Modelling of CO and NO emissions in MILD combustion systems

Abstract

Recently large attention is paid to Moderate and Intense Low Oxygen Dilution (MILD) combustion [1], also denoted as flameless combustion, as it allows limiting pollutants formation, in particular that of CO and NOx. MILD combustion is generally achieved through a massive recirculation of exhaust gases in the reaction region, which ensures dilution of reactants before reaction can occur. In this manner oxidation takes place across a relatively large volume rather than restricted to a flame front, thus reducing temperature peaks and NO formation via thermal mechanism. A requirement to obtain a stable combustion is that reactants are heated up above their self-ignition temperature. MILD combustion has been exploited for industrial processes that require a high and homogeneous temperature distribution within the combustion chamber, as in the glass and ceramic industry, as well as in steel thermal treatments. Moreover MILD combustion is receiving increasing interest because it allows fuel flexibility, it being suited also for low-calorific value fuels, highcalorific industrial wastes and industrial byproducts (.e.g coke oven gas) [2]. The modeling of MILD combustion systems generally requires an accurate description of the gas-phase oxidation. However to what extent a kinetic scheme can be reduced is under debate. To this purpose the comparison of predicted and measured CO emissions is crucial. The present paper summarizes and discusses the modeling of CO and NO emissions in terms of kinetic schemes and pollutant formation routes, comparing predictions to measurements taken in three different MILD combustion devices: a self-recuperative burner, a lab-scale burner and a micro-CHP system. All systems are fed with methane and methane/hydrogen mixtures, but with different H2 concentrations and air excess. In this manner a wide range of operating conditions and burner geometries is covered. The level of confidence and the error budget associated to the computations, to better discriminate the limits and potential of the proposed CFD model are evaluated following the Verification and Validation methodology [3].