NOx Concentration Prediction for Cement Rotary Kiln Under Multiworking Conditions Based on an Echo State Network With Modular Output

Cement industry is the third-largest source of nitrogen oxides (NOx) emissions. Accurately predicting the concentration of NOx in cement kilns is crucial for reducing pollutant emissions through optimizing the calcination process of cement clinker. However, due to the complex physical and chemical reactions and variable working conditions inside rotary kilns, it is difficult to predict the NOx concentration accurately. To address this problem, this paper develops a prediction model based on a novel echo state network with modular output (MOESN) for NOx concentration prediction in cement rotary kilns under different working conditions. First, k-means++ clustering method is used for working condition recognition according to the calcination process data. Then, a novel ESN with modular outputs is proposed to predict the NOx concentration in the cement rotary kiln under different working conditions. Meanwhile, the hyperparameters of the proposed model are optimized using Particle Swarm Optimization (PSO) algorithm to determine the best network configuration to different working conditions. Finally, the effectiveness of the prediction model is evaluated using a dataset from an actual cement clinker calcination process. Experimental results show that the proposed model achieves higher prediction accuracy compared to several widely recognized NOx prediction models.