Combustion optimization is an effective and economical approach for reducing nitrogen oxide (NO_X) emissions from coal-fired utility boilers. To implement an online reduction in NO_X, a precise and rapid NO_X emissions model is required. This study establishes an efficient NO_X emission model based on the principle component analysis (PCA) and support vector regression (SVR). Modeling performance comparisons were also conducted using a traditional artificial neural network (ANN) and SVR. A considerable amount of worthwhile real data was acquired from a 1000-MW coal-fired power plant to train and validate the PCA-SVR model, as well as the traditional ANN and SVR models. The predictive accuracy of the PCA-SVR model is considerably greater than that of the ANN and SVR models. The time consumed in the establishment of the PCA-SVR model is also shorter compared with that of the other two models. The proposed PCA-SVR model may be a better choice for the online or real-time modeling of NO_X emissions in achieving a reduction of NO_X emissions from coal-fired power plants.