Prediction of ash fusion temperatures of municipal solid waste incinerator ash based on support vector regression

Abstract

The ash fusion temperatures (AFTs) of the municipal solid waste incinerator (MSWI) ash and slags is the main indicator to judge its slagging characteristics. Support vector regression (SVR) algorithm is more appropriate for creating an AFT prediction model than the classic regression method because it can adjust to a wider range of ash sample types in a wider range of areas by avoiding underfitting and overfitting. In this paper, SVR prediction models for the deformation temperature (IDT), softening temperature (ST), hemisphere temperature (HT), and flow temperature (FT) of ash are constructed using the ash chemical composition and AFTs dataset of coal ash and biomass ash and other ash. The AFTs of fly ash, bottom slag, and other ash in the municipal solid waste incinerator are predicted. With average relative errors of 1.98 % and 3.56 % and maximum relative errors of 5.21 % and 5.2 %, respectively, the results demonstrate that ST and FT have greater prediction accuracy. IDT and HT both provide poor predictions, with maximum relative errors of 14.08 % and 9.36 %, respectively, and average relative errors of 10.06 % and 5.95 %, respectively. Additionally, the impact of removing various numbers of outlier samples from training data on prediction accuracy, and the impact of test sample measures, such as ash composition analysis and AFTs determination, on prediction accuracy are analyzed and examined.