Comprehensive Analytics for Massive and Diverse Li-Ion Battery Aging Datasets

In the world’s recent push to electrify transportation, the adoption of lithium-ion batteries (LiBs) in vehicles emerges as one of the key enablers for emission-free modes of transportation. As with many complex systems, Lithium Ion Batteries (LiBs) alike, the availability of large, high quality datasets with rich information allows for the development of diagnostic and predictive tools that aid their adoption and performance. Tracking the internal state of health (SOH) metrics of a LiB, such as the electrode-specific composition, lithium inventory, and impedance over the course of its lifetime, is central to understanding degradation and failure prediction. In normal operation, such information is often not directly accessible, and largely hidden behind nonlinear dynamics, device variability and diverse operating conditions. By combining electrochemical aging cycles with periodic diagnostic check-up cycles, we generate an extensive battery degradation dataset containing both thermodynamic and kinetic features directly tied to the battery’s SOH throughout its lifetime. The data consists of nearly 500 high energy density commercial NCA/Gr + SiOx cylindrical cells taken from the same electric vehicle, cycled under almost 250 unique cycling protocols with 30 protocols containing more than 4 repeats. With this methodology and data we are able to make insights on initial variation, dominant degradation modes, and relationships between cycling conditions and SOH metrics of cells. We present the experimental results of this study as a benchmarking dataset for both academics and industry.