Atmosphere-driven mechanisms in biogas residue chemical looping pyrolysis: Insights from kinetic characteristic and machine learning prediction: Insights from kinetic characteristic and machine learning prediction

Chemical looping pyrolysis (CLPy) is a promising technology for the high-value utilization of biogas residue (BR). In this study, the effects of atmosphere condition ( N₂ and CO₂), heating rates, and oxygen carriers (OCs) loading of CLPy characteristics were investigated by thermogravimetric (TG) and derivative thermogravimetric (DTG) analyses. The master plots method was employed to determine kinetic parameters and identify the most appropriate reaction model for the CLPy process of BR. Additionally, machine learning techniques, including artificial neural networks (ANN), decision tree regression, and support vector regression, were employed to predict the TG curves. Results indicate that the TG curves in the CLPy process exhibit unique distribution characteristics. Compared to traditional pyrolysis methods, the TG curves in the CLPy process display multiple stages of reaction rate changes, with significant shifts in the onset and peak temperatures due to the involvement of the OCs. At temperatures above 600 °C, the DTG curves with OCs addition show a pronounced weight loss peak under N₂ atmosphere, which intensifies with increasing heating rate. In contrast, the GLPy characteristics of BR are suppressed under CO₂ atmosphere, with no clear weight loss observed. Deconvolution of overlapping peaks further clarified the individual reaction mechanisms, and kinetic model revealed that the D3 and D4 diffusion models closely align with experimental data. The ANN model demonstrated the highest predictive accuracy (R² = 0.999, RMSE = 0.391) for forecasting TG curves. These findings provide valuable insights into optimizing CLPy processes and underscore the potential of advanced predictive model to enhance pyrolysis efficiency under varied atmospheric conditions.