Machine learning classification methods for battery cathode material crystal system

Abstract

The optimization of lithium-ion battery material design is becoming increasingly pivotal in the context of the transition to new energy sources. This research delves into the classification of the crystal structures of 339 sets of lithium silicate cathode materials, depending on the chemical and physical parameters of the materials to classify the crystal system type. Five classification algorithms were employed in this study, and the classification accuracy was enhanced through algorithm optimization. Notably, the accuracy of the random forest model with PCA (Principal component analysis) =5 reached 74% accuracy with 80% of the data used for training. Simultaneously, this study optimized the model accuracy by fine-tuning parameters such as depth, k-fold, and learning rate. Through this research, the classification accuracy of cathode materials in batteries was further elevated, bearing significant implications for the development of new, suitable battery materials. This research aids in expediting the screening process for researchers and facilitating industrial-scale experimentation, thereby accelerating the application of next-generation batteries.