Machine Learning for Perovskite-based Oxygen Carriers Development in Chemical Looping Hydrogen Production
Why is this research needed?
Chemical looping is a next generation technology for CO2 capture. Oxygen carriers (OCs) constitute the cornerstone of chemical looping processes. Perovskite-based oxygen carriers have recently attracted a great interest in applying in chemical looping processes due to their excellent redox properties, high oxygen mobility, thermal stability, good oxygen suppliers, and the high selectivity to synthesis gas. Especially in CLHP, using the non-stoichiometric perovskite-based OCs (e.g. La0.6Sr0.4FeO3) could overcome the equilibrium limitations of traditional water gas-shift reaction for high-purity and low-carbon H2 production (Reaction 1 and 2).
Reaction 1: H2O + [OC] = H2 + O[OC] Reaction 2: CO + O[OC] = CO2 + [OC]
The traditional process of developing the desired perovskite-based OCs is based on trial-and-error synthesis and characterisation, which is costly and time-consuming. To overcome this shortcoming, Machine-learning (ML) approaches that can learn and build a model to make predictions for future OC materials, can be applied as a powerful tool to assist in the design and screening of perovskite-based OCs for CLHP.
What is this research investigating?
The aim of this research proposal is to develop a new framework for finding promising perovskite-based oxygen carrier candidates by ML that can be used in Chemical Looping Hydrogen Production (CLHP). In short, ML algorithms trained by historical experimental data will be used to:
(i) find promising perovskite-based oxides,
(ii) determine the key parameters to affect the performance of perovskite-based oxygen carriers, and finally
(iii) optimal design of perovskite-based oxygen carriers informed by ML models.
Experimentally, candidates will be synthesized, and tested under operando conditions, and provide further guidance to the computational method. If successful, the suggested framework will find a new generation of oxygen carriers (beyond the common mono- and binary oxides frequently used) with designed properties, which will make CLHP even more promising.
In this project, a database for perovskite-based OCs, including the compositions, physical and chemical properties, mechanical strength, production methods, kinetics, oxygen transfer capacity over multiple cycles, operating conditions, reactivity with fuels and, if available, cost will be established. This data will then be fed into the ML algorithms developed by Python and in-house codes to predict the kinetics, mechanical strength, oxygen transfer capacity, and reactivity with fuels based on the remaining data as inputs. It will be possible to predict the performance of new OCs and learn the relationship between their compositions, structures, and properties by using the trained ML model that will lead to the prediction of new OCs with the greatest performance of perovskite-based oxygen carriers for CLHP.
What does the research hope to achieve?
The results from this multi-disciplinary research will benefit others in research areas including chemical looping, carbon capture, hydrogen production, metal oxides synthesis techniques, and machine learning in materials design. Investigators will learn of a new framework for screening and designing suitable metal oxides for low-carbon hydrogen production.
Furthermore, academics in chemical engineering, chemistry, materials science and artificial intelligence will develop their understanding of the redox activity of new mixed metal oxides, and the reaction kinetics of metal oxides in chemical looping hydrogen production.