Data-based Modeling And Scheduling For Blast Furnace Gas System

Steel industry involves high energy consumption and emission, and its related energy resource conservation and pollution reduction are the key issues of national industrial development. As an important byproduct energy, studying the scheduling problems of the blast furnace gas (BFG) could help to optimize the energy assignment and energy management for the steel industry. It has important significance for realizing industrial energy conservation and emissions reduction. This dissertation studies the data-based modeling and scheduling methods for the BFG system, considering the fact that it is rather difficult to establish the system model by using the process mechanisms because of their phsical complexity. The specific contents are as follows.As for the data missing problem exhibited in industrial time series, an imputation method based on manufacturing procedure characteristics is designed here, which takes into account the data correlation under same operating conditions in production practice. Considering that the production rhythm is changeable, a correlation analysis method for non-equal-length granules is proposed in this dissertation to construct the correlation of unfixed-cycle time series data. For further searching the relationships between sample sequence and target one, an estimation of distribution algorithm (EDA) based model is presented by transforming the calculation of correlation into the evolution of probability matrix in the solution space.In view of the BFG system modeling, given the fluctuation characteristic of the BFG generation and high level noises and outliers mixed in the original industrial data, a quantile regression-based echo state network ensemble (QR-ESNE) is modeled to describe the prediction intervals (Pis) of the BFG generation. In the process of network training, a linear regression model of the output matrix is reported by the quantile regression to improve the generalization ability. Then, in view of the practical demands on reliability and further improving prediction accuracy, a bootstrap strategy based on the QR-ESN units is designed for confidence intervals and prediction ones via combining with the regression models of various quantiles. When establishing a fuzzy T-S model for the gas tank level, the high level noises in industrial data and the disturbances in training samples could lead to overfitting phenomenon. A fuzzy subset fusion combined with a rule reduction method is proposed in this dissertation to simplify the structure of the rule base and enhance the generalization ability of the fuzzy model. The proposed methods could help to conquer the overfitting problem and exhibits strong robustness and generalization when modeling the industrial data with high level noises and outliers.When solving the scheduling problem of BFG byproduct energy, based on the interval estimation model of BFG generation flow and the fuzzy model of BFG tank, a hybrid method which combines community finding (CF) of a complex network with case-based reasoning (CBR) is reported here. The proposed CF-CBR method extractes typical scheduling cases from the constructed communities by using the CF method with a new evaluation criterion for the vertex combination degree. In the reasoning process, the matched cases are retrieved according to the similarity, and the corresponding credibility of each case is calculated by using the similarity and the community distribution, which could provide solid basis for the scheduling work.To verify the performance of the proposed data imputation, modeling and scheduling methods, the practical data coming from the BFG system of a steel plant in China are employed. The experiment results indicate that the proposed methods exhibit high accuracy and reliability for the industrial data. Furthermore, based on the proposed methods, an application software system for the BFG scheduling and management is developed and applied to the practice of this plant. The practical operation shows that, the work of this dissertation could provide important guidance for the scheduling work, and has practical significance for the energy conservation and emissions reduction.