Data-Driven Strategies for Next-Generation Aviation Batteries: Improving Energy Density and Cycle Stability in Electric Propulsion Systems

With the increasing adoption of battery technology in scientific and industrial applications, fully electric and hybridelectric propulsion aircraft are becoming viable. However, the relatively low energy density of current battery materials compared to traditional fuels limits their potential. To address this, we propose an innovative prediction model using the random forest machine learning algorithm. By integrating key physicochemical properties (e.g., crystal structure, elemental composition, conductivity) as input features, our model accurately predicts critical performance metrics such as specific capacity, cycle stability, and volume expansion rate. Compared to traditional experimental methods, our approach significantly improves prediction efficiency and reduces development costs. Validation shows that the random forest model achieves an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R}^{\mathbf{2}}$</tex> of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{0. 9 5}$</tex>, demonstrating its effectiveness in feature importance analysis and nonlinear relationship modeling. This enhances both prediction accuracy and understanding of complex datasets. Our data-driven methodology accelerates the screening and optimization of battery materials, promoting the development of high-performance energy storage solutions and advancing electric propulsion technologies. This work paves the way for future advancements in aviation and other high-performance batterydependent fields.