This paper investigates the construction of a simulation model for the emission of nitrogen oxide (NO/sub x/) from a coal-fired power generation plant. A polynomial NARX model has been identified using a novel identification algorithm which includes a model selection procedure based on simulation error minimization, a model reduction phase and filtering to improve the accuracy in a desired bandwidth. The method improves model robustness and yields a model simple to implement and with a relatively high accuracy. Finally, this model can be integrated in a advanced real-time control system to reduce NO/sub x/ emissions.