Prediction NOx emission from sintering plant with a radial basis function and back propagation hybrid neural network

NOx emission in iron and steel enterprises mainly comes from sintering flue gas. Traditional artificial neural networks cannot meet the requirements of computational accuracy and convergence speed. In order to reduce and predict NOx emission, back propagation neural network model and radial basis function neural network model were established to predict the NOx in sintering flue gas. On this basis, combined with the characteristics of strong adaptive learning ability of back propagation model and rapid convergence of radial basis function model, the method of optimizing model structure and setting up connection layer were used, a new hybrid neural network model was constructed for NOx prediction research, and the prediction results of the three models were compared and analyzed. The results showed that the hybrid neural network based on connection layer structure optimization had the best evaluation indexes among the three models, which root mean square error and mean absolute error reached 11.37 and 7.14 mg/m3, respectively. The maximum and minimum absolute errors reduced to 35.47 and 0.0083 mg/m3, respectively. The hybrid neural network model had better stability of prediction data, higher fitting degree of results and convergence speed was the fastest. Predicting NOx by using mixed model can effectively eliminate the feedback delay caused by NOx and provide guidance for the control of sintering site process.