Event-Triggered Prediction of Blast Furnace Gas Generation Based on a Trend Self-Adaption Scheme

The prediction of blast furnace gas (BFG) generation in ironmaking processes is crucial for the BFG scheduling work. Due to frequent switching of working conditions of the ironmaking process, the generation of BFG under each working condition fluctuates greatly. It is difficult to predict accurately during the transition of working conditions. In order to address these problems, this paper proposes a data-driven BFG generation prediction method based on a combination of an event-triggered scheme and a trend self-adaption scheme. In this method, different working conditions are divided by events such as the change of supply of hot air and oxygen. Considering the process characteristics of air reduction and restoration, the decrease and increase trends are predicted by a parameter adaption-based linear logarithmic regression. The model parameters are corrected online to realize trend self-discrimination. Besides, as for the residual sequence after trend fitting, it is modeled by using the long short-term memory (LSTM) network. To validate the effectiveness of the proposed method, actual industrial field data of a steel plant in China are utilized. Simulation experimental results show that the proposed method significantly improves prediction accuracy of the BFG generation flow.