Abstract: A model of support vector machine (SVM) forecasting system was built up for state prediction of slagging on coal-fired boilers based on fuzzy C-means clustering (FCM) data preprocessing, which takes the softening temperature, alkali-acid ratio, SiO2-Al2O3 ratio, percentage of silicon content, dimensionless average furnace temperature and dimensionless diameter of actual tangential flow circle as input variables, and the slagging degree as output variable. Slagging characteristics of 10 boilers were evaluated with the optimized model. Results show that the model is able to avoid over-fitting of training sample sets, and it has relatively strong generalization capability. The prediction accuracy with FCM-SVM model used in this experiment is 100%, indicating high prediction accuracy of boiler slagging characteristics.