Study On The Prediction Method For BFG Generation In Steel Mill

Gas consumption and balance in steel industry are the major issues related to production efficiency of enterprise and environmental pollution. One of the important by-product generated in the steel smelting is the blast furnace gas which is an important secondary energy, while it is also the main sources of pollution of the atmospheric with the characteristic of low heating value, big fluctuations in production and consumption with the state of process. Sudden gas excess or shortage is prone to cause the shutdown of equipment and user, and poor adjustment on the gas scheduling and balancing make enterprise release it into the atmosphere. This would result in environmental pollution, so predicting the output and consumption of BFG generation accurately, making scientific and reasonable rebalance, improving energy efficiency, and protecting the environment have become a significant research and development project.This paper studies echo state network, multiple regression analysis, Gaussian process regression and other methods aimed at the problem of BFG generation’s long-term forecast. At first, the blast furnace process will be introduced, blast furnace’s utilization value and significance for forecasting was expounded later, and ESN’s principle and characteristic was described in details, then Matlab was used as a simulation tool to model and forecast BFG generation. According to the practical experience, we determine the parameters and analyze the problem of traditional ESN. For existing problem, ESN improved finally achieved a stable long-term forecast for BFG generation. Before prediction, three data were studied with an improved model of grey T’s correlation degree, the greater influence factors will be as inputs of prediction model. And threshold de-noising method based on empirical mode decomposition was used for BFG generation. At last, the method of Gaussian process regression and multiple regression analysis for prediction were introduced briefly and used to forecast for BFG generation. All the methods mentioned above were compared, so the advantages of the modified ESN can be confirmed.