Machine learning towards automated discovery of organic molecules as active materials in non-aqueous redox flow batteries
Loading...
Date
2024-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of New Brunswick
Abstract
With increasing demand for energy and the resources needed to provide this energy, redox flow batteries (RFBs) have shown potential as large-scale electrochemical energy storage systems. In this project, interest lies in the automated discovery of organic redox-active materials that undergo both oxidation and reduction reactions for symmetric non-aqueous RFBs. Machine learning methods were applied to automate the generation of organic molecules followed by the application of a genetic algorithm (GA) to improve the generated population.
A set of molecules were constructed through a series of random choices under set structural parameters. Multiple GA generations were run on a selected population where two randomly chosen molecules combine their structural features to generate new molecules. All molecules were characterized computationally to determine their cell potential, stability, and solubility values used to assess their capabilities as redox-active materials. A set of top-ranking molecules have been proposed as potential candidates for non-aqueous RFBs.