Combustion Rate of Biomass Char under Low Oxygen Concentration and Its Machine Learning-based Prediction

The combustion rate of palm kernel shell (PKS) char was measured under 3–5 vol% low oxygen concentration. The combustion rate increased with rising ash content, temperature, and oxygen concentration. Demineralizing the PKS char with hydrochloric acid resulted in a reduced combustion rate, while loading the PKS with 1.36 wt% Fe2O3 led to an increase. The physical properties (proximate analysis values, ultimate analysis values, and ash content) of the PKS, along with combustion conditions (combustion temperature and oxygen concentration), were used as explanatory variables, and the combustion rate, as the response variable, was predicted through machine learning across 5 model types. In the Gaussian process regression model (squared exponential kernel function), the root mean square, mean square, and mean absolute errors were closest to zero, while the coefficient of determination (R2) was closest to one. This study accurately predicted the combustion rate in the rate-determining step of a chemical reaction at low temperatures using the Gaussian process regression model (squared exponential). In addition, it was found that among the explanatory variables, the combustion temperature, volatile matter, oxygen concentration, Fe2O3 content, hydrogen content, and carbon content had a large effect on the combustion rate.