The low-temperature operation of lithium-ion batteries (LIBs) is a challenge in achieving high-stability battery technology. Moreover, the design and analysis of low-temperature electrolytes are impeded by the limited understanding of various solvent components and their combinations. In this study, we present a data-driven strategy to design electrolytes with high ionic conductivity at low temperature using various machine-learning algorithms, such as random forest and feedforward neural networks. To establish a link between prediction of electrolyte chemistry and cell performance of LIBs, we performed parameter-free molecular dynamics (MD) prediction of various salt concentrations and temperatures for target solvents. Finally, electrochemical modeling was performed using these properties as the required material parameters. Combining works of the fully parameterized Newman models, parameter-free MD, and data-driven prediction of electrolyte chemistry can help measure the discharge voltage of batteries and enable in silico engineering of electrolyte development for realizing low-temperature operation of LIBs.

Low Temperature Electrolyte Machine Learning Multi-scale Simulation Molecular Dynamics Ion-conductivity