Abstract: To address issues of unstable working conditions within waste incinerator selective non-catalytic reduction (SNCR) denitrification system, including many factors affecting outlet NO_x concentration and the inability to timely and accurately measure outlet NO_x concentration, a dynamic soft measurement model for NO_x concentration at the outlet of SNCR denitrification system was proposed based on variable selection and the pelican optimization algorithm-nonlinear autoregressive(POA-NARX). The factors affecting NO_x concentration at the outlet of SNCR denitrification system were firstly analyzed by mechanism and the characteristic variables were selected. An improved fast correlation-based filter (FCBF) algorithm was then used to select highly correlated variables and remove redundant ones. Time delay estimation was carried out by using data trend analysis method and mutual information algorithm. Finally, pelican optimization algorithm was used to determine the optimal order of the system variables and establish a dynamic soft measurement model of NO_x concentration at the outlet of SNCR denitrification system. The experimental results show that the accuracy of the NARX dynamic model after variable filtering and time lag analysis is significantly improved, and the prediction effect of the POA-NARX model is significantly better than other soft measurement models.