Flue gas enhanced water leaching: Prediction and condition optimization of demineralization effect from coal via machine learning algorithm

Abstract

A novel “Flue Gas Enhanced Water Leaching” method was previously proposed to efficiently demineralize high alkali and alkaline earth metals coals. Yet the optimized treatment conditions are unclear in consideration of the balance between removal effect and costs. Consequently, the back propagation neural network algorithm was proposed in this study to analyze the effect of characteristic parameters (particle size, ultimate analysis, proximate analysis and ash compositions) of coals and operational conditions (temperature and time) on the removal rates of Na (RNa) and Ca (RCa). Results showed that the back propagation neural network models can precisely predict RNa and RCa with respective correlation coefficient (R2) of 0.9854 and 0.9777. The coal with higher oxygen content tends to have higher RNa and RCa owing to their high organic Na/Ca contents and good hydrophilicity, and minimizing coal particles contributes to the removal of Na/Ca. Longer leaching time results in higher RNa and RCa, while higher leaching temperature contributes to RNa but shows complex effect on RCa. For practical considerations, recommended operational conditions are room temperature, within 2 h and particle size of smaller than 2 mm, and the RNa and RCa of higher than 75% and 30% can be obtained under these conditions, respectively. Therefore, this study not only provides an effective method for the prediction of RNa and RCa under “Flue Gas Enhanced Water Leaching” treatment, but also contributes to the condition optimization in its industrial applications.