State Prediction Of Slagging On Coal-fired Boilers Based On Clustering Algorithm And Support Vector Machine

Currently, the domestic coal market was less than demand, the burning coal of power plant boiler was always changing and getting worse. The slagging problems seriously affected the safety and economy of boiler operation. Slagging of coal-fired boiler was a complex physical and chemical process, was not only related with the property of coal itself, also with the size of furnace structure, the temperature field and the aerodynamic field. The prediction of boiler slagging timely and accurate was very important.This paper introduced the support vector machine (SVM) and clustering algorithm, clustering algorithm used in this paper were fuzzy C means clustering (FCM) and grey clustering algorithm (GCA).This article introduced thirty nine samples, first twenty nine as training samples, last ten as testing samples. The input vectors were softening temperature (tST), alkali-acidratio(B/A), SiO2-Al2O3ratio(w(SiO2)/w(Al2O3)), percentage of silicon content(G), the dimensionless average temperature furnace((pt) and the dimensionless inscribed circle diameter furnace(φd), and the output vectors were slagging degree. In this model, the complex training sample set was viewed as a mixture of multiple populations, and each population built a single SVM model. The prediction problem of the large data set was viewed as a multiple model prediction problem. Support vector machine algorithm retained the sample point near optimal separating hyper planed and removed the distant sample points, which can reduce the training sample sets fitness and improve the prediction accuracy. The feasibility of this models proved by the result of predicting the state of slag on the ten coal-fired boilers, and FCM-SVM prediction accuracy rate is100%, and GCA-SVM prediction accuracy rate is90%. The SVM forecasting system based on clustering algorithm data preprocess can accurately predict the state of slag on coal-fired boilers.