Interval Prediction Method For Endpoint Temperature Of Liquid Steel In LF

The temperature prediction can provide a reference for endpoint control of temperature in Ladle Furnace(LF).The information of uncertainty isn’t included in point prediction results.The interval prediction can provide the range of the actual temperature of liquid steel,and the confidence that the actual value of temperature are covered in the intervals,lowering the risk caused by prediction results.Therefore,interval prediction method for endpoint temperature of liquid steel in LF is researched in this paper.The main work in this thesis can be summarized as follows:Firstly,the uncertainty in prediction results is analyzed,and the prediction intervals are constructed based on uncertainty.According to the current interval prediction methods,the total uncertainty caused by the model and the data in prediction results is qualified.Afterwards,the prediction intervals of the endpoint temperature of liquid steel in LF are constructed based on the total uncertainty.Secondly,the directions of improvement are proposed according to the problems in interval prediction based on uncertainty.From two evaluation interval quality indicators,the interval coverage and the mean prediction interval width,the prediction intervals are not ideal which are calculated based on the uncertainty.There are two reasons for this result.On one hand,the interval prediction method based on uncertainty is assumed that the prediction results have a t distribution.However,the temperature of liquid steel in LF refining process are influenced by multiple factors.It is difficult to determine the exact relationship between the temperature of liquid steel and the factors.Therefore,it is difficult to prove whether the hypothesis is reasonable.On the other hand,the prediction intervals are constructed based on "the point prediction model and analyzing the uncertainty",nor aim to improving the quality of prediction intervals.According to the two reasons,the directions of improvement are proposed.First,avoiding the assumption of prediction result distribution.Second,focusing on improving the quality of prediction intervals.Thirdly,the structure of prediction model for the liquid steel temperature in LF is restructured according to the direction of improvement.The optimization objective function under the novel structure is determined.And PSO algorithm is used to determine the parameters of the model.The parameters in PSO is determined by Uniform Design test.Eventually,the better performance of the interval prediction model for liquid steel temperature in LF is verified by simulation results.