Prediction of coke CSR using time series model in Coke Plant

Abstract

Coke is raw material for blast furnace for production of hot metal. Good quality raw material produces low cost hot metal and one of the important quality parameter for blast furnace is Coke Strength after Reaction (CSR), as it refers to coke “hot” strength, generally a quality reference in a simulated reaction condition in an industrial blast furnace. In this research, the effects of coal properties and process parameters on the Coke CSR were studied by Time Series forecasting and artificial neural network models. In this method, historical data of last 3 years was used to estimate the CSR value. In this investigation, thirty-four input parameters such as moisture, volatile matter, ash, fluidity, battery temperature etc. were used. An Unobserved Component Model of Time Series was found to be optimum with nine parameters of coal properties and other process parameters with maximum accuracy of 76%, on the validation dataset respectively. The operating range of coal properties and controllable process parameters is derived from model developed to operate for consistent Coke CSR of 65.5% and above. The potential saving from this modelling initiative is INR 120 Million/USD 1.75 Million.