Paper published: Closed-loop optimization of fast-charging protocols for batteries with machine learning
I’m excited to announce that our paper (co-led with Aditya Grover) on using machine learning to optimize battery fast charging was recently published in Nature! This work has been featured by Stanford News and the Nature podcast.
Many battery development challenges are slow due to both the length and the number of required experiments. The above video highlights our approach as applied to battery fast charging. To reduce the length of each experiment, we use the early prediction approach from our previous work. To reduce the number of required experiments, we use Bayesian optimization, which helps us intelligently choose the next protocols to test. This approach allowed us to efficiently search over 224 charging protocols to find those with high lifetime.
I’m excited by this work because it illustrates how machine learning methods can dramatically reduce testing times; optimistically, machine learning can speed up the scientific method itself. I’m convinced that exciting times are ahead for the burgeoning AI + science field!