CAREER: An Ecologically Inspired Approach to Battery Lifetime Analysis and Testing

随着运输和电网应用对电池的依赖性增加，与电池操作和对个体上下文环境的老化依赖性相关的挑战仍然存在。这是一个特别相关的问题，因为电池在每个应用中执行多个任务（例如，驾驶、充电、电网服务等）。这会以不同的方式导致其老化。此外，电池不仅在单个应用中执行多个任务，而且作为第二寿命电池迁移到第二应用。该CAREER提案旨在将电池老化动态理解为上下文相关，并提供统一的理论和建模，可以将上下文事件和寿命与电池和模块老化事件联系起来。这将有利于所有电池应用和新兴的电池再利用领域，提供切实可行的方法来改善电池测试、评估和管理。作为教育部分，该项目将为本科生和研究生提供新的实践分布式实验室能力，以探索电网和车辆应用背景下的电池技术。外展活动包括通过密歇根大学和大学合作伙伴关系接待西班牙裔女学生，并参加西班牙裔专业工程师协会的指导和会议。电池受到高度不确定的情况下，取决于他们的上下文，目前电池模块和包装的变化，由于其空间和功能分布以及不同规模的不同监测能力。本CAREER提案将考虑这些条件与生态系统（如渔业、林业等）具有可比性，并且电池寿命和老化应在生态方法和方法下解决多尺度和多寿命问题。为此，测试方法将包括低成本的大规模分布式测试，这些测试将在不同的环境变化和不同的生活中对电池的个体和群体进行实验性探测。来自这些测试的数据将用于开发概率推理网络，以将电池老化的因果关系联系起来，并将提供建立跨寿命监测和数据需求的能力。为了制定电池老化和寿命模型，该提案将重点研究由电池老化引起的种内性状变异（物种内部的变异）。为此，将通过建立斑块等级来识别组合群体，以在个体、亚群体和群体水平上识别每个斑块的种内性状的结构和功能分布。将使用基于个体的混合模型和在生态系统中用于模拟种群变化的积分投影模型对每个斑块的种内性状进行建模。这些方法将提供一个整个人群的概率模型。然而，由于电池数量是以不同的规模进行监测的（电池组和电池稀疏地取决于技术），因此模型将考虑不完整的数据可用性并开发比例阶梯。这些阶梯将把种内性状模型从个体扩展到群体，反之亦然，以适应不同的数据可用性和混合方法。这些模型将在电池管理系统中实现，以从收集的数据中学习特征模型。还将制定和部署特质过滤器，以识别和建模内部和外部因素，这些因素将决定每个生命的特质变化。所获得的模型和基于生态学的理论将通过大规模的人口测试和真实的电动汽车和电网规模的电池部署进行实验验证。

As transportation and grid applications increase their dependency on batteries, challenges related to battery operation and aging dependency on the individual context circumstances remain. This is a particularly relevant problem as batteries perform multiple tasks in each application (e.g. driving, recharging, grid services, etc.) which can contribute to its aging differently. Furthermore, batteries not only perform multiple tasks in a single application, but migrate to a second application as a second life battery. This CAREER proposal aims to understand battery aging dynamics as context-dependent and to provide a unified theory and modeling that can link context events and lives with cell and module aging events. This will benefit all battery applications and the emerging battery repurposing sector by providing tangible methods to improve battery testing, estimation and management. As educational components, this project will propose new hands-on distributed laboratory capabilities for undergraduate and graduate students to explore battery technologies in the context of grid and vehicle applications. Outreach includes hosting female Hispanic students through the Michigan College and University Partnership, and also participating in the Society for Hispanic Professional Engineering mentoring and conferences.Batteries are subjected to highly uncertain scenarios depending on their context, present cell to module and pack variations due to its space and function distribution and different monitoring capabilities at different scales. This CAREER proposal will consider that these conditions are comparable to ecological systems, such as fishery, forestry, etc. and that battery lifetime and aging should tackle the multi-scale and multi-life problem under ecological approaches and methods. For this, testing methods will include low-cost large-scale distributed testing that will experimentally probe individuals and populations of batteries across context variations and different lives. Data from these tests will be used to develop probabilistic reasoning networks to link causality for battery aging and will provide the ability to establish monitoring and data needs across lives. To formulate the battery aging and life modeling, the proposal will focus on studying intraspecific trait variations (variations inside a species) that arise from battery aging. For this purpose, populations of batteries will be identified through the establishment of a patch hierarchy to identify the structure and functional distribution of intraspecific trait per patch at the individual, sub-population and population level. The intraspecific traits will be modeled for each patch using individual-based, mixed models and integral projection models that are used in ecological systems to model population variations. These approaches will provide a probabilistic model across the population. However, as battery populations are monitored at different scales (pack and cells sparsely depending on the technology), the models will consider incomplete data availability and develop scaling ladders. These ladders will scale the intraspecific trait models from individuals to populations and vice versa to adapt to different data availability and mixed approaches. These models will be implemented in battery management systems to learn the traits models from scavenged data. Trait filters will also be formulated and deployed to identify and model internal and external factors that will determine the trait variations for each life. The models and ecology-based theory obtained will be experimentally validated through the large-scale population testing and real electric vehicle and grid-scale battery deployments.