Data reduction processing method and device for prediction model of silicon content in blast furnace molten iron

Abstract

The invention discloses a data reduction processing method and device for a blast furnace molten iron silicon content prediction model. According to the method, process characteristics, data characteristics and practical experience of workers of blast furnace molten iron smelting are introduced and converted into property priority sequences, a necessary set conflict detection mechanism and a recursive method are utilized, and therefore it is guaranteed that solved reduction is unique and is most matched with an application. The set necessary set BS guarantees non-redundancy of properties in a reduction set R and is used for judging a core property conflict; a recursive mode is adopted to solve the most matching reduction, and by use of site protection of recursion, the simplest mode is directly used to return to the last state one by one when the core property conflict is encountered without the need for performing state recovery; and the whole processing process is easy to operate, the calculation result is accurate, and the method has high popularization value.

Description

technical field [0001] The invention relates to a data reduction processing method and device for a blast furnace molten iron silicon content prediction model. Background technique [0002] Prediction of silicon content in blast furnace hot metal is a key step in the optimal control of blast furnace smelting. The forecast results have a decisive influence on the adjustment of operating parameters, and are a necessary way to improve the quality, output, resource and energy utilization of molten iron. Due to the complexity of the process mechanism, the prediction method based on data modeling has become the mainstream of silicon content prediction. However, due to the different degrees of coupling between the input parameters, the existing model input parameter sets based on correlation solutions are difficult to achieve The best match of the prediction model, which will affect the prediction hit rate. Therefore, it is necessary to adopt a reasonable method to calculate the ...

AI Technical Summary

Problems solved by technology

[0004] However, due to the limitations of the definition of reduction, there are multiple reduction results in an original data set, and traditional reduction algorithms often only randomly calculate one result

Images