Abstract
Silicon content in hot metal ([Si]) is a chief indicator of in-furnace thermal status (IFTS) for the blast furnace. Adjustment of PCI velocity (pulverized coal injection volume per unit time, t/h) is the most common way to control IFTS because it is very quick and flexible. As PCI velocity adjustment need a long continuous accumulation time (CAT) to affect the thermal status and [Si], a long-term prediction model is proposed, and the recommended value of PCI velocity is calculated based on the predicted [Si]. In order to predict the long-term results, the delay times of the input variables are analyzed, and a prediction model based on Auto-regressive eXogenous (ARX) and Akaike Information Criterion (AIC) is established. Principal component regression (PCR) is used to compute future [Si] to reduce dimension and noise filtering. The system of [Si] prediction and PCI recommendation is very helpful for IFTS online control, due to the system design principle is the early prediction and early adjustment. It can reduce the tremendous workload of the operator from online monitoring, and it is especially effective to avoid misoperation. The practical application results show the effectiveness of the system.