The introduction of strongly non-linear chemistry effects in numerical simulation of flows in Selective Non Catalytic Reduction (SNCR) DeNO(x) is addressed. In these systems of large dimensions, NOx in flue gas is reduced by injecting either ammonia or urea as a reducing agent. First, an analysis is performed to seek out with detailed chemistry (207 elementary reactions, 33 species by Klippenstein et al. (2011) [8]) the importance of global parameters in the performance of NO removal, such as temperature and major species concentration levels. Then, this detailed reaction mechanism needs to be simplified for its subsequent introduction in flow simulations. In this paper, two different methods relying on automated optimization tools for reducing the cost of chemistry are discussed. The first one is based on the tabulation of the detailed chemical response from canonical problems, using automatically defined progress variables. In the second one, a large sample set of detailed chemistry solution points is processed by an iterative optimization procedure, leading to a reduced two-step chemistry reproducing the response of the global parameters characterizing NO removal. (C) 2013 Elsevier Ltd. All rights reserved.