Hybrid Model for NOx Prediction in Cement Clinker Calcination Process

Accurate measurement of nitrogen oxide (NOx) concentration from cement clinker calcination process is of great importance for atmospheric environmental protection. However, the NOx concentration is difficult to measure by hardware sensors due to the harsh environment of the calcination process. In this paper, a hybrid model is developed for N Ox concentration prediction in the cement clinker calcination process. First, an improved attention mechanism is introduced into the hybrid model to improve the quality of the input signal. Then, an echo state network (ESN) is employed to develop a nonlinear model for NOx concentration prediction. Meanwhile, a regularization constraint is used to eliminate redundant information, which improves the generalization ability of ESN. The proposed hybrid model is applied to predict the NOx concentration from a real cement clinker calcination process. The experimental results show that the hybrid model is effective for NOx concentration prediction.