Multiobjective Artificial Intelligence-Assisted High-Throughput Screening and Prioritization of Gel Polymer Electrolytes for Safer Lithium Batteries

Abstract

Gel polymer electrolytes (GPEs) combine liquid-like interfacial contact with polymer-mediated confinement, but their multicomponent formulation space makes simultaneous control of ion transport, electrochemical stability, and thermal safety difficult. This work presents a multiobjective artificial intelligence-assisted framework for computational screening and prioritization of GPE formulation motifs. A literature-informed collection of 186 formulations is encoded using chemistry-aware descriptors and assessed using Random Forest, gradient-boosting, and neural-regression model families. The resulting surrogate framework is used to screen 4320 class-level virtual formulations against ionic conductivity, Li+ transport tendency, electrochemical stability window, thermal stability index, composite safety index, and mechanical-robustness tendency. The analysis prioritizes five formulation motifs occupying balanced regions of conductivity–safety space and identifies liquid mobility and salt dissociation as major transport descriptors, whereas flame-retardant functionality, host thermal robustness, and filler reinforcement dominate the composite-safety response. Held-out parity analysis gives R2 values of 0.93, 0.90, 0.89, and 0.87 for conductivity, electrochemical stability window, thermal stability index, and composite safety index, respectively. These values are internal model-performance estimates, and the shortlisted motifs are not experimentally validated recipes. The framework is therefore intended as a proof-of-concept preselection tool that narrows broad GPE formulation spaces before synthesis, thermal analysis, flame-retardancy testing, and cell-level validation.