Lithium-ion battery state of health estimation

电动汽车(EVS)在减少移动部门的碳足迹方面发挥着关键作用。然而，它们的高前期成本、有限的续航里程和缓慢的充电速度是提高电动汽车使用量的障碍。降低成本和改进电动汽车锂离子(Li-ion)电池可以减少这些障碍。然而，人们对锂离子电池的安全运行和降解率的了解有限。这在很大程度上是由于人们对其复杂的电化学机制还没有很好的了解。此外，较大的工作范围(温度、充电速度等)在其生命周期内，需要资源密集型测试来将半经验模型参数化。因此，电池的操作非常保守，导致电池尺寸过大，操作条件不佳，导致效率低下和成本较高。本PHD旨在提供锂离子电池的最佳测试策略和准确的模型，以便为更有效的操作策略(如快速充电)提供信息。这将通过先进的实验设计(DOE)、建模和机器学习相结合来实现。PHD的核心将集中在评估不同工作范围内电池健康状态的方法上。PHD的最初部分将专注于建立一个基于数据驱动神经网络的模型结构。然后将使用文献中现有的电池数据和工业合作伙伴提供的数据来评估该模型的准确性。然后将设计和实施一项试验性测试活动，试图有效地对电池模型进行参数化。得到的电池模型将为提高电池的安全运行范围提供重要信息。更有效的测试方法和电池退化的准确建模将加快电动汽车产品设计阶段的开发时间。它还将促进开发新的电池架构和更有效的操作方案(例如，改进的快速充电策略)。这两项发展都将降低电动汽车电池的总体成本。这可能会降低电动汽车的总体成本，加速消费者对电动汽车的接受，从而减少移动部门的整体碳影响。

Electric vehicles (EVs) play a key role in decreasing the carbon footprint of the mobility sector. Their high upfront cost, limited range and slow charging speed are however a barrier to increased EV uptake. Reducing the cost and improving the EV Lithium-ion (Li-ion) battery could reduce these barriers. There is, however, limited knowledge in the safe operation and degradation rate of Li-ion batteries. This is largely due to the complex electrochemical mechanisms not being well understood. Furthermore, the large operating envelope (temperature, charging speed etc.) over its lifetime require resource intensive testing to parameterize semi-empirical models. The battery is therefore operated very conservatively, resulting in oversizing the battery and sub-optimal operating conditions resulting in inefficiencies and higher costs.This PhD aims to provide optimal testing strategies and accurate modelling of Li-ion batteries in order to provide information to facilitate more efficient operating strategies (e.g. fast charging). This will be achieved by a combination of advanced design of experiments (DOE), modelling and machine learning. The core of the PhD, will focus on methods to estimate the state of health of the battery cells over various operating ranges. The initial part of the PhD will focus on building a model structure which is based on a data driven neural network. The accuracy of this model will then be assessed using existing battery data in literature and data provided by the industrial partner. An experimental test campaign will then be designed and implemented, in an attempt to efficiently parameterize the battery models. The resultant battery models would then provide important information to improve the safe operation range of the battery.More efficient testing methods and accurate modelling of battery degradation will speed up development time in the design phase of the Electric Vehicle production. It will also facilitate development of new battery architectures and more efficient operating schemes (e.g. improved fast charging strategies). Both of these developments would decrease the overall cost of batteries in EVs. This could decrease overall EV cost and accelerate their uptake by consumers, thereby reducing the overall carbon impact of the mobility sector.