Blend Coal Ash Softening Temperature (ST) Model Building and Application

Abstract

Many power plants use blend coals to replace design coal for burning due to economic reasons. These blend coals are usually blended by two types of off-design coal, so there is a need to select the off-design coals and to optimize the blend ratio. In this process, an important problem is how to estimate the properties of the blend coal, especially for the softening temperature (ST) of coal ash. To deal with this issue, this article combined support vector machines (SVM) and genetic algorithms (GA) to predict coal ash's softening temperature (ST) from a given ash composition. An SVM model was built for predicting the ST of coal ash, and its parameters were optimized using GA. The SVM model was trained and verified using experimental data from 12 coal samples and subsequently was tested using data from 17 blended coal samples. The results confirmed the validity of the SVM model, which accurately predicted the ST for all these coal samples. Based on the SVM model, a coal-blending system was developed for a given 300 MW power plant boiler. The blend coals designed by this system met the firing requirements of the boiler and achieved high thermal efficiency.