Abstract: In order to improve the prediction accuracy of NO_x emission concentration at boiler outlet of thermal power plant, this paper proposes a Stacking ensemble learning model via combined time domain features. First, with the aim of mining the deep information of the data, time series analysis, CEEMDAN and statistical calculation of data standard deviation, skewness and other characteristics are used for combination, and domain feature extraction to construct the reconstructed data. Then, considering the influence of redundant variables in the reconstructed data on the accuracy of the model, the GA is used to reduce the feature dimension of the reconstructed data. Finally, in order to make the most of the advantages of each model to improve the prediction accuracy of the model, the paper constructs a Stacking integrated learning NO_x emission concentration prediction model, utilizing ELM, DNN, MLP, and XGBoost as base models, and ESN as the meta-model. The experimental results show that the prediction model has a good prediction effect under different data sets, and the prediction error is less than 2%, which can accurately predict the NOx emission concentration of the boiler.