This paper introduces a novel on-line monitoring performance method of coal-fired power unit. Support vector machine (SVM) is used to predict the unburned carbon content of fly ash in the boiler and the exhaust steam enthalpy in turbine, which are two difficulties in the real time economic performance calculation model in coal-fired power plant. Comparison between the output of SVM modeling and the experimental data shows a good agreement, and compared with conventional artificial neural network techniques, SVM can achieve better accuracy and generalization. This presented monitoring method is proven by the results of application cases in a practical coal-fired power plant.