Intervals prediction of molten steel temperature in ladle furnace

Temperature prediction is a key factor in the steel-making process control of Ladle Furnace because molten steel temperature can´t be measured continually. To obtain the reliability of the temperature prediction model, a model based on single hidden layer feed-forward networks with extreme learning machine algorithm is applied to establish a model of steel temperature in the steel-making process ladle furnace. And a statistical method is used to construct the prediction intervals based on the simple calculation. The model misspecification variance and data noise variance are considered to obtain accurate prediction intervals. The efficiency of the method is verified by simulation.