Design of High Voltage and Capacity Electrode for Lithium-ion Battery using Machine Learning

Electric vehicles (EVs) present a possible answer in the backdrop of a growing increase in greenhouse gas emissions during the past five decades, primarily due to industries, growing populations and the transport sector. The need for EVs with high ranges necessitates the development of advanced batteries with high capacity and voltage. Conventional approaches, including experimental methods and Density Functional Theory, are accurate but time and resource-intensive. This study leverages machine learning (ML) techniques to supplement these conventional approaches, aiming to rapidly identify high-voltage, high-capacity lithium-ion battery materials crucial for next-generation batteries. Initially, the Materials Project dataset was reduced from 4000 to 2360 instances, setting the stage for enhanced exploration. Applying an Artificial Neural Network model for voltage prediction and clustering algorithms for optimal range identification, the methodology enabled a significant expansion of the original battery database from 2360 to 52000 instances. Within this expanded database, the model revealed 83 potential battery candidates, including 23 unique and novel compounds that had not yet been explored as battery material, thus guiding the design of materials. The results demonstrate a data-driven strategy that expedites the discovery and optimization of battery materials, highlighting the potency of ML in revolutionizing electrochemical energy storage research.