Deconstruction Analysis and Regeneration Optimization of Solid Electrolyte Components for Energy Storage Based on the Finite Data Statistics

Solid-state electrolytes have a superior theoretical energy density, higher safety, and longer cycle stability. However, researching solid-state electrolyte materials with high performance remains difficult. Machine learning methods can accurately predict performance, speed up the screening process, reduce expensive and time-consuming experimental trials, and have significant advantages in the optimization and fabrication of solid-state electrolyte materials. This research constructed a solid-state electrolyte database based on existing data, trained a random forest (RF) algorithm model to achieve performance prediction of Na superionic conductor (NASICON) solid-state electrolytes by using only the component information, and analyzed the effect of different components. The optimization results screened and provided the optimal ratio of 61 doping elements, and meanwhile verified the accuracy and optimization of the model prediction results. The effects of different components on the ionic conductivity of solid-state electrolytes, including the descriptors used and types of doping elements, were distinguished. This study provides a reference for the optimization direction of solid-state electrolyte materials and the selection of components, which contribute to the exploration and development of high-performance solid-state electrolyte materials.