关键词: 高炉, 铁水温度, 支持向量回归机, K-means 聚类

Abstract:

As a key parameter in blast furnace smelting process，the temperature of molten iron is of importance for smooth operation of
blast furnace and the energy consumption． This paper studies on the important indicator for heat state of the blast furnace，namely molten
iron temperature． By taking advantages of both method of K-means clustering and support vector machine( SVM) ，a K-means clustering
－ based SVM model is proposed for predicting the temperature of molten iron． Firstly，the training sample data are divided into m
classes and m SVM regression prediction models are established accordingly． At the same time，a particle swarm optimization algorithm
is utilized to optimize the model parameters． Then，m discriminant functions are established to recognize which class the sample data
belongs to． Finally，the sample data are put into the corresponding class of regression model to predict temperature． Compared to the
standard SVM － based prediction method，the proposed method predict the molten iron temperature with a higher accuracy．

Key words: blast furnace, hot metal temperature, support vector machine, K-means clustering