Determining the Key Performance Factors in Lithium-Oxygen Batteries Using Machine Learning
Aysegul Kilic, Damla Eroğlu, Ramazan Yıldırım
Abstract
Lithium-oxygen (Li-O 2 ) batteries are among the most prominent alternative battery chemistries to lithium-ion batteries with their high theoretical capacities. However, attaining their high theoretical capacity is difficult due to the poor cell design and insufficient cell materials. In this study, machine learning algorithms are used to determine the effective cell design factors and the most promising materials for reaching high discharge capacities and voltages. Association rule mining (ARM) and decision tree (DT) algorithms show that bulk cathode materials, especially N-doped carbons, graphene and porous carbons, are beneficial for achieving high performances. Moreover, ARM analysis indicates that cathode ingredients, namely LaFe oxides and Ni oxides, should be utilized for high discharge capacities. In addition, the choice of the electrolyte solvent seems to be highly influential on the discharge capacities. Dimethyl sulfoxide (DMSO) is shown to be one of the best options for high cell voltages and discharge capacities.