e.g., turbulence model, source representation and mesh sensitivity) were able to capture the evolution of chemical reactive plumes from power plants in high accuracy, however, with high computational cost and thus limited applicable spatial range.