Multi-operation Conditions Prediction Based on Least Square Support Vector Machine for Blast Furnace Gas System in Steel Industry

Abstract

It is very significant to predict the trend of the level of gas-holder in the blast furnace gas (BFG) system for controlling and scheduling the energy system in steel enterprises. A multi-operation conditions prediction approach for the level of gasholder is proposed in this paper, where the correlation coefficients between the difference of energy generation and consumption and the first difference of the level of gas-holder are calculated to cluster training samples. The proposed algorithm involves two phases. At the first phase, the training samples are clustered by calculating the correlation coefficients iteratively. The second phase contains the training of the subsets of the samples using the least square support vector machine (LSSVM). Finally, the predicted values of the test samples are estimated by the sub models after it is classified by the correlation coefficients measurement. For verifying the performance of the proposed method, a series of comparison experiments is carried out using the realistic data from an iron and steel industry plants. The experimental results proves that the proposed method produces higher accuracy than other methods, which can provide an effective plan for balancing the pipe network and dispatching the BFG system.