The ever-growing energy demand has necessitated the exploration of renewable energy resources, such as wind and solar energy, which are expected to contribute a great proportion to the total energy supply in future. However, these renewable energies are highly intermittent. To facilitate the rapid deployment of renewable energy, efficient energy storage technologies are urgently required. Redox flow batteries (RFBs) have great potential to overcome these drawbacks. Currently, in RFBs, vanadium RFB (VRFB) is the representative and commercial energy storage system. However, the rise in the cost of vanadium and the limited volumetric energy density from VRFBs have motivated researchers to develop new kinds of redox active molecules. The aqueous redox-active organic materials have recently attracted intense research attention as alternative energy materials for achieving high-energy-density, cost-effective RFBs. The organic redox species also provide ample advantages for molecular engineering to improve the physical/chemical properties and ensure electrochemical/chemical stability via simple and economical synthetic methods. For the energy density of RFB, Solubility, cell voltage and multiple numbers of electron transfer are three core parameters. Molecular design of the redox-active organic molecules can tune the solubility using molecular engineering methods. However, it is still not very clear about the relationship between these properties and the chemical structure of molecules, In this work, we present a machine learning (ML) model that can learn and predict the physicochemical property of organic molecule using Gaussian process regression method based on a molecular graph kernel. We demonstrate that our ML model can predict the physicochemical properties of organic molecules as solvation free energy and pKa at chemical accuracy. To solve the general data scarcity problem for a graph-based ML model, we propose a dimension reduction algorithm based on the distance between molecular graphs, which can be used to examine the diversity of the molecular data set. It provides a promising way to build a minimum training set to improve prediction for certain test sets where the space of molecular structures is predetermined.

Artificial intelligence based property predictions and molecule design in organic redox flow battery

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.