Prediction Model of SO 2 Distribution in Boiler Based on Deep Neural Network

The prediction of SO 2 distribution in the boiler is an important basis for accurate control of the combustion process. In order to better grasp the distribution of SO 2 inside the boiler, and at the same time solve the problems of long time-consuming CFD numerical simulation process and narrow coverage of typical working conditions, a deep neural network (DNN)-based SO 2 prediction model in the furnace is proposed. The model uses the Lasso method to select the characteristics of SO 2 related variables, and on this basis, selects the DNN algorithm for predictive modeling. Experimental results show that DNN has higher prediction accuracy compared with common data modeling methods, and the average absolute error is reduced by about 36.66% and 76.38%.