Prediction of Silicon Content in Molten Iron Based on EMD-GA-LSTM

Abstract

Stable and accurate prediction of molten iron silicon content in blast furnace is of much concern to production scheduling and stable, safe, and efficient operation of blast furnace ironmaking. To make the most of the information gathered during process of the blast furnace ironmaking and improve the stability and validity of predictions, this paper proposes a new model that combines empirical mode decomposition (EMD), genetic algorithm (GA), and long short-term memory neural network (LSTM) to predict the silicon content in molten iron. First, EMD algorithm is used to decompose the original silicon content sequence; respectively, several IMF decomposition components are obtained. Then, GA algorithm is used to optimize the batch size parameters and neuron parameters of LSTM network, so as to predict each IMF component. Finally, the predicted value of IMF component is reconstructed to obtain the final forecast result. This paper uses the measured data for verification. The results prove that the proposed method can make an accurate prediction of the multivariable time series of molten iron and silicon content and have higher prediction accuracy and less margin of error than autoregressive integrated moving average model (ARIMA), support vector regression model (SVR), and standard LSTM neural network model.