Semi–Empirical Correlation of NOx Emissions From Combustion Turbines With Water/Steam Injection

Abstract

Inert (water or steam) injection is commonly used to reduce NOx emissions in stationary gas turbine combustors, both lean premixed when oil–fired and conventional. Thus, having an accurate phenomenological model to predict these reductions could be useful in both design and implementation for low emissions. In this work, the semi–empirical characteristic time model (CTM), which has been validated for thermal NOx emissions from conventional, diffusion flame combustors, is modified to account for inert injection effects. Measured NOx data from two heavy–duty, utility combustion turbines operating on natural gas and fuel oil #2, both dry and with water or steam injection, are correlated. Inert injection is modeled as thermal, and two limiting cases are proposed which successfully bound the measured data. An empirically selected effective inert injection flame temperature was substituted for the stoichiometric flame temperature used to estimate the thermal NO formation rate in the CTM. This procedure correlated all of the measured data from both combustors for both fuels with a standard deviation of 1.02 g NO2/kg fuel. The high standard deviation results from systematic trends in the dry data for one combustor which propagate through the lower NOx values of the inert injection data. Removing these trends empirically improves the combined correlation to a standard deviation of 0.28 g/kg (approximately 3.2 ppmvd at 15% O2).