A Novel Exhaust Gas Temperature Prediction Method of Hot Blast Stove

Hot blast Stoves (HBSs) is a piece of important equipment to provide hot air for the blast furnace. In order to control its operation process and save resources, it is necessary to predict its exhaust gas temperature. Considering a large amount of production data in this system, a novel prediction method based on echo state network (ESN), correlation analysis and differential evolution (DE) algorithm are presented in the paper. ESN is a special kind of recurrent neural network. Due to its simple training and fast convergence, the ESN model is widely used in time series prediction of a nonlinear system. To improve the availability of samples and the accuracy of results, we analyze the correlation between different features of samples for time hysteresis search before prediction. Besides, the differential evolution algorithm is used to search optimal solutions for crucial parameters of the echo state network. Simulation results show that the proposed method can effectively predict the exhaust gas temperature in a short time.