Prediction Of Silicon Content In Hot Metal Of BF Based On Fuzzy Time Series

As the pillar industry of national economy, Steel industry plays an important role in the process of industrialization and urbanization. Iron steel industry process is the upstream of the steel industry, it takes important effect to the quantity and quality of the steel production in the whole process. At the same time, the energy consumption of the process takes two-thirds of the total energy consumption, therefore, the iron-making process is the key link to the energy conservation and emissions reduction in the iron and steel industry. The blast furnace smelting process is a highly complex process, and its operation mechanism often has the characteristics of nonlinear, large delay and big noise, distribution parameters, and so on. In the process of blast furnace ironmaking, to guarantee the blast furnace process keeps smoothly, the effective prediction and control of the temperature is necessary, and we often use the molten iron content of silicon [Si] to represent the blast furnace temperature. Therefore the research on the prediction to the molten iron content of silicon has been praised by the people’s attention.Because of the complexity of the blast furnace, along with the improvement and development of the fuzzy mathematical theory, blast furnace temperature prediction model based on the fuzzy mathematics theory is then set up. This paper mainly use the fuzzy mathematics methods to establish the blast furnace temperature prediction model, with the data collected from No.6BF at Baotou Steel to analysis, inspect and then forecasting.Firstly, the present situation of the blast furnace temperature prediction model were reviewed in this paper, secondly as in this paper the main theoretical basis for modeling, it introduces the basic knowledge of the fuzzy mathematics theory. thirdly, makes statistical analysis to the data to find out the distribution and the data range of different parameters, further doing correlation analysis between various parameters and the dependent variable to choose the appropriate independent variable. At the end of this article is to establish the molten iron blast furnace of silicon content prediction model, considering the complexity of the blast furnace process, it cannot get good result of prediction of silicon content by simple time series model, so this paper first presents the multiple fuzzy time series model. Furthermore, in order to make more fully use of the history information, it uses the weight coefficients calculated by the membership degree to make prediction instead of fuzzifying one data into one single fuzzy set. As the second prediction model proposed in the paper, the fuzzy regression model makes regress to the number of the fuzzy sets and the factual silicon content. In the second model we put the focus on the process of fuzzification, we use the evolution algorithm and evolution strategy to search the best result. Both of the two models considered the other influencing factors and use the fuzziness of the blast furnace data, and they both achieved the hit-rate of86%and84%respectively.