A Trend Prediction Method Based on Fusion Model and its Application

The change trend of silicon content in molten iron is one of the important indexes for evaluating the condition of blast furnace. An accurate is necessary to realize fine control. However, due to the closed smelting process, high temperature and other harsh environments, which makes the silicon content cannot be detected online in real time, the accurate prediction change trend is more difficult. This paper proposes a fusion model to predict the change trend of silicon content, which integrates extreme gradient boosting (XGboost) with long short-term memory (LSTM) model. Firstly, the XGboost is designed to capture the feature representation of input data, and the trend of silicon content is extracted by regression fitting in sliding window data. Then, the processed data is inputted to the two models in the first stage, and the output results as new features are inputted to the second stage to complete the training of the fusion model. The fusion model is applied to predict the change trend of silicon content in blast furnace molten iron. The validity and feasibility of the proposed model are verified by field data, and the prediction results can provide reliable reference for the blast furnace operator to judge the varying trend of furnace condition and the direction and amplitude of regulation.