Accelerating Battery Material Discovery: Integrating Computational Workflows and Machine Learning
Juan Maria García Lastra
DTU Energy. Danmarks Tekniske Universitet
The need for new materials, particularly electrodes and electrolytes, to enhance battery performance is critical for advancing energy storage technologies. However, relying solely on experimental methods to discover these materials is an overwhelming and time-consuming endeavor. Computational simulations offer a complementary approach by efficiently calculating properties like material stability through first-principles methods. Despite this, certain key properties, such as ionic diffusivity, remain challenging to predict.
To address this, computational workflows are essential in managing simulations that integrate both straightforward and complex property calculations. The vast space of potential materials for battery components further complicates this task, making the search virtually infinite. Consequently, integrating machine learning (ML) algorithms into these workflows is necessary to accelerate material discovery. Two primary categories of ML algorithms are valuable in this context: those that enhance the efficiency of simulations, enabling the exploration of larger time and length scales, and more advanced conditional generative models that can guide the search for materials in specific directions.
At DTU, we follow this approach by developing and implementing computational workflows incorporating ML algorithms to streamline the discovery of novel battery materials. In this talk, I will present our progress in integrating these techniques to advance energy storage solutions.