Application And Improvement Of Blast Furnace Temperature Prediction Basing On LSSVM

Steel industry plays an important role in the development of the national economy as the lifeblood. Blast furnace is the important part of steel industry, which directly determines the development of the steel industry. The blast furnace temperature is an important indicator in the production. Keeping the temperature at a reasonable level could guarantee the blast furnace smoothly and successfully. At the same time, the reasonable temperature is the ensurance of high yield and low power of blast furnace. The blast furnace system has the characteristics, such as complex nonlinear, large hysteresis and big noise. So the furnace temperature control should be based on the accurate prediction of furnace temperature.The restriction or promotion exists among parameters associated with the blast furnace temperature from the perspective of system modeling. In addition, the change of furnace temperature also has a big lag because of the large hysteresis of blast furnace. The parameters of the blast furnace always might bring errors in the process of collecting, so it makes greater interference to the system analysis. Therefore, the selection and treatment of parameters are the keys of the establishment of furnace temperature prediction model.Since the least squares support vector machines could solve problems of complex nonlinear, small sample and local extremum, the method of least squares support vector machines is used to predict blast furnace temperature. The correlation of parameters(such as feed speed, pressure, PCI) of blast furnace could be analyzed based on the correlation theory by one thousand sets of data from the blast furnace of Baosteel. After that, the parameters of blast furnace temperature and the lag time for each parameter could be determined. According to the analysis, the method of the least squares support vector machines is optimized. The fuzzy c-means clustering(FCM) and support vector domain description(SVDD) are introduced to process the samples separately. These methods achieve better prediction result than the simple method basing on the least squares support vector machines.The software of the blast furnace temperature prediction system is developed based on the completion of the blast furnace temperature prediction model. Furnace temperature prediction system contains the interface and background. The system is mainly to show changes of the relevant data and the results of a variety of furnace temperature forecasting methods. It also obtains other functions, such as printing the report of data, providing the relevant operating recommendation.