Optimizing dielectric properties of corannulene nanomaterial for enhanced performance of next-generation electric vehicle batteries: A Machine learning and Nelder-Mead optimization approach

Abstract

Enhancing the performance of next-generation electric vehicle (EV) batteries relies on advancements in energy storage technologies. This study explores the potential of corannulene, a nanomaterial renowned for its exceptional electronic properties, to optimize battery performance by fine-tuning its dielectric properties. Traditionally, such optimization methods have been limited to conventional approaches. However, this research adopts a novel machine learning methodology utilizing an experimental dataset encompassing five key features: “frequency,” “real part,” “imaginary part,” “dielectric strength,” and “dielectric loss.” These features are employed to train various machine learning models, including baseline, ensemble-tree, and boosting techniques, with the aim of predicting optimal values for dielectric loss (to minimize negative impact) and dielectric strength (to maximize efficiency). Subsequently, a Nelder-Mead optimization algorithm, coupled with a stacking model, is employed to determine the optimal range for these features, thereby enhancing battery performance. Promising results are obtained, with Ensemble Tree Regression (ETR) and Stacking models achieving remarkable R² scores of 0.9973 and 0.9995 for predicting dielectric loss and dielectric strength, respectively. The Nelder-Mead optimization, guided by these machine learning models, effectively recommends optimal ranges for corannulene nanomaterial properties. Notably, ETR and Stacking based optimization outperform other models. This integrated machine learning and optimization approach represents a significant step toward designing not only more efficient but also more sustainable EV batteries, thereby accelerating the automotive industry’s transition to a greener future.