Otimização de transportadores de oxigênio à base de manganês para recirculação química através de análise bibliométrica, aprendizado de máquinas e validação experimental

Abstract

Anthropogenic emissions of CO2 into the Earth's atmosphere, mainly by thermal power generation industries and means of transportation, are the main causes of global warming. As such, carbon capture and storage (CCS) technologies have emerged as a very interesting alternative, since they use fossil fuels to generate energy with low or zero CO2 emissions into the atmosphere. Chemical Looping (CL) technologies have therefore emerged as promising options for the thermal energy generation sectors, since they can capture CO2 with a low energy penalty and low cost. However, for these technologies to perform properly, the oxygen carrier (OC) must be chosen appropriately. Manganese-based oxygen carriers are very interesting because they are environmentally safe and low cost. In this way, this master's thesis aims to determine optimized manganese-based oxygen carriers for application in chemical recirculation processes using bibliometric analysis, machine learning algorithms and an experimental validation. VOSviewer, Excel and Web of Science were used to carry out the bibliometric analysis of the 65 articles in the portfolio selected using the Proknow-C method. The Random Forest and XGBoost polynomial regression algorithms were used to process the data from the portfolio articles using machine learning. In the experimental validation, the microwave combustion method was used to synthesize only one OC indicated by machine learning, due to the short time to carry out the experimental part. XRD, SEM, EDS and TG analyses were carried out to characterize the OC. The bibliometric analysis contributed to the optimization of the physicochemical properties of oxygen carriers, since it considered the influence of the type of active phase and the support in reactivity tests, oxygen transport capacity, friction rate and agglomeration in continuous fluidized bed reactors in CL processes. With regard to the processing of the data in Machine Learning, it was found that the data relating to Oxygen Carrying Capacity (Roc) fitted the Random Forest and XGBoost regression models very well, with highly accurate predictions, so it was possible to determine three oxygen carriers that have good oxygen carrying capacity and good mechanical resistance: OC_MnFe, OC_MnFeTi and OC_MnMg. Based on OC_MnMg, sample S-MgMn was synthesized, with active phase MgMn2O4 and inert phase MgO. It was found that the sample's Roc values were 7.95 wt% in the first redox cycle, 8.09 wt% in the second redox cycle and 8.14 wt% in the third redox cycle, falling within the Roc range stipulated by Machine Learning for Chemical Looping Combustion processes (Roc between 6.5 and 8.2 wt%). Finally, the sample was very reactive with H2 and air and could be considered a promising material for application in Chemical Looping Combustion technology and its reactivity with other types of fuel (e.g. CH4, CO, coal) should be studied.