Statistical and Machine Learning-Based Durability-Testing Strategies for Energy Storage

There is considerable interest in developing new energy storage<br/>technologies for the electric grid, but economic viability will require<br/>that manufacturers provide warranties guaranteeing 15+ years of<br/>life. Although there are extensive efforts to make early predictions<br/>for the expected life of new storage technologies, we argue here<br/>that for the purposes of pricing warranties and valuing second-life<br/>potential—considerations that are crucial to whether the technologies<br/>can be commercialized—the full failure probability distribution,<br/>not just the expected life, is required. We use published battery<br/>cycle-life data to suggest efficient statistical and machine learningbased<br/>testing and analysis strategies that can rapidly estimate and<br/>also take advantage of the failure probability distribution. One<br/>approach is a Weibull analysis, which can (1) reduce the number of<br/>testing machine hours required for setting a warranty, (2) quickly<br/>determine whether a new technology is better than a baseline technology,<br/>and (3) estimate the maximum intensity of testing acceleration<br/>that does not change the failure mode. A second approach is<br/>driven by the idea that all measured data—such as capacity or energy<br/>as a function of time or cycle number—are valuable and generated<br/>by an underlying latent function. This analysis employs a Gaussian<br/>process to find the underlying latent function, together with its uncertainties,<br/>which can be used to calculate the failure distribution.