Keywords: 
• support vector machine (SVM) /
• multi-model /
• carbon content of fly ash /
• soft sensor

Abstract: In view of the present fly ash carbon content measurement method for thermal power plant coal-fired boiler exists the problems such as time delay and low precision, on the basis of the boiler fly ash carbon content influencing factors' analysis, using the thought that the combination of multiple models can improve the model accuracy and robustness, we proposed a multi-model dynamic soft sensor modeling method based on support vector machine (SVM) fusion. This modeling method uses the time series data to build a model, each submodel express the estimate of the output in one condition, and each submodel's predicted output realizes the variable weighting fusion through SVM method. We did a modeling study of fly ash carbon content soft sensor, using the historical data of thermal power plant, and the result shows that the method can achieve better measurement effect.