Support Vector Machine-Based Prediction for Mercury Speciation in Combustion Flue Gases

Mercury emission from coal combustion has become a global environmental problem. In order to accurately reveal the complexly non-linear relationships between mercury emissions characteristics in flue gas and boiler type as well as coal properties, a advanced artificial intelligence (AI) regression models, Support Vector Machine (SVM), are developed and employed to simulate the mercury speciation (elemental, oxidized and particulate) and concentration in flue gases from coal combustion. Based on the normalization method and random sampling method for dataset, and the optimized search technique with 10-folds cross validation for determining algorithm parameters, the configured SVM model are trained and tested by simulated results. Model performance is evaluated according to predicted accuracy and generalized capability. As a result, it is found that, for the ratio of training and testing sample size being equal to 80%:20%, the SVM is able to provide good prediction performances with the mean squared error of 0.0095 with correlation coefficient of 0.9164 on comparison with the experimental data.