O-14　ガス化炉内クリンカの形成抑制に関する深層学習を用いた考察

抄録

This study investigated modelling the relationship between the chemical composition of Japanese cedar ash and its fusion temperature using a neural network (NN) for the stable operation of small-scale biomass gasification combined heat and power (CHP) systems. The NN was trained on experimental data and demonstrated high accuracy in predicting ash fusion temperatures, excluding the softening point. Analysis confirmed that the CaCO₃ and K₂CO₃ ratio has a significant impact on the melting point. Adjusting ash composition to achieve optimal ratios could effectively suppress slag formation and enhance the stability of system operation, with potential applicability to other tree species. These findings suggest that NN-based prediction models can serve as a valuable tool in optimising fuel composition.