Application of K-NN regression for predicting coal mill related variables

Modern coal fired power plants are required to handle a variety of coal types and accommodate large load changes. Coal mills grind the coal to required fineness and primary air dries and supplies the pulverized fuel to the burners. The dynamic response of coal mills is poor due to simple controls and various faults occurring inside the milling system. In this paper, an approach for time series prediction of n-step ahead values of important variables associated with the milling system is provided. A simple, data driven, non parametric technique i.e. k-NN regression is used for the prediction. The prediction of mill variables is helpful for improving controls and optimizing the mill operation. The proposed approach is applied for 5 minute ahead prediction and validated using the actual data obtained from a coal fired power plant in Gujarat, India.