Clustering LS-SVM models for the prediction of unburned carbon content in fly ash

This paper investigates factors that influences the unburned carbon content in fly ash and selects the optical factors from the original characteristics by means of minimal-redundancy-maximal-relevance criterion (mRMR). And on this basis, this paper proposes a novel model called clustering least squares support vector machine (CLS-SVM) to predict the unburned carbon content in fly ash. In this CLS-SVM model, a fuzzy c-means cluster algorithm (FCM) is adopted to decompose the original data into three different sub data sets. Taking advantage of both theory of clustering algorithm and advanced statistical learning methodology, CLS-SVM models are built specifically for each different sub data sets. Then the CLS-SVM models are developed to predict the key parameter - unburned carbon content, which is verified through operation data of a 300MW generating unit.