Online prediction of hot metal temperature using transient model and moving horizon estimation

Abstract

Precise control of hot metal temperature (HMT) is crucial for achieving stable operation of a blast furnace, but it is difficult due to the sluggish dynamics caused by the huge heat capacity. To cope with such difficulty, this work aims at developing a method that can predict future HMT by adopting moving horizon estimation (MHE) based on a one-dimensional transient model. MHE is useful to successively adjust model parameters so that the undesirable influence of past disturbances on the prediction is minimized. The real application result demonstrated that the root mean square error (RMSE) of HMT of eight-hour-ahead prediction was only 11.6°C. The high-performance prediction enables operators to realize the efficient operation of the blast furnace.