Research On Interval Prediction Method Of Sulfur Content Of LF

Refining desulfurization is an important part of the deep desulfurization of molten steel.LF is the main equipment for refining desulfurization.Establishing the sulftur content in molten steel prediction models can make up for the lack of the samples sulfur content testing and provide more technical support for the refining desulfurization control.But todays forecasting models can only provide a forecast value with the deviation of the actual value.When the deviations between the predicted and actual sulfur content are large,those models being applied in industrial production sites may causing negative impacts on LF refining desulfurization control.Just because of this,in this paper,a new forecasting method which can provide the trustiness of the prediction results called interval prediction was proposed.The interval prediction can offer not only the prediction of molten steel sulfur content but also the confidence level associated with the point prediction as expected in modern industry.The interval prediction model can guarantee that the true values are contained in the prediction intervals(PIs)with a certain probability.It can provide more information for the refining desulfurization control,and avoid the control risk caused by large prediction error.In this paper,based on the in-depth understanding of the refining process,the principle of desulfurization,the uncertainty analysis of traditional LF molten steel sulfur content prediction model and the statistical knowledge,the LF molten steel sulfur content interval prediction model is established.The main work of this paper is summarized as follows:(1)In order to get the uncertainty of the sulfur prediction models,the Following works had been done.First of all,through the study of the desulfurization mechanism,two important formulas which were desulfurization thermodynamic equation and desulfurization kinetic equation were obtained to describe the sulfur content of the molten steel.Then the sulfur prediction model was established by those two formulas.Finally,through the uncertainty analysis of the prediction model,a conclusion that the overall uncertainty is summed up by the model uncertainty and the data uncertainty was got.(2)In order to construct the prediction interval based on the uncertainty of prediction model,the Following works had been done.Model uncertainty and data uncertainty unbiased estimation of the results obtained,then the prediction intervals were constructed by the calculation results.Using field data to verify the effects of interval forecasting and analyzing the interval widths which are obtained from the unbiased estimation which is not changed with the sample.This methodology cannot be used in predicting the content of Sulfur in refining desulfurization process.(3)In order to obtain the forecasting interval of sulfur content which satisfies the desulfurization process,based on the resampling method to get the uncertainty of the data.Then using the uncertainty of the network fitting model to get the prediction interval.The prediction interval can make up for the result of the previous method in some degree,however,there would still be some abnormal values.This methodology cannot also be used in predicting the content of Sulfur in refining desulfurization process.(4)Although the previous methods have their respective defects leading to the prediction of sulfur content not suitable for refining desulfurization,through the analysis found that the shortcomings of the two can make up for each other.So a new method called interval fusion method which fusing advantages of these two interval prediction methods.In order to be able to combine the advantages of those two methods,the new prediction interval is settled by weighted sum of overall uncertainty of these two methods.The weight represented by a multi input single output network is a variable that is related to the input.The objective function was constructed based on the idea that in the premise of ensuring coverage,the smaller the interval,the better.Through the verification,analysis and comparison,this method combines the advantages of the two methods,make up for the deficiencies of the previous,suitable for the sulfur content of refined desulfurization...