The optimization of the MgO/MgCO3 decarbonation process and machine learning-based improved reactor design approach

MgO/MgCO3 reaction system is considered as a promising candidate of thermochemical heat storage. Most relevant studies focus on the improving the material chemical reaction performance. The reaction process inside the reactor has been less studied. In this study, a comprehensive multi-physics model is developed to analyze the decarbonization process of the MgO/MgCO3 reaction system. Impacts of various factors on the reaction process, including the injected energy flux density, reactor height and configuration of the thermal conductive plate (represented by thermal conductive area ratio) were analyzed. The heat transfer process is identified as the primary hindrance. By increasing the thermal conductive area ratio from 0.085 to 1.16, overall reaction time is decreased by 77 % and temperature difference from the top to bottom is decreased by 86 K. Additionally, a prediction model is developed using machine learning method and the prediction error is less than 5 %. Based on the predictive model, a design guideline was further developed so that various critical factors in reactor design could be explored and optimal choices could be obtained without consuming excessive computational resources. The identified influence laws of key factors and the proposed machine learning-based reactor design guideline may provide valuable insights for future thermochemical reactor designs.