Research on lime rotary kiln temperature prediction by multi-model fusion neural network based on dynamic time delay analysis

Abstract

The lime rotary kiln systems are widely used in the metallurgical industry, where the combustion state is exceptionally complex, and it is difficult to predict and control the calcined zone's temperature. The lime rotary kiln system uses the entropy and grey correlation model, combining the lime rotary kiln operation process to determine the input and output characteristics of the model. Then, it analyzes the time lag and inertia in the lime rotary kiln combustion system to compensate for the temperature prediction in the lime rotary kiln by using the CNN-BILSTM-OC model. Correcting the expected output results with the actual situation. The experimental analysis shows that the proposed model has a higher prediction accuracy than others. The maximum relative error calculated for the future temperature prediction is 0.2098%, while the generalized average of the root mean square error of the model under different working conditions is 0.9639. The generalized average of the mean absolute error is 0.6683, which shows that the model has a strong generalization ability to meet practical applications.