Integration of fuzzy information granulation and support vector machine for prediction alumina concentration

Abstract

There is often a lot of redundant information in observed values of alumina concentration to result in large computation and affect the predictive validity. A prediction method based on fuzzy information granulation and support vector machine (FIG-SVM) for alumina concentration is proposed to solve the problem that prediction model can not be established accurately while there were strong correlations in many factors of aluminum reduction cells. In the proposed approach, Theory of fuzzy information granulation is used to granulate time-series data of alumina cell. Granulated data can not only reflect the characteristics of original but also reduce redundant information. Support vector machine can be used to forecast short-term alumina concentration. By using real data of 170KA operating aluminum cell from a factory, the method in which the computation time is reduced effectively can surely accuracy of parameter estimation.