Physics inspired model for estimating 'cycles to failure' as a function of depth of discharge for lithium ion batteries

摘要

Estimating the life of lithium ion batteries is a longstanding issue for electric vehicles as well as energy storage applications. For grid scale storage applications, this is particularly pertinent given that the commercial viability of projects is closely correlated with the accuracy of battery degradation estimations. A large volume of literature therefore is devoted to understanding various degradation mechanisms in lithium ion batteries and developing both diagnostic and prognostic degradation models. In this context, estimating calendar aging resulting from the chemical decomposition of electrolyte solution is relatively well established. Developing a cycle life counterpart however has been more challenging. The convoluted nature of interactions between various degradation mechanisms during cycling results in complex physics-based life models; coupled with a lack of detailed battery life testing data, these models are often difficult to adopt in application, both in academic and industrial settings. To this end, empirical relations such as analogues of the Wohler curve are used to estimate the 'cycles to failure' for battery cycling with various 'Depth of discharges (DODs)'. In a previous publication, Deshpande et al. [1] proposed that at lower discharge/charge rates of battery operation, SEI cracking and reforming is a dominant mechanism of cell capacity loss for continuous cycling. These ideas are further exploited in this article to develop a pragmatic physics-based model for estimating Li-ion battery cycle life combining the effects of SEI growth and SEI cracking-reforming mechanisms. Such a model is of particular interest to grid scale energy storage applications where the operating C-rates are relatively mild and operational life is long. Beyond proposing a physics-based model to estimate 'cycles to failure' for various DODs, we also validate the model against long duration cycling datasets from the battery energy storage literature. The simplicity of the model and its adaptability for batteries with sparsely available datasets makes it highly useful.