Coke Production Process, The Quality Of Coke And Coking Energy Consumption Prediction Model

The coking-blending is a complex industrial process.The key to control the coking plant process is to get the quality of coke and heat consumption timely. In order to measure the quality of coke and heat consumption, a quality prediction model of coke and an intelligent integrated prediction model of heat consumption is built, based on the intensive analysis of coking process technology.As the quality of coke is difficult to measure timely in coking plant production process, a prediction model based on principal component analysis (PCA) and radial basis function (RBF) neural networks is built First, effecting factors on the quality of coke is determined by the analysis of mechanism, including blended coal and the parameters of coking plant production process.Then,the PCA method is adopted to eliminate the linear correlation of blended coal and to reduce the input data of RBF neural networks.Finally, the k-means clustering method is applied to determine the hidden layer parameters of RBF neural networks, and the parameters of the linear output layer are determined by means of Least Squares Algorithm.Meantime, an integrated model combining principal component regression and supervised distributed compound neural networks based on the features of coke oven is proposed. On the basis of the effecting factors of heat consumption in coking plant process, a principal component regression model of heat consumption is built. To increase the generalization ability, after supervised clustering the samples,compound neural networks is built to describe every group based on the supervised clustering, then the prediction value of the heat consumption is received through the fuzzy compose.Meantime, the heat consumption is got through the iterative algorithm of entropy of information which is used to integrate the outputs of principal component regression and supervised distributed compound neural networks.The prediction results of coke quality and heat consumption indicate that the two models have an advantage of high prediction ability. The application of the two models will lay foundation for the control of coking plant process.