Subspace Identification Algorithm For Prediction Of Silicon Content Researching

With the growing energy shortages and increasingly stringent environmental requirements, the blast furnace steel production system as the largest part of energy consumption, a higher level of automation will bring great economic and social benefits. Most control methods are a mathematical model of the process, however, due to the complexity of the blast furnace process, the mathematical model is difficult to get through the mechanism modeling, so more and more need to get through the process model identification modeling methods. For multiple-input multiple-output identification model, based on subspace system identification methods under control in recent years, industry experts and scholars of great concern, as obtained by subspace identification algorithm for state space model most commonly used in modern control theory, more reveal the internal characteristics of the system, especially for multiple-input multiple-output system, etc., has control of complex industrial applications achieved good results.Automatic control of blast furnace smelting process the core problem is the prediction and control of temperature, therefore, a good prediction model can guide the actual production of an important value, if we can get an effective control model, its significance is even more significant. In this paper, characteristics of blast furnace smelting, the subspace identification algorithm in the blast furnace process modeling research. Details are as follows:â’ˆIn a lot of reading literature on the basis of previous research results are summarized, learning and understanding of blast furnace production process technology, based on iron and on-site expert advice to understand the principle of the blast furnace operation, and obtain the field data needed for identification;â’‰For most of the data collected blast furnace site has deficiencies in value, abnormal values, different magnitude, stochastic noise, multi-dimension and so on, in order to create a more accurate model of data pretreatment. In which including statistical knowledge, also have smooth filtering, on-site iron-making expert consultation get empirical knowledge.â’ŠSystematically introduced and summarized the subspace identification method elementary theory and the algorithm process. Includes the subspace identification specifically the geometry foundation, the spatial projection theory, counts the tool, the system mode space description, the subspace data matrix constitution, as well as subspace parameter identification principle and algorithm. â’‹On the basis of the measured data, select the state with the closely related heat furnace coal injection, air temperature, etc. 8 input variables, output variables as the silicon content, the use of subspace identification algorithm for establishing a 8 input 1 output of silicon content prediction model.Finally, the validity of model is tested by the measured data and the satisfied results are obtained. Which show that the subspace methods have the advantages of simple operation and good robustness, and prove that it is an excellent method for the identification of high order and multivariable complicated system.