Dynamic NOx emission prediction for a 660 MW super‐critical coal‐fired boiler using integrated time series network with adaptive filtering and deep learning

Accurate prediction of NOx emissions is crucial for optimizing combustion control and improving environmental compliance in coal‐fired power plants. This study proposes an integrated time series network with adaptive filtering and deep learning (ITSN‐AFDL) to realize NOx emission forecasting. The model incorporates interactive convolutional feature extraction, adaptive spectral analysis, and long‐term temporal modelling. It can effectively capture both short‐term fluctuations and long‐term trends. Comparative experiments are conducted against traditional machine learning and deep learning models. The results demonstrate that ITSN‐AFDL achieves the highest predictive accuracy and generalization capability, with R 2 = 0.9832, RMSE = 3.3759 mg/m 3 , MAE = 2.5188 mg/m 3 , MSE = 11.3971, and MAPE = 0.0078%. These findings confirm that the proposed framework significantly outperforms existing methods, offering a robust, high‐precision solution for real‐time NOx emission prediction.