<p>In the present study, the coal burnout during pressurized oxy-fuel combustion is analyzed by combining experiment and machine learning. According to the experimental results, it is shown that the coal conversion increases significantly with the increase of oxygen concentration, temperature and pressure. At relatively low oxygen concentration, CO₂ gasification expedites the coal consumption, so the coal burnout in O₂/N₂ environment is lower compared to that in O₂/CO₂ environment, especially at elevated temperature and pressure. As the oxygen concentration increases, the influence of CO₂ gasification weakens and on account of the CO₂ physical properties, the coal burnout in O₂/N₂ condition is higher than that in O₂/CO₂ condition. Furthermore, the impact of gasification on the conversion of coal with low reactivity is obvious at high pressure. On the basis of the experimental data, four independent machine learning algorithms and one integrated machine learning algorithm are employed to propose a prediction model of pulverized coal burnout. The Pearson correlation coefficient between the characteristic parameters (oxygen concentration, heating temperature, combustion length, and reaction atmosphere) and the output variable (coal burnout) is calculated. And then the input data are independently trained by combining four different base models to generate a prediction of the target variable (coal burnout) which is used as the input characteristics of the meta-model. Moreover, the Gradient Boosting Regression (GBR) model is selected as the meta-model of Stacking ensemble learning to construct the accurate prediction model of coal burnout. Finally, the application of SHAP analysis is employed to assess the interpretability of the forecasted outcomes for coal combustion in the experimental samples, thereby significantly enhancing the explanatory capacity of the predicted results.</p>