Open AccessData‐driven recursive subspace identification based online modelling for prediction and control of molten iron quality in blast furnace ironmaking
Author(s) -
Zhou Ping,
Dai Peng,
Song Heda,
Chai Tianyou
Publication year - 2017
Publication title -
iet control theory and applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.059
H-Index - 108
eISSN - 1751-8652
pISSN - 1751-8644
DOI - 10.1049/iet-cta.2016.1474
Subject(s) - subspace topology , model predictive control , computer science , controller (irrigation) , identification (biology) , blast furnace , process (computing) , data mining , predictive modelling , control theory (sociology) , control engineering , artificial intelligence , control (management) , engineering , machine learning , chemistry , organic chemistry , agronomy , operating system , botany , biology
In blast furnace ironmaking operation, the molten iron temperature and the silicon content ([Si]) are two key molten iron quality (MIQ) indices. The measurement, modelling and control of these indices have always been of importance in metallurgic engineering and automation. In this study, data‐driven methods for online prediction and control of multivariate MIQ indices are proposed by integrating hybrid modelling and control techniques together. First, a data‐driven hybrid method that combines canonical correlation analysis and correlation analysis is proposed to identify the most influential controllable variables as the modelling inputs from multitudinous factors. Then a data‐driven online model for prediction of MIQ is established by recursive subspace identification (R‐SI) with forgetting factor. Unlike the conventional SI modelling, the proposed R‐SI‐based online modelling only identifies subspace matrices for a data‐driven input–output model without explicitly estimating the system matrices, which can reduce the computation complexity. Finally, a predictive controller is designed to maintain the MIQ indices at an expected level by using the developed MIQ prediction model as an online predictor. Since the parameters of the predictor are updated adaptively by the latest process data, the predictive controller can produce more reliable and stable control performance. Experiments using industrial data have verified the superiority and practicability of the proposed methods.