A neural network model for quantitative prediction of oxide scale in water walls of an operating Indian coal fired boiler

Application of artificial neural network (ANN) for predicting industrial process behavior has been increasingly popular in the power sector industry. In this paper, a multi-layer perceptron (MLP) based ANN model has been developed to predict the deposition rate of oxide scale on
waterwall tubes of a coal fired boiler. The input parameters in the ANN model are boiler water chemistry and relevant operating parameters, namely, pH, alkalinity, total dissolved solids, specific conductivity, iron and dissolved oxygen concentration of the feed water and heat flux of a
typical 250 MW coal fired boiler. The neural network architecture has been optimized using an efficient gradient based network optimization algorithm to minimize the training and testing errors rapidly during simulation runs. The parametric sensitivity of heat flux, iron content, pH and the concentrations of total dissolved solids in feed water and other operating variables on the scale deposition behavior has also been investigated. It has been observed that heat flux, iron content and pH of the feed water have a relatively predominant influence on the oxide scale deposition phenomenon. There has been very good agreement between ANN model predictions and the measured values of oxide scale deposition rate substantiated by the regression fit between these values. @2014 Trade Science Inc. - INDIA