Selection of oxygen carrier for chemical looping combustion of natural gas and syngas fuels – A machine learning approach

This research focuses on selecting suitable oxygen carriers (OCs) using data driven modeling in order to prevent operational issues such as agglomeration, attrition, and sintering, which are challenges in chemical looping combustion (CLC) operations. The complexity of choosing effective OCs arises from the diverse compositions of natural ores and synthetic compounds used in the process. In this work, eight machine learning techniques were employed to predict the performance of oxygen carriers using a parameter known as gas yield under different operating temperatures for gaseous fuels primarily natural gas and syngas. A comprehensive dataset including experimental data from the literature for various carriers were used to train multiple machine learning models. The models predicted gas yield with knowledge of reactor operating temperature, fuel composition, and the elemental makeup of oxygen carriers. Cross-validation and bootstrap techniques were employed to ensure model robustness and minimize prediction error. The results demonstrate that the GBR and CatBoost have been the best-performing model achieving a high coefficient of determination 0.820 and 0.822 value respectively and same low mean error value of 0.015. It was observed that Fe and Mn based mixed oxide performed as good OCs with their reactivity increasing with Fe to Mn ratio. This study highlights the potential of machine learning in optimizing oxygen carrier performance and accelerating advancements in CLC technology.